Liquid Film Coating

LIQUID FILM COATING
JOIN US ON THE INTERNET VIA WWW, GOPHER, FTP OR EMAIL:

WWW: http://www.thomson.com
GOPHER: gopher.thomson.com A ^ Q f ^
FTP: ftp.thomson.com w
EMAIL: [email protected]
LIQUID FILM
COATING
Scientific principles and their
technological implications
Edited by
Stephan F. Kistler
Imation Corporation
Oakdale
Minnesota
USA
and
Peter M. Schweizer
ILFORD AG
Fribourg
Switzerland
SPRINGER-SCIENCE+BUSINESS MEDIA, B.V.

First edition 1997
© 1997 Springer Science+Business Media Dordrecht

Originally published by Chapman & Hall in 1997
Typeset in 10/12pt Times by AFS Image Setters Ltd, Glasgow
ISBN 978-94-010-6246-6 ISBN 978-94-011-5342-3 (eBook)

DOI 10.1007/978-94-011-5342-3
Apart from any fair dealing for the purposes of research or private
study, or criticism or review, as permitted under the U K Copyright
Designs and Patents Act, 1988, this publication may not be reproduced,
stored, or transmitted, in any form or by any means, without the prior
permission in writing of the publishers, or in the case of reprographic
reproduction only in accordance with the terms of the licences issued
by the Copyright Licensing Agency in the U K , or in accordance with
the terms of licences issued by the appropriate Reproduction Rights
Organization outside the U K . Enquiries concerning reproduction
outside the terms stated here should be sent to the publishers at the
London address printed on this page.
The publisher makes no representation, express or implied, with
regard to the accuracy of the information contained in this book and
cannot accept any legal responsibility or liability for any errors or
omissions that may be made.
A catalogue record for this book is available from the British Library
Library of Congress Catalog Card Number: 96-86101
® Printed on permanent acid-free text paper, manufactured in

accordance with ANSI/NISO Z39.48-1992 (Permanence of Paper).
CONTENTS
Colour plates appear between pages 228 and 229
List of contributors vii
Notation ix
Introduction 1
1. Coating science and technology: an overview 3

Stephan F. Kistler and Peter M. Schweizer
Part One: Physics and material interactions of coating processes 17

2. Capillary hydrodynamics and interfacial phenomena 19
Steven J. Weinstein and Harvey J. Palmer
3. Wetting: static and dynamic contact lines 63

Terence D. Blake and Kenneth J. Ruschak
4. Surfactants: static and dynamic surface tension 99

Yves-M. Tricot
5. Coating rheology: component influence on the rheological response and

performance of water-borne coatings in roll applications 137
J. Edward Glass and Robert K. Prud'homme
6. The fate of thin liquid films after coating 183

Haroon S. Kheshgi
Part Two: Methods of investigating coating processes 207
7. Experimental methods 209

Peter M. Schweizer
8. Asymptotic methods for the mathematical analysis of coating flows 251

Marc K. Smith
9. Advances in computational methods for free-surface flows 297

Kostas N. Christodoulou, Stephan F. Kistler and P. Randall Schunk
vi Contents
Part Three: Theory and practice of coating processes 367
10. Analysis and design of internal coating die cavities 369

Robert B. Secor
Chapter II: Premetered coating processes 399

11a Slot coating 401
Franz Durst and Hans-Giinter Wagner
11 b Slide coating 427
Jules Hens and Willy Van Abbenyen
11c Curtain coating 463
Kimiaki Miyamoto and Yoshinobu Katagiri
lId Surfactant effects in coating processes 495
P. Randall Schunk and L. E. Scriven
Chapter 12: Self-metered coating processes 537

12a Knife and roll coating 539
Dennis J. Coyle
12b Meniscus roll coating 573
Philip H. Gaskell and Michael D. Savage
12c Elastohydrodynamic coating systems 599
Ferdinand R. Pranckh and Dennis J. Coyle
12d High-speed blade coating 637
Cyrus K. Aidun and Nick G. Triantafillopoulos
13. Free-meniscus coating processes 673

P. Randall Schunk, Alan J. Hurd and C. Jeffrey Brinker
14. Spin coating 709

Ronald G. Larson and Timothy J. Rehg
15. Control and optimization of coating processes 735

Peter M. Schweizer
Index 769
CONTRIBUTORS
CYRUS K. AIDUN J. EDWARD GLASS

Georgia Institute of Technology, North Dakota State University,
Institute of Paper Science and Technology, Polymers and Coatings Department,
Atlanta, Georgia, USA. Fargo, North Dakota, USA.
JULES HENS
TERENCE D. BLAKE Agfa-Gevaert N.V.,
Kodak Limited, Research Laboratories,
Research Division, Mortsel, Belgium
Harrow, UK. (retired).
ALAN 1. HURD
C. JEFFREY BRINKER Sandia National Laboratories,
Sandia National Laboratories, Ceramics Processing Science Department,
Ceramics Synthesis and Inorganic Chemistry Albuquerque, New Mexico, USA.
Department,
Albuquerque, New Mexico, USA. YOSHINOBU KATAGIRI
Fuji Photo Film Co., Ltd.,
Production Engineering & Development
KOSTAS N. CHRISTODOULOU
Division,
E.!. Du Pont De Nemours & Co.,
Minami-ashigara-shi, Kanagawa-ken, Japan.
Scientific Computing Division,
Wilmington, Delaware, USA. HAROON S. KHESHGI
Exxon Research and Engineering Company,
DENNIS J. COYLE Corporate Research,
General Electric Co., Annandale, New Jersey, USA.
Corporate R&D,
STEPHAN F. KISTLER
Niskayuna, New York, USA.
3M Company,
Data Storage Products Division,
FRANZ DURST St. Paul, Minnesota, USA.
Friedrich-Alexander-UniversiHit, (Current affiliation: Imation Corp.,
Lehrstuhl fUr Stromungsmechanik, Oakdale, Minnesota, USA.)
Erlangen, Germany.
RONALD G. LARSON
Lucent Technologies, Bell Laboratories,
PHILIP H. GASKELL Murray Hill, New Jersey, USA.
The University of Leeds, (Current affiliation: University of Michigan,
Department of Mechanical Engineering, Department of Chemical Engineering,
Leeds, UK. Ann Arbor, Michigan, USA.)
Vlll Contributors
KIMIAKI MIYAMOTO L. E. (SKIP) SCRIVEN

Fuji Photo Film Co., Ltd., University of Minnesota,
Production Engineering & Development Division, Department of Chemical Engineering and
Minami-ashigara-shi, Kanagawa-ken, Japan. Materials Science,
Minneapolis, Minnesota, USA.
HARVEY J. PALMER
University of Rochester,
Department of Chemical Engineering, ROBERT B. SECOR
Rochester, New York, USA. 3M Company,
Engineering Systems and Technology
FERDINAND R. PRANCKH Laboratory,
Avery Dennison, S1. Paul, Minnesota, USA.
Corporate Process Technology,
Pasadena, California, USA. MARC K. SMITH
(Current affiliation: HPM Corp., Georgia Institute of Technology,
Mount Gilead, Ohio, USA.) The George W. WoodrutfSchool of Mechanical
Engineering,
ROBERT K. PRUD'HOMME
Atlanta, Georgia, USA.
Princeton University,
Department of Chemical Engineering,
Princeton, New Jersey, USA. NICK G. TRIANTAFILLOPOULOS
GenCorp Polymer Products,
TIMOTHY J. REHG Latex Laboratories,
Allied Signal Inc., Akron, Ohio, USA.
Aerospace Equipment Systems,
Torrance, California, USA.
YVES-M. TRICOT
KENNETH J. RUSCHAK ILFORD AG,
Eastman Kodak Company, R&D Department,
Emulsion Coating Technologies, Fribourg, Switzerland.
Rochester, New York, USA.
MICHAEL D. SAVAGE WILLY VAN ABBENYEN

The University of Leeds, Agfa-Gevaert N.V.,
School of Mathematics, Research Laboratories,
Leeds, UK. Mortsel, Belgium.
P. RANDALL SCHUNK HANS-GUNTER WAGNER

Sandia National Laboratories, BASF AG,
Department of Computational Fluid Dynamics, Ludwigshafen, Germany.
Albuquerque, New Mexico, USA.
PETER M. SCHWEIZER STEVEN J. WEINSTEIN

ILFORD AG, Eastman Kodak Company,
Engineering Department, Emulsion Coating Technologies,
Fribourg, Switzerland. Rochester, New York, USA.
NOTATION
In each chapter, all the symbols are defined E Young's modulus [N/m 2 J or
in the text where they first appear. The dilational surface elasticity modulus
following list defines the symbols most [N/m]
commonly used throughout the book. The El elasticity number [El;: IlUL 2 /D =
list is by no means complete, and a particular 12(1 - V 2 )IlUL 2 /{Ec5 3 }]
symbol may have more than one meaning, Es surface elasticity number
depending on its local definition. Also, given [Es == 9lTr*/IlU]
the large number of authors and the wide F force [NJ; may also be used as a
range of subject matters covered, some general-purpose function
inconsistencies were nearly impossible to f focal length [m]
avoid. As a general rule, upper-case letters f activity coefficient [ - J
denote dimensional variables, and lower- 9 gravitational acceleration of
case letters are used for dimensionless [g = 9.81 m/s 2 J
quantities; dimensionless groups are given G Gibbs free energy [JJ or shear
by two-letter symbols, the first one being modulus [N/m 2 J
upper case. Apart from a few well-marked G' loss modulus [N/m 2 J
exceptions, SI units are used throughout the G" elastic modulus [N/m 2 J
book. h dimensionless film thickness
[h == H/LJ
A surface area [m 2 J or Hamacker hoo dimensionless final wet thickness of
constant [J] coated film [h oo == H oo/LJ
a Langmuir constant [mol/cm 3 ] H film thickness or height [mJ
Bo Bond number [Bo ;: pgL 2 /a] H oo final wet thickness ofcoated film [mJ
C concentration [mol/cm 3 J or HG coating gap [m]
capacitance [A s/VJ Ho reference thickness (e.g., film thickness
Cj concentration of species i [mol/cm 3 J down an inclined slide, half gap in
Co bulk concentration [mol/cm 3 J roll coating, etc.) [mJ
Cs sub-surface concentration [mol/cm 3 ] Jf mean curvature of surface [m - 1 J
C* reference concentration [mol/cm 3 J 10 incident radiant energy [JJ
Ci dimensionless concentration ofspecies It transient radiant energy [JJ
i [c i ;: CJC*J J Jacobian
Ca capillary number [Ca == IlU /aJ J Jacobian matrix
Ca AE critical capillary number at onset of L characteristic, macroscopic length
air entrainment [Ca AE == IlU AE/aJ scale [mJ
D diffusion coefficient [m 2 /sJ or bending Ld length of boundary layer [mJ
stiffness [D == Ec5 3 /{12(1 - v2 )}] La capillary length [La == Ja/pgJ
~j binary diffusion coefficient of species M change of momentum per unit width
i [m 2 /sJ [kg/s 2 J
x Notation
M mass matrix V(t) volume of fluid body [m 3 ]

n power law index, or index ofrefraction ~ surface velocity [m/s]
n unit normal to a surface V* volumetric flow rate [m 3 /s]
p dimensionless pressure [p := PL/p'u] W coating width in cross-web direction
lip dimensionless pressure difference (e.g., em]
bead vacuum) [lip:= liPL/p'u] WA work of adhesion [J/m 2]
P pressure [~/m2] We Weber number [We := pU 2 L/a]
liP pressure difference (e.g., bead vacuum) X coordinate in main flow direction [m]
[~/m2] y coordinate normal to film flow em]
PA ambient pressure [~/m2] Z transverse (cross-web) coordinate em]
Pv sub-ambient pressure underneath x, y, z dimensionless coordinates [X/L, Y/ L,
coating bead [~/m2] Z/L]
Po liquid property number
[Po:= a(p/j.l4 g)1 /3]
q dimensionless flow rate per unit
coating width [q:= Q/(UL)] Greek letters
Q volumetric flow rate per unit coating (X eigenvalue, impingement angle or
width [m 2/s] application angle
R residual f3 angle of inclination (measured from
R residual vector horizontal)
R radius em] '}' angle of solid corner (e.g., at static
Rh hydraulic radius em] separation line)
Rm mean radius of curvature em] Y shear rate [S-I]
91 universal gas constant Yo reference shear rate [s - 1]
[91 = 8.31JK- 1 mol- 1] r surface excess concentration
Re Reynolds number [Re:= pUL/j.l] (adsorption density) [moljm 2]
S arc length [m] or spreading coefficient r i surface concentration (adsorption
[m~/m] density) of species i [moljm 2]
s dimensionless arc length [s:= S/L] r o equilibrium adsorption density
SCi Schmidt number for species i [mol/m 2]
[SC i := j.l/p~J r* reference surface concentration
SCs surface Schmidt number [mol/m 2]
[Scs := j.l/p~s] roo maximum surface excess
St Stokes number CSt = pgL 2 /j.lU] concentration (adsorption density)
t time or surface age [s] [moljm 2]
t unit tangent to a surface () thickness of substrate, blade or
T absolute temperature [K] boundary layer em]
T total stress tensor B small parameter
u, v dimensionless velocity components f. extension rate [s - 1]
[-] B o dielectric constant of vacuum
U coating/web speed [m/s] [A s/(Vm)]
U AE critical coating speed at onset of air Bs dielectric constant of material
entrainment [m/s] [As/(Vm)]
V alternative characteristic velocity (e.g., 11 non-~ewtonian viscosity CPa s]
impingement velocity in curtain 110 non- ~ewtonian viscosity at reference
coating, metering roll velocity in shear rate CPa s]
reverse roll coating, etc.) [m/s] K curvature of curve [m - 1]
Notation xi
K dimensionless curvature of curve Young's equilibrium contact angle

[K = KL] °E
OR recently receded/receding static
Km dimensionless mean curvature of contact angle
surface [K m = YfL] 00 apparent static contact angle
A. wavelength Em] or dimensionless (dynamic value extrapolated to zero
flow rate in roll coating [ - ] speed)
A." viscosity ratio [A." == 1l2/1l1] Ow local contact angle imposed as
A. p density ratio [Ap == P2/P1] mathematical boundary condition
Il Newtonian viscosity [Pas] (J surface tension [N/m]
Ils surface dilational viscosity [N s/m] (Jeq equilibrium surface tension [N/m]
v Poisson's ratio [ - ] or frequency (Js solvent surface tension [N/m]
[S-l] (J12 interfacial tension between two liquids
1t surface pressure [J/m 2 ] [N/m]
II disjoining pressure [N/m 2 ] (JT tension in substrate or blade
P density [kg/m 3 ] [N/m]
<pi
•
,
finite-element basis function viscous stress [N/m 2 ]
t/Ji finite-element basis function
contact angle ·w
~,
wall shear stress [N/m 2 ]
isoparametric coordinates
°°A recently advanced/advancing static n frequency [S-l]
contact angle OJ dimensionless frequency [OJ == nL/U]
00 apparent dynamic contact angle
INTRODUCTION
COATING SCIENCE AND TECHNOLOGY: 1
AN OVERVIEW
Stephan F. Kistler and Peter M. Schweizer
In industrial coating processes, one or several in the sub-micron to Angstrom range are applied
liquid layers are deposited on a solid substrate, by sputtering or chemical vapor deposition (CVD)
often a flexible web, and are subsequently dried to produce, for example, anti-reflection coatings
or cured to form solid films that serve specific on glass, one-way mirrors, etc.
functions. Figure 1.1 shows a much simplified Coating processes as considered in this book
schematic with the most essential components are employed in many industries that manufacture
of an industrial coating line. At the heart of such a vast number of different products, including
processes is the displacement of a gas previously paper of various grades and surface finishes,
in contact with the substrate by a liquid film. packaging materials, printed matter like books,
The wet thickness of the coated films is typically newspapers, catalogues, etc., photoreceptors for
on the order of 1 to tOO microns, or more. This xerographic reproduction processes, multilayer
range of wet film thickness explains the choice silver-halide photographic films and papers, non-
of the book title, namely Liquid Film Coating. silver imaging products, magnetic storage media
The purpose is to distinguish the coating processes such as audio and video tapes as well as computer
discussed in the volume at hand from another discs, optical storage media such as audio and
class of coating processes often referred to as photo compact discs, fibers and wires, labels,
'thin film coating'. There, films with thicknesses protective and decorative paintings, photoresist
LIQUID SUPPLY COATER DRYER/CURE
Figure 1.1 Simplified schematic of industrial liquid film coating process.
Liquid Film Coating. Edited by S. F. Kistler and P. M. Schweizer. Published in 1997 by Chapman & Hall,
London. ISBN 0 412 06481 2.
4 Coating science and technology: an overview
coatings for the manufacture of microelectronics tension and static pressure forces in liquid film
such as printed circuits, solar cells, ceramic deposition as it may occur in dip or extrusion
components made from sol/gels, sensors, medical coating. By matching the film profile downstream
products like transdermic systems which release to a static meniscus region upstream, they were
active substances into the skin from a thin able to derive a simple relationship for the final
band-aid-like applicator, textiles, sand paper, film thickness as a function of liquid properties
adhesive tapes and many more. and coating speed. A major thrust of coating
Liquid films may be applied in one or several research began in the early 1970s under the
layers, either simultaneously or in multiple coating leadership of Professor L. E. (Skip) Scriven at
passes. Moreover, the substrate may be coated the University of Minnesota. He not only started
on just one or on both sides, again, either in one teaching about interfacial phenomena, but he
or in subsequent passes. Coating solutions are also initiated a comprehensive research program
based on water or organic solvents. The latter are of various coating methods, thereby emphasizing
no longer attractive for ecological reasons. However, and developing experimental methods, theoretical
in spite of enormous efforts by various coating models, as well as analytical and numerical
industries to substitute water for organic solvents, techniques for solving the models. It was perhaps
many desired product properties can, to date, only the availability of more advanced mathematical
be achieved with organic solvents which, for many methods, more potent computers, and more
specialty applications, remain in common use. refined measurement tools that triggered an
The economic importance of the U.S. coating avalanche effect in coating research and devel-
industry has been estimated by the US Department opment in the early 1980s.
of Commerce (1994) to be about $20000000 Today, industrial and academic research lab-
million per annum. This immense figure represents oratories all over the world communicate and
the value of coated products, coating process collaborate with each other, as is evident in the
hardware, and other products in which coating international coating conferences organized, for
is a key technology. The large value generated by example, by the Technical Association ofthe Pulp
the coating industry and the ever increasing and Paper Industry (TAPPI), the American Institute
demands on film thickness uniformity as well as of Chemical Engineers (AIChE), the Society for
on productivity and yield of the coating processes Imaging Science and Technology (IS&T) and the
justify the need for a sound and basic under- Society of Chemical Engineers in Japan (SCEJ);
standing of coating technology. short courses and workshops on various topics
The origins of coating technology are rooted related to coating technology, offered, for example,
in different industry segments associated with by AIChE's Continuing Education Department,
different product lines. Moreover, coating is a the Center for Interfacial Engineering at the
complex multi-disciplinary matter, comprising University of Minnesota, and the Universities of
aspects of wetting, spreading, adhesion, fluid Erlangen (Germany) and Bradford (England);
mechanics, rheology, mathematics and others. It close collaboration between universities and
is therefore not surprising that a consolidation industrial companies, manifested by engineering
of disciplines and open communications between research centers such as the Center for Interfacial
various industry segments did not take place for Engineering at the University of Minnesota;
a long time. Consequently, coating technology technical journals devoted to liquid film coating
remained an art for the better part of the 20th technology, including the Journal of Coating
century. First signs of turning coating from art Technology, Coating (in German), and Industrial
into science appeared in the land-mark paper by Coating Research (in Japanese and English); and
Landau and Levich (1942). For small capillary a rapidly growing number of publications in other
numbers, these authors derived a rigorous professional journals related to fluid mechanics,
mathematical analysis of the viscous, surface colloid and surface science, rheology, etc.
Introduction 5
In spite of all the activity in coating process wide range of available methods or to invent a
research, there is no preceding comprehensive new one. Then, of course, the details of the
text that summarizes results obtained in various coating equipment must be specified and designed,
places around the world and published in different and feasibility of the process selected must be
technical journals. In particular, there is no single demonstrated for a desired range of operating
text that systematically interprets all the physical conditions. For an existing product, on the other
mechanisms that control coating processes, and hand, the most pressing issues are often to
that explains implications of scientific principles expand the range of successful operation, for
on industrial coating applications. This is not to instance to higher coating speeds, thinner wet
say that no books or reviews on coating technology layers, or more layers coated simultaneously; to
have been written in the past. However, they all eliminate coating imperfections or defects that
focus on either coating equipment (Higgins 1965; degrade product quality; and to increase the
Booth 1970; Weiss 1977; Satas 1984); on just a yield, or the fraction of salable material from an
few particular coating methods (Middleman 1977; entire lot coated. Coating engineers, together
Ruschak 1985); or on particular aspects ofcoating with formulation chemists, also respond to
technology such as numerical methods (Kistler never-ending quests to improve the performance
and Scriven 1983), process control (Frost and of the final products which, in some instances in
Gutoff 1991), or wettability (Blake 1984; Berg very subtle ways and in others to a significant
1993). The recent volumes edited by Cohen and extent, depend on careful control of the micro-
Gutoff(1992) and Benkreira (1993) emerged from structure or surface properties imparted on the
short courses on coating technology. They cover coated layers by the coating flow process upstream.
a wide range of topics, but make no attempt at Altogether, most of the activities of industrial
being comprehensive or consistent in approach coating-process research and development amount
and thus provide only snap-shots of recent to a gigantic optimization procedure, with
progress made in the selected subject areas. seemingly countless independent variables for
The multi-author book at hand, in contrast, the design and operation of the process as well
seeks to provide a systematic up-to-date guide as the chemical composition of the coated products,
to the literature, a critical assessment of the and with many, often conflicting target functions
current state of knowledge, and an outlook on for both manufacturability and ultimate product
directions for future developments, all written by performance. Rapidly intensifying competitive
recognized experts in their respective fields. The pressures on a world-wide basis, furthermore,
book is intended primarily for scientists, engineers demand that the optimization addresses not only
and technical managers at industrial companies, technological refinement, but also economical
universities and government laboratories who viability.
are engaged in research and development of All in all, the challenges listed make apparent
coating processes, coated products and advanced that future refinements in coating technology
materials. The book is also intended to be become increasingly intricate and difficult. In
valuable as a text for self-study, or as a reference addition, the challenges must be met in ever
for class-room lectures. shorter time periods. All this puts a premium on
The overall aim of the book is to provide the being able to identify and analyze the physical
interested researcher or engineer with the infor- mechanisms that decide between success and
mation needed to address, in a systematic and failure of a coating process rather than relying
scientifically sound manner, the challenges most on an empirical approach based mostly on trial
commonly faced in industrial coating practice. and error. In liquid film coating of the sort
When a new product is being developed or an considered in the book at hand, many of the
existing one is modified in a significant way, the essential mechanisms are those of viscous free-
challenge is to select a coating method from a surface flows, and these are the main subject of
FREE SURFACES
WITH SURFACE TENSION
(often influenced by surfactants)
LIQUID/LIQUID INTERFACES
EXTERNAL FORCES
(e.g. electrostatic)
-
DYNAMIC WETTING
- AIR ENTRAINMENT
Figure 1.2 Cursory overview of physical mechanisms involved in liquid film coating (the figure shows a
hypothetical coating process for two-layer coating).
the book. Figure 1.2 shows a schematic of a Catastrophic coating failure may also stem from
free-surface flow that is prototypic of the flows the inability of the coating liquid to displace
encountered in liquid film coating. It includes sufficient amounts of air previously in contact
many of the features that contribute to the with the substrate, a process that takes place at
technological complexity ofcoating-flow processes what is called a dynamic wetting line. Both flow
yet, at the same time, make for interesting and stability and successful air displacement may,
rewarding fluid mechanics problems. Because of quite often, be enhanced by the application of
the small thickness scales of the coated wet films, external forces, such as those from an electrostatic
Reynolds numbers rarely exceed 100 and so the field or an external pressure difference applied
difficulties associated with turbulence are usually between two free surfaces of a coating flow. Even
not of concern, at least in the liquid phase. when the flow is stable, however, imperfections
Significant complications and resulting math- in coated liquid films may arise from a wide
ematical nonlinearities, however, arise from the variety of mechanisms. These include in-
free surfaces between the coating flow and the homogeneities, foreign matter or contaminants
surrounding gas, and from interfaces between arriving with the coating solution or the substrate;
stratified layers in multilayer coating. Primarily disturbances induced by equipment vibrations,
because ofthe free surfaces and interfaces, coating uncontrolled air flow, or evaporation; various
flows are susceptible to numerous hydrodynamic wetting phenomena such as the uneven advance
instabilities that, all too often, lead to catastrophic of a static contact line past a sharp corner, or
failure of the process to deposit a uniform liquid patchy charge distributions arriving on a non-
layer. The stability of coating flows is also conductive substrate. Coating flaws may also
extremely sensitive to non-Newtonian behavior stem from flow instabilities that grow only to
of the coating solutions, sometimes strongly modest amplitudes during liquid film application
influenced by small traces of a non-Newtonian and partially decay further downstream. In some
component whose effect can barely be detected instances, coating imperfections can be alleviated
in rheological tests. Likewise, surface-active by spontaneous self-leveling of the liquid film
substances, either added on purpose or present after coating. In other cases, however, small
because of contamination, can drastically alter imperfections or defects may grow between coating
the equilibrium of forces at the free surfaces and and drying to gross flaws in the film or even dry
thereby affect coating flows and their stability. patches. In either case, surface-active components
Introduction 7
can have a significant effect on the formation into some of the key mechanisms at work in
and fate of defects. The ultimate properties of liquid film coating. For a more phenomenological
the coated product, finally, depend not only on approach to interfacial hydrodynamics, the reader
the coating flow mechanics upstream, but also may also want to consult the review by Berg
on the mechanisms that constitute drying and (1982). Issues of wettability, associated with
solidification downstream, including diffusion, contact lines where liquid/gas interfaces meet
evaporation and other forms of mass transfer, solid surfaces, are touched upon in Chapter 2
phase behavior, particle migration and orientation, but are treated in depth in Chapter 3 by Blake
film-internal stress buildup, and many more. and Ruschak. These authors give an authoritative
The goal of the book at hand is not only to review and assessment of the current level of
review the mechanisms that control the deposition understanding of both static and dynamic wetting.
of liquid films on solid substrates, but also to At the same time, they keenly focus on the issues
bridge the gap between scientific principles and most relevant in industrial coating practice,
practical implementation on an industrial scale. namely how to control the pinning and uniformity
Part One (Chapters 2-6) reviews the underlying ofstatic contact lines, and how to avoid deleterious
physical principles and important material air entrainment at dynamic wetting lines. Chapter
properties. The discussion is at a fairly advanced 4 by Tricot addresses the subject of surfactant
level, but is complemented with plenty of references effects in coating, for which Chapter 2 also lays
to introductory texts. Part Two (Chapters 7-9) some of the foundations. Apart from an overview
gives a comprehensive treatise of the most of the classical concepts in surfactancy, the main
sophisticated experimental and mathematical focus in Chapter 4 is on material properties of
methods available today to investigate coating aqueous surfactant formulations. In particular,
processes. The techniques are described in sufficient the chapter discusses experimental methods to
detail to give the reader a realistic picture for accurately measure static and dynamic surface
what is involved, and a helpful guide for where tension, and the application ofsuch measurements
to look for more details. Part Three (Chapters to systematically optimize surfactant systems for
10-15) finally provides an in-depth overview of improved coating performance (the issue of how
a wide range of industrial coating processes. surfactants actually affect the coating flow
Special emphasis is placed on combining scientific mechanics is taken up in detail in Chapter lid
principles with results from sophisticated methods of Part Three). Chapter 5 by Glass and
of analysis to give the reader a solid knowledge Prud'homme covers the sometimes dramatic
of the operating behavior of various coating effects of non-Newtonian liquid behavior in
techniques, and an appreciation for how many coating flows. All in all, this subject is still rather
phenomena that commonly prevent successful poorly understood, perhaps because of the
coating can be traced back to fundamental complicated deformation fields encountered in
mechanisms. most coating processes, and the complex rheologi-
In Part One, Chapter 2 by Weinstein and cal behavior of many coating formulations. The
Palmer recounts the capillary hydrodynamics chapter reviews some of the most important
and interfacial phenomena that are all-important non-Newtonian material properties and related
in coating flows. The main emphasis is on a constitutive models, and illustrates the interactions
careful derivation of the complete set ofgoverning between material structure, rheological behavior
equations and boundary conditions for a rigorous and actual coating performance for selected
mathematical analysis of liquid film coating; coating systems. For additional, more general
identification of dimensionless groups and closely information, many excellent texts are available
related, limiting flow regimes; and a set of on constitutive theory (e.g., Larson 1988), non-
well-chosen sample problems that are worked Newtonian fluid mechanics (Bird, Armstrong
out in detail to reinforce the physical insights and Hassager 1987), and rheological measurements
(Barnes, Hutton and Walters 1989; Macosko complementary rather than competitive. Experi-
1994). Chapter 6 by Kheshgi concludes Part One ments on pilot coaters are indispensable to
with a concise overview of the mechanisms that demonstrate feasibility ofa process or to generate
control the fate ofliquid films after coating. These sample product. Experiments, either on pilot
include surface leveling which can heal coating coaters or cleverly designed table-top devices,
irregularities; de-wetting which can result from are also invaluable to identify the mechanisms
the nucleation and growth ofdry patches; diffusion, by which coating processes fail or defects form,
mass transfer, and phase transitions which are for they provide the ultimate validation for any
associated with drying; and stress build-up from hypothesis or theory. Experiments alone, however,
solvent removal and other mechanisms that can often cannot fully explain cause-and-effect rela-
cause cracking, crazing, peeling and curling of tionships. They can also be very costly and, in
coated films. some instances, visualizing essential small-scale
In Part Two on methodology for coating features of a coating flow can be exceedingly
process investigations, Chapter 7 by Schweizer difficult ifnot impossible. Conventional methods
discusses experimental equipment and design, of mathematical analysis, such as simple lubri-
flow visualization and measurement instrumen- cation-flow models or more systematic methods
tation, and on-line inspection techniques. The based on asymptotic expansions, have a keen
focus is on methods suitable for coating process ability to zero in on the essential mechanisms
studies in an R&D environment as well as while, at the same time, indicating which mech-
methods used for scale-up and production anisms may be secondary and can be ignored.
purposes, and on illustrating the application of For problems that are simple enough, classical
these methods for particular case studies. Chapter methods of flow analysis furthermore produce
8 by Smith introduces approximate methods of closed form solutions which reveal important
conventional mathematical analysis of coating trends from simple inspection of the mathematical
flows. It starts with simple lubrication-flow analysis expressions, or from quick numerical evaluation.
and then treats in depth more systematic The applicability of these methods, on the other
perturbation methods that seek to exploit, by hand, is usually restricted to idealized flow
way ofasymptotic expansions, the simplifications configurations that may have to be much simpler
made possible by limiting parameter regimes or than the coating process of practical interest. In
simple flow geometries. Carefully developed addition, the validity ofthe asymptotic expansions
examples show when and how asymptotic methods is usually restricted to limiting regimes. These
can be applied to analyze steady flow fields as limitations are not a factor with computational
well as transient behavior and related stability. methods. In their most sophisticated form available
Chapter 9 by Christodoulou, Kistler, and Schunk today, such methods not only predict the mostly
reviews recent advances in numerical methods two-dimensional base flows in successful coater
of computer-aided mathematical analysis of operation at realistic conditions, but also the
coating flows. The main emphasis is on finite- stability of the computed flow states and the
element methods which, over the last fifteen years, susceptibility of stable coating flows to external
have become remarkably powerful and which, at disturbances. Computational methods - and
present, are the preferred choice. The chapter also asymptotic methods where applicable - have the
highlights alternative methods for computational ability to visualize details of the flow fields that
free-surface flow analysis which, for certain are difficult to measure in experiments; unravel
applications, may offer distinct advantages. the nonlinear behavior of free-surface flows and,
As has been pointed by many others before thereby, isolate key mechanisms and stability
(e.g., Scriven (1984); Churchill (1992)) - and as limits that are difficult to understand from
is apparent throughout the volume at hand - experimental data; evaluate many more designs
the three methods covered in Part Two are and configurations than could realistically be
Introduction 9
tested; and explore wider parameter ranges than with illustrative examples and practical design
may be accessible with existing equipment. In considerations for some of the most common die
spite of their power and promise, however, geometries.
computational methods have limitations as well. Chapter 11 covers pre-metered coating processes
For one, difficulties associated with dynamic in which the amount ofliquid applied to the web
wetting lines, non-Newtonian flow behavior, and per unit area is predetermined by a fluid metering
other specific material effects still seriously hamper device upstream, such as a precision gear pump,
predictive capabilities. In addition, systematic and the remaining task of the coating device is
exploration of wide parameter ranges can be to distribute that amount as uniformly as possible
costly and time-consuming, and a large number in both the down-web and cross-web direction.
of output files and color plots do not guarantee In slot or extrusion coating, discussed by Durst
that important trends are recognized, especially and Wagner in Chapter 11a, one or several layers
by inexperienced users with limited background of coating liquid emerge from a coating die in
in coating flow mechanics. close proximity to the moving web and bridge
Part Three covers the theory and practice of the gap between the die and web to form a coated
a wide range of industrially significant coating film. In slide coating, covered by Hens and Van
processes. A casual survey of various coating Abbenyen in Chapter 11 b, one or several liquid
configurations might suggest that the possibilities layers are metered onto an inclined plane to form
of inventing or modifying coating systems are a falling-film flow that is picked up by the moving
limitless. A more thoughtful analysis, however, web across a narrow gap between the lip at the
reveals that all existing coating methods - and, end of the incline and the web. In curtain coating,
most likely, all that will be developed from here which Miyamoto and Katagiri review in Chapter
on - are merely permutations of the sequence 11c, the coating die is mounted at a considerable
or geometric arrangement in which the key tasks height above the point ofcoating and the coating
ofa coater are performed (Scriven 1991; Benjamin liquid(s) fall as an unsupported sheet onto the
and Scriven 1991). These tasks are the distribution moving web. The key features that differentiate
of the liquid in the cross-web direction and the three pre-metered coating methods are the
associated control of the coated film thickness length of the liquid bridge that needs to be
in that direction; the control of the coating established between the coating device and the
thickness in down-web direction; the deposition moving substrate; the orientation of gravity with
of the liquid on the substrate, which includes the respect to the main direction of flow in the liquid
displacement ofthe air previously in contact with bridge; and the amount ofmomentum the coating
the substrate; and the formation of the free liquid carries toward the substrate. Resulting
surface that bounds the coated film. from these distinguishing features are different
In a wide range of coating configurations, the operating limits and characteristic coating defects,
first task is performed by the internal distribution to be explained in detail in Chapters 11a through
cavity and metering slot of a coating die or 11c. In spite of significant physical and practical
another feed device that spreads the coating differences, however, the three pre-metered coating
liquid from a single inlet port to a slot-shaped methods share many similarities. In particular,
outlet that extends in the cross-web direction. they involve two free surfaces, two static contact
Chapter 10 by Secor discusses the analysis and lines, and one dynamic wetting line (for an
design of die internals. The chapter surveys example, see Fig. 1.2 above). One free surface
methods of die design, focusing on a fairly simple, forms the liquid film; it separates at a static
yet very effective approximate method that readily contact line from a stationary solid piece that is
accounts for non-Newtonian effects; addresses part of the coating die. The other free surface
mechanical and thermal aspects of die design; also starts at a static contact line on the coating
identifies various flow regimes; and concludes die and terminates at a dynamic wetting line
where most of the air previously in contact with metered coating processes, in turn, can be divided
the substrate is displaced by the coating liquid. into those in which the coated film thickness is
All three pre-metered coating processes are metered primarily by a narrow gap between the
furthermore suitable for single- and multi-layer web and a solid member upstream of the film
capability; and the stability of the flow is often forming region; and processes in which the
assisted by externally applied forces, most meniscus at which the film forms also performs
commonly an ambient air pressure difference the metering task via the action of capillarity.
across the liquid bridge. Self-metered coating processes with a narrow
Because of the many similarities between slot, metering gap can be subdivided further into
slide, and curtain coating, Chapters 11a, 11 b, those where the solid boundaries of the applicator
and 11c all follow in essence the same structure: are rigid; and those where one or more solid
they first specify the design and other important members are flexible and, hence, deform in
aspects of the coater set-up; they then discuss response to fluid pressure or external loads. The
the hydrodynamics that make up successful latter class is referred to as elastohydrodynamic.
coater operation; they move on to identify Chapter 12a by Coyle describes knife, forward-
operability limits past which the process fails roll, and reverse-roll coating. These processes are
catastrophically; they assess susceptibility to prototypical of configurations with narrow, rigid
external disturbances within the window of metering gaps in which the coating thickness
successful operation, which can have a significant depends mainly on the interplay of viscous forces
effect on final coating quality; and they conclude with the geometry of the coating device and the
with an overview of practical considerations, relative speed of its surfaces. The chapter shows
showing how specific solutions proposed in the how much of the operating behavior can be
patent literature or elsewhere can be rationalized understood from the classical theory of hy-
from a fundamental understanding ofthe process. drodynamic lubrication between slowly converg-
Chapter lId by Schunk and Scriven discusses ing and diverging surfaces (Reynolds 1886; see
surfactant effects in coating flows. This chapter also Chapter 8). The chapter also explains the
is included as part of Chapter 11 because it physical cause and mathematical analysis of
focuses primarily on pre-metered coating processes various flow instabilities - including the much
ofthe sort used in the manufacture of water-based celebrated ribbing instability that can arise in a
photographic materials. Building on the infor- wide range of coating methods. The chapter
mation given in Chapter 4, Chapter lId presents furthermore assembles operating diagrams for
a comprehensive mathematical theory of the the three processes covered; and discusses how
action of surface-active materials, and reviews coating non-uniformities may arise from various
analyses of surfactant effects in coating and disturbances, contaminants, and also ancillary
related free-surface flows. The bulk ofthe chapter, flows such as those associated with feeding.
however, is dedicated to three particular case Chapter 12c by Pranckh and Coyle follows a
studies intended to show how care must be taken similar program for elastohydrodynamic coating
in interpreting dynamic surface tension data, systems, addressing blade, membrane, tensioned-
how surfactants influence falling film-flow stability, web, squeeze-roll and gravure coating. In addition,
and how they alter the local force balance in the Chapter 12c provides the theoretical background
dynamic wetting region of a slide coating flow. in solid mechanics needed to understand elas-
Chapters 12 and 13 treat self-metered coating tohydrodynamic coating systems (see also Dowson
processes in which the final wet thickness of the and Higginson (1966)). Chapter 12d by Aidun
coated film is mostly controlled by the interactions and Triantafillopoulos focuses on blade coating
of the fluid flow with the coating applicator, in greater depth, specializing on paper coating
rather than the amount of liquid supplied to the in which speeds in excess of 25 mls are not
coater head as in pre-metered processes. Self- unheard of. The chapter deals with the special
Introduction 11
issues that arise from the rough, porous substrate transport in both the liquid and the gas phase,
and the high speeds, in particular as they pertain including the volume changes associated with
to applying the fluid, managing hydrodynamic evaporation and condensation. The results reveal
instabilities, and avoiding air entrainment. The possible effects of the drying environment on the
chapter furthermore addresses specific coating microstructure of the final sol-gel films, and are
defects and microstructures that arise from the exemplary of the type of future work needed to
particulate coating solutions and their viscoelastic expand the scope of coating flow mechanics from
properties, as well as from the porous and the deposition of uniform films to the formation
strongly hydrophilic substrate. of films with controlled microstructure.
Chapter 12b by Gaskell and Savage, like Chapter 14 by Larson and Rehg reviews
Chapter 12a, considers coating with rigid rollers, another specialized coating process for advanced
but specializes in a particular regime referred to materials processing, namely spin coating. The
as meniscus roll coating in which the liquid film chapter identifies the relative importance of
arriving at the nip between the rollers is much hydrodynamics and evaporation, explains why
thinner than the roll separation. In this regime, on flat substrate surfaces uniform films are rather
capillarity plays a dominant role. The coated easy to achieve, and isolates the key factors that
film thickness is not metered by the gap between favor leveling or planarization of spin-coated
the rollers, but the process is self-metered in the films on surfaces with textured topography. The
sense that the final film thickness depends on the last case is of particular importance in the
speed ratio between the rollers, and is influenced microelectronics industry, and the chapter gives
by capillarity. Chapter 12b uncovers a wealth of due consideration to the specific material prop-
structures in meniscus-roll-coating flows, and erties of the photoresist solutions used in that
focuses on identifying different flow regimes and particular application.
related flow transitions in-between. Chapter 15 by Schweizer, finally, seeks to put
Meniscus roll coating is in some ways closely much ofthe information contained in the volume
related to dip coating, which is the subject of at hand into perspective by consolidating it into
Chapter 13 by Schunk, Hurd and Brinker. Dip the frameworks of control and optimization. The
coating is the withdrawal of a substrate, either chapter briefly reviews some of the elementary
a continuous web or individual pieces ofsubstrate concepts of control theory and argues for a more
material, from a liquid pool. It is the archetype systematic application of these concepts in coating
of a self-metered coating process, for the coated technology. The chapter also recounts the
film thickness is controlled entirely by the capillary fundamental principles behind optimization, and
hydrodynamics involved in the film deposition. illustrates how they can and should be applied
Chapter 13 begins with a review of the classical at various levels, including micro-optimization
and well-known results on dip coating, but then of a single detail of a coating flow, such as a
stresses that, at low coating speeds, the mass wetting line position; overall optimization of an
transport by evaporation of volatile components entire process step, such as flow distribution
can compete with the hydrodynamics in con- within a coating die; global optimization of an
trolling the meniscus profile and setting the final entire manufacturing process to maximize, for
film thickness. The chapter furthermore establishes instance, productivity or yields; combined process
important links between coating flow processes and product optimization; and last, but not least,
and resulting microstructures by considering economic optimization addressing unit costs of
modern applications of dip coating in advanced the coated product.
materials processing, namely the deposition of
Langmuir-Blodgett films and sol-gel films. For Even though the book at hand treats a wide
the latter, the chapter summarizes a comprehensive range of coating configurations and associated
computational analysis of fluid flow and mass physical mechanisms, it does not cover all aspects
of coating technology. As mentioned above, the work that is documented in numerous patents
book focuses on the fundamentals of the actual rather than scientific investigations. Noteworthy
deposition process and, for that reason, ignores exceptions are the recent papers by Katagiri
all ancillary technologies, such as fluid preparation (1992), who applied the finite-element method to
and delivery, web handling and cleaning, air flow predict the three-dimensional flow along an edge
management and clean-room conditioning, or guide of a falling curtain (see Chapter 11c), and
calendering and other forms of post-processing. by Brandon et al. (1994), who successfully
For a practical guide on how to condition, pump, employed finite-element methods for transient
filter, and de-gass coating solutions, the reader and stability analysis to guide the invention of
should consult the review by Schweizer (1992). an improved procedure for splice passage in slide
Kheboian (1992) gives practical advise on web- coating.
handling but, unfortunately, his treatise does not Another mechanism that is not covered in
attempt to review the growing scientific literature sufficient depth is the all-important effect of
that is available on the mechanics involved in electrostatic charges. Coating practitioners have
web transport and winding (e.g. Good 1993). long known about the deleterious effects on
Similarly, for most other supporting technologies coating quality from uncontrolled charges carried
involved in coating, there exist no preeminent by polyester webs and other substrates - as well
scientific reviews, and the reader must be satisfied as their safety hazards - and have devised
with the general overviews in some of the books practical means to cope with the problem (e.g.
cited above (Booth 1970; Weiss 1977; Satas 1984), Keers 1984). On the other hand, electrostatic
or dig into the literature themselves. fields applied on purpose can also be used to
Even within its intended scope of treating the great advantage to assist dynamic wetting and
scientific principles of liquid film coating, the other functions of a coating process, as described
book at hand is not exhaustive in its coverage in a vast number of patents following the
of coating technology, for it ignores some pioneering works of Chappel (1934) and Nadeau
specialized, yet very common application methods, (1956). Scientific investigations into the physical
such as spray coating (for a cursory introduction, mechanisms at work have a well-established
see Satas 1984). This process involves interesting foundation in the theory ofelectrohydrodynamics
fluid-flow, heat- and mass-transfer phenomena (e.g., Melcher and Taylor (1969)), but are only
associated with atomization, evaporation, im- now beginning to appear in the public domain
pingement, spreading and coagulation of liquid (e.g., Feng and Scriven (1992)).
droplets (e.g. Lefebvre (1988); Trapaga and Szekely A somewhat related aspect of liquid film
(1991); Coeling (1993); Fukai et al. (1993)). coating that the book fails to address in an
In addition, the book pays insufficient attention up-front manner is the influence of the substrate
to some important mechanisms in liquid film surface. Those with industrial coating experience
coating - mostly due to the lack of a sufficiently know all too well that some substrates simply
large body of scientific literature at the time the 'coat much better' than others. They also know
outline of the book was planned back in 1990. of various attempts to influence the success of a
One class of such mechanisms is related to the coating process by applying 'priming' or 'subbing'
three-dimensional phenomena associated with layers to the film, or devising other on-line
the edges of coated layers, and the transient treatments such as corona discharge. Curiously,
phenomena that arise during splice passage, with the exception perhaps of Chapter 3, the
start-up and termination of coating processes. discussion of these effects is confined to a few
Because these phenomena are very complex, parenthetical remarks well-hidden in the main
much of the know-how on the quality of coated text. The reason is probably our incomplete
edges, coatability of splices, and controlled start- understanding of the physics of dynamic wetting,
up of liquid film coating is based on empirical which remains one of the most prominent
Introduction 13
unresolved issues in coating flow hydro- never-ending refinements in experimental and

dynamics. analytical instrumentation and rapid advances
Another aspect of liquid film coating that, in computer power, should see dramatic progress
given recent developments, receives insufficient in the near future on both the experimental and
attention in the present collection of review the theoretical side.
articles is that of drying and other methods of A further aspect of liquid film coating that is
solidification. Chapter 6 by Kheshgi provides a not treated in a separate chapter in the book at
concise and insightful summary of the most hand is the issue ofcoating defects. Most chapters
important mechanisms and identifies different make reference to various hydrodynamic in-
drying regimes. And Chapters 13 and 14 analyze stabilities that produce gross defects. Some also
in detail specific coating flows accompanied by discuss coating nonuniformities induced by
simultaneous evaporation. None of the chapters, ongoing disturbances, such as mechanical vibra-
however, does full justice to the significant progress tions, pump pulsations, speed fluctuations, etc.
that has been made in recent years in advancing Many chapters furthermore mention various
the basic understanding of solvent removal, characteristics of coating flows, such as enclosed
polymerization and other cross-linking reactions, recirculations or poorly controlled static wetting
or changes in colloidal stability and phase behavior lines, that supposedly cause streaks or other
in liquid films undergoing drying and/or solidi- down-web striations. With the exception of
fication. For a lucid assessment of the key issues Chapters 6 and 15, however, there are very few
involved, results available today, and future explicit explanations of specific mechanisms that
research directions, the interested reader may cause the many seemingly random defects that
want to look at the recent paper by Cairncross, plague every-day coating practice, and
Francis and Scriven (1996). On the other hand, predicatably foul up 'quality circles' in their
the reader looking for a description of drying industrious quest for making 'defect-free coating'
equipment or advice on how to dimension dryer a reality. Hansen and Pierce (1974) and Gutoff
ovens based on macroscopic mass and energy and Cohen (1995) provide helpful overviews of
balances may want to consult Cohen (1992). defects in liquid film coating.
Closely related to drying and solidification is, In fact, it is perhaps the biggest shortcoming
of course, the formation of microstructure, i.e. of the hydrodynamic principles behind liquid
the distribution and physico-chemical state of film coating - which constitute the bulk of the
individual chemical compounds, the concentration, material in this book - that, at best, they explain
agglomeration, flocculation and orientation of various defects but that, in many cases, they are
particles, porosity, surface texture, and many unable to predict the frequency at which defects
other attributes of coated films that are essential occur or recommend fool-proof remedies. This
for their ultimate performance as a useful product. is not to say that hydrodynamic studies of
Chapter 13 reports a special case study in which coating are not extremely helpful. They must be
the final film microstructure has been linked to performed, however, with the humbling awareness
drying and even coating upstream. More generally, that some aspects ofliquid film coating technology
however, none of the chapters in the book at may evade rigorous prediction from physical
hand - nor any other review we are aware of - principles for some time to come - and thereby
synthesizes a coherent picture of the many intricate remain, for the moment, being art rather than
ways in which the materials in a coating science. Still, a considerable amount of funda-
formulation, and also the substrate, interact with mental knowledge in liquid film coating has been
the coating and drying process to form the accumulated over the years, and we hope this
ultimate microstructure. Controlled microstruc- book makes a contribution by providing a
ture formation is an exciting research area of comprehensive consolidation of the knowledge
coating process fundamentals which, due to available today.
ACKNOWLEDGMENTS Encyclopedia of Chemical Technology, Volume 6,
4th edn, eds J. I. Kroschwitz and M. Howe-Grant,
We are grateful to Marjorie Spencer from Van pp. 661-669. New York: Wiley.
Nostrand Reinhold for her initiative and encour- Cohen, E. D. 1992. Thin film drying. In Modern
agement to write this book and for her valuable Coating and Drying Technology, eds E. D. Cohen
guidance during the early stages of the project. and E. B. Gutoff, pp.266-298. New York: VCH.
Cohen, E. D. and Gutoff, E. B. 1992. Introduction. In
We are thankful to all participating authors for Modern Coating and Drying Technology, eds E. D.
their fine contributions and for their exemplary Cohen and E. B. Gutoff, pp. xv-xx. New York: VCH.
commitment to carry this large project to the Cohen, E. D. and Gutoff, E. B. (eds). 1992. Modern
end. We appreciate the generosity of 3M Company Coating and Drying Technology. New York: VCH.
and Ilford AG for providing the means for Dowson, D. and Higginson, G. R. 1966. Elasto-
Hydrodynamic Lubrication. Oxford: Pergamon Press.
extensive communications between Minnesota Feng,1. Q. and Scriven, L. E. 1992. Electrostatic effects
and Switzerland by phone, fax and express mail. in continuous liquid coating. Paper read at AIChE
Most of all, we thank our wives, Karen and Sue, Spring National Meeting, New Orleans, LA, March
for their patience during six long years, and for 30-ApriI2.
their encouragement to keep on going, particularly Frost, P. J. and Gutoff, E. B. 1991. The Application of
Statistical Process Control to Roll Products. North
when we were tempted not to do so. Quincy: P. J. Associates.
Fukai, J., Zhao, Z., Poulikakos, D., Megaridis, C. M.
REFERENCES and Miyatake, O. 1993. Modeling of the deformation
of a liquid droplet impinging on a flat surface. Phys.
Barnes, H. A., Hutton, 1. F. and Walters, K. 1989.An Fluids A - Fluid Dynamics 5: 2588-2599.
Introduction to Rheology. Amsterdam and New Good, J. K. (ed.) 1993. Proceedings of the First
York: Elsevier. International Conference on Web Handling (Web
Benjamin, D. F., Scriven, L. E. and colleagues. 1991. Handling Research Center, Oklahoma State Uni-
Coating flows: form and function. Ind. Coating Res. versity, May 19-22, 1991). Stillwater, Oklahoma:
1: 1-37. Oklahoma State University.
Benkreira, H. (ed.) 1993. Thin Film Coating. Boca Gutoff, E.B. and Cohen, E.D. 1995. Coating and Drying
Raton: CRC Press. Defects. New York: John Wiley.
Berg, J. C. 1982. Interfacial hydrodynamics: an overview. Hansen, C. M. and Pierce, P. E. 1974. Surface effects
Canad. Metall. Quart. 21: 121-136. in coating processes. Ind. Eng. Chem. Prod. Res.
Berg, J. C. (ed.) 1993. Wettability. New York: Marcel Develop. 13: 218-225.
Dekker. Higgins, D. G. 1965. Coating methods. In Encyclopedia
Bird, R. B., Armstrong, R. C. and Hassager, O. 1987. of Polymer Science and Technology, Volume 3, eds
Dynamics of Polymeric Liquids - Volume 1: Fluid H. F. Mark, N. G. Gaylord and V. M. Bikales,
Dynamics. 2nd edn. New York: Wiley. pp.765-807. New York: Wiley.
Blake, T. D. 1984. Wetting. In Surfactants, ed. T. F. Katagiri, Y. 1992. Analysis of edge effects in curtain
Tadros, pp. 221-275. London: Academic Press. coating. Paper read at AIChE Spring National
Booth, G. L. 1970. Coating Equipment and Processes. Meeting, New Orleans, LA, March 29-April 2.
New York: Lockwood. Keers,1. 1. 1984. Static electricity. In Web Processing
Brandon, N. c., Hackler, M. A., Hirshburg, R. I. and and Converting Technology and Equipment, ed. D.
Christodoulou, K. N. 1994. Coatable splice process Satas, pp.511-523. New York: Van Nostrand
in photographic film coating: theoretical prediction Reinhold.
and experimental investigation of process latitude. Kheboian, G. I. 1992. Drive systems for off-machine
Paper read at AIChE Spring National Meeting, coaters. In Modern Coating and Drying Technology,
Atlanta, GA, April 17-21. eds E. D. Cohen and E. B. Gutoff, pp. 193-266. New
Caimcross, R. A., Francis, L.F. and Scriven, L.E. 1996. York: VCH.
Predicting drying in coatings that react and gel: Kistler, S. F. and Scriven, L. E. 1983. Coating flows.
Drying regime maps. AIChE. J.42: 55-67. In Computational Analysis of Polymer Processing,
Chappell, F. R. 1934. Spreading, extruding, or like eds 1. R. A. Pearson and S. M. Richardson,
operations. US Patent 2,052,131. pp. 243-299. London and New York: Applied Science
Churchill, S. W. 1992. The role of analysis in rate Publishers.
processes. Ind. Eng. Chem. Res. 31: 643-658. Landau, L. and Levich, B. 1942 Dragging of a liquid by
Coeling, K. 1. 1993. Spray coating. In Kirk-Othmer a moving plate. Acta Physico-Chim. USSR 17: 42-54.
References 15
Larson, R. G.1988. Constitutive Equationsfor Polymer Schweizer, P. M. 1992. Fluid handling and preparation.
Melts and Solutions. Boston: Butterworths. In Modern Coating and Drying Technology, eds E.
Lefebvre, A. H. 1988. Atomization and Sprays. New D. Cohen and E. B. Gutoff, pp.23-61. New York:
York: Hemisphere. VCR.
Macosko, C. W. (ed.) 1994. Rheology: Principles, Scriven, L. E. 1984. Coating flows and coating rheology
Measurements, and Applications. New York: research. Introductory remarks read at Annual
VCR. Review of Coating Flows and Coating Rheology
Melcher, J. R. and Taylor, G. I. 1969. Electrohyd- Research Program, University of Minnesota, Min-
rodynamics: a review of the role of interfacial shear neapolis, September 20 and 21.
stresses. Ann. Rev. Fluid Mech. 1: 111-146. Scriven, L. E. 1991. Bringing fundamentals to bear on
Middleman, S. 1977. Fundamentals of Polymer Pro- choosing a coating method. In Proceedings of the
cessing. New York: McGraw-Hill. IS&T Symposium on Coating Technologiesfor Imaging
Nadeau, G. F. 1956. Method for coating a liquid Materials (St. Paul, Minnesota, May 12-17), eds J.
photographic emulsion on the surface of a support. G. Truong and E. D. Cohen, pp. 79-85. Springfield,
US Patent 2,952,559. Virginia: IS&T - The Society for Imaging Science
Reynolds, O. 1886. On the theory of lubrication and and Technology.
its application to Mr. Beauchamp Tower's experi- Trapaga, G. and Szekely, 1. 1991. Mathematical
ments, including an experimental determination of modeling of the isothermal impingement of liquid
the viscosity of olive oils. Phi/os. Trans. Roy. Soc. droplets in spraying processes. Metall Trans. 22B:
Lond. A. 177: 157-234. 901-914.
Ruschak, K. 1. 1985. Coating flows. Ann. Rev. Fluid U.S. Department of Commerce. 1994. 1994 U.S.
Mech. 17: 65-89. Industrial Outlook. Bureau ofIndustrial Economics,
Satas, D. (ed.) 1984. Web Processing and Converting U.S. Government Printing Office.
Technology and Equipment. New York: Van Nostrand Weiss, H. L. 1977. Coating and Laminating Machines.
Rheinhold. Milwaukee, WI: Converting Technology Co.
PART ONE
PHYSICS AND MATERIAL INTERACTIONS

OF COATING PROCESSES
CAPILLARY HYDRODYNAMICS AND 2
INTERFACIAL PHENOMENA
Steven J. Weinstein and Harvey J. Palmer
2.1 INTRODUCTION
attention in this chapter will be given to the
In coating processes, one or more liquid layers specific physics involved with their appearance.
is typically applied, either individually or simul-
taneously, onto a substrate; after coating, these
2.2 PRELIMINARY CONCEPTS
liquid layers are ultimately transformed into a
solid coated structure via chilling, drying, or Before considering the conservation laws which
some other means. For much of a coating govern fluid flow, it is necessary to define what
process, then, the layers are in liquid form. To is meant by a fluid. A precise, all-encompassing
understand the advantages and limitations of definition of a fluid itself is difficult, in that the
such processes, it is imperative to have a particular physical configuration in which a
fundamental understanding of the physical material is found may impact the way in which
principles governing fluid flow. In many cases, it behaves from a classification point of view. We
these principles can be expressed concisely in define a fluid in the following way. The material
mathematical form, allowing the suitability of in question is placed between two parallel plates,
various coating operations for a desired product and a constant external force, F, is applied to
application to be assessed from a theoretical the top plate tangent to its surface while the
point of view. Once the agreement between bottom plate is held stationary. We define a fluid
theory and experiment is established for a given as a material which continually deforms in such
process, the need to perform experiments to a configuration, albeit with some resistance. This
explore the effect of process variables is greatly is in contrast to a solid, for which the distortion
diminished; the use of a theoretical analysis can of a material eventually ceases even though the
often represent significant savings in time and constant external force is applied. Some materials
money in evaluating various coating options. have characteristics of both fluids and solids,
In this chapter, the underlying physics of fluid such as those fluids that also exhibit elastic
flow is discussed, and the associated governing behavior, according to the above definition.
equations and boundary conditions are presented. Generally speaking, the equations of fluid
A distinguishing feature of coating flows, as mechanics can also handle these materials as
compared with many other flow configurations, fluids, provided the correct relation between
is the existence of fluid-fluid interfaces (i.e. force and distortion is employed. For practical
gas-liquid or liquid-liquid). Interfaces have a coating flows, those materials being identified by
marked effect on fluid flows, and therefore much phase as liquids and gases are fluids.
London. ISBN 0 412 06481 2.
20 Capillary hydrodynamics and interfacial phenomena
Wall
Ao~
Auid2
F10~ Fluid 1 Fl~
t
Wall
Fluid-Fluid
Interface
Figure 2.1 Multiphase flow configuration used to motivate the governing equations and boundary conditions.
In the above phenomenological approach to proceed to identify the equations and boundary
material characterization, the observed distortion conditions which govern the interfacial fluid
of the material determines its classification. This flow. The equations to be presented are quite
is consistent with the fact that fluid mechanics general and can be used for any flow configuration;
provides a macroscopic description offluid motion, they are not restricted to the flow domain shown
where average properties such as fluid density in Fig. 2.1.
and viscosity have physical meaning. Thus, Multiphase flows, such as shown in Fig. 2.1,
attention is confined to length scales larger than are a combination offlows of single fluid phases,·
those of molecular dimension. Once the length which are linked through contact at interfaces.
scale of a flow field is reduced to approximately Thus, in defining the governing equations offluid
ten times the distance between the molecules mechanics, which embody conservation of mass
comprising a material, the appropriateness of and Newton's laws of motion, the flow of a
modeling the material as a fluid is in question. single-fluid phase is examined. The unique
The smallest 'body' of fluid which can be character imparted to the flow by the interface
envisioned, but still within the constraints of a and solid boundaries enters the fluid mechanical
macroscopic continuum, is called a fluid point. description as boundary conditions, which are
The equations of fluid mechanics govern the presented subsequent to the consideration of the
motion of fluid points. governing conservation equations for a single
fluid phase.
2.3 MATHEMATICAL DESCRIPTION OF

FLUID FLOW 2.3.1 CONSERVATION EQUATIONS
To help motivate the mathematical treatment of 2.3.1.1 Conservation of mass

coating flows, let us consider the physical
configuration shown in Fig. 2.1, which is meant All the equations of fluid mechanics can be
to embody some arbitrary portion of a coating derived by applying conservation laws to collec-
process. Here, two immiscible fluid layers are tions offluid points, called material bodies, which
flowing in the domain as shown, where either of by definition always consist of the same fluid
the top or bottom walls may be moving. We now points as the boundaries of the material body
distort and translate with time (Whitaker 1968).
* In a fluid flow consisting of more than one fluid, it is Consider the motion of an arbitrary material
common to refer to each fluid as a phase, regardless of body, which can be viewed as being imbedded
whether it is a gas or a liquid. entirely in one of the fluid layers shown in
Mathematical description of fluid flow 21
Material Body Fluid
Set}
Vet}
Figure 2.2 Arbitrary material body for a single fluid phase.
Fig. 2.1. Let the volume of the material body be in space with equation (2.1) requires knowledge
denoted as V(t), where t denotes time (Fig. 2.2). of where the boundary of the body was at a
Further, let the fluid density be denoted by p, previous time and where it will be at a later time.
which is parameterized as p = p(X,t) where the Rather, it is usually more desirable to examine
spatial location within the material body is the consequences of conservation of mass for
denoted by the coordinates of the vector X; we volumes whose surfaces can deform arbitrarily
note here that the convention to be used in what in time (which also includes the case where the
follows is that quantities in boldface denote boundaries are fixed in time). Such arbitrarily
vectors and tensors. For a given differential deforming volumes are often referred to as control
volume element of the body, dV, the associated volumes.
mass is given by p d V; thus, the total mass M of The change in perspective from a material
the body is obtained as the sum of those differential body to a control volume is afforded by the
elements as well-known Reynolds transport theorem
(Whitaker 1968):
f pdV
M =
Vet)
As by definition, all the fluid points always

:t f
Vet)
«DdV= f ~~ f
Vet)
dV+
Set)
(f)U·ndS (2.2)
remain in the material body, it is clear that

where S(t) denotes the boundary of V(t), U is the
dM
<It =
d
dt
f pdV= 0 (2.1)
velocity vector of the fluid parameterized as
U = U(X,t), n is the unit outward normal to the
surface S(t), and «D is a quantity (scalar or vector)
Vet)
to which the theorem is applied. With the aid of
Equation (2.1) is an integral form of conservation (2.2), equation (2.1) can be rewritten as
of mass.
To apply equation (2.1), it is necessary to know
how the boundaries of the material body change
with time, as is evident from the time-dependent
f~ f
Vet)
dV +
Set)
pU·ndS =0 (2.3)
integration domain. This form, however, is often

inconvenient, as an assessment of the density which relates the rate of change of the mass of
distribution and velocity field for a specific region an arbitrary control volume of interest, V(t), to
the rate of material flow in and out through its 2.3.1.2 Conservation of linear momentum
boundary S(t).
Conservation of mass places one constraint on
The local differential form of the mass conser-
fluid motion, but to achieve flow there must be
vation law can be derived directly from (2.3) with
some driving force. The relation between applied
the help of the divergence theorem:
force and fluid flow is described by Newton's
f
S(/)
r'ndS = f
V(/)
V'rdV (2.4)
second law of motion, which states that the time
rate of change of the momentum of a body, L,
is equal to the net force on the body, F oel '
dL
where r is any vector. Applying (2.4) to the surface
integral in (2.3) produces the following result:
<it = F Oel (2.9)
f [~
As Newton's second law is valid for both
deformable and undeformable material bodies,
+ V'pU ]dV= 0 (2.5)
the equations of fluid motion can be derived, as
V(/)
for conservation of mass, by summing external
To obtain the final differential form, we argue forces on an arbitrary deformable material body
that, for a given fluid, the choice of V(t) can be as it moves in space. Referring again to Fig. 2.1,
made arbitrarily, and thus all material bodies attention is confined to a material body moving
must satisfy (2.5). This implies that the integrand entirely in one of the fluid layers. Note that it is
in (2.5) is identically zero. Thus, the local differential assumed that the motion of the material body
form of conservation of mass is given by is observed in a frame of reference which is not
accelerating, as (2.9) is valid only for such cases.
ap Using the material body notations shown in
-+V'pu=o (2.6)
at Fig. 2.2, the linear momentum associated with a
differential element, d V, of an arbitrary material
Equation (2.6) is commonly referred to as the body, V(t), is given by pU d Vand the total linear
continuity equation. momentum for the body, L, is thus given by
For coating flows, the density of the fluids
involved can be considered to be constant. In
the constant-density case, equation (2.6) becomes
V'U=o (2.7)
L= f
V(/)
pUdV (2.10)
The physical interpretation of this local equation Our material body may be acted upon by both
can best be made by utilizing the integral forms, body forces, F B , and surface forces, F s . Although
from which (2.7) was derived. For a constant there are several possible body forces, such as
density fluid, equation (2.1) yields those due to magnetic and electrical fields acting
on ferro-magnetic fluids, attention will be restricted
dV(t) = 0 here to the most common of such body forces
(2.8)
dt in coating flows, namely that due to gravity.
Denoting g as the constant gravitational accel-
Equation (2.8) indicates that the volume of any
eration vector, the body force is
arbitrary material body in the fluid does not
change with time; that is, the fluid is incompressible.
Thus, we see that (2.7) embodies the fact that
velocity components must be related in a unique
FB = f
V(/)
pgdV (2.11)
way to achieve a constant volume as a material

body is distorted. The surface force, F s , acts on the bounding
surface of the material body, and depends upon perspective from a material body (whose boundary
the identity of the fluid in contact with the distorts so as to always contain the same fluid
material body, the type of motion the fluid is points) to that ofa control volume (whose surface
experiencing (i.e., static, laminar, or turbulent), can deform arbitrarily in time). Note that this
and the orientation and shape of the surface S(t). process leads to a convection term, which in this
The surface force can be expressed in terms of case is the convection of momentum through the
the surface traction vector, ts ' which gives the boundaries of the arbitrary volume (the surface
local force per area on a differential surface integral on the left-hand side of (2.14)). Thus, we
element, as see that convective transport of momentum is
an artifact of changing from a balance of forces
on a material body, to a balance of forces on an
(2.12) arbitrary region of space.
The next step is to apply the divergence
theorem (2.4) to the two surface integrals in (2.14)
Upon substitution of (2.10), (2.11), and (2.12) into
to convert them to volume integrals (the divergence
(2.9), we obtain a general form of conservation
theorem is valid for vectors and tensors alike)
of linear momentum for a material body:
and to rewrite (2.14) under a single volume integral:
:t f pU dV= f tsdS + f pgdV (2.13a)

f [o~~ +V'PUU-V'T-pg]dV=O
V(I) S(I) V(I)
V(I)
A more useful form of the above conservation
equation can be obtained by explicitly determining (2.15)
the form of the surface traction vector. It can be Then, as in conservation of mass, it is argued
shown that there exists a stress-tensor, T, which that the choice of V(t) within the fluid is arbitrary,
when operated on the unit outward normal, D, and all such bodies satisfy (2.15). Consequently,
yields the surface traction vector (Whitaker 1968) as the integrand of (2.15) is identically zero:
t s = D'T (2.13b) opU
- + V'pUU = V'T + pg (2.16)
As for the case of conservation of mass, it is ot
desirable to have a differential equation which
A more widely used local form of conservation
describes the response of the fluid to applied
oflinear momentum can be obtained by employing
forces pointwise in a given domain. The local
the vector identity V' pUU = pU' VU +
form can be derived from the integral form by
UV' pU to the left-hand side of (2.16), and
following the approach used to obtain the local
simplifying the result by invoking conservation
form for conservation ofmass. First, the Reynolds
of mass (2.6):
transport theorem (2.2) is applied to the volume
integral on the left-hand side of (2. t 3a), to yield DU
p-=V'T+pg (2.17a)
Dt
f o~~ dV+ f pUU'DdS = f D'TdS where the operator DjDt, commonly referred to
V(I) S(I) S(I) as the substantial derivative, is defined as
+ f pgdV (2.14)
D
-=-+U'V
Dt ot
0
(2.17b)
V(I)
At this point, it is important to note that the
Recall that the transport theorem changes our physical meaning of the differential terms in
(2.17a) may be deduced by relating them to the final local form of conservation of linear mo-
corresponding integral terms in (2.13a). In mentum:
u.vuJ =
particular, recall that the integral term on the
left-hand side of (2.13a) gives the rate of change
of momentum in the body. Thus, the left-hand
p[aa~ + -VP + V'r + pg (2.19)
side of (2.17a) represents the time rate of change
One commonly used constitutive relation, which
of momentum of a fluid point, i.e. mass times
is often used in the examination of coating flows,
the acceleration, where we identify the total
is that ofa Newtonian fluid ofconstant viscosity, Jl:
derivative in (2. 17a) as the acceleration of a fluid
point as its trajectory is followed through space. 2
The correspondence between the integral and r= -3 Jl(V' U) I + Jl[VU + (VU)T] (2.20a)
differential terms containing the stress tensor
indicates that forces exerted on the surface of a where the superscript 'T' denotes the transpose of
material body as it moves through space are the quantity in parentheses. Because, as stated
translated into local variations in the internal previously, coating operations typically involve
stress of that body. Furthermore, according to liquids whose densities are constant, equation
the differential and integral balances, these stresses (2.7) is valid, and thus (2.20a) becomes
can give rise to a localized fluid motion inside
the material body. The gravity term on the (2.20b)
right-hand side of(2. 17a), which incorporates the The quantity in brackets is the rate of strain
local influence of the gravity body force exerted tensor; thus, (2.20b) states that the local viscous
on a material body, provides an additional stress on a fluid point is proportional to the local
source of fluid motion as well. rate of strain. Substituting this relationship into
In order to use the differential or integral forms (2.19) yields the well-known Navier-Stokes
ofconservation oflinear momentum, it is necessary equation (which is valid for a fluid of constant
to relate the stress tensor T to the type of fluid density and viscosity from the above assumptions):
u.vuJ =
and its motion. Such a relationship is called a
constitutive relation. Generally speaking, these
relations are determined experimentally. Any
p[')a~ + -VP + JlV 2 U + pg
constitutive relation can be postulated; however,
(2.21)
the justification for such a postulate ultimately
lies in the agreement between the resulting There are other important constitutive rela-
theoretical predictions and the experimental tionships which are relevant to coating flows,
observations for the particular fluid. It turns out although they typically are more complicated
that in all constitutive relations relevant to than (2.20). For example, fluids may exhibit
coating flows the surface forces are a sum of shear-thinning behavior in one or more parts of
pressure forces and viscous forces. For a positive a coating process; for such cases the general form
pressure, the pressure force on a differential given by (2.20) is valid, although the viscosity
element acts inward and normal to the element. itself is dependent on the magnitude of the local
Thus, all stress tensors can be written as rate of strain. Different constitutive equations
may also be necessary when fluid elements
T = -PI +r (2.18)
exhibit viscoelastic behavior as fluid elements are
where I and r are the identity and viscous stress stretched in a coating process. These cases are
tensors, respectively. As a consequence, efforts considered in more detail in Chapter 5. Note
to identify constitutive relations are focused on that the conservation equation (2.19) is valid for
determining the appropriate form of the viscous any of these cases, once the appropriate form of
stress. Substituting (2.18) into (2.17) yields the r is utilized.
2.3.1.3 Various limiting cases of governing flow regimes associated with the governing
equations equations (2.7) and (2.19); the presentation of
boundary conditions, as well as their simplifica-
The differential equations derived above, namely tions in various limits are left until Sections 2.3.2
conservation of mass and momentum, describe and 2.4, respectively (for a more comprehensive
the motion of each individual fluid phase found examination ofvarious flow regimes, see Schlicht-
in a coating process. The basic forces which can ing 1979).
affect fluid flow in a given fluid phase are To motivate our discussion, consider the
embedded in the conservation of momentum simplified coating process shown in Fig. 2.3.
equation (2.19); that is, surface forces due to Here, a large fluid reservoir is connected to a
pressure and viscous stresses, and body forces coating die (hopper) through a pipe of diameter
due to gravity, where the net sum of these forces D as shown. The fluid reservoir is open to the
determines the inertia of a material body of fluid. atmosphere having pressure P A' and the fluid
There are often portions of a coating flow field level in the reservoir is maintained at a height
in which one or more of these forces dominates H. The hopper itself is shown in cross-section,
over others; in other parts of the same flow field, and has a width W which is oriented into Fig.
these same forces may become negligible compared 2.3. Fluid enters the cavity of the coating hopper
with those previously neglected. It is thus often through the inlet, is distributed widthwise in the
possible to consider the flow field found in a cavity, flows through the slot, and forms a
complex coating process as a combination of two-dimensional liquid sheet (henceforth called
simpler, less complex flow fields having different a curtain), also of width J¥, which is subsequently
balances of these forces. Such an approach is coated on a substrate moving at speed U. We
useful because it allows for a coating flow field assume that the fluid being delivered is Newtonian,
to be understood and examined as a series of and so we utilize the Navier-Stokes equation
component parts, for which the physics is more (2.21) instead of (2.19), although the general
clear; this can endow the practicing coating arguments can be used when other constitutive
engineer with intuition needed to modify coating relations are employed. We furthermore will
processes to achieve a desired result. Furthermore, assume that the flow is at a steady state, i.e.,
from a mathematical point of view, numerical there is no time dependence in (2.21).
techniques are often necessary to solve (2.7) and Let us begin by considering the flow in the
(2.19) with the appropriate boundary conditions reservoir itself. As indicated in Fig. 2.3, the
(which have yet to be described). Numerical constant height of the fluid in the reservoir, as
solutions are often quite time consuming to well as the reservoir width, are extremely large
obtain owing to the nonlinear terms found in compared with the radius of the pipe connected
(2.19), the interdependence of the velocity com- to the reservoir. Thus, there is almost no apparent
ponents through mass conservation (2.7), and the fluid motion in the reservoir, except in the
complex flow geometries often encountered in a vicinity of the pipe. Under such conditions, we
coating process. By identifying regions of the can expect that the flow terms in (2.21) are small
flow in which certain forces dominate, small compared with those due to pressure and gravity,
terms in (2.19) can be neglected, and less-intensive and thus the equation governing the flow in the
numerical calculations can be used to solve the reservoir simplifies to
approximate equations; in some cases, these
equations are simple enough that analytic solutions
VP = pg (2.22a)
can be obtained. We note here that approximations This equation can be integrated directly to
must be made to both governing equations and obtain the pressure field in the reservoir:
boundary conditions in a self-consistent manner;
in what follows, we only consider some different P=PA+pgX (2.22b)
Atmosphere PA
t
Fluid Reservoir
H
t
Entrance Length
x (Developing Flow Region)
Pipe_ L
Hopper
I
Distribution
~Wi+-"";:;" Cavity ... c
8.,~
oo.Oll
..,
.'Holl¥7'n+-- Slot ::ccx
- - - - - - - - - - - - -t-t -
Liquid
Curtain
Entrainment Region
u------<·~
Substrate
Figure 2.3 The simplified coating process used to motivate the discussion of various limiting forms of the
governing equations.
where X represents the vertical distance below aspects of the flow to be considered. To begin,
the liquid surface (as shown in Fig. 2.3). In we recall that the flow of a Newtonian fluid in
obtaining (2.22b), the relation g = gj has been a straight pipe can be described by the well-known
employed, where g is the magnitude of the gravity Poiseuille relation between volumetric flow rate,
vector, j is the unit vector in the X -direction, and Q, and pressure drop across the pipe, !iP, which
X = 0 is the location of the air-fluid interface. is given by
Equation (2.22b) is the well-known expression
for the hydrostatic pressure in a fluid, which
Q = nR 4 (!iP + pgL) (2.23)
arises due to its weight. 8Jl.L
In the pipe region of Fig. 2.3, there are a few where L is the axial pipe length, and R = 0/2 is
the pipe radius. Such a flow is characterized by relationship between the pressure gradient and
a precise balance between the pressure, viscous the volumetric flow rate is different from that
and gravity terms in (2.21); the absence of density predicted by the Poiseuille relation (2.23). We
in (2.23), except in the hydrostatic pressure term, note here that developing flows will arise in any
indicates that fluid inertia does not influence the portion of a flow process in which the geometry
flow. The Poiseuille relation provides a valid suddenly changes; in the current example, they
description of the flow field only if the fluid flow arise in the vicinity of the junction between every
is laminar, i.e., the flow proceeds in an orderly region in Fig. 2.3.
fashion, in which fluid elements proceed along Let us now consider the forces at play in the
parallel streamlines. This is the case when the specific developing flow region associated with
pipe Reynolds number, Re, is less than a minimum the flow into the pipe from the reservoir. An
critical value: expanded view of this region is shown in Fig.
2.4a. In this region, the pressure, viscous and
Re = p VD < 2300 (2.24) gravity forces must be important, since we know
J-l that the developing flow field must ultimately
approach the limit of fully developed Poiseuille
where V is the average of the fluid velocity in flow at the end of the developing region, in which
the pipe and D is the inner diameter of the pipe. these forces precisely balance. However, in the
For flows which violate this criteria, small entrance region to the pipe, fluid streamlines are
disturbances which naturally arise in a process converging and the flow is not unidirectional;
(such as vibrations) can induce instabilities in the that is, there is both an axial and a radial
laminar flow; these instabilities grow, and the component of fluid velocity, denoted by U x and
orderly streamline pattern oflaminar flow is lost. U r , respectively. The interrelationship between
In its place, a time-fluctuating turbulent flow these components is given by the continuity
arises. We note here that the time-dependent equation (2.7) expressed in cylindrical coordinates
nature of turbulence often precludes the use of
extremely high Reynolds number flows in a t 0 oU x
coating process. For example, if the turbulence -;: or (rU r ) = - oX (2.25)
occurs in a region of the coating process which
is near the location where fluid layers are deposited where r is the radial location in the pipe. Equation
on the substrate, the time-dependent flow can (2.25) shows that the presence of a radial
induce undesirable thickness variations in the component of velocity causes the velocity in the
coated fluid layers. axial direction, U x , to vary with axial location
Another important assumption in using the X; that is, the fluid will accelerate or decelerate.
Poiseuille relation (2.23) is that the pipe is long Thus the inertial terms in (2.21), which quantify
enough for the laminar flow to arrange itself into the acceleration of fluid points (see discussion
a fully developed configuration; when fully following (2.17», may be important in the
developed, the flow is in the X-direction only developing flow region. Downstream, where the
and the velocity field is independent of axial radial component of velocity Ur approaches zero
distance X from the reservoir. Near the inlet to and the flow becomes fully developed, equation
the pipe in Fig. 2.3, of course, there is some axial (2.25) indicates that U x is independent of axial
distance, called the entrance length, over which location X; this indicates that the flow is no
the flow adjusts to the influence of the pipe walls longer accelerating, and inertial terms in (2.21)
as it enters the pipe from the reservoir. Here, the disappear.
fluid velocity profile does indeed change with The relative importance of the viscous and
axial location, and the flow is termed a developing inertial terms in the developing flow region can
flow. Because of this flow rearrangement, the be inferred from a simple dimensional analysis
of the viscous and inertia terms in (2.21). To this this entrance length and the associated flow can
end, we scale U ~ V, V ~ liD and thus obtain be greatly influenced by fluid inertia. Under such
circumstances thin boundary layers form along
pU·VU the walls of the pipe, in which the viscous drag
Re ~ jlV 2 U (2.26)
of the walls affects the flow (Fig. 2.4a). In these
regions, all of the forces represented in (2.21) are
where Re is the Reynolds number given by (2.24).
important, although the velocity gradients in the
Thus, for small Reynolds numbers, fluid inertia
radial direction dominate the magnitude of the
is negligible compared with viscous forces; for
viscous terms. In the interior of the pipe away
large Reynolds numbers, the opposite is true.
from the influence of these boundary layers,
When Re is small, an approximate form of(2.21)
viscous effects are negligible, and thus the
can be used to analyze the developing flow, in
governing equation (2.21) becomes
which the inertial terms of (2.21) are neglected:
(2.27) pU·VU = -VP + pg (2.28)
Flow fields satisfying (2.27) are termed creeping Equation (2.28) is called Euler's equation, which
flows. For a creeping flow, the length of the governs the flow of fluids considered to be
developing flow region is relatively small, being inviscid. It is not until the influence of the viscous
on the order of a few pipe diameters. forces actually reach the center of the pipe, i.e.
On the other hand, for high Reynolds numbers, the boundary layer thickness is on the order of
Flow in
Free-Fall
Reservoir
Flow
\ ~ I I
-.+--- Cunain
\ I
\ Inviscid Region I Inviscid Flow Atmosphere
\ I Pipe Wall Region
\ I
\ I
I I Boundary Layer
Boundary Layer
Region
,
,
, I Region Boundary Layer Region
II
I
r'
X
r
I
I
u-----
I
, - Plug Flow
I I
Substrate
(a) (b)
Figure 2.4 Details of the coating process illustrated in Fig. 2.3: (a) the developing flow region found as fluid
moves from the reservoir into the pipe under conditions of high Reynolds number, where thin boundary layers
form along the pipe walls; (b) the entrainment region near the moving substrate, under conditions of relatively
high substrate speed relative to the impingement speed of the curtain.
the pipe radius and the inviscid region disappears, Note that once the velocity potential is determined
that the flow can rearrange and become fully by solving to (2.3Oc), the pressure in the flow can
developed. Because the boundary layers are be extracted directly using (2.29); this pressure
initially quite thin, the distance it can take for provides a driving force for the flow in the viscous
the flow to become fully developed can be boundary layers near the pipe walls at high
significantly larger than for the creeping flow case. Reynolds numbers.
We now consider some important simplifica- Returning to Fig. 2.3, after flowing through
tions to Euler's equation which are relevant to the pipe region, the fluid enters the hopper
the developing flow. If we choose the perspective region. In the hopper itself, the role of the cavity
of moving along a fluid streamline, which by is to distribute fluid uniformly widthwise (i.e.,
definition is everywhere tangent to the velocity into Fig. 2.3) so that the local volumetric flow
vector, then Euler's equation can be integrated rate does not vary across the hopper; this is
along that given streamline in the inviscid region because flow non-uniformities ultimately can
to obtain Bernoulli's equation: lead to undesirable thickness variations in fluid
layers on a substrate after coating. To accomplish
V2 P this, a hopper is typically designed such that the
-+--gX=K (2.29)
2 p resistance to flow in the slot is much higher than
that in the cavity itself so that any lateral pressure
where V is the magnitude of the velocity vector variations incurred as the flow distributes
U, and K is a constant. In general, the constant widthwise in the cavity will be a negligible
in (2.29) varies with position in the fluid, depending percentage of the mean pressure in the cavity.
only on the particular streamline. For the current Thus, a hopper typically has a cavity of large
problem, however, we note that in the reservoir cross-sectional area, a small slot height, b, and
itself the hydrostatic equation (2.22) is identical a large slot length, L H • Furthermore, the width
in form to (2.29) when V is small; as the inviscid of the hopper, W; which determines the ultimate
region in the pipe does link directly with the flow coated width of the fluid on the substrate, is
in the pressure reservoir, we deduce that the typically on the order of L H or larger. In the ideal
constant in Bernoulli's equation is precisely the case where the cavity pressure is constant, a
atmospheric pressure PA for all locations in the constant pressure drop tiP is applied across the
central inviscid region of the pipe. slot in the flow direction, as the hopper is opened
When the Bernoulli constant K does not vary to the atmosphere. For such a flow, the volumetric
from streamline to streamline in the fluid, as in flow rate, Q, is related to the constant pressure
the current case, the flow itself is irrotational, drop as
i.e., it satisfies
VxU=O (2.30a) (2.31)

which implies the existence of a scalar velocity
potential 4J as We note here that this expression is the planar
analogue ofthe fully-developed Poiseuille relation,
U= V4J (2.30b) which gives a balance between viscous, pressure
and gravitational forces. As b is very small, the
Combining this equation with conservation of
pressures generated inside the hopper can be
mass (2.7), it is seen that the velocity potential
quite large for moderate flows.
satisfies Laplace's equation, i.e.
We acknowlege here that the flow in the
(2.3Oc) hopper is almost never perfectly uniform, owing
in part to the fact that there is always a pressure
Flows satisfying (2.30) are termed potential flows. variation in the hopper cavity as the fluid is
distributed widthwise*. When this occurs, there and assembly of the hopper. When the slope of
is a widthwise component of velocity which is the opposing surfaces of the slot is small (e.g., b
imparted to the fluid in the slot, and the flow is varies slowly with X), the flow field in the hopper
no longer unidirectional in the X -direction, as can again be described by the creeping flow
described by (2.31). As discussed previously, once equations (2.27), with the simplification that the
a flow is no longer unidirectional, the prospect viscous terms are dominated by the velocity
arises for inertial effects to become important in gradients across the gap. Furthermore, the flow
the flow field. We now deduce the magnitude of profile across the gap itselflooks identical to that
inertial effects in the slot. It is intuitive that in a slot of constant height, except that the
velocity gradients across the narrow gap will be location of the slot walls is adjusted locally. This
much larger than those in the flow direction itself, local variation in wall location yields a modifi-
owing to the viscous drag of the walls. In fact, cation to the pressure gradients in the slot as
for a characteristic slot flow velocity V, the well. These flow characteristics are predicted by
viscous terms in (2.21) will have magnitude utilizing the hydrodynamic theory oflubrication,
Jl Y/b 2 , where b is the slot height, while the inertial which is examined extensively in Chapter 8. We
terms will have magnitude py 2 /L H , where LH is comment here that, more generally, this theory
the slot length. Consequently, the Reynolds is widely used to analyze various coating flows
number characterizing such a flow, as defined by with fluid-fluid interfaces, which typically have
(2.26) is given by small characteristic aspect ratios similar to those
for the slot described above, and for which the
pVb 2 slope of the fluid streamlines is small.
Re=-- (2.32)
~Jl After emanating from the hopper, the fluid
enters the curtain region, in which a liquid
Based on the previously described slot geometry, curtain of width W (extending into Fig. 2.3) is
the Reynolds number given by (2.32) will be small formed. In this region, the viscous drag of the
in a typical hopper, indicating that inertial effects air on the fluid is negligible under typical
are negligible in the slot. Thus the flow in the conditions, and thus there is no mechanism by
slot is governed most generally by the creeping which velocity gradients perpendicular to the
flow equations, given by (2.27), although these streamwise flow can be maintained. As is typical
equations can be simplified further by noting of all transitions in flow geometry, the fluid flow
that viscous terms in (2.27) are dominated by the issuing from the hopper must adjust to this new
velocity gradients across the gap. Note that a environment, with its absence of viscous drag at
detailed discussion of the design of coating the boundaries. Consequently, there is an entrance
hoppers is considered in Chapter 10. length equal to several times the hopper slot
Before leaving our discussion of the hopper height b over which the flow rearranges itself
slot, we consider the complications to the flow from the parabolic, Poiseuille flow in the hopper
which can arise when the hopper slot height, b, slot to the fully-developed, flat velocity profile
varies from position to position. Such a variation (no velocity gradients) in the curtain. This flat
might occur due to the internal pressure loading velocity profile is termed plug flow.
on the hopper itself, causing a distortion in the Because the fully developed curtain flow is
slot cross-section, or due to errors in the fabrication plug at all streamwise locations, the flow field in
the curtain region of Fig. 2.3 is effectively inviscid
* Although the internal hopper geometry can be adjusted to
counteract these pressure variations, such adjustments can and is governed by Euler's equation (2.28), as
only assure uniformity for one particular flow rate and fluid well as Bernoulli's equation (2.29) along each
type. A single hopper may be called on to deliver many streamline. Let us reposition the coordinate
different fluids and/or flows; the effect of pressure variations
corresponding to these different conditions cannot be cancelled system from the reservoir region (as shown in
by the same hopper geometry. Fig. 2.3) into the curtain region, such that X = 0
is located at the position below the hopper where ance of various terms can often be identified by
the curtain flow first achieves plug flow. If we utilizing a dimensional analysis, such as that used
denote the curtain speed there as Vo, then the in obtaining estimates for the Reynolds number
Bernoulli constant in (2.29) will be in (2.26) and (2.32) in the above discussion, and
assessing the magnitude of the resulting dimen-
v~ PA sionless groups. Such an approach is also utilized
K=-+- (2.33a) in Section 2.4, where we examine various limiting
2 P
cases of the boundary conditions at fluid-fluid
which is the same for all streamlines at X = 0, interfaces and their potential impact on coating
and hence, throughout the whole curtain. We flows. A formal framework for approximating
also note that the pressure experienced by the dimensionless equations in various limits is
fluid in the curtain is essentially atmospheric afforded by the powerful technique ofasymptotics,
across the thickness of the curtain; i.e., P = P A which is discussed extensively in Chapter 8.
everywhere. Substituting this into Bernoulli's
equation (2.29) and utilizing the Bernoulli constant
(2.33a) results in the following relationship for
2.3.2 BOUNDARY CONDITIONS
the velocity in the curtain:
In the previous discussion, we described how a
V 2 = V~ + 2gX (2.33b) flow field can change depending upon the relative
importance of various terms in the mass and
Flows which satisfy (2.33b) are said to be in free-fall. momentum conservation equations. In that
The curtain itself ultimately impinges on the discussion, we determined the magnitude of these
moving substrate in the entrainment region, as terms by using arguments which implicitly
shown in Fig. 2.3. Under conditions where the depended upon the interaction of the fluid flow
substrate speed, U, is significantly greater than with the solid walls and air which bounded the
the impingement speed of the curtain, the viscous flow domain. This underscores the physical
effects of the substrate motion are confined to a importance that the boundaries of a flow domain
boundary layer in the vicinity of the web, as play on a fluid flow itself. To emphasize this
shown in Fig. 2.4b. Outside of this boundary point further, we note that the integral and
layer, the fluid flow in inviscid and smoothly differential forms ofthe conservation laws derived
links with the free-fall flow valid in the curtain thus far are generally applicable to each phase
region above the web. We have already discussed of a single-phase flow. Thus, for the domain
the features of boundary layer flows and inviscid shown in Fig. 2.1, conservation of mass and
flows in the context of the developing flow in the conservation of linear momentum are applied to
pipe, and the basic physics is essentially the same each layer separately. Clearly then, for the case
in this impingement region. Eventually, all the in which two distinct fluid phases come in
fluid flow in the curtain is entrained in the contact, the unique character of the resulting
boundary layer (i.e., the inviscid region disappears), fluid flow is determined by the conditions applied
and ultimately, downstream on the web, all of along the fluid-fluid interface, such as shown in
the fluid moves at precisely the web speed U. Fig. 2.1. The physical information introduced via
This ultimate plug flow profile arises due to the the boundaries fulfills a mathematical requirement
lack of viscous drag of the air on the fluid. that boundary conditions are necessary to uniquely
The above example indicates how the import- solve the governing differential equations of fluid
ance of various terms of the complete governing mechanics. In what follows, we consider the
equation (2.21) can change throughout a coating various boundary conditions which are generally
process, and how markedly different flows can applicable to both single-phase and multi-phase
arise when different terms dominate. The import- coating flows.
2.3.2.1 The kinematic boundary condition:

conservation of mass across the boundaries of
the flow domain
The kinematic boundary condition provides a

constraint on a fluid flow such that conservation
+ :t f P2 dV + :t f rdS = 0 (2.34)
V2(t) A(t)
of mass is satisfied across the boundaries of the
where Vt(t) and V2 (t) are the volumes occupied by
flow domain. The general condition is obtained
phases 1 and 2 within the material body at time
in a manner that resembles our derivation of the
t, A(t) is the interfacial area in V(t), and r
continuity equation (2.6). In particular, we start
represents any excess mass per unit area associated
by focusing our attention on an arbitrary deform-
with the interface itself. Although V(t) is a
able material body V(t) as it moves in space.
material body, Vt(t) and V2 (t) are not, because
However, in contrast to the previous approach
generally speaking mass can transfer across the
to the conservation laws, the body is now positioned
interface (e.g., the case of solvent evaporation
such that it straddles the interface between two
from the coating during flow). For such arbitrary
fluids, as shown in Fig. 2.5 (Slattery 1967).
volumes, the time derivatives in (2.34) can be
Because V(t) is a material body, its mass M
brought inside each integral with the help of the
is independent of time, and thus the ~as~
general transport theorem (Whitaker 1968), which
conservation condition (2.1) is necessarily satisfied.
is written for the density of each phase as
f f ~.a/ f
Unfortunately, because the density is discontinu-
ous in V(t), we cannot use the Reynolds transport
theorem (2.2) to adapt this equation to the d
dt PjdV= dV + Pjwj'ndS
perspective ofan arbitrary control volume. Instead, Vjlt) Vjlt) SP)
we must rewrite (2.1) as the sum of three integrals: (2.35)
one for phase 1, one for phase 2, and one for the
interface itself: where for the current problemj = 1 or 2, and w·J
Auid2
Fluid 1
Figure 2.5 Arbitrary material body straddling an interface B(t) that separates two fluid phases.
is the velocity of the surfaces Sit), which bounds defined in the direction normal to the surface;
°
J.j(t), as it deforms in time. In addition, is the thus V" = U"o where U" is the magnitude of the
unit outward normal to the surface Sj(t). A vector V". Using these facts, we can separate
comparison of (2.35) with (2.2) reveals that the each surface integral in (2.36) into two parts, to
Reynolds transport theorem is a special case of obtain the following result:
the general transport theorem; the Reynolds
transport theorem applies to a material body
whose surface, by definition, moves with the local
f a~t f
VIet)
dV+
A/(t)
PtVt oodS + f
A(l)
PtV"OOt dS
velocity of the fluid (i.e., wj = V j ) so that the
dV+ f
body itself always consists of the same fluid
points, while the general transport theorem applies + f ap2
at P2 V 2 °O dS
to a body whose boundary motion is, in general,
f
V2(t) A2(t)
different from that of the local fluid velocity.

Employing (2.35) in the integrals in (2.34), we + P2 V ,,002 dS =O (2.37)
obtain A(t)
°
where is the outward unit normal to the surface
of the material body S(t), and 0t and 02 are the
unit normals to the interface A(t) pointing into
f
phases 2 and 1, respectively, where 0\ = -02 (as
indicated in Fig. 2.5).
+ P2w200dS=O (2.36)
To reduce equation (2.37) to its local, differential
S2(t) form, the volume integrals are rewritten by
employing the continuity equation (2.6) and the
where we have neglected the mass of the interface,
divergence theorem (2.4) to obtain
as it is small relative to the terms shown in (2.36).
Note that St(t) can be represented as the sum of
the surface area of V(t) that resides in phase 1,
denoted as At(t), plus the interfacial area, A(t),
separating the two phases in the material body.
Similarly, S2(t) equals A 2(t) plus A(t), where Ait)
is that portion of the surface of the body residing
f
A/(l)
ptVtOodS- f
A(t)
PtVtOotdS (2.38a)
in phase 2. Using this definition, we note that

along the surfaces At(t) and A 2 (t), the boundary
velocity is equal to the local fluid velocity, i.e.
wj = V j in each phase; this is because, taken
together, these surfaces comprise the outer
boundary of the material body across which no
fluid points can move (Fig. 2.5). On the other
f
A2(t)
P2V200dS - f
A(t)
P2V2002dS (2.38b)
hand, the interface A(t) is not in general a

If we now combine (2.38) with (2.37) and rearrange,
material boundary (i.e., fluid may pass across it).
the following equation is obtained:
f
Thus, the local fluid velocity at the interface may
be different from the rate of displacement of the
interface itself. For this reason, we designate the {Pt(V t - V,,)oOt
rate of displacement of the interface as V" to A(t)
distinguish it from the local fluid velocity at the
interface, V j • Consequently, wt = W 2 = V"along
A(t). The rate of displacement of the interface is Because the above equation must hold for arbitrary
choices of the material volume, it follows that where i, j, and j{ are unit vectors in the X, Y,
the integrand must be identically zero: and Z directions, respectively. Also, the magnitude
of the rate of displacement of the surface, Va' is
P1(U 1 - U a )o01 - P2(U 2 - U a )o01 = 0 given by the relationship
at the interface. (2.40)
aB
where in writing (2.40), we have explicitly used
Va = -
aH}_j o01 == {aB}_
{at at
at j o01 == IVBI (2.43)
the fact that 01 = -02 along A(t).
Equation (2.40) is the general form of the
kinematic boundary condition (also called the Combining (2.4 t) with (2.42) and (2.43), we obtain
jump mass balance), and it applies at both the following general expression for the kinematic
permeable and impermeable boundaries. If, for condition at a fluid interface with negligible mass
example, phase t is an impenetrable solid, its transfer:
normal component of velocity, U 10°1' will be
identical to the normal component of the rate U
2
0VB = aB
at (2.44a)
of displacement of the surface, U a 0 1 = Va' o
Thus, (2.40) simplifies to the intuitively pleasing

result that (2.44b)
U 2 °0 1 = Va at the surface, (2.4ta) In component form, the kinematic boundary

meaning that the fluid velocity normal to the condition (2.44) is thus written as
solid surface is the same as the rate ofdisplacement aH - V aH aH
of the solid surface itself. If, on the other hand, Vy - Vxax z az = at at X = H(Y,Z,t)
phase 1 is also a fluid, so that the solid boundary
is replaced with a fluid-fluid interface across (2.45)
which there is no mass transfer, then (2.40) where Vx , Vy , and Vz are the components of
indicates that phase 1 satisfies U in the i, j, and k directions, respectively. In
U 10°1 = Va at the interface, (2.4 t b) writing (2.45), the subscripts denoting the phase
have been neglected, as the result is generally
while for phase 2, equation (2.4ta) is still valid. valid for either phase. Equation (2.45) places a
Let us now consider the application of (2.41) restriction on the fluid flow adjacent to a deforming
to an impermeable fluid-fluid interface with a interface, consistent with its impenetrability. Notice
shape as shown in Fig. 2.1. Here, the thickness that interfacial movement, which superficially
of fluid 1 is parameterized as Y = H(X, Z, t), appears as a rate of displacement in the j
where X is the horizontal coordinate measured direction, aH/at, actually resolves itself into a
in the flow direction, Y is the vertical coordinate fluid flow that has X, Y and Z components of
measured upwards from the bottom wall, and velocity at the interface. It is only when the fluid
the Z direction is oriented out of Fig. 2.1. From interface is horizontal, that the vertical component
vector calculus, we know that the normal vector of fluid velocity equals the rate of vertical
to the surface is determined by the relation 01 = VB/ displacement of the interface, aH/at.
IVBI, where the vertical bars denote the length Generally speaking, the kinematic boundary
of the vector, and B = Y - H(X, Z, t) = O. Thus, condition for an impenetrable surface (2.44) is
adequate for describing coating flows, even if
j_ aH i _ aH j{ interphase mass transfer is taking place, as long
VB ax az as the mass transfer process does not significantly
(aH)2 (aH)2J1 /2
(2.42)
°1 = IVBI = [ alter the thickness of the coating flow in the
1 + ax + az region of interest. When the local interphase
mass flux is significant, (2.40) provides a complete restricted to the region in the close vicinity of
description of the interrelationship between the the line of contact between the interface and the
local fluid velocities normal to the interface and solid surface. As fluid-fluid interfaces are imbed-
the rate of interfacial displacement. In particular, ded in virtually all coating flows, the mathematical
if we consider a flat interface that appears difficulties associated with the use of (2.47) are
stationary from our frame of reference, U" equals quite relevant. This issue will be discussed in
zero and (2.40) simplifies to the following result: more detail in Chapter 3.
P2(U 2 'n 1) = Pl(U1'n 1) (2.46)
2.3.2.3 Combined application of kinematic
which states that the mass flux leaving phase 1
and no-slip conditions to an impermeable
must equal the mass flux entering phase 2.
fluid-fluid interface
The preceding two boundary conditions, in
2.3.2.2 The no-slip condition
conjunction with the previously derived local
Thus far we have discussed a relationship between conservation equations, are all that is necessary
the normal component of velocity in each phase to describe flow of a single fluid phase in contact
at an interface, and its relationship to the rate with a rigid solid boundary. For multiple phase
of displacement of the interface itself. There is flow, these boundary conditions apply as well.
also a need to constrain the motion of fluid points Consider for example the fluid-fluid interface
in each phase along an interface. A widely used shown in Fig. 2.1. Because the no-slip condition
assumption in viscous flows is that the tangential applies at the interface, equation (2.47) holds.
component of velocity is continuous across Furthermore, the kinematic boundary condition
bounding surfaces; this assumption is applied to must apply for both fluids at the fluid-fluid
both solid surfaces and fluids alike. In other interface. In multilayer coating flows, such as
words, the tangential component of velocity for that implied by Fig. 2.1, the rate of mass transfer,
phase 1 is assumed to equal the tangential if any, between the various layers is usually small
component of velocity of phase 2 on their surface enough to have a negligible impact on the flow
of contact. That is, profiles, and thus the kinematic condition for an
impermeable surface is given by (2.41). Combining
(2.47)
(2.41a and b), we obtain
where t is a tangent to the interface. If the
bounding surface is a solid, typically the motion
U2 •n = U1•n at the interface, (2.49)
of the solid is prescribed as Us, and (2.47) is where we have removed the subscript on the
written as: normal vector written in (2.41); it is understood
here that n is a unit normal to the surface. The
(2.48)
conditions (2.47) and (2.49) together imply that
We comment here that this condition yields flow
predictions in excellent agreement with experiment
U1 = U 2 at the interface, (2.50)
for a wide range of physical configurations. There i.e., the velocity must be continuous across the
is one notable exception, however, where math- impermeable interface. Note, however, that (2.50)
ematical analyses of flow which utilize this says nothing about the impenetrability restriction
boundary condition run into difficulties; namely, which must arise at the fluid-fluid interface.
flow fields involving a moving fluid - fluid interface Thus, the kinematic boundary condition for one
where the interface makes contact with a solid of the phases (given by (2.41a) or (2.41b)) must
wall. Even in such cases, however, the no-slip be applied together with (2.50) to provide essential
condition can be employed over the bulk of the information about the movement of the interface
domain, the mathematical difficulties being as well as the two bulk phases at the interface.
2.3.2.4 The dynamic boundary condition: the line of contact between the interface and the
the balance of interfacial stresses boundary of the material body. Denoting B(t) as
the location of the interface, the line of contact
Thus far, the boundary conditions discussed is given by the intersection of B(t) with the surface
have addressed purely kinematic (motion) con- of the material body S(t); this intersection is
siderations. The way in which stresses transmit denoted by C(t) in Fig. 2.5.
across an interface or, equivalently, the way in The force, Fa' that the interface exerts at the
which forces balance on an interface needs to be line of intersection C(t), can be expressed in terms
considered as well. In what follows, the balance of the line traction vector, t a , which gives the
of forces on a material body containing two local force per unit length on a differential line
distinct fluid phases is considered. The local or element, dC:
differential form of such a force balance yields
the desired stress boundary condition at the (2.51)
interface.
The approach taken is identical to that for
conservation oflinear momentum. External forces We note here that the expression for the line
are summed on an arbitrary deformable material force Fa is entirely analogous to that for the
body as it moves in space. However, because we surface force given by (2.12). The total force
want to develop a stress balance at an interface, balance can be written directly, by employing
we position our material body such that it (2.9) and including as an additional term (2.51)
straddles the interface between two fluids (see in the right-hand side of (2.13):
Fig. 2.5 and our derivation of the kinematic
boundary condition). Those forces already identi-
fied for a material body consisting of a single
:t f pUdV= f n·TdS
f f
V(I) S(I)
phase, namely the body force and surface force
given by (2.11) and (2.12) respectively, are of + pg d V + t a dC (2.52a)
identical form for the new material body; similarly,
the momentum accumulation term given by (2.9) V(I) C(I)
and (2.10) is of identical form as well. However, The equation above is an integral form of
because our material body now consists of two conservation of linear momentum for a body
separate phases, each with its own distinct physical consisting of two fluid phases and an interface.
properties, an evaluation of the applicable integral A more useful form of this conservation equation
terms require that each integral be divided into can be obtained by explicitly determining the
three integrals, one for each phase and one for form of the line traction vector. In an analogous
the interface itself. way to the surface traction vector defined in
If the interface does not exert some sort of (2. 13b), it can be shown that there exists an
force on the material body shown in Fig. 2.5, interfacial stress tensor, T a , which when operated
then the single-phase integral conservation law on the unit vector, m, yields the line traction
(2.13) is valid. Indeed, there are a number of vector as
physical situations for which this is true, such as
(2.52b)
in fluid-fluid systems where the same solvent is
employed in each phase. More generally, however, where m is perpendicular to C(t) and tangent to
flow and interfacial phenomena are observed the interface B(t) (see Fig. 2.5).
that can be explained only if the interface is The initial motivation for obtaining (2.52) was
presumed to exert a force on the material body. to derive a boundary condition to apply at the
To account for this force, the balance (2.13) must interface when considering the local differential
include an additional term, which acts only along equations governing fluid flow. To this end, we
M athematieal description of fluid flow 37
are led to derive a local form of (2.52), with which yields
attention focused on the interface itself. A general
expression for the local form of the balance of
stresses at the interface can be obtained by
f [01'(T2 -T 1)+VII 'T"JdS=0 (2.54)
A(r)
following the procedure used in the previous
section for the derivation of the kinematic Here, VII is the divergence operator resolved in
boundary condition; that is, by breaking up the the tangential plane, which is related to the three-
integrals of (2.52) into parts that apply to each dimensional divergence operator as
of the two phases and the interface, and then
(2.55)
using the general transport theorem in conjunction
with the momentum equation (2.17) for each phase. Then, it is argued that as A(t) is arbitrary, the
An alternative, intuitive approach is to apply integrand must be zero such that (2.54) holds for
(2.52) to a material body drawn about the all material bodies that incorporate part of the
interface, whose local thickness is given by An, interface. Consequently,
and whose top and bottom surfaces have areas
A 2 (t) and A 1(t), respectively, as shown in Fig. 2.6. ° 1 ' (T2 - T 1) + VII' T" = 0 at the interface.
Now, consider the limit as An-O. In such a (2.56)
limit, A 1(t), A 2 (t)-A(t), where A(t) is the area
of the resulting surface, with sides 1 and 2 The result (2.56) is called the 'dynamic boundary
embedded in the interface B(t). In this limit, the condition', so called because its origin lies in a
material body has no volume, and thus the force balance. Basically, it says that if the interface
volume integral contributions to (2.52) are zero*. has the ability to exert a force on a material
Equation (2.52) can be written for a material body, which is characterized by a non-zero value
body by denoting 01 and 02 as the normal to of T", then a jump in stress across the interface
the interface which points into fluids 2 and 1, arises. On the other hand, if the interface has no
respectively (see Fig. 2.6), and noting that 02 = -01 force-inducing properties, then T" = 0 and (2.56)
at each point along A(t). The result is becomes
f °1'(T 2 -T 1)dS+ f m'T"dC=O (2.53)

01'(T 2 - T 1) = 0 at the interface,
which indicates that the stress is continuous
(2.57)
A(I) C(I)
across the interface.
The procedure to obtain the local form of (2.53) As in the case of the surface stress tensor T, a
is to combine both integrals under a single one. constitutive relation must be defined to relate T"
This can be done with the divergence theorem to physical quantities associated with the interface
for a surface (eJ, Scriven (1960), and the justification for any
f
ql)
m'T"dC =
A(I)
f Vn'T"dS
such relation lies again in the agreement between
theory and experiment. The simplest of such
constitutive relations is given by
• Note that the left-hand side of equation (2.52) is negligible (2.58a)

only if either the mass density of the two phases are the same
or the rate of mass transfer across the interface is negligible. where (J is the surface tension and III is the
Otherwise, there will be a finite rate of change in the total identity tensor in the tangential plane, defined
linear momentum of the body due to the phase change
associated with interphase mass transfer, as the size of one in terms of the three-dimensional identity tensor,
phase in V(t) grows at the expense of the other, even for I, as
vanishingly small values of An. In ignoring the body force
term, we are assuming that the excess mass at the interface (2.58b)
associated with the interfacial adsorption of surfactant is
insufficient to contribute measurably to the force at the interface. Substituting (2.58) into (2.56), applying various
Figure 2.6 Material body used to derive the local form of the dynamic boundary condition.
tensor identities, and employing (2.18) yields intrinsic viscous and elastic properties. These
features can be accounted for through more
complex constitutive relations for T a .
At this point, let us give a physical interpretation
at the interface. (2.59a) of the mean radius of curvature R m in (2.59). The
In (2.59), R m is the mean radius of curvature of mean radius of curvature of an interface at a
the interface, which is related to the unit outward given point may be visualized by first erecting
normal as two perpendicular planes whose line ofintersection
is coincident with the outward normal at that
1 point (Fig. 2.7). The intersection of the interface
Ii: = VII· 01 (2.59b)
m with each plane yields a curve. The mean radius
For the case of zero flow, the only nonzero of curvature is related to the curvature of these
terms in (2.59a) are in the normal direction, and two curves at the point of interest. The curvature
(2.59a) becomes of each curve is determined by fitting a circle to
the curve at the point (called the osculating circle)
a
PI - P 2 = Ii: at the interface. (2.60) such that the first and second derivatives of the
m circle at the point equal the derivatives for the
The simplest constitutive relation (2.58) accounts curve. If we denote the radii of the two osculating
for the effect of surface tension on the pressure circles as R 1 and R 2 , then the mean radius of
jump across a curved interface, as originally curvature at the point of interest is given as
described by Young and Laplace (see Adamson
1990). Without any further generalization, the
(2.61)
constitutive relation (2.58) also explains the origin
of Marangoni flows through the inherent de-
pendence of a on temperature and the interfacial where the appropriate choice of sign is discussed
concentration of surfactants (this is discussed in below. A special feature of this definition is that
section 2.3.2.5), which gives rise to gradients in the value of R m is independent of the choice of
surface tension (i.e., nonzero values of VIla in the two perpendicular planes, as long as their
(2.59a)). Sufficient experimental evidence exists line of intersection incorporates the outward
to suggest that the influence of surfactants on normal to the surface at the point of interest.
interfacial flows can go beyond that suggested Thus, the orientation of these two planes can be
by the VIla term in equation (2.59a). Namely, chosen arbitrarily, in a manner that simplifies
surface-active agents can impart to the interface the evaluation of R m .
r------------------~
I I
1 O,culating Circle I
1 Lying in Plane I I
1 1 1
1 ' I Plane 2
: / - - - ~- - - - - OscUiaiingCirCle Ly~£in P'ane2 - - - - /- - "7

1 // /
1 / //
1/ R2 /
/
/1" /
// I /
/ I Intersection of Plane 2 / /
/ I and Interface /
/ I I /
~---I------- ------r----------------/
I 1
i
1
i
Intersection of Plane I
:
1
------------------r-
I and Interface 1
Plane 1
Figure 2.7 The principal radii of curvature, R 1 and R 2 , used to determine the mean radius of curvature.
Let us now apply (2.61) to a few simple cases. (see Adamson 1990), the internal air pressure is
For the case of a sphere of radius R, the above higher than the external liquid pressure. To
procedure yields R 1 = R 2 = R at every point on choose the correct sign of R m for this case, we
the spherical surface; thus R m = ± R/2. For the embed it in (2.60). Noting the direction of the
case of an infinitely long right circular cylinder pressure and surface tension forces on a material
of radius R, orient one of the planes in the axial body drawn to include a portion of the spherical
direction, so that it includes the axis of the interface (see Fig. 2.8), it is clear that the surface
cylinder. Consequently, the perpendicular plane tension forces must oppose the higher pressure
intersects the cylindrical surface on a circle of inside the bubble. Thus, the positive sign for R m
radius R, while the axial plane intersects the is chosen:
surface on a straight line (R = (0). Thus, R 1 = R
and R 2 = 00, yielding R m = ± R. Finally, for the
case of a flat surface, the intersection of all
perpendicular planes with the surface yields a
straight line. Therefore, R, = R 2 = 00 and We note here that in any physical configuration,
I/R m = O. even those including fluid flow, the correct sign
The sign of (2.61) depends on the configuration of the mean radius of curvature can always be
ofthe problem in question. For example, consider identified by imagining that the problem is static
the case of a spherical, static air bubble of radius and making sure that the pressure drop across
R in a liquid, where the constant internal and the interface is consistent with its curvature.
external pressures are P 1 and P 2' respectively. Alternatively, as the mean radius of curvature is
According to the equation of Young and Laplace defined in terms of the normal vector by (2.59b),
this problem is that the soap film is exposed to

a constant atmospheric pressure PA everywhere
(i.e., no films span the circular openings at either
end). Therefore, the interior and exterior pressures,
PI and P 2' are identical; thus 1/R m = 0, in
accordance with (2.60). But it is obvious, from
the axisymmetric geometry of the soap film, that
the film is curved. To resolve the dilemma,
consider in particular the point D on the film,
which is equidistant from the two rings. Now
construct two perpendicular planes at D such
that one contains the vertical axis of the film
while the other is parallel to the wire rings. The
Interface
intersection of the first plane with the film results
in a curved line that is visualized in Fig. 2.9 as
Figure 2.8 Force balance on a portion of a spherical the 'soap film', whose radius of curvature at D
interface. Here, FP, and F p, are the pressure forces is denoted as R l' The intersection of the second
exerted on the body due to the internal and external plane with the soap film results in a circle whose
pressures PI and P 2' respectively, and F denotes the
radius, R 2 , is smaller than that of the rings. The
0'
surface tension forces exerted on the body.

important point to recognize is that the sense of
each curve is opposite to the other: the first curves
inward while the second curves outward. Thus,
the correct sign will naturally arise by using this the condition I/R m = 0 is met because R 2 = - R 1
definition. at D, as well anywhere else along the surface*.
The importance of considering both principal As a consequence, soap films freely suspended
radii in the mean radius of curvature is seen by from curved wires invariably have saddle-like
considering the case of an axisymmetric soap shapes.
film pinned to the surface of two wire rings, as The mathematical relationship between the
illustrated in Fig. 2.9. The important feature of mean radius of curvature and the parameterization
of a given surface is a complicated nonlinear
function of the interface shape. For example, for
Wire a surface parameterized as Y = H(X, Z, t), the
corresponding mean radius of curvature is given
by (2.62).
oap Film
* Note that for all other points on the film, the second plane
will make an oblique angle with the planes containing the
wire rings, as the second plane must intersect the axial plane
along the outward normal to the film at the point of interest.
Wire Thus, the intersection of the soap film with this second plane
will yield a curve whose curvature at each point, 1/R 2 , will
vary with vertical position in a complex way. Note that,
Figure 2.9 Shape of an axisymmetric soap film because R 2 varies with vertical position, the shape of the
having identical internal and external pressures PI soap film in the vertical plane, as illustrated in Figure 2.9,
and P 2' respectively. is not circular.
1
02 H[1 + (OH)2J
aX2
02 H[ (OH)2J o2H oH oH
iii + iii2 1 + ax - 2a-xaz ax iii
(2.62)
R=±
m [ (OH)2 (OH)2J
1 + oX + oZ
3{2
For the simple cylindrical, spherical, and flat of temperature and surface concentration gradients
interface cases described above where I/R m is along the interface:
constant, (2.62) can be solved exactly for the
OU OU
surface parameterization, despite its nonlinearity.
For most other cases, however, the appearance of
Vnu = oTVnT + or. Vnrj
I
(2.63)
such a formidable mathematical expression in The common assumption is that u in (2.59a) and
the dynamic boundary condition makes interfacial (2.63) is the equilibrium tension, and thus its
systems under both static and dynamic conditions dependence on T and r j pointwise along the
extremely difficult to solve exactly. interface can be determined from equilibrium
surface tension measurements. In particular, the
2.3.2.5 Surface stresses derived from gradients equilibrium surface tension decreases linearly
in surface tension with increasing temperature, typical values of
(ou/oT) being of the order of -1 x 10- 4 N/m 0c.
As mentioned above, the surface stress balance For dilute solutions of surface-active solutes, the
presented in equation (2.59) is based on the Gibbs adsorption equation (Adamson 1990)
simplest constitutive relationship for the surface provides the following relationship between surface
stress tensor (2.58); that is, an interface whose tension and surface excess:
sole mechanical property is interfacial tension u.
Nevertheless, such an interface can have a
profound effect on the fluid dynamics of multiphase r =
j
-(~~)(:~)T (2.64)
systems through the inherent dependence of
where R is the gas constant and C j is the molar
interfacial tension on temperature and solute
concentration of species i in the bulk fluid.
concentration. In problems involving heat and/or
Furthermore, for dilute solutions, the surface
mass transfer, for example, the possibility exists
tension decreases linearly with increasing bulk
for local spatial variations in temperature and/or
solute concentration. Thus, it follows from (2.64)
concentration to develop along the interface,
that
which in turn will create local gradients in the
interfacial tension (i.e., nonzero values of Vnu (2.65)
in (2.59a». As can be seen from (2.59a), such
gradients in u appear as tangential surface stresses and
that can drive fluid flows in the adjacent phases.
In particular, this is the mechanism by which
ou = -RT>:::! -2500Nm/mole
or. (2.66)
Marangoni flows are initiated during interphase I
heat and/or mass transfer in fluid-fluid systems. Note that f3 is a constant which represents the
Gradients in the interfacial tension occur distribution coefficient for species i between the
because the local value of the interfacial tension interface and the bulk phase, its value being
u depends upon the local interfacial conditions; strongly dependent on the molecular nature of
namely the local surface temperature and the the solute. For a mildly surface-active substance
local value of the surface excess of solute i such as octanol in water (with C j <O.1 mM at
adsorbed at the interface, rj' Thus, with the aid 25°C), f3 equals 1.9 x 10- 5 m;for aqueous solutions
of partial differentiation, we can rewrite the of a homologous series of organic solutes, the
interfacial tension gradient term as a combination value of f3 increases by a factor of '" 3 for each
CH 2 group added to the hydrocarbon chain Recall that U a is the rate of displacement of the
(Traube's Rule; see Adamson 1990). interface as defined in the context of equation
One of the most important effects ofsurfactants (2.37). The velocity Un in (2.67a) is the velocity
on fluid flows results from the convective along the interface which is given by
redistribution ofsurfactant in the interface. When
an interface is accelerated or stretched in the (2.67b)
direction of flow, surfactant material in the where In is defined in (2.58b), and U is the fluid
interface is swept downstream, thereby creating velocity in either phase at the interface. We note
a region of higher tension upstream. The net here that for the case of negligible surface
effect is an interfacial tension gradient that concentration, (2.67a) reduces to (2.40) for each
opposes the flow (see Section 2.4.2 for a model component i. From left to right, the terms in
problem illustrating this effect). Because of the (2.67a) account for the local rate of accumulation
strong dependence of a on r i , only modest of solute at the interface, the rate of change in
amounts of convective redistribution are needed the local surface concentration due to local
to produce large forces which tend to oppose increases/decreases in the interfacial area owing
flow. Thus, it is not unusual to discover interfaces to the motion of a curved surface normal to itself,
that resist viscous stresses and remain immobile, the rate of convective redistribution of solute in
in a fashion that resembles a solid boundary, the interface, the rate of surface diffusion, the
even when only small amounts of surfactant are rate of convective transport to/from the bulk
present in the system. To account for the effect phases to the interface, and the local diffusive
of convective redistribution of surfactant in the flux to the interface from the bulk phases.
interface, an interfacial mass balance on the In the vast majority ofcoating flow applications,
solute is needed. This solute balance is obtained the flow is steady and the surface is not deforming
in precisely the same way as the kinematic normal to itself. Furthermore, the bulk concen-
boundary condition (2.40), except that the tration ofsurfactant is extremely low. Thus, (2.67)
accumulation of solute at the interface must be can be simplified, to
retained in the final form of the integral equation.
The result is: Un,vnri = -rivn'U n + (J i \ - J i2 )'OI
(2.68)
or. r. oa
a; + 2~ ot + Vn,(riUn) + Ds Vn
2
ri
where the vector identity Vu ' (riUU) =
uII'Vilri + riVil 'Un has been utilized in obtaining
= Pi\(U\ - U a )'OI - Pi2(U2 - U a )'OI this form. Equation (2.68) is usually the starting
+ (J i1 - J i2 )' 01 (2.67a) point for all analyses of surfactant effects on
coating flows. According to (2.68), gradients in
where a is the determinant of the surface metric interfacial concentration caused by convective
tensor which characterizes the shape of the redistribution (i.e., nonzero values of Vn' Un) are
surface, D s is the coefficient of diffusive transport moderated by mass transfer from the bulk phases:
within the interface itself, and J il and Ji, are the the faster the rate of exchange between the
respective diffusive fluxes of species i in phases interface and the bulk phases, the more uniform
1 and 2 at the interface. In addition, Pil and Pi2 the interfacial concentration and the smaller the
are the mass concentrations of species i in phases effect of the surfactant on the flow field.
1 and 2 at the interface. We note here that Pi in In summary, the effects of surfactants on
either phase (written generally for either phase coating flows are accounted for by the Vna term
without the subscript 1 or 2) equals the corre- in the interfacial stress balance (2.59a). The local
sponding molar concentration, Ci (introduced in value of this term requires knowledge of the
(2.64», times the molecular weight of species i. gradient in r i (see (2.63», which in turn is
described by an interfacial mass balance (2.67) don-van der Waals dispersion forces and elec-
or (2.68). In applying these equations, knowledge trostatic forces.
of bulk diffusive fluxes in each phase to/from the Consider as an example, the air-liquid-solid
interface are needed, and these are obtained from system shown in Fig. 2.lOa, which indicates the
Fick's law location of the boundary regions. The thickness
of a boundary region is the distance over which
(2.69a) long-range intermolecular forces act. Since the
coupled with the equation of solute conservation intermolecular forces vary with distance as either
for each phase the air or solid is approached, physical properties
such as density and viscosity will also vary, and
op. will be different from those measured in the bulk
-'
ot + U'Vp, = I
-V·J.
"
+ r· (2.69b)
liquid phase. Consequently, the pressure tensor
is anisotropic in the boundary regions; i.e., the
Here, Di is the molecular diffusivity of species i, pressure measured relative to a plane parallel to
and r i is the rate of production of species i per the solid is different from the pressure measured
unit volume by chemical reaction. To solve (2.69), relative to a plane perpendicular to the solid.
a boundary condition on Pi is needed at the Outside of these boundary regions, physical
interface. Either local equilibrium at the interface properties are simply bulk liquid properties in
is assumed and an equilibrium adsorption the classical sense, and the pressure tensor is
relationship such as (2.65) is applied, or the flux isotropic.
of solute into the interface is specified according For the vast majority ofcoating flow problems,
to an adsorption/desorption reaction rate model the length scale of a typical liquid film is orders
such as of magnitude greater than the thickness of these
diffusive flux = Jil'n l = klPi - k_ll i
Air
where k I and k _ 1 are adsorption and desorption
rate constants. A more detailed discussion of Air-Liquid
Inlerface--.~'l;7;%r:?
/,'7;%7/';7':?/,"'/';7;%7?/,n/';'7;%7;%7':?/,'7~7/';7/"/,n:;.-:;7
Z Boundary
surfactant effects on coating flows can be found regIon
in Chapter 11 d. Liquid Bulk Phase
2.3.2.6 Disjoining pressure: modification to the Solid

dynamic boundary condition for ultra-thin films (a)
The body forces, surface forces, and line forces

discussed thus far involve measurable macroscopic Air
Air-Liquid
quantities such as density, viscosity and surface tnterface-_ _ove.rlapPing
tension. As mentioned previously, a continuum regIOn
Solid ----·.....lIIIijI..lIIIijI..lIII
approach inherently considers length scales larger
than that of molecular dimension so that these
(b)
properties have meaning. However, we must
acknowledge that these fluid properties do have
a molecular origin, and in particular, in the Figure 2.10 The origin of disjoining pressure: (a) a
vicinity of a fluid-solid or fluid-fluid boundary, thin fluid layer whose thickness is greater than the
range of intermolecular forces; (b) a fluid layer that
the molecular interactions will be different from
has been thinned to the point where the interfacial
those in the bulk fluid away from these boundaries. boundary regions (hatched) have begun to overlap,
These interactions can be due to, for example, thereby eliminating the region in the film that has
long-range intermolecular forces such as Lon- bulk fluid properties.
boundary regions. On this length scale, the

macroscopic manifestation of the anisotropic
boundary region near the air-liquid interface is
interfacial tension (Derjaguin, Churaev and Muller
1987), which is lumped into the interfacial stress
tensor, T a , in the dynamic boundary condition
(2.56); furthermore, the properties of the liquid
are interpreted in the classical sense. On the other Figure 2.11 Schematic diagram of an ultra-thin fluid
hand, if the liquid layer is made progressively film in equilibrium with its bulk phase.
thinner to the point where its dimension ap-
proaches the length scale of long-range inter- R m = (0) for most of the region between the
molecular forces (on the order of 0.1 pm or less), supports, there is a pressure drop across each
the boundary regions themselves constitute a interface. Namely, a pressure, P l ' in excess ofthe
significant portion of the domain, and may begin fluid pressure, P 2' must be applied to maintain
to overlap (Fig. 2.10b). In this case, the concept a given degree of separation, h, between the two
of bulk properties will no longer be meaningful. interfaces (assuming that the excess force is
Even in such cases, however, the total thickness repulsive).
of the liquid layer often is large enough that it As the thickness of the liquid layer determines
can be modelled in the context of the continuum the degree to which the boundary regions near
assumption. the solid and air-liquid interface will interact,
The net effect of the overlap of these boundary the disjoining pressure is a function of the film
regions is an attractive or repulsive force between thickness as well as the temperature, pressure,
the air-liquid interface and the solid. When the and composition of the system. In addition, the
force is attractive, fluid has a tendency to be disjoining pressure is a function of the particular
squeezed out from between the solid and the forces being exerted in the layer, i.e., dispersion
interface. If the force is repulsive, fluid has a and electrostatic.lfthe interface and solid surfaces
tendency to accumulate between the solid and are uncharged, so that the net force between the
air-liquid interface. The net force per unit area interfaces is due solely to the London-van der
is called the disjoining pressure (Derjaguin, Waals dispersion force, the relationship between
Churaev and Muller 1987), and it can be lumped disjoining pressure and layer thickness, h, is given
into the dynamic boundary condition as an extra by
pressure term. In particular, for a static interface A
configuration, the pressure drop given by (2.60) n = 6nh 3 (2.71)
can be modified to incorporate the disjoining
pressure, n, as where A is the Hamaker constant, whose magnitude
(J
depends on the degree of polarizability of the
P1 - P2 - n=- at the interface (2.70) molecules in the interfacial region. Other relation-
Rm ships are available to incorporate additional effects
The pressure relationship given by (2.70) is (Derjaguin, Churaev and Muller 1987). In addition,
typically the starting point for analyses involving it should be noted here that once an overlap of
the flow occurring in ultra-thin films. Its physical the boundary regions occurs, the surface tension
significance can be elucidated by considering the itself will be dependent on the film thickness.
situation illustrated in Fig. 2.11. Here, a static From a practical point of view, however, ther-
ultra-thin layer of fluid is suspended between two modynamic arguments typically indicate that this
supports and is in direct communication with a thickness effect is small, and it is reasonable to
bulk phase of the same fluid. Although the interpret the surface tension in (2.70) as that in
opposing interfaces of the layer are flat (i.e., the conventional, macroscopic sense.
Measurement indicates that the Hamaker Interface

constant is typically on the order of 10- 20 J.
Air Liquid
Thus, from equations (2.70) and (2.71), we can
deduce that disjoining pressure will be significant Contact
Line"
only for film thicknesses less than approximately
0.1 J.lm. Consequently, its effect need not be
considered in most hydrodynamic models of Solid
coating flows. On the other hand, at high coating (a)
speeds, the potential exists for entraining an
ultrathin layer of air between the coating and 9
the substrate. The tendency of this air layer to
collapse or break up into tiny air bubbles is
acutely influenced by the disjoining pressure of
the layer (see Chapter 3). Similarly, the process
Contact
by which a continuous liquid coating on a solid Angle
substrate breaks up into a patchwork of wet and } Hysteresis
dry areas inevitably involves the formation of an 9R
ultrathin film as a crucial intermediate step (see
Chapter 6). Thus, the disjoining pressure also
plays a key role in the hydrodynamic stability
ofthin liquid films, with applications that include
not only coatings but also soap films and
Speed of Contact line, U
coalescence phenomena. A more thorough treat-
ment of disjoining pressure and its role in coating (b)
flows will be presented in Chapter 6.
Figure 2.12 The contact angle: (a) cross-section of
an air-liquid-solid system, indicating the contact
2.3.2.7 Contact angles angle, 8; (b) the dependence of the contact angle on
the speed of the contact line.
The mathematical expression for the curvature,
such as that shown in (2.62), has interfacial
derivatives, implying that additional boundary is valid for any fluid-fluid system whose interface
conditions are needed to determine the interfacial exhibits interfacial tension.
parameterization. In coating processes, it is In a static case in which there is no fluid flow,
common for one or more fluid-fluid interface to the contact angle is intimately affected by the
come in contact with a solid boundary, such as identity of the solid surface (including its finish),
is shown in Fig. 2.1. The line of such contact is the fluids and the interface. Thus, changing the
called a contact line, and the angle of intersection, material properties of the system is likely to
e, is called the contact angle; both of these are change the value of this angle. In a case where
shown for an air-liquid system in Figure 2.12a. the contact line moves, the contact angle (which
The location of the contact line, or the variation is now referred to as a dynamic contact angle)
in the contact angle along this line, provide is still affected by the material properties. However,
necessary boundary conditions to solve for experimentally observed dynamic contact angles,
interface shapes (Dussan V. 1979). We now often called apparent contact angles, are also
discuss the basic properties of the contact angle. dependent upon the flow field found in the
For the purposes of discussion, we will restrict measurement domain. There have been some
attention to air-liquid systems, as they are attempts at obtaining submicroscopic contact
common in coating flows, although the discussion angles, which are also called actual contact
angles, which are supposed to be truly material extent. The implication ofcontact angle hysteresis
properties that are independent of the macroscopic is profound. Force balances show that it would
fluid flow (so they can be utilized in other be impossible for stationary drops to exist on
geometrical configurations). However, as these inclined surfaces without hysteresis. Furthermore,
contact angles cannot be experimentally observed, for a given static horizontal drop, the wide
mathematical analyses which make some as- variety of shapes often seen are also attributable
sumption about the flow in the vicinity of the to the existence of hysteresis.
contact line must be employed to extract their An implication of the contact-angle-versus-
values from the observed interface shapes. It is speed curve is that the mathematical solution to
thus difficult to assess whether the assumptions a static contact line problem may be obtained
made in such analyses are correct. An extensive for a large range of values of (J, but not all of
discussion of these issues is presented in Chapter those solutions are physically realizable. In order
3; in that chapter, it is shown that from a practical to achieve a static contact line, it is necessary
point of view, the apparent contact angle is that all contact angles along the contact line lie
typically most useful to model coating flows. within the hysteresis range; if they do not, then
We now discuss the properties of the apparent there will be contact line modon in those sections
contact angle, which is henceforth simply called whose contact angles lie outside the hysteresis
the contact angle. For a given material system region. Thus, when solving for an interface shape
and flow geometry, it has been found experimentally in which no contact line motion is intended, it
that the speed of the contact line, U, is related to is important to specify contact angles which lie
the contact angle by a curve as shown in Fig. in the hysteresis region. In some cases, the
2.12b. By convention, positive values ofthe contact location of the contact line along a solid is used
line speed correspond to cases where liquid is as a boundary condition instead of the contact
displacing air, the contact angle being measured angle. To assure that the interface is static for
through the liquid as shown in Fig. 2.12a. The these cases, it is important that the resultant
contact line in such a case is called an advancing angles along the contact line, extracted from the
contact line. On the other hand, negative contact mathematical solution, fall within the hysteresis
line speeds are associated with receding contact region.
lines, for which air displaces the liquid. It should Finally, we close this section by emphasizing
be noted that, in either case, the contact line that the need to specify constraints on the
motion is always defined normal to the contact interface itself, such as the value of the contact
line in the plane of the solid, and the local value angle or the location of the contact line, is
of the contact angle in this normal direction intimately related to the appearance of the
determines the local speed of the contact line. curvature term in the dynamic boundary condition.
In Fig. 2.12b, there are two angles which In those cases where the interface has zero surface
characterize the advancing and receding contact tension, or in cases where the curvature operator
lines. The first, (J A' is called the critical advancing is neglected through approximation, an equation
contact angle, and is the angle above which the system will be overspecified if contact angle or
contact line moves towards the air. The second contact line boundary conditions are imposed
angle is the critical receding contact angle, denoted on the interface without relaxing constraints
as (JR' which is the angle below which the contact elsewhere in the domain.
line will move towards the liquid. An important
feature of such a curve is that (J A =F (JR; i.e., there
2.3.2.8 Additional constraints for a well-posed
exists a range of angles over which there is no
problem
contact line motion. This range of angles is
termed contact angle hysteresis, and it exists in The boundary conditions presented above are
virtually all material systems, at least to some fundamental to the flow of multiple fluid phases.
Model problems: the effects of fluid-fluid interfaces on flow 47
Let us now summarize, in general form, the physical implications of the presence ofan interface
system of equations and boundary conditions when there is fluid motion. Generally, flow
which are applicable to the flow configuration calculations involving interfaces are quite com-
shown in Fig. 2.1. For each fluid phase, the plicated, owing to the fact that the interface
governing equations which apply are the continuity location is intimately coupled with the flow field,
equation (2.6) and conservation of linear mo- as can be seen from the boundary conditions
mentum (2.19). At the solid surfaces, the kinematic derived previously. An interface location is not
condition (2.40) and no-slip condition (2.48) known a priori and must be determined as part
apply. At the fluid-fluid interface, the kinematic of the solution to the problem. The complex
boundary condition (2.40), the no-slip condition mathematical structure of such problems can
(2.47), and dynamic boundary condition (2.56) often obscure the relatively simple physics
are imposed. If, in addition, the interface has associated with the presence of the interface. In
surface tension, a contact angle condition or analyzing complex flow situations, it is often
contact line location must be specified to constrain desirable to view the flow field as a combination
interfacial shape. of much simpler flow fields in which the physical
It should be noted that the governing conser- mechanisms involved are clear. In addition, it is
vation equations, in their most general form, also important to examine limiting cases for
require that boundary conditions be identified which the mathematical solution of fluid flow is
on all boundaries of the domain, and as of yet, relatively simple, and to apply the intuition
we have not specified constraints along the gained to cases where such limits are not strictly
inflow and outflow boundaries of the domain valid but the underlying physical mechanisms
shown in Fig. 2.1. Boundary conditions applied are essentially the same. The goal of the remaining
at these locations can take various forms which part of this chapter is to provide a physical
depend upon the particular problem posed. We understanding of the basic concepts associated
might treat the domain in the flow direction as with interfacial fluid flow using both of these
being infinite, and such inflow and outflow approaches.
boundary conditions would assure the correct The unique physical character of flow fields
limiting form of the flow field for the problem involving interfaces enters through the dynamic
of interest. In other cases, for the purpose of boundary condition (2.59a and b), and so it is
computation, we might decide to approximate natural to begin the discussion witt~ this condition.
the domain of interest as being finite, and in this To this end, it is useful to resolve (2.59a) into its
case, appropriate inflow and outflow boundary normal and tangential components. For any
conditions would be obtained by looking at the point along an interface between phases 1 and
fully developed flow fields upstream and down- 2, it takes one unit normal vector 01 (which is
stream of the domain of interest, but applying defined, following the convention established in
them at a finite location for which the flow Fig. 2.6, to be the unit outward normal to phase
solution is still insensitive to the precise location 1) and two linearly independent tangent vectors,
of the conditions. Additional conditions are ta and t p , to characterize the surface. Thus, (2.59a)
required ifthe flow is transient, as initial conditions can be written in component form as
for the fluid and interface must be specified as
well. °1: (Pi - P ) + [°1 '(T - T )]'01 = R
(J
2 2 1
m
2.4 MODEL PROBLEMS: THE EFFECTS OF at the interface (2.72a)

FLUID-FLUID INTERFACES ON FLOW
ta : [01'(T 2 - T 1) + VII(J]·ta = 0
With the equations and boundary conditions
identified, we can now further examine the at the interface (2.72b)
t p : [n l '(r 2 - 1"1) + Vila] ·t p = 0

at the interface (2.72c)
where the definition of VII' given by (2.55), has
been used in eliminating the del operator in (2.72a).
2.4.1 IMPLICATIONS OF THE NORMAL

STRESS BALANCE AT THE INTERFACE
We have already seen that in a static problem, X=H(y)
surface tension is responsible for creating a
pressure drop across a curved interface, as given
by (2.60). The physical effect of surface tension
in statics is confined to (2.72a), where viscous Plale Liquid Plate
stresses are zero when there is no fluid motion.
When viscous stresses are significant, the rela-
tionship between the pressure drop and the mean b
radius of curvature is altered. Given the intimate
linkage between the flow field and the interface Figure 2.13 Air-liquid interface moving between
location, it is important to ask under what parallel planes at a speed U.
conditions the interface has a static-looking
shape. In the context of such a case, given the
complex form of the mean radius of curvature such a frame of reference, the plates move at a
(as shown in 2.62), it is also useful to identify velocity - Vi, where i is the unit vector in the
those conditions under which the mean radius X-direction. We assume that the viscous effects
of curvature is essentially constant, as the of the air on the film are negligible, i.e. 1"2 = 0 in
evaluation of the pressure drop across such an (2.72); we denote the constant air pressure as PA'
interface shape is easily determined by the dynamic As all the fluid dynamics is thus limited to the
boundary condition. Both of these cases are liquid phase, we will not explicitly write the
examined in what follows, in the context of two subscripts on the flow variables.
important dimensionless groups - the capillary It is clear from (2.72a) that, if the viscous
and the Bond numbers. stresses in the liquid are negligible, the relationship
between the pressure drop across the interface
and the mean radius of curvature is of the same
2.4.1.1 The limit of small capillary numbers
form as for a static problem. This does not
Consider a meniscus between air and a Newtonian necessarily imply, however, that the interface will
liquid formed between two parallel plates separated have a static shape; rather, a static shape will be
by a distance b, as shown in Fig. 2. t 3, where the obtained only ifthe pressure drop obtained from
interface is moving upwards with a constant the equation of motion is hydrostatic; i.e., it is
shape at a steady speed V. The coordinate system governed by (2.22). With this limit in mind, let
is oriented as shown, where the plates are infinitely us determine the order of magnitude of terms in
long in the Z-direction (which is oriented into (2.72a) and (2.21) by choosing appropriate scales
the figure), and gravity acts in the - X-direction. for all the variables as follows. The velocity and
With the coordinate system fixed on the interface, length are scaled with V and b, respectively, and
the interface location is time independent. It is the viscous stresses with Jl.V/b. In anticipation
denoted as X = H( Y), and Y = ± b/2 denotes the ofa limit where a static-looking shape is obtained,
locations of the surfaces of the two plates. In the pressures are scaled with a/b. Then, with
dimensionless variables defined as As the Weber number is independent of viscosity,

it is the preferred dimensionless group for
interfacial problems in which viscous effects are
small and the resulting flow is dominated by fluid
inertia. For problems in which the flow is relatively
slow, so that viscous effects and surface tension
effects are most important (as in the sample
where problem of Fig. 2.13), the Reynolds number as
written explicitly in (2.73c) is often more conven-
h(y) == h(~) == H~Y) tional, as the relative sizes of viscous and inertial
forces are compared in this single dimensionless
the normal component of the dynamic boundary group.
condition (2.72a) becomes From (2.73), it is clear that in the limit as
Ca~O, with all other groups held fixed,
1
(2.73a) p - PA =- at x = h(y) (2.74a)
rm
where Vp = -(Bo)i (2.74b)
Ca=j).U (2.73b) The system (2.74) is a dimensionless version of
a (2.22) and (2.60); thus, in the limit of small
The equation of motion (2.21) becomes capillary number, the shape of the interface can
be calculated as if it were static. The capillary
ou -] -
[ot+ u·Vu = -Vp + CaV u -
-2
ReCa (Bo)i number quantifies the relative importance of
viscous forces (due to fluid flow) to surface
(2.73c) tension forces in distorting the interface away
from its static-looking shape. We note here that
where the gravity vector, whose magnitude is
there is an important distinction between the
denoted by g, has been written as g = - gi, and
dynamic problem and the static problem in the
pUb pgb 2 small Ca limit. At small but nonzero capillary
Re =-, Bo =- (2.73d)
numbers, the interface may be moving at some
It (J
speed even though the interface has a static-

The dimensionless group defined in (2.73b),
looking shape; the dynamic contact angle indicates
namely Ca, is called the capillary number, which
that the interface is moving. For the configuration
is a ratio of viscous to surface tension forces. In
shown in Fig. 2.13, the appropriate contact angle
(2.73d), Re is the Reynolds number, which is a
corresponding to the speed U is chosen from a
ratio of inertial forces to viscous forces, and Bo
relationship such as that shown in Fig. 2. t 2b,
is the Bond number, which is a ratio of gravi-
where the contact angle resides along the
tational forces to surface tension forces. Inspection
advancing portion of the curve.
of the equation of motion (2.73c) shows that the
quantity Re Ca actually characterizes the relative
effect ofinertia, in that it determines the magnitude 2.4.1.2 The effect of the Bond number for small
of the left-hand side of (2.73c). This quantity is capillary numbers
called the Weber number, We, which is thus
defined as We now consider those conditions under which
the interface can have a constant mean radius
pU 2 b of curvature. To continue our arguments, let us
We= ReCa=-- (2.73e)
a simplify the system (2.74). Integration of (2.74b)
yields which, as written, explicitly indicates that Bo is

a ratio of gravitational force to surface tension
p = PR(f) - (Bo)x (2.75a) force. A small Bond number implies that the
where PR(t) is in general a time-dependent function surface tension forces dominate, with the interfacial
which arises from integration, and which acts as shape tending towards constant curvature (see
a reference pressure in the fluid. Substitution of 2.75b)), while large Bond numbers are charac-
(2.75a) into (2.74a) yields the equation that is teristic of systems in which gravity tends to
solved for the interface shape: distort the interface shape away from such a
condition.
1 A natural question to ask at this point is to
PR(f) - PA - (Bo)h(y) = - (2.75b)
rm what shape does gravity tend to distort the
interface at low capillary numbers? An answer
Inspection of this equation reveals that the
to this question can be obtained mathematically
reference pressure PR in the fluid must be an
by looking at the large Bond number limit, where
absolute constant in order for r m to be time-
the gravity forces dominate the surface tension
independent. Furthermore, equation (2.75b),
forces as characterized by (2.76). A large Bond
obtained in the limit of small capillary number,
number can be obtained in our illustrative example
indicates that it is not possible, as long as gravity
(Fig. 2.13) by increasing the spacing of the plates.
is acting, to obtain an air-liquid interface of
For such a limit, the scaling of the pressures used
constant mean radius ofcurvature at low capillary
to arrive at (2.75) is not appropriate, as gravity
numbers. Rather, the interface shape approaches
forces will dominate and pressure differences will
a constant curvature only when the Bond number
be more closely associated with hydrostatic
approaches zero. In the present example, this can
effects rather than capillary effects. Thus, we
be accomplished, according to the definition of
redefine the dimensionless pressures as
the Bond number in (2.73d), by decreasing the
spacing between the two plates while holding the
physical properties of the system fixed. On the
other hand, it is evident from (2.73a) that the
limit Bo------.O is insufficient to ensure that the
interface will have a constant mean radius of causing equation (2.75b) to become
curvature as fluid flow also distorts the shape of
1
the interface, through viscous stresses in proportion PR - PA - h(y) = -- (2.77)
to the capillary number. However, for many Borm
problems of practical interest, the capillary number
is usually low enough to permit the assumption where the fact that PR is a constant has been
that r m is constant, provided the Bond number used. Now, as Bo------. 00 , we see that h(y) approaches
is also small. It is important to realize that in a constant, as the right-hand side of (2.77) drops
such cases the shape of the interface (the value out in that limit. In other words, the interface is
of r m) will not be the shape obtained for the case driven towards a constant height, flattening out
of zero fluid motion because, with finite fluid as gravity forces increase their dominance over
flow, the dynamic contact angle at solid boundaries surface tension forces.
will inevitably be different from that in the static We have already stated that problems in which
state. an interface makes contact with a solid boundary,
The Bond number can be interpreted physically a contact angle-speed relationship must be
by rewriting it as imposed provided that the interface has a tension.
For large but finite Bond number, the interface
Bo = pgb (2.76) flattens according to (2.77), but it cannot flatten
u/b uniformly near the wall unless the contact angle
that the interface makes with the wall fortuitously written in dimensional form as
equals 90°. Thus, in the vicinity of the contact
line at the wall, there will be a region of high
curvature. We will now consider a relatively 0= f -PndS+ f pgdV+ f amdC
simple analysis of this region, which leads to an S(I) V(I) C(I)
important length scale often used in coating flows (2.78)

and which provides an alternative interpretation
of the Bond number. In obtaining (2.78), we have utilized the constitutive
Let us focus on the region of fluid near the relation (2.58) which introduces the surface tension.
plate at Y = -b/2 in Fig. 2.13, where angle of We now proceed to evaluate the terms in the
contact with the wall is specified. This region is force balance. In order to determine the first
shown in more detail in Fig. 2.14, for a large integral in (2.78), the local pressure field given
Bond number. Note that from the wall, the by (2.75a) is employed, which is rewritten in
interface becomes horizontal and is coincident dimensional form as
with the X = 0 plane, consistent with the above
P = PR(t) - (pg)X (2.79a)
physical description of the large Bond number
limit. Consider a force balance on the rectangular where PR(t) is a time-dependent function arising
material body shown in Fig. 2.14. The body is as a result of integration. This function is evaluated
drawn with sufficient length to meet the location by using (2.75b), noting that the interface far
where the interface becomes horizontal. Further- away from the wall has zero curvature and is
more, the body has width W in the Z-direction coincident with the X = 0 plane. In that region,
(oriented into the figure) and is of height d in the I/R m = 0, and thus
X -direction, where d is the height that the
(2.79b)
meniscus rises above the X = 0 plane. Note that
the body is drawn close to, but not intersecting, We now evaluate the pressure integral, using
the solid plate, and so it lies completely within the notation shown in Fig. 2.14. The pressure
the liquid and air. The general force balance for integrals on sides 3 and 4 cancel as they are both
the body in Fig. 2.14 is given by (2.52) which, to at atmospheric pressure and act in opposite
be consistent with the small capillary limit, is directions. The pressure on side 2 is atmospheric
Material Body
CD
CD
I Air PA
CD
X
X:: 0 Plane
y
-b Liquid +
Y::~
CD
Figure 2.14 Material body used to analyze end effect in the large Bond number, low capillary number limit.
acting in the - Y-direction, while on side 1 the the material body experiences a hydrostatic
pressure is hydrostatic given by (2.79), acting in pressure less than atmospheric for X> 0, as can
the + Y-direction. The resulting pressure integral be seen from (2.79). This indicates that the
is interface itself must curve inward as drawn in
Fig. 2.14, to be consistent with the way in which
f
S(I)
- PndS
1
= -"2pgd WJ
2 'l'
(2.80) surface tension moderates a pressure drop,
although the precise shape must be determined
by solving the local equations (see Princen 1969).
where j is the unit vector oriented in the + Y- For a given angle f), the first square root in
direction. Let the volume of liquid which is lifted (2.83b) determines the height of rise of the fluid
above the X = 0 plane be denoted as V,,; with up the wall. The quantity
this definition, the volume integral in (2.78)
becomes r;;
,,- -Jpg
L -
f
(2.84)
pgdV = -pgV"i (2.81)

is called the capillary length, and is often used
V(I) as a length scale in flow problems involving
Finally, we evaluate the line integral due to interfaces. With the definition (2.84), the Bond
surface tension in (2.78). At the intersection of number in (2.73d) can be written as
side 2 with the interface, the vector m = j. At the
intersection of side 1 with the interface, the vector
m is given in terms of the contact angle f) (as
Bo=(LY (2.85)
shown in Fig. 2.14): where b again denotes the gap spacing in Fig.
m = i cos f) - ]sin f) 2.13. We have already indicated that when the
Bond number is large, the interface flattens, at
Both line integrals act along a line of length Jv, least away from the walls. For a given material
to yield system, the height of rise along the plates shown
f <TmdC = <TW(j(1 - sin f) + icosf) (2.82)

in Fig. 2.14 is approximately the same regardless
of the gap, as characterized by the capillary
length L" in (2.84), provided that b» L" (i.e.,
ql)
Bo » 1in (2.85». As the plate spacing is decreased
Substituting (2.80)-(2.82) into equation (2.78)
and separating the i and j components yields the
Bo
such that b---+L" (i.e., is of order 1 in (2.85»
the end effects at each plate increasingly interact,
following result: and the region between them can no longer be
<TW cos f)
approximated as being flat; thus, the reference
i: v=
" pg
(2.83a) pressure P R in the fluid changes. Eventually, as
b is further decreased such that b« L" (i.e.
Bo« 1in (2.85», surface tension effects dominate
j: d = J"tg.J2(1 - sin f) (2.83b) and the circular interface shape is approached.
pg
These limits on the interface shapes, for high and
The force balance and result (2.83) indicate low Bond numbers at low capillary number, are
that it is the action of surface tension which shown in Fig. 2.15.
provides the force necessary to oppose gravity
and lift the fluid above the X = 0 plane shown
2.4.1.3 The limit of large capillary numbers
in Fig. 2.14. Note that, in the derivation of(2.83),
the shape of the interface was not considered; The above discussion on the effect of the capillary
however, the fluid underneath the interface inside number and Bond numbers has been restricted
(a) (b)
Air Air
Liquid Liquid
,......- - - - - - - b - - - - - -....~I
Figure 2.15 The effect of the Bond number on the interface shape in the low capillary number limit: (a)
Bo» 1; (b) Bo« I.
to low capillary numbers. For high capillary number, the contact-angle boundary condition
numbers, viscous forces dominate over surface can still be imposed, but its effects are confined
tension forces, and thus the pressure scale chosen largely to the vicinity of the contact line, as it is
in (2.73) is not appropriate. Instead, the new in the large Bond number case discussed above.
dimensionless pressures are written as Finally, we note that for large capillary number,
the pressure field can be quite complex and is
_ Pb _ PAb far from hydrostatic, being governed by (2.87b).
P-j1.U,PA-j1.U (2.86)
and with (2.86), equations (2.73a) and (2.73c) 2.4.1.4 Surface tension-induced pressure variations
become as a source of flow
1 For many practical cases, the flow field being
(P - PAl - [D 1 • r]· D 1 = Ca r at the interface analyzed does not fall into one of the two
m
capillary and Bond number regimes just discussed.
(2.87a) However, despite this fact, we can use the above
au - ] - -2
discussion to assess qualitatively the effect of
Re [ af+u·Vu =-Vp+Vu-Sti (2.87b) surface tension on the pressure in a fluid. Consider,
for example, the case of a wavy air-liquid
where St = pgb 2 /(j1.U) is called the Stokes number, interface with surface tension as shown in Fig.
which gives a ratio of gravitational to viscous 2.16. From (2.73a), it is evident that at moderate
forces. The important point to note in (2.87a) is capillary numbers, viscous effects can be viewed
that in the limit of infinite capillary number, the as modifying the effect of the static-looking
term including the mean radius of curvature interfacial curvature on the pressure drop across
disappears. For cases where the surface tension the interface. Nevertheless, we can often quali-
is identically zero (corresponding to Ca being tatively determine the pressure field in the fluid
infinite), we have already noted that the loss in relative to atmospheric pressure by examining
the mean radius of curvature term dictates that the curvature of the interface and viewing the
boundary conditions at the contact line cannot pressure drop as being due to a static interface.
be imposed. However, for large but finite capillary For example, in Fig. 2.16, the pressures relative
Interface having Surface Tension Air
Pressure-Induced Pressure-Induced
Flow -----i.~ P<PA ......t - - - -Flow
Liquid
Wall
Figure 2016 Qualitative pressures underneath a curved interface and the corresponding induced flow.
to atmospheric will be as indicated, even at the two ends of the domain will give rise to a
moderate capillary numbers, and the resulting surface tension gradient along the surface. Our
pressure gradients induced by such curvatures goal is to assess the effect ofsuch a surface tension
will provide the driving force for the indicated gradient on the fluid flow; for purposes of
fluid motion. discussion, we will restrict attention to incom-
pressible Newtonian fluids governed by the
constitutive relation (2.20b). For the domain as
2.4.2 IMPLICATIONS OF THE TANGENTIAL specified, we denote the unit outward normal
STRESS BALANCE AT THE INTERFACE vector to the liquid phase in the + Y-direction
To understand the physics associated with the as J, and the two independent unit tangent
tangential components of the interfacial stress vectors in the + X and + Z-directions as i and
balance (2.72b,c), let us apply the dynamic k, respectively. To simplify the analysis, we will
boundary condition to a completely flat air-liquid assume that the fluid is unbounded in the
interface. The coordinate system is as shown in Z-direction and that liquid temperature varies
Fig. 2.17 (the Z-direction, not shown, extends only in the X -direction; thus there is no Z
out of the figure), the interface is assumed to be dependence in the problem.
located for all time at Y = 0, and there is no mass Since the interface is flat, there are no curvature
transfer across the interface. We will assume that effects to motivate consideration of the normal
the left-hand side of the interface experiences a stress balance at the interface. Instead, the heart
higher temperature than the right-hand side, as of this model problem is the balance of the
shown in the figure. Furthermore, we will assume tangential stresses in the i-direction, as described
that the viscosity of the air is negligible and the by (2.72b) with t" = i. The terms in this equation
air pressure is a constant, P A' Let us denote the are evaluated as follows. Because the viscosity
liquid as phase 1 in (2.72) and the air as phase of the air is assumed to be negligible, '2 = 0;
2. Where clarity is necessary, variables in the thus, the jump in tangential stress across the
liquid phase are subscripted with a 1; otherwise, interface in Fig. 2.17 becomes
as all the dynamics is limited to the liquid phase,
variables will be unsubscripted. [° 1 °('2 - 'I)Jo. = [j°(-'1)]0. =
Recall that the surface tension of the liquid-air
interface is a function of temperature, T, where
-['t x'! + 'Iry no. = -'In (2.88a)
typically au/aT < 0 (see (2.63) and associated where ' I n is a component of the viscous stress
discussion). Thus, the different temperatures at tensor, which is expressed in terms of the. and
Air
. . :-Fo~~L--
Material Body
Interface
F X X+~
X
Hot - ---l.-.....~-----~O~_f=~~=~=~~=3--- Cold
FI!
Direction of Material
Body Motion -
Liquid
Surface Ten ion

Induced Flow
Figure 2017 Flat interface on which dynamic boundary condition is applied. The action of forces in the
X-direction is shown for a differential element of a surface. Here, F" and F /l denote the surface tension and
shear forces acting on the material body, where the direction and magnitude of the forces are as indicated by
the vectors.
j velocity components denoted respectively by The result (2.89) indicates that viscous stresses
U x and U y , as balance the surface tension gradient along the
free surface. To see this more clearly, consider a
iJUx iJUy) force balance on an element of the flat surface
'Ixr=Ji. ( iJY + iJX (2.88b)
of length i1X and width i1Z, as shown in Fig.
Equation (2.88b) can be simplified by using the 2.17, where the thickness of the body, i1n, is
kinematic boundary condition (2.40). Along Y = 0, negligible. For such a body, the force balance
the kinematic condition yields U y = 0 for all X, (2.53) applies; with the constitutive relation (2.58),
implying that auylax is equal to zero. Using the force balance is written explicity as
f f f
the definition (2.55), we obtain z+t.z x+t.x z+t.z
__ iJ iJ_ J-(T 2 - TI)dXdZ + erl x(-i)dZ
VII = V - j(joV) = iJX i + iJZk
z x z
f
Thus, the contribution ofsurface tension gradients z+t.z
to the tangential stress balance in the i direction + erl XHX idZ = 0 (2.90a)
becomes
z
(2.88c) Let us now simplify (2.90a) focusing attention on

the tangential balance of forces, i.e., those in the
X-direction. By assumption, there are no variations
Equating (2.88b) and (2.88c) in accordance with
in flow in the Z-direction. Thus, as in (2.88) above,
(2.72b) yields the desired balance of tangential
stresses at the interface: [J'(T 2 - TI)]oi = [(PI - PA)j + -jo'I]oi
Y=O (2.89) = -'I xr = -Ji. ( iJy

iJU x) (2.90b)
where we have utilized (2.18) in writing this motion will convect the solute in the system, and
expression. Substituting (2.90b) into (2.90a), the the effect of this convective transport may be
i component of (2.90a) can be rewritten as assessed with the aid of (2.68). For the current
(UIXHX~Z)i - (ulx~Z)i
problem, we are assuming that the interface is
flat, there is no flow in the Z-direction and the
f
X+.1.X
surfactant does not transfer into the air (i.e., the
=( Jl a~; dX )~Zi (2.90c) surfactant is non-volatile); thus (2.68) is written as
x
By inspection, we see that the result (2.90c) is an (2.91)
integrated form of the local condition (2.89). This
result can be simply expressed as
where the third term of(2.91) is the diffusive flux
F"l xHx - F"lx = -F~ (2.90d) of species i to the interface from the liquid phase,
where Fu and FJl are the surface tension and obtained by employing Fick's law (2.69a) (after
viscous shear forces in the X-direction acting on noting that the mass concentration Pi is related
the material body, respectively, and are defined to the molar concentration C j by the molecular
by the terms in corresponding locations in (2.9Oc). weight of the species i).
As indicated by (2.90d) and as indicated by the The surfactant affects the resultant fluid flow
vectors drawn in Fig. 2.17, the net surface tension only if the right-hand side of (2.91) is nonzero,
force on the material body must be opposed by as variations of r j in the X-direction are required
the retarding shear force due to fluid viscosity. to adjust the magnitude of the surface tension
Such a shear force can be generated only if there gradient au/ax in (2.89) through equation (2.63).
is a gradient in fluid velocity in (2.90c), which Since the system was originally at chemical
implies that fluid flow occurs. equilibrium, C j was independent of position.
Because surfactants reduce the surface tension Thus, nonzero values of arJax must derive from
ofthe interface, just as an increase in temperature accelerations or decelerations of the interface (i.e.,
does (see (2.63)-(2.66) and associated discussion), nonzero values of au x/ax). Based on this
we can use the above reasoning to explain how observation, we can conclude that the presence of
the addition of a surfactant to a locality on the surfactant will have no effect on the flow if the
interface can also induce interfacial motion radially fluid layer is unbounded (so that r j is at its
outward from the point where the surfactant is equilibrium concentration at either end of the
introduced. But in coating flows, surfactants are system) and the flow in the X-direction is uniform.
more often responsible for damping out flow In practice, coating flows are usually bounded
disturbances, by imparting a rigidity to the or involve regions where fluid-fluid interfaces
interface with respect to tangential motion. are stretched or compressed. Even in the case
The retardation of interfacial flows by surfac- where a liquid layer issues from a two-dimensional
tants can be explained by considering what slot and forms a liquid curtain, such as shown
happens if our fluid layer in Fig. 2.17 contains in Fig. 2.3, the interface is bounded in that it
some surface-active solute ofmolar concentration originates at the slot exit. In the curtain itself
C j • Before a temperature gradient is imposed and close to the slot, surfactant is adsorbing into
along the interface to induce flow, the interface the interface from the liquid phase, because the
is in chemical equilibrium with the bulk liquid, interface has not had an opportunity to equilibrate
which means that the surface concentration of with the bulk liquid, and the interface is
solute, r j , is in equilibrium with the bulk accelerating in reaction to the sudden transition
concentration, and a relationship such as (2.65) from the no-slip condition in the slot to zero
is satisfied. When the temperature gradient is viscous stresses from the air.
imposed on the interface, the resulting fluid Returning to Fig. 2.17, we can introduce a
retarding effect of surfactants on the flow by Now, with the force balance ideas used in Fig.
adding a solid wall to our system, downstream 2.17, let us examine how this concentration
of the material body, as shown in Fig. 2.18. Now, gradient can impart a rigidity to the interface
as the interface and the liquid below move in the through its effect on surface tension. Consider
positive X-direction due to the imposed tem- again a thin material body with sides ilX and
perature gradient, surfactant is carried downstream ilZ in the interface, now positioned near the end
where it accumulates near the solid wall. The wall as shown in Fig. 2.18. As the geometry and
explicit details of the spatial variation ofsurfactant assumptions are the same as in the previous
in the interface (or j aX) near the wall is determined example, the general force balance given in (2.90)
by the recirculation flow that must occur in this is valid. The one feature that is new in this
region. For the purposes of our discussion, it is problem (as compared with the discussion of Fig.
only important to recognize that surfactant 2.17) is that surface tension is now a function of
accumulates because the interface must decelerate both temperature and interfacial concentration,
as it approaches the wall. Thus, au x lax is and both are functions of X. If we assume that
negative and arJax is positive in accordance the length of the body ilX is small, then we can
with (2.91). If the surfactant is soluble, its high approximate the surface tension at X + ilX by
interfacial concentration near the wall will promote a Taylor series expansion:
desorption of surfactant into the bulk liquid to
compensate for the steady convective transport
of surfactant in the interface from upstream. By
this mechanism, a steady-state concentration
0-1 x+ 6X '" o-Ix + (~~)x il T + (:;1 ilrj (2.92a)
where
profile of surfactant is established in the interface
where arJax >0.
Air
Low Concentration of High Concentration of

Surfactant Surfactant
Material Body
y ...-- Surfactant
Fo,.
Hot Cold
~-----F~
Vector FI! without

Surfactant Addition
Direction of Material
Liquid
Body Motion -
•
End Wall
Figure 2.18 Force balance on a differential element of the surface, demonstrating the effect of surfactant
addition when there is a temperature-induced surface tension gradient. Here, FaT and Far denote the net
surface tension forces due to temperature and surfactant gradients, respectively, while FJl is the shear force
acting on the material body.
Thus, equation (2.9Oc) becomes which produces a dilation or compression of the

interface will produce a redistribution ofsurfactant
au\ LlTLlZi + (au) LlrjLlZi
( aT)x and thus generate surface-restoring forces that
ar x j
will tend to damp out these disturbances. We
close this discussion by noting that in some cases
= Jle~;)x LlX LlZi (2.92b) surface contamination which may occur, for
example, due to airborne particles, can play the
which, upon inspection, is seen to be identical
role of a surfactant in a system. The presence of
to (2.89) once the chain rule of differentiation is
such contaminants may lead to an unanticipated
applied and terms are rearranged.
Marangoni flow in a coating process while the
We can express the result (2.92b) simply as
fluid layers are still mobile; this flow can ultimately
F"T + F"r = -FI' (2.92c) disturb the final layer thicknesses on the moving
where F"T and F"r are the net surface tension substrate. Thus, while we can take advantage of
forces due to temperature and surfactant gradients the surface-tension dependence on surfactant
acting on the material body in the X -direction, concentration to enhance the quality of a coating
respectively, and are defined by the terms in by damping out undesired flows, this same
corresponding locations in (2.92b). Note that it sensitivity requires that surface contamination
is the use of the force balance (2.90c), coupled be controlled in a coating process.
with the Taylor series expansion (2.92a), that
allows the interpretation of F" and F" as being
• T ~ r 2.4.3 COMBINED EFFECT OF THE NORMAL
net forces. As an increase In surlactant concen-
AND TANGENTIAL COMPONENTS OF THE
tration decreases surface tension, the quantity
INTERFACIAL STRESS BALANCE
au/ar; is negative; recall that the quantity au/aT
is negative as well. Furthermore, Ll T < 0, indicating The model problems presented in Sections 2.4.1
that the first term on the left-hand side of (2.92b) and 2.4.2 illustrate how the interface acts in two
is a positive quantity; consequently F"T is oriented principal ways, as embodied in the dynamic
in the direction as shown in Fig. 2.18. The boundary condition. First, the normal component
temperature gradient thus provides the driving of the dynamic boundary condition indicates
force for flow. On the other hand, Llrj > 0, and that surface tension causes a jump in pressure
thus the second term on the left-hand side of across a curved interface, where the sense of the
(2.92b) is negative. pressure drop can be obtained by using intuition
Consequently, F" is oriented in the direction about the curvature of a pressurized container
as shown in Fig. 2.Ur, indicating that the presence (or by rigorously using the curvature definition
of surfactant gives rise to a surface tension force in (2.59b) in conjunction with (2.59a)). The addition
which opposes F "T. Thus, the shear force F 1" and of fluid motion complicates, but does not
the resulting temperature-induced flow, is reduced invalidate, this physical interpretation. In regard
through surfactant addition. Because surface to the pressure drop induced by surface tension,
tension is several orders of magnitude more the interface can be viewed, in some sense, as an
sensitive to changes in surfactant concentration elastic impermeable membrane. Furthermore, as
than to changes in temperature, it takes very suggested in Fig. 2.16, interfacial curvatures in
little surfactant to diminish significantly or even a problem can act to produce a pressure gradient
eliminate a flow produced by a temperature and induce motion. We have also seen that
gradient. surface tension gradients can act to induce fluid
It should be noted that the importance of flow or to oppose an existing fluid motion. In
surfactants to coating flows is not restricted to this respect, it is the fact that the fluid points are
damping out surface-tension-driven flows due to free to move along the interface which gives rise
temperature gradients. Any local disturbance to such effects.
Thus far, the normal and tangential components Fig. 2.19. Let us also assume that the surface
of the dynamic boundary condition have been tension decreases linearly with increasing tem-
examined separately in different model problems. perature as
We now examine a model problem which explicitly
demonstrates how these components operate
together and affect fluid flow. Let us consider where f3 is a positive constant, and thus
the case where a Newtonian liquid layer of
constant thickness covers a stationary horizontal (1 = (10 + f3KX (2.93)
solid surface, as shown in Fig. 2.19. Let the
interface be located at Y = H for all time, where The goal is to determine the motion within the
the coordinate system is as indicated in the figure liquid induced by such a surface tension gradient.
(the Z-direction, not shown, is oriented out of We will assume that the length of the domain,
the figure), and there is no mass transfer across L, is large compared with the film thickness H.
the interface. A linear temperature gradient is Thus, on the length scale H, the domain looks
applied to the bottom surface, which in turn infinite in the X-direction, and it is reasonable
induces a temperature variation in the liquid and to assume that the motion of the fluid is everywhere
the interface. Let the fluid be of high thermal parallel to the solid surface, i.e. the component
diffusivity (heat transfers extremely easily by ofvelocity in the Y-direction, U y, is zero everywhere
conduction), and assume that the flow which in the liquid. Furthermore, we will assume that
occurs is slow, so that convective heat transport there are no velocity variations in the Z-direction,
is negligible relative to conduction. In this limiting and the velocity component in that direction,
case, the flow field is uncoupled from the U z, is zero. At this point, we will not specify
temperature field, and it can be shown that the constraints on the ends of the domain in the
temperature in the fluid and along the air-liquid X-direction, but instead these will be examined
interface is identical to the linear temperature through the course of the problem.
variation along the wall. Thus, we write the Based on the above assumptions, we begin by
temperature variation everywhere in the domain as writing Ux = Ux(X, Y). The continuity equation
(2.7) for this case is
T=To-KX
where it is assumed that the constant K is greater
than zero, i.e., the left-hand side of the domain
is hotter than the right-hand side, as shown in which implies that Ux = Ux(Y). With this sim-
Air
T Hot ,
Interface
1 Y~-:..!:H:!....- C:::old
JL """"""",~::~:"",l~"~"""",r"~""",,,,,",.
.. T=To- KX ..
1--1->---------- L - - - - - - - _ .I
Figure 2.19 Liquid layer of constant thickness over a stationary, horizontal solid surface which is also heated.
plification, the Navier-Stokes equations (2.21) where we have used the results for the flat
become, in component form, interface (2.89) derived previously, as well as
(2.93). We note here that, because the flow is
oP d2 U x
i: (2.94a) always parallel to the solid surface and interface,
oX=fJ. dy 2
the kinematic boundary condition at both surfaces
oP is automatically satisfied, and so it is not explicitly
j: - = -pg (2.94b) written.
oy
We begin by solving for the pressure field.
oP =0 From (2.94c), the pressure is not Z-dependent,
k: (2.94c) and thus P = P(X, Y). Integrating (2.94b) and
oZ
applying (2.95b), we find that the pressure field
The no-slip condition at the solid surface is is purely hydrostatic:
Ux =0 at y= 0 (2.95a)
P = PA + pg(H - Y) (2.96a)
and the dynamic boundary condition at the free
surface is, Thus, (2.94a) becomes
2
P = PA at Y= H (2.95b) d Ux = 0
dy 2
fJ. o~; = ;; = 13K at Y= H (2.95c)
which when integrated, subject to (2.95a) and
(a)
Air PA
Interface Control Volume
Hot
+ '\. Cold
'f
'{D
Liquid
( X
1---(1)
1
L ~
@--I
1
.J
I..
• T=To-KX -
L -I
(b)
Low Pressure solil Wall High pr,ssure

I u-
t
1C
Aow~1
I..
Liquid
f X
L •I
Figure 2.20 Effect of placing end walls on the domain shown in Fig. 2.19: (a) control volume on which mass
is conserved; (b) configuration where a solid plate replaces the interface as a means for driving flow.
(2.95c), yields the desired result: for the pressure field and velocity field under
such conditions yields
pK
Vx=-Y (2.96b)
J1. 6VJ1.
P=-X+P
H2 R
-pgY (2.98a)
We note here that if, in the same geometry, a
solid plate moving in the X -direction with a and
speed V replaces the air-liquid interface, and
the pressure at the two ends of the domain in Vx = %2 (3y 2 - 2HY) (2.98b)
the X-direction are equal and hydrostatic, then
the solution (2.96b) is valid with the quantity
V/H replacing the group PKIJ1. in (2.96b). Thus, where P R is a constant reference pressure in the
the action of a surface tension gradient along the fluid. This result indicates that, when a solid plate
surface acts as a solid plate which drags fluid replaces the interface, the pressure in the fluid
along. Let us suppose, however, that we consider increases with X when the domain is closed at
a domain which is closed at the ends as shown the ends. This is consistent with the physical
in Fig. 2.20a. Although we viewed the domain intuition that fluid must turn around near the
in this problem as being infinite on the length ends of the domain, where a higher pressure
scale H, we acknowledge that the domain does affords the force necessary to do so (Fig. 2.20b).
indeed have a finite length, and the fact that the When a flat interface is present (previous case),
domain is closed on the ends will have an effect it is not possible for such a pressure profile to
on the velocity field. The effect of end walls can develop, as the dynamic condition (2.95b) will
be incorporated via a mass balance on control not allow pressure variations in the X-direction.
volume extending up to the end wall as shown The interface in the above cases sets both the
in Fig. 2.20a. Here, the kinematic boundary pressure in the fluid (through the normal
conditions along sides 1, 2 and 3 of the domain component of the dynamic condition) as well as
dictate that the normal velocity at every point the motion of the fluid (through the tangential
along these sides is identically zero; consequently, component). On the other hand, the solid surface
to conserve mass, there is no net flow along side does not moderate a pressure drop and only
4 of the body, and induces motion. This is an important distinction
between interface and confined flow problems.
f
H
The question then arises as to how the interface
VxdY= 0 (2.97) compensates for the requirement of fluid turning
around when there are end constraints in the
o
above problem. The answer to the question is
Clearly, (2.96b) does not satisfy (2.97) unless there that our assumption about the shape and
is no flow at all. Thus, (2.96b) is not an appropriate orientation of the interface is in error. Because
solution for the case where ends are placed on the channel is long relative to the thickness of
the problem. the film, it is generally not feasible to develop
To understand the implication of this fact interfacial curvatures that can provide the pressure
more fully, let us consider the identical configur- gradient needed to drive the recirculation through
ation where the air-liquid interface is replaced the action of surface tension in the normal force
with a solid wall, moving at a speed V, for the balance. Rather, the interface remains relatively
case where the domain is closed on the ends (see flat, but it becomes inclined to the horizontal as
Fig. 2.20b). The governing equations for such a shown in Fig. 2.21, the film thickness being
case are given by (2.94), subject to (2.95a), the greater at the colder end of the channel, with the
no-slip condition at the moving wall (i.e. Vx = V), slope of the interface providing the gradient in
and the no-flux constraint (2.97). The solution hydrostatic pressure needed to drive the
Air
Flat Interface
1"S:,rr~,~~:~,,,,"""~~:::~""""~:~'"::~~2r
• •
Figure 2.21 Interfacial distortion which can induce backflow to conserve mass for the domain shown in Fig. 2.20a.
recirculation through the action of gravity. It is viscous forces in the equation of motion, which
possible, however, to imagine a physical situation also compete with surface tension forces through
similar to Fig. 2.19 where interfacial curvature the dynamic boundary condition.
and surface tension provide the needed pressure
gradient to produce the recirculatory flow. For
REFERENCES
example, in the absence of gravity, such as on a
space station, any recirculating flow process Adamson, A. W. 1990. Physical Chemistry ofSurfaces.
involving liquids with fluid interfaces will contain 5th edn. New York: Wiley-Interscience.
Derjaguin, B. V., Churaev, N. V. and Muller, V. M.
such interfacial curvature effects. We note here 1987. Surface Forces. New York: Plenum. See also
that if the slope of this interface is small, the Derjaguin, B. V. 1989. Theory ofStability ofColloids
above described physics can be predicted by and Thin Films. New York: Plenum.
utilizing the hydrodynamic theory oflubrication, Dussan, V. E. B. 1979. On the spreading of liquids on
introduced in Section 2.3.1, and presented in solid surfaces: static and dynamic contact lines. Ann.
more detail in Chapter 8. Rev. Fluid Mech. 11: 371-400.
From a more practical viewpoint, in a gravi- Princen, H. M. 1969. Shapes of interfaces, drops, and
bubbles. In Surface and Colloid Science, Volume 2,
tational field, the role of interfacial curvature will ed. E. Matijevic, pp.I-84. New York: Wiley-
become important in situations where the Bond Interscience.
number becomes small (i.e., Bo < 1); in other Schlichting, H. 1979. Boundary Layer Theory. 7th edn.
words, the lateral bounds on the fluid flow would New York: McGraw-HilI.
have to be ofthe same order of magnitude as the Scriven, L. E. 1960. Dynamics of a fluid interface:
thickness of the layer. This situation is realized equation of motion for Newtonian surface fluids.
in coating flows with interfacial waves whose Chern. Engng Sci. 12: 98-108.
Slattery, J. C. 1967. General balance equation for a
wavelength is of the same scale as the film phase interface. Indust. Engng Chern. Fund. 6: 108-115;
thickness. For more complicated flow fields, the ibid. 1968. 7: 672.
effect of gravity in the above arguments is Whitaker, S. 1968. Introduction to Fluid Mechanics.
augmented via competition with fluid inertia and New Jersey: Prentice-Hall.
WETTING: STATIC AND DYNAMIC 3
CONTACT LINES
Terence D. Blake and Kenneth J. Ruschak
3.1 STATIC AND DYNAMIC WETTING IN

applications, that ofdynamic wetting is uncertain.
COATING PROCESSES
In practical coating processes, the speed limits
Wetting is basic to coating. Initially air contacts imposed by dynamic wetting failure are usually
the solid, and during coating the liquid displaces determined empirically. Nevertheless, the known
the air from the moving solid surface so that features and trends of dynamic wetting can be
none is visible in the coated film. Thus, coating useful.
is a process of dynamic wetting. For uniform Static and dynamic wetting have been reviewed
coating, the wetting line must remain straight from a more fundamental perspective by Dussan
and advance steadily. At sufficiently high speeds, V. (1979), de Gennes (1985), Schrader and Loeb
however, the wetting line becomes segmented (1992), and Berg (1993). Adamson (1982)is a basic
and unsteady as a thin air film forms between resource for general surface chemistry. The closely
the solid and liquid. The air film disrupts the related topic of capillary hydrodynamics is
uniformity of the coated film, and often air extensively developed in books by Miller and
bubbles appear in the coating. Dynamic wetting Neogi (1985) and Levich (1962).
failure limits coating speed. Coating flows are of sufficiently small scale
Static wetting also affects the coating process, that capillary forces are usually important. The
though less directly. There is a need to control surface tension of the liquid/air interface, a,
where the liquid spreads. Leakage of liquid from characterizes the magnitude of capillary forces.
the applicator or coated surface can produce If R is the characteristic radius of curvature of
contamination and waste. If the wetting lines at a mensicus, then pressures due to surface tension
the edges of the coated film migrate, the width are oforder a/ R. Pressures due to the gravitational
changes and the uniformity at the edges degrades. field are of order pgR, where p is the density of
The intent of this chapter is to present results the liquid and 9 the gravitational acceleration.
for static and dynamic wetting that are applicable If capillary and gravitational pressures are in
to coating processes, to illustrate their use, and balance, then the characteristic radius ofcurvature
to indicate their limitations. Attention is restricted is the capillary length, La = .j(a/pg). The length
to liquid/air/solid systems. The essentials of static scales R and La are often compared through the
wetting are covered first, because the theory for Bondnumber,Bo = pgR 2/a = (R/L a )2. Forlarge
dynamic wetting has developed as a generalization Bo, surface tension has negligible influence on
of the static case. Whereas the physics of static meniscus shape. For small Bo, the variation of
wetting is well enough developed for many hydrostatic pressure has negligible influence. If
London. ISBN 0 41206481 2.
64 Wetting: static and dynamic contact lines
there is flow with a characteristic speed V, the As a result of the curvature term, the
viscous pressures are of order /lV/ R, where /l is Young- Laplace equation is a nonlinear, second-
viscosity, and capillary forces are important order, partial differential equation. One boundary
when the capillary number, Ca = /lV/a, is of condition is required at the edge of a meniscus.
order unity or smaller. Similarly, inertial pressures Usually, the static contact angle is prescribed
are oforder pV 2,and capillary forces are important where the meniscus intersects a solid surface.
when the Weber number, We = pV 2 R/a, is of
order unity or smaller. In a coating flow, Bo is
3.2.1 THE STATIC CONTACT ANGLE
not likely to exceed unity, but Ca and We can
be large or small, with significant implications Under ideal conditions, a liquid at rest intersects
for dynamic wetting. a solid at a unique angle (Fig. 3.1). The static
contact angle BE is the angle at which the
meniscus intersects the solid as measured through
3.2 STATIC MENISCI AND STATIONARY
the liquid. The angle is measured in the plane
WETTING LINES
defined by the unit vector normal to the meniscus
When there is no flow, the Young-Laplace and that normal to the solid. Young's equation
equation governs meniscus shape (Chapter 2). relates BE and the surface tension of the liquid
Surface tension supports a pressure difference to the surface tensions of the solid/liquid and
across a curved mensicus, with the higher pressure solid/air interfaces for conditions of ther-
on the concave side. modynamic equilibrium and a perfectly flat,
homogeneous solid surface. With a SL the surface
(3.1)
tension of the solidjliquid interface and a SA that
Here P is the pressure in the liquid, P A the of the solid/air interface, Young's equation is
pressure in the air, and ;j(' the mean curvature,
acos(BE) = aSA - a SL (3.2)
positive for liquid on the concave side of the
meniscus. The pressure in the liquid is hydrostatic, The contact angle thus defined is that for which
increasing linearly in the direction of gravity. The no change in free energy accompanies a virtual
pressure in the less dense air is sensibly uniform. displacement of the wetting line.
If Bo « 1, then the pressure in the liquid can be If the surface tension of the solid/air interface
considered uniform as well, and the meniscus has exceeds that of the solid/liquid interface by more
a uniform mean curvature. than the surface tension of the liquid, then the
Physical insight into the Young-Laplace liquid spontaneously spreads on the solid to form
equation follows from considering the pressure a film. Such behavior, often called complete
inside a small liquid drop of radius R. For wetting, is an indication of strong intermolecular
Bo « 1, the drop is spherical. A balance of forces forces of attraction between the liquid and solid.
on one hemisphere of the drop yields the These forces give rise to a pressure, the disjoining
relationship between the pressure inside the drop
and its radius. The pull of surface tension is
tangent to the meniscus at the equator and so
is 2nRa, the circumference at the equator times
the value ofsurface tension. The ambient pressure
exerts no net force, and so the net pressure force
on the hemisphere is (P - PA )nR 2 , the gage
pressure in the liquid times the cross-sectional (JA
area at the equator. Equating these forces gives

(P - PA ) = 2a/R, where I/Ris the mean curvature
of the sphere. Figure 3.1 Equilibrium wetting line and contact angle.
Static menisci and stationary wetting lines 65
pressure, which opposes film thinning. Disjoining measurement of the so-called advancing and
pressures may have components arising from receding contact angles. These terms imply motion
both short-range and long-range molecular forces. and are misleading when applied to the static
The former include polar interactions such as case (Huh and Scriven 1971), but the usage is
hydrogen bonds, while the latter comprise van firmly established. The advancing angle () A is the
der Waals dispersion forces. A completely wetting largest contact angle achievable before the wetting
liquid spontaneously spreads by means of a line begins to move in the direction of the air
precursor film that moves ahead of the main phase. The receding contact angle ()R is the
body of the liquid. For low-viscosity liquids like smallest contact angle achievable before the
water on strongly polar substrates like glass, wetting line begins to move in the direction of
disjoining pressures are high and the precursor the liquid phase. A measurement is usually made
film may spread at velocities up to 0.1 m/s. by causing the wetting line to move and deter-
Nevertheless, such rapid spreading is uncommon, mining the angle once movement has ceased.
and precursor films are therefore of limited Young's equation (3.2) provides a theoretical
significance to industrial coating processes. basis for the contact angle but is of little direct
The contact angle can be measured directly practical use because the surface tension of the
using low-power optics or computed from solid cannot be measured directly and is usually
measurement of the force exerted on the solid subject to considerable uncertainty. In addition,
by the meniscus. It can also be inferred by fitting the solids encountered in practice seldom meet
a measured meniscus profile to a solution of the the equation's assumptions, as evidenced by the
Young-Laplace equation and computing the nonuniqueness of the contact angle. So, ()A and
angle of intersection with the surface of the solid. ()R are usually the contact angles that are
°°
experimentally determined, while 0E can be difficult
or impossible to determine and relate to A and
3.2.2 CONTACT-ANGLE HYSTERESIS
OR' Even experimental measurement of A and
According to Young's equation, the contact OR is sometimes troublesome: chemical species
angle is a unique property of the solid/liquid/air such as surfactants may diffuse among the phases,
system under consideration. In practice, however, porous surfaces may draw in liquid by capillary
a range of contact angles is usually possible, and action, or the liquid and solid may react chemically.
this effect is called contact-angle hysteresis. Two As an example, we have attempted to measure
generally accepted causes of hysteresis are the contact angle of water on a dried coating of
roughness and surface heterogeneity, as reviewed gelatin, which sorbs water in a fraction of a
by Blake and Haynes (1973) and, more recently, second (Joos 1992). The wetted gelatin swells,
by Johnson and Dettre (1993). For rough surfaces, and the wetting line, at the boundary between
Young's equation may still apply on the scale of the dry and wetted areas, sticks and shows
the roughness, but the apparent contact angle massive hysteresis. Once it breaks free, the wetting
measured on the macroscopic scale usually differs line moves a considerable distance before coming
from that existing on the microscopic scale. For to rest. In this case, the measurement and
heterogeneous surfaces, there cannot be a unique interpretation of the advancing contact angle are
contact angle. Microscopic surface roughness at best uncertain.
and heterogeneity make possible multiple equi- Even on uncomplicated surfaces, poor measur-
librium configurations which appear as hysteresis ing technique may overestimate hysteresis.
on the macroscopic scale. Jansons (1985) has Equilibrium must be achieved before measurement
developed the mathematics relating the micro- is made. A common mistake is to take measure-
scopic and macroscopic contact angles for nonideal ments while the wetting line is in slow but
surfaces. detectable motion. Wetting line speeds as small
Quantifying the extent of hysteresis requires as 10 Ilm/s can cause the contact angle to change
significantly from its static value and become a 3.2.3 EXAMPLES OF STATIC MENISCI WITH
dynamic contact angle. A WETTING LINE
In general, hysteresis reduces wetting line
Except for special cases, the Young-Laplace
mobility. The wetting line does not move toward
equation cannot be solved in closed form. Two
the air if the contact angle is less than the
applications follow showing how the static contact
advancing contact angle or toward the liquid if
angle can affect the equilibrium configuration of
the contact angle is greater than the receding
the liquid. In the first example, the contact angle
angle. At the wetting line of a static mensicus,
determines the extent of spreading of a liquid.
the contact angle () can therefore assume any
In the second example, the contact angle can be
value between ()A and ()R: inconsistent with meniscus shape so that no
()R ~ () ~ ()A (3.3) equilibrium configuration is possible; the wetting
If the solid has an edge, the wetting line may line therefore moves giving rise to a dynamic
preferentially locate along it (Oliver, Huh and contact angle.
Mason 1977). In effect, an edge augments hysteresis
and thereby further reduces wetting-line mobility.
Fig. 3.2 shows a wetting line positioned on an 3.2.3.1 Application to capillary rise and wicking
edge with included angle y. The wetting line
A simple example ofcapillary hydrostatics relevant
remains on the edge for a range of contact angles
given by Gibbs' inequality to coating operations is meniscus rise in a
capillary tube (Fig. 3.3). The tube is inclined at
()R ~ () ~ ()A + 180· - y (3.4) an angle f3 to the horizontal, and one end
The sharper the angle, the greater the apparent contacts a pool of liquid. The inside radius of
hysteresis. In the design of coating applicators, the capillary R c is sufficiently small (Bo « 1) that
edges are commonly exploited to 'pin' a wetting the variation in hydrostatic pressure along the
line. meniscus is negligible. In this case, the meniscus
pinned
meniscus
.... ....
.... ....
\
range for hinging effect
I
I
I
I
I
I
.... .... I
I
I
Figure 3.2 Conditions for a wetting line to advance or retreat from an edge.
Static menisci and stationary wetting lines 67
a circular cylinder is given below by equation

(3.12). Tightly fitting parts reduce flow rate but
do not stem flow (unless flow rate is so low that
solidification occurs by drying or some other
mechanism, creating a total blockage). Tightly
fitting parts can also be expensive and difficult
to use and maintain. Capillary wicking can be
prevented by making the hardware from non-
wetting material or by applying a nonwetting
coating to parts so that eA > 90·, but surface
contamination may defeat this approach. Capillary
breaks can be designed into hardware to halt
Figure 3.3 The rise of a meniscus in a capillary tube. spreading. In this case, geometry is such that the
mensicus achieves an equilibrium configuration,
is a spherical cap of radius R, and the and the liquid progresses no farther.
Young- Laplace equation (3.1) is simply
2ulR = P - PA' The contact angle, presumed
constant along the wetting line, determines the
3.2.3.2 Application to a meniscus that is asym-
radius as Rdcos e, where the contact angle e
ptotically planar
satisfies condition (3.3). So, the pressure in the
liquid at the meniscus is below atmospheric by Amore complicated example is a two-dimensional
2u cos eiRe' The equilibrium distance of the meniscus that is planar at large distances from
meniscus along the capillary from the surface of the wetting line. The plane approached by the
the pool, Leo, must be consistent with the meniscus is initially presumed to be horizontal;
hydrostatic pressure field in the liquid. Equating how this restriction can be relaxed is described
the pressure drop across the meniscus to the later. The meniscus can represent the surface of
hydrostatic pressure drop between the pool a pool or the edge of a film of uniform thickness.
surface and the meniscus gives Coordinate Y is measured vertically from the
plane. The horizontal coordinate of the meniscus
2ucos e is F(Y), and the Young-Laplace equation may
L =---- (3.5) be written
00 pgResinp
Note that the meniscus is above the surface of F yy

the pool only for e < 90·, in which case the liquid
is said to wet the surface of the capillary. Because
of contact-angle hysteresis, Leo is larger if the
meniscus falls toward its equilibrium position The term on the left is meniscus curvature, and
than if it rises; this is because eR < eA' The that on the right is the hydrostatic pressure in
displacement of the meniscus along the capillary the liquid. Upon introducing the dimensionless
is large for low angles of inclination and is variables y = YILa and f = FILa' where La is the
unbounded for a horizontal capillary. The capillary length, this equation becomes
displacement also increases with decreasing
capillary radius.
The wicking of liquid through narrow cracks
and crevices is a mechanism by which liquid can
spontaneously move long distances and leak
from coating hardware. The rate of wicking for A first integration, for y > a and with boundary
condition/y- - 00, y-O+, gives A given film thickness corresponds to one contact
y2 angle, which must satisfy condition (3.3) for
--1 equilibrium. Otherwise, 8 is a dynamic contact
2
J( 1-~2) (0 < Y < 2) (3.8) angle, and the wetting line moves (see Section
/y = 3.3). Low levels of contact-angle hysteresis favor
Y wetting-line motion and an unsteady meniscus.
In the absence of significant hysteresis, the edges
A second integration, with the arbitrary condition of coatings tend to be dynamic.
that f = 0, y = J2 (the meniscus is invariant to If the solid surface is inclined at an angle p to
a horizontal translation), completes the solution. the horizontal as shown in Fig. 3.6, then the
component of gravity parallel to the incline,
f = J2 - J(4 - y2) + In[~(2 + J(4 - y2))J 9 sin p, drives a rectilinear flow parallel to the
wetting line that exerts no stress on the meniscus.
-In[1 + J2] (3.9) The component of gravity perpendicular to the
The meniscus shape is plotted in Fig. 3.4. incline, 9 cos p, produces a hydrostatic pressure
A solution for y < 0 is f( - y), which is a field. So, meniscus shape is determined by equation
reflection across the horizontal axis. This solution (3.6) with 9 replaced by 9 cos p. This combination
can describe the edge of a film resting on a of rectilinear flow and a static meniscus represents
horizontal, solid surface (Fig. 3.5). The contact the edge of a liquid layer on the inclined surface
angle and the dimensionless asymptotic film of a slide coater (Chapter 11). The flow-rate
thickness, h oo = H oo/L", are related as follows distribution near the edge is affected by the
(see equation (3.8»: contact angle through the meniscus shape.
Stationary wetting lines can exist when there
is flow, as in this example. In such cases, boundary
h~ _ 1
1t 2 condition (3.3) for a static contact angle still
8=2+ tan - 1 (3.10) applies, even when meniscus shape is not
determined by the Young-Laplace equation
because of hydrodynamic stresses.
2
3.3 DYNAMIC MENISCI AND WETTING LINES
Liquid displaces air from a solid surface in a
Y1 coating process. Inevitably, there is a wetting line
that moves with respect to the solid.
In the flow of liquid onto the solid, viscous
2 3 4 5 6 7 stresses and dynamic pressure generally influence
f (y) meniscus shape. The meniscus shape and flow
Figure 3.4 Plot of equation (3.9), the shape of a static

meniscus that asymptotes to a horizontal plane.
Figure 3.6 The edge of a layer in fully developed flow

Figure 3.5 The edge of a horizontal layer. on an inclined plane.
Dynamic menisci and wetting lines 69
field are coupled, and surface tension acting in factors influencing the dynamic contact angle are
a curved meniscus supports a net stress normal not well studied.
to the meniscus. The methods used to measure fJ o are similar
to those used to measure the static contact angle.
In many experiments, eo is measured directly
3.3.1 DYNAMIC WETTING LINES
through low-power optics. However, if the flow
Ablett (1923) was the first to find that the contact is so weak (small Ca and We) that the meniscus
angle changes with the speed of the wetting line is essentially a static meniscus, then fJ o can also
relative to the solid. The contact angle observed be inferred by fitting a solution of the
experimentally at a moving wetting line is called Young- Laplace equation to the observed profile,
the dynamic contact angle, fJ o' It usually increases and mathematically extrapolating the solution
steadily with speed and ultimately approaches to intersect the surface of the solid.
its maximum possible value of 180°. Except for Examples ofdynamic contact angles are shown
extremely low speeds and special cases, the in the photographs reproduced in Figs 3.7 and
dynamic contact angle differs appreciably from 3.8. Figure 3.7 shows a vertical tape entering a
its static advancing or receding values. Unfortu- pool of liquid. Figure 3.8 shows a liquid curtain
nately, there is no dependable way of predicting impinging on a horizontal tape (see Section 11.3).
the dynamic contact angle in flows where it has Through such observations, the contact angle
not been measured. In addition, some of the can be determined at a distance of, at best, about
wetting line
curved tape
Figure 3.7 Photograph of a poly(ethylene terephthalate) tape entering a pool of aqueous glycerol (viscosity
77 mPa s and surface tension 63 mN m - 1). The tape is bent into an arc for a clear view of the meniscus profile
near the wetting line; a portion of the smooth wetting line is visible. At the speed of 0.05 m s - 1, the dynamic
contact angle is about 143°. The air appears dark due to total internal reflection at the interface.
.........----1 frlm ----.~...
wlJtllcal
cuf1$.ln
+u- hortzont:etl substrA!Q
Figure 3.8 Photograph by A. Clarke of the impingement of a curtain of 15% aqueous gelatin on a moving
tape (nominal viscosity 63 mPa s). The curtain height is 3 cm, and the volumetric flow rate is 4.3 cm 3 s- t per
cm of curtain width. At the speed of 0.83 m s- t, the dynamic contact angle is about 156°.
0.01 mm from the solid. Little information is result in a changing contact angle. In that case,
available at smaller distances. In this chapter, it the contact angle is determined by some molecular
is this measured or apparent contact angle that rate process. This mechanism would seem to be
is referred to as the dynamic contact angle. operating, for example, in cases where the dynamic
The dynamic contact angle is not a material contact angle increases with speed from the static
property and does not have the fundamental contact angle and then levels out over some
significance and general applicability of the static higher range of speeds. The second general
contact angle. From the practical point of view, mechanism is hydrodynamic shaping of the
the angle is mainly important because dynamic meniscus. This mechanism would seem to be
wetting failure occurs as its value approaches operating in cases where the dynamic contact
180· (Section 3.4). Mathematically, the dynamic angle depends on some macroscopic flow variable,
contact angle is required as a boundary condition such as flow rate. Hydrodynamic shaping of the
for modeling problems in capillary hydrodynamics, meniscus is most likely at high capillary and
just as the static contact angle is required for Weber numbers where the influence of capillarity
modeling problems in capillary hydrostatics; the becomes weak. Kistler (1983) computed flow
curvature term of equation (3.1) appears in the fields in the (singular) limit of infinite capillary
hydrodynamic equations. number. Here, the meniscus-curvature term is
Contact-angle variation with speed is usually lost from the equations of hydrodynamics, and
attributed to one or both of two general no contact-angle boundary condition is needed
mechanisms. An upset of the thermodynamic to compute the macroscopic flow; rather, the
equilibrium expressed by Young's equation may dynamic contact angle is computed together with
the flow field. Capillary forces are important very scale becomes the problem of the unknown
near the wetting line, but the macroscopic flow contact angle on the submicroscopic scale.
is unaffected because the flow of information is Nonetheless, hydrodynamic bending of the
toward the wetting line. Ostensibly, hy- meniscus on a submicroscopic scale provides a
drodynamics should not affect the dynamic contact possible hydrodynamic mechanism for the change
angle at small capillary number. However, it will of the dynamic contact angle with speed that can
be seen that the high stresses expected near the operate at capillary numbers significantly less
wetting line provide a roundabout way for than unity.
hydrodynamics to influence 0D at capillary The force singularity can also be dealt with
numbers signficantly below unity. by modifying the flow equations and boundary
Beyond the uncertainty surrounding the dy- conditions to alleviate the singularity and
namic contact angle, prospects for modeling are continuing the solution up to the wetting line.
further clouded by a singularity that arises at the A popular modification is to relax the no-slip
wetting line when the equations and boundary boundary condition of classical hydrodynamics
conditions ofclassical hydrodynamics are solved. at the solid surface to eliminate the velocity
The hydrodynamic problem posed in the cus- discontinuity (Huh and Scriven 1971; Huh and
tomary manner does not produce a physically Mason 1977; Dussan V. 1979). Very near the
acceptable solution (Huh and Scriven 1971; wetting line, the speed of the liquid adjacent to
Dussan V. and Davis 1974). Stresses are un- the solid is assumed to differ from that of the
bounded near the wetting line, and the force solid so that the limiting velocity as the wetting
exerted by the fluids on the solid is also unbounded. line is approached along the solid is zero. The
The origin of the singularity is a discontinuity force exerted on the solid is then finite, although
in velocity at the wetting line. In the frame of stresses in the liquid are still unbounded. So far,
reference in which the wetting line is stationary, however, no specific slip model has been estab-
the limiting velocity as the wetting line is lished, and the physical validity of this approach
approached along the meniscus is zero, but remains unproven. The hydrodynamic model,
because of the no-slip boundary condition, the with or without slip, predicts significant bending
limiting velocity as the wetting line is approached of the meniscus on the submicroscopic scale, and
along the solid is the velocity of the solid. a submicroscopic contact angle Ow differing from
Despite much attention, the best way to cope the dynamic contact angle is required.
mathematically with dynamic wetting lines has It is sometimes assumed that the submicroscopic
not been settled. The solution can be truncated contact angle needed in hydrodynamic models
at a distance from the wetting line that is small is the advancing static contact angle. However,
compared with the macroscopic scale of the flow because of hysteresis, the static angle itself may
but large enough to be observationally accessible. not have significance at the submicroscopic level.
The dynamic contact angle must then be deter- Furthermore, because of the risk that a dynamic
mined by experiment. Alternatively, the solution wetting line is not close to thermodynamic
can be truncated on the molecular scale, where equilibrium, the possibility that Ow depends upon
the continuum description breaks down (Voinov speed must also be considered.
1976). When this is tried, it is found that When the air or liquid extends to a thin film
hydrodynamic forces cause the meniscus to bend having a thickness of the order of 1 ~m or less,
on a submicroscopic scale. In consequence, a as is the case when the dynamic contact angle
contact angle differing from the observed contact is near O· or 180·, long-range molecular forces
angle needs to be specified at the cut-off point. must be taken into account. Teletzke, Davis and
Inevitably, this microscopic contact angle, Ow, is Scriven (1988) modified the hydrodynamic equa-
observationally inaccessible, and the problem of tions by including these effects as disjoining
the unknown contact angle on the macroscopic pressures. This modification does not resolve the
problems at the dynamic wetting line, but empirical or at least semi-empirical approach
disjoining pressures evidently do affect meniscus must be taken.
shape over submicroscopic distances from the
surface ofthe solid. de Gennes, Hua and Levinson
(1990) applied the modified equations to a dynamic 3.3.2 EXAMPLES OF FLOW WITH A DYNAMIC
wetting line for a small dynamic contact angle. WETTING LINE
They concluded that van der Waals forces
determine the cut-off for the classical hy- Three applications follow that illustrate how the
drodynamic model. Inclusion of long-range dynamic-contact-angle boundary condition can
molecular forces allowed the solution to be affect flows. In the first example, the contact
continued to a distance from the wetting line of angle drives the flow; in the second, it determines
the order of molecular size, at which scale the the pressure to apply to a coating bead; in the
continuum description must break down. de third, it determines the migration of the wetting
Gennes, Hua and Levinson argued that slip can line as speed changes.
be expected only at the molecular scale.
A difficulty inherent in extending the hy- 3.3.2.1 Meniscus advancing in a capillary tube
drodynamic solution to the wetting line or to
molecular distances from the wetting line is that When the capillary tube of Fig. 3.3 is first touched
the scale over which the meniscus bends is very to the liquid (time t = 0), the meniscus begins
small compared with the scale of the macrosopic moving up the capillary toward its equilibrium
flow. This disparity complicates the solution of position. If the capillary and Weber numbers are
flow problems. However, the macroscopic solution small, then the flow does not affect meniscus
can be required to conform to the asymptotic shape, which is a spherical cap as previously
behavior of the flow as the wetting line is discussed. This type of flow problem was con-
approached, if that can be determined. A classical sidered by Washburn (1921). For the case of
mathematical device is to use matched asymptotic wetting, the pressure in the liquid at the meniscus
expansions to exploit the small parameter that is 2acos(Oo)/Rc below atmospheric pressure.
is the ratio of the submicroscopic length scale to The pressure at the level of the pool is atmospheric,
the macroscopic length scale. Alternatively, if the and consequently there is a negative pressure
problem is solved by large-scale numerical analysis, gradient that drives the liquid up the capillary.
then substantial grid refinement is necessary near With L the distance of the meniscus along
the wetting line (for example, Huh and Mason the capillary, this pressure gradient is
1977; Tilton 1988). - 2a cos (Oo)/RcL. If inertial effects are negligible,
In addressing a coating flow problem, a central the pressure gradient and the volumetric flow rate
question is the importance ofthe dynamic contact are those for Poiseuille flow, except within about
angle to the issue at hand. If the dependence of a tube diameter of the meniscus. The volumetric
the response of interest on the dynamic contact flow rate, in turn, determines the rate of meniscus
angle is weak, then it may be sufficient to truncate advance, dL/dt, through a mass balance
the solution away from the wetting line and apply 8J1. dL 2acos(Oo) .
a reasonable dynamic contact angle. In high- R~ dt = RcL - pg sm IX (3.11)
speed coating flows, the dynamic angle is always
well over 100°. Experimental measurements can For mathematical simplicity, wetting line speed
perhaps supply values where greater precision is presumed so slow that the dynamic contact
matters. For many problems this is enough. If angle does not change much from the static
the contact angle is central to the issue at hand, advancing contact angle. Because 00 increases
as for dynamic wetting failure (Section 3.4), then with speed and cos(Oo) decreases, this simplifi-
modeling is probably not an option, and an cation overpredicts meniscus speed in the more
likely case that eo does increase significantly.
Eventually, the meniscus reaches the equilibrium
position L oo given by equation (3.5). The solution
to the differential equation is
t= 8~L~ {_~
pgRcsmp L
-In(1 - ~)}
L
(3.12)
oo oo
At small times, or at all times for the special case

of a horizontal capillary,
R c (J cos(e A)t
L~ t-O (3.13)
2Jl.
and hence the common observation that L oc Jt.
On the other hand, at large times the meniscus
approaches its equilibrium position exponentially
-L ~
1 - exp (pg
- R~ sin pt ) t - 00 H,
L oo 8Jl.L oo '
(3.14)
In each case, surface tension drives the flow via
the dynamic-contact-angle boundary condition.
Figure 3.9 Slide-hopper bead coating.
3.3.2.2 Wetting meniscus of a coating bead

In many coating methods, the applicator is applicator at a static wetting line that is presumed
spaced a small distance H G from the substrate. to be pinned at an edge. To a first approximation,
The liquid bridge spanning this gap is sometimes the shape ofthe film-forming meniscus and hence
called a coating bead. The bead has two menisci: PB are determined by speed, coating thickness,
one forms the coated film, and the other wets and the properties of the liquid. The capability
the moving solid surface. The slide hopper bead for setting Ps to subatmospheric pressure becomes
coating method (Section 11.2) is shown schemati- essential to attaining equilibrium as PB falls. Such
cally in Fig. 3.9. The upper, film-forming meniscus is the case, for example, as thinner films are formed.
of the bead is curved such that the pressure in The configurations for the wetting meniscus
the bead, PB' is below atmospheric pressure. can be determined with reference to the circles
Application to the lower, wetting meniscus of a of Figs 3. lOa and b. The dynamic wetting line is
pressure Ps that is below atmospheric and of the point 1; point 1 in Fig. 3.10b is 180· from point
same magnitude as PB permits thinner films to 1 in Fig. 3.10a. The dynamic contact angle
be coated at speeds higher than would otherwise indicates the side of the meniscus that is the
be possible (Beguin 1954). liquid; hence P B is less than P s in Fig. 3.10a and
Under conditions of small capillary, Weber greater than Ps in Fig. 3. lOb. Points 2a and 2b
and Bond numbers, the wetting meniscus is a meet the gap requirement and are possible
circular arc of radius R. Ruschak (1976) modeled locations for the pinned wetting line. The largest
the bead under these circumstances. A dynamic possible gap occurs when points 2a and 2b
contact angle eo applies at the moving solid coincide with point 3. Figure 3.10a shows that
surface. The wetting meniscus contacts the the gap can range from 0 to a maximum of
31-:---H--':"'"
3 I--...L------f~__:_..Joj
(a) (b)
Figure 3.10 Possible configurations for the circular wetting meniscus of a coating bead.
R + Rcos(Oo) for Ps > PB' and Fig. 3.10b from allows a broader range of bead pressures, and
o to R - Rcos(Oo) for Ps < PB. Thus, hence coating conditions, to be consistent with
the configurations available to the wetting
H
o ~ Ii:o ~ 1 + cos(Oo) (Ps> PB) (3.15) meniscus. The versatility of the wetting meniscus
also provides important process latitude, as
otherwise P s would have to be set at precisely
one value. Small gaps are necessary in bead
coating because, as H o increases, the latitude for
Although these inequalities were derived with R Ps to vary around P B becomes smaller and
fixed and H o varying, they hold also for H o fixed approaches zero. Evidently, the capability to
and R varying. adjust Ps to a value less than P B is particularly
An inequality for the pressure drop across the important for dynamic contact angles near 180°
wetting meniscus follows by substituting into that are typical of high-speed coating. Without
equations (3.15) and (3.16) the relationship between suction (p s = P A)' P B - P A is negative, but the
pressure drop and curvature, equation (3.1). configurations available for the wetting meniscus
limit PB to values just slightly below atmospheric
-1 - cos(Oo) ~ (PB - Ps)H o ~ 1 - cos(Oo)
pressure. Indeed, in the limit as 00 approaches
(J
180°, no values for P B are possible without
(3.17)
suction, and the method fails.
If this relation is violated, a circular wetting Setting Ps below atmospheric pressure can
meniscus is not possible, and a failure mode such also shift to higher speeds the onset of dynamic
as bead breakup occurs. For a specified dynamic wetting failure (Section 3.4), which occurs as 00
contact angle, a range of pressure drops is approaches 180°. In bead coating, the extent by
possible because the wetting meniscus can adopt which Ps can be reduced for this purpose is
different curvatures. The application of suction limited, as equation (3.17) makes plain.
3.3.2.3 Tape entering a pool the meniscus shape under static conditions
depends upon the static contact angle. When
Frequently, dynamic contact angles are measured
the tape is at rest, the wetting line is above
by plunging a tape vertically into a pool ofliquid
the surface of the pool if the liquid wets the web
(Fig. 3.11). Usually, air entrainment occurs for
(8 A < 90°). As tape speed increases, the
capillary numbers above a critical value of order
dynamic contact angle increases, and the wetting-
0.1. With capillary and Weber numbers smaller
line position moves downward. For example, the
than unity, the meniscus has roughly the
meniscus is essentially flat and horizontal when
Young-Laplace shape given by equation (3.9).
the dynamic contact angle is 90°. Wetting-line
The meniscus shape under dynamic conditions
position and the dynamic contact angle are
depends upon the dynamic contact angle just as
related by
tan(8 D ) =
x
J( x 2 )'
1--
4
x ../2> x > -../2, x#-O (3.18)
o As x passes through zero, 8D passes through 90°.

Equation (3.18) is plotted in Fig. 3.12. The main
effect of the motion of the tape is to increase the
dynamic contact angle, which in turn changes
the shape of the static meniscus describing the
pool surface through the contact-angle boundary
condition. As the capillary number is small, the
viscous drag of the tape on the liquid is not
Figure 3.11 The meniscus for a tape entering a pool directly responsible for the migration of the
vertically. wetting line.
1,5
~
0,5 ~
x 0 ~
-0.5 ~
., ~
-1,5
~
o 30 60 120 '50 180
deg
Figure 3.12 Wetting-line position versus dynamic contact angle for a tape entering a pool vertically, as
predicted by equation (3.18).
3.3.3 THEORIES FOR THE DYNAMIC fluids on the motion of molecules (Cherry and
CONTACT ANGLE Holmes, 1969) can also be taken into account,
There is, as yet, no complete and experimentally and a nonhydrodynamic effect of viscosity on
verified theory for the dynamic contact angle. the contact angle then results. The final expression
relating the wetting line speed U to the dynamic
Two different approaches to the problem are
contact angle is
outlined here. According to the first, the motion
of the wetting line disturbs the thermodynamic 2K s L M hp
U=-"----'''----'-
equilibrium that gives rise to the static contact J1.v
angle. Dynamic wetting is viewed as a molecular-
rate process, and the dynamic contact angle is sinh(2:kT[COS(OE) - cos(0 0 )] ) (3.19)
considered to be the actual contact angle. The
second approach is hydrodynamic, and the Here K s is the frequency of molecular displace-
dependence of the dynamic contact angle on ments at a stationary wetting line, LM the length
speed is attributed entirely, or at least in part, of an individual displacement, hp the Planck
to hydrodynamic bending of the meniscus so constant, v the molecular flow volume, N the
close to the wetting line as not to be visible. In number of adsorption sites per unit area, k the
this case, the dynamic contact angle differs from Boltzmann constant, and T the absolute tem-
the actual contact angle on the submicroscopic perature. An estimate for LM is 1/..1N. Hence,
scale. there are effectively just two parameters, K s and
N, that are determined by fitting experimental
data. For small angles,
3.3.3.1 Molecular-kinetic theory
The molecular-kinetic theory of dynamic wetting
01- oi ;: ; Ca ( VNkT) (3.20)
KsLMh p
is based upon the methods developed by Eyring that is, U oc.01.
and others (Glasstone, Laidler and Eyring 1941).
The theory views wetting-line movement statis-
3.3.3.2 Hydrodynamic theory
tically as a stress-modified rate process composed
of individual molecular displacements (Cherry A second approach, first investigated by Hansen
and Holmes 1969; Blake and Haynes 1969; Blake and Toong (1971), is to account for much or all
1993). In the simplest case, the velocity dependence of the change in the dynamic contact angle
of the contact angle is due to the disturbance of through hydrodynamic bending of the meniscus
adsorption equilibria at the wetting line. The at distances so close to the wetting line as not
driving force that alters the energy barriers to to be visible. Voinov (1976) developed the
molecular displacements, so that the wetting line mathematics of this approach for a liquid
can move, may have components arising from displacing air. He considered conditions for
both surface tension and viscous shear stress. which the slope of the meniscus changes slowly
According to Blake and Haynes, surface tension with distance from the solid. The flow field local
provides the primary driving force for wetting-line to a point on the interface can then be approxi-
motion: specifically, O'[COS(OE) - cos(Oo)]. This mated by flow in a wedge with an angle determined
is equivalent to O'SA - O'SL - 0' cos(Oo).lf 00 > OE' by the solid surface and the local meniscus slope.
then the state ofequilibrium described by Young's An integration of these wedge flows yields the
equation is not possible, and the wetting line meniscus shape. Voinov avoids the wetting line
moves. Although molecular interactions between singularity by truncating the solution at molecular
the liquid and solid provide a major barrier to dimensions. He acknowledged but did not treat
wetting line movement, the retarding influence situations where long-range molecular forces
of the viscosities of the advancing and retreating must be taken into account.
Voinov's primary result is Ls (inner region), slip is important. The flow
domain is wedge shaped, with an included angle
x(O) - X(Ow) = Ca In(Y/Yw) (3.21) equal to the submicroscopic contact angle Ow.
where Y is the distance from the solid of a point As distance from the wetting line increases, the
on the meniscus and Yw the value of 0 at Yat flow comes to resemble that for no slip at the
which the solution is truncated, 0 the angle solid. To lowest order, the only information
through the liquid that the tangent to the meniscus communicated outward from this region is the
makes with the solid, and Ow the value of 0 at submicroscopic contact angle.
Y= Yw ' The function X(O) is defined by At distances from the wetting line of order
Lexp( -1/Ca), hydrodynamic bending of the
18 meniscus takes place. This intermediate region
X(O) = 2J[O/sin(O) - cos(O)]dO (3.22) connects the inner and outer regions and in effect
o
converts the submicroscopic contact angle into
For 0 < 3n/4, the integrand may be approximated the dynamic contact angle. In this region, the
by 20 2 /3, and equation (3.21) becomes no-slip condition is appropriate at the solid. To
lowest order in capillary number, the apparent
03 - O~ = 9Caln(Y/Yw ), 0 < 3n/4 (3.23)
contact angle is given by
Voinov recognized that the angle near the wall,
x(Oo) - X(Ow) = Ca In(L/Ls) (3.24)
Ow, may depend upon speed.
To estimate 00 , suitable values for Y and Yw where, for the case in which the displaced fluid
are required. With Yw = 10- 7 cm, the order of is inviscid, X(O) is given by equation (3.22);
molecular size, and Y = 10- 3 cm, the approximate otherwise, the complete expression given by Cox
distance from the wetting line at which a contact must be used. For liquid displacing air, equation
angle can be measured, In(Y/Yw ) is estimated to (3.22) applies except as 00 nears 180·, in which
be on the order of 10. Later it will be seen that case the viscosity of the air becomes important.
this rough estimate is in agreement with values For small and moderate values of 00 , equation
obtained by fitting experimental data. (3.23) applies with In(Y/Yw ) replaced by In(L/Ls).
More often in hydrodynamic models, the force This linear variation of O~ with Ca was also
singularity at the wetting line is eliminated by arrived at by Tanner (1979) and de Gennes
permitting local slip. Dussan V. (1976), Huh and (1985). Kalliadasis and Chang (1994) considered
Mason (1977), Hocking and Rivers (1982) and the case of perfect wetting, where a precursor
others have developed this method. Cox (1986) film advances faster than the meniscus. For a
treated both fluids as viscous and appears to thick precursor film, they determined that
have obtained the most general results. For the 0= 7.48 Ca 1/3 • They also obtained a small
purpose of eliminating the force singularity, slip correction, arising from disjoining pressure, for
is permitted over a distance L s from the wetting thin precursor films and very small capillary
line that is small compared with the macroscopic numbers.
scale of the flow L. The capillary number is Interpreted literally, the term In(L/Ls) implies
presumed small, so that away from the wetting that the dynamic contact angle depends upon
line the meniscus has a shape given by the the outer (macroscopic) geometry. The term is
Young- Laplace equation. The angle determined basically an adjustable parameter because a
by extrapolating this static meniscus to the solid precise value cannot be assigned. There is no
is the dynamic contact angle. proven model for the slip region, but the basic
Cox used the method of matched asymptotic result, equation (3.24), depends neither on the
expansions to take advantage of the two small particular form of the slip boundary condition
parameters, the capillary number and the ratio (Dussan V. 1976) nor indeed on the model for
Ls/L. At distances from the wetting line of order the inner region, and so is likely to be true
regardless of the physics on the submicroscopic 3.3.4 EXPERIMENTAL DATA FOR DYNAMIC
scale. This insensitivity to the inner-region model CONTACT ANGLES
arises because only the submicroscopic contact
angle is communicated outward to the intermedi- Plentiful experimental data are available for the
ate region. So, although the inner region is dynamic angle at small capillary numbers for the
explicitly taken into account in slip models, what case where the displaced fluid is air. At extremely
is gained by continuing the solution to the low speeds, the dynamic contact angle may
wetting line over Voinov's truncated solution remain equal to the advancing static angle
seems minimal. (Schwartz and Tejada 1972; Johnson, Dettre and
Indeed, Dussan V., Rame and Garoff (1991) Brandreth 1977), but in many systems the dynamic
proposed a form of the hydrodynamic theory for contact angle begins to increase from its static
the intermediate region, common to both the value by speeds as low as 10 ~m/s. After this
inner and outer regions, that is not expressed in initial rise, the dynamic angle may become
terms of a slip length and a macroscopic length sensibly constant over some higher range of
scale. Instead, they introduced the slope of the speeds, as found by Ablett (1923), Elliott and
interface at some small distance from the wetting Riddiford (1967), and Cain et al. (1983). In all
line within the intermediate region, OJ. This angle cases, however, further increases in speed event-
might depend upon speed, but otherwise would ually result in the dynamic contact angle ap-
be a material property of the system. A universal proaching 180°.
form for the variation of interface slope in the Inverarity (1969a,b) measured dynamic contact
intermediate region supplies an asymptotic angles for glycerol/water solutions on E-glass
boundary condition for the interface slope 0, for filaments. Data for 95% glycerol are shown in
use in solving the flow in the outer region: Fig. 3.13. The curve for the molecular-kinetic
theory represents the data well. The curve for
x(O) - X(O,) ~ caln(:) (3.25)
the hydrodynamic theory does not represent the
data as well, and the combined theory does not
improve the fit over molecular-kinetic theory
Here, R is the distance of a point on the interface alone. In this case, molecular-kinetic theory can
from the wetting line, and R , is a value of R in adequately represent the data.
the intermediate region where 0 = OJ. The Hoffman (1975) measured apparent contact
intermediate and outer solutions can be combined angles in glass capillary tubes for liquids ranging
to form a composite solution, and the value of in viscosity from 0.96 to 2430 Pa s. He found that
R j is chosen such that the outer solution has no his data for liquids with zero contact angles,
significant effect on the composite solution at plotted as angle versus capillary number, fell on
that location. It is an advantage that OJ depends a single curve Ca = F(OD). His data for liquids
only upon material properties and perhaps speed, with nonzero contact angles also fell on this curve
but the angle is submicroscopic and must be if a shift factor was used: Ca = F(OD) - F(O A).
inferred. A comparison of equations (3.25) and Hoffman's data are shown in Fig. 3.14.
(3.21) shows that the end result is substantially Cox (1986) applied his theory to Hoffman's
the same as Voinov's. experimental results and concluded that if Ow = 0A'
Petrov and Petrov (1992) allowed the submic- then F(O) = x(O)jln(L/Ls). Hoffman's shift factor
roscopic contact angle in Cox's hydrodynamic for producing a master curve appears naturally
theory to vary with speed according to the in Cox's theory. By fitting Hoffman's data, Cox
molecular-kinetic theory. This gave them three determined that In(L/Ls ) = 9.21. The resulting
adjustable parameters. In this manner they were curve represents the data well and is plotted in
able to improve the fit between theory and Fig. 3.14. On the other hand, Jiang, Oh and
certain experimental data. Slattery (1979) determined an empirical equation
180
deg
• /0
//
150
rtl
120
eo 90
V
60
~~ ,
/!
30
y ...
.A
~
a
0.001 0.01 0.1 10 100
J1U mN/m
Figure 3.13 Data of Inverarity plotted as (JD versus fJ.U for E-glass filaments entering a pool of aqueous
glycerol. Viscosity, 410 mPa s; surface tension, 65 mN/m. - - - - molecular-kinetic theory with
N = 2.5 x 10 14 cm- Z and K s = 8.8 x 109 s-I;-----hydrodynamictheory with In (L/L s) = 2.8;--combined
theory.
180
deg
150
120
80 90
60
30
a
o. 00001 0.001 0.1 10
shifted Ca
shifted Ca
Figure 3.14 Data of Hoffman plotted as (JD versus shifted Ca for liquids advancing in a capillary. ------
hydrodynamic theory with In(L/Ls) = 9.21; - - empirical equation (3.26).
which fits the data equally well: and Joos (1989), using their data for solid strips
dipped into a pool at capillary numbers greater
than about 0.002, determined
cos(O A) - cos(Oo) = 2 Ca O. 5 (3.27)

This curve is also shown in Fig. 3.14. Note that cos(OA) + 1
a contact angle of 180· is reached only asym-
ptotically as capillary number is increased. Their liquids were aqueous glycerol, aqueous
Molecular-kinetic theory can fit Hoffman's ethylene-glycol, and corn oil; their solids were
data if the liquids are considered singly. This is poly(ethylene) and poly(ethyene terephthalate).
necessary because the molecular flow volumes Seebergh and Berg (1992) found nearly the same
are not precisely known. Fig. 3.15 shows Hoffman's result for strips of low-density and high-density
data for Brookfield standard viscosity oil. The poly(ethylene) dipped into silicone oil:
curve for molecular-kinetic theory does not
cos(O A) - cos(Oo) = 2.24 Ca O. 54 (3.28)
represent the data as effectively as hydrodynamic
cos(O A) + 1
theory, but the curve for the combined theory
represents the data as well as hydrodynamic Capillary numbers ranged between about 0.001
theory alone, with the change in Ow attributed and 0.02.
to molecular-kinetic theory accounting for about There are also data for the dynamic contact
80% of the change in 00 , angle suggesting that more than one mechanism
There are other empirical correlations for the is at work and that Hoffman's correlation is not
dynamic contact angle besides those of Hoffman, universally applicable. Blake (1993) reported an
and Jiang, Oh and Slattery. Bracke, De Voeght extensive set of data (further augmented here)
180
deg a ~
150
~~
120
~/
eD 90
V
60 ~
30 rid
V
0
0.Q1 0.1 1 10 100
pU mN/m
Figure 3.15 Data of Hoffman plotted as IJ D versus J1.U for Brookfield Standard Viscosity Fluid advancing in
a capillary. Viscosity, 98.8 Pa s; surface tension, 21.7 mN/m; static contact angle, 0·. - - - hydrodynamic
theory with In (L/L s) = 8.68; ----- combined theory with N=1.2x10 14 cm- 2 , K s = 1.6 x lO 11 s- l , and
In(L/Ls) = 4.3; - - molecular-kinetic theory with N = 1.3 x 10 14 cm- 2 and K s = 9.2 X 1O IO s- l .
for smooth poly(ethylene terephthalate) tape at dynamic contact angles in the vicinity of 130·.
drawn vertically into a pool of aqueous glycerol. Nevertheless, Hoffman's data tail off to higher
Viscosities ranged from 1 to 670 mPa s; the range capillary numbers as the contact angle approaches
of speeds was correspondingly broad. Data for 180·, and Blake's data do not correlate with
five viscosities are shown in Fig. 3.16. Plotted in capillary number as the dynamic angle approaches
this way, the data make plain the major influence the static contact angle but lie systematically
of viscosity on the dynamic contact angle. The above Hoffman's results.
complete set ofdata is plotted as dynamic contact The fact that Hoffman's correlation does not
angle versus capillary number in Fig. 3.17. The fit Blake's data at low capillary numbers is shown
data collapse to a single curve at capillary more clearly in Fig. 3.19 where Blake's results
numbers above about 0.005, except at contact for 0.67 and 0.10 Pa s aqueous glycerol are plotted
angles near 180·. At lower capillary numbers, the with those for two of Hoffman's liquids, 11.2 Pa s
data diverge systematically, even though the Santicizer 405, and 109.3 Pa s Admex 760. Al-
advancing contact angles differ by only a few though the advancing static contact angles of the
degrees. This divergence is evidence that Hoffman's liquids are nearly the same (Hoffman's static
correlation is not universal. In Fig. 3.18, some contact angles are presumed to be advancing)
ofBlake's and Hoffman's data are plotted together. and the data are in agreement at capillary
The data have been shifted in capillary number numbers above about 0.01, the data diverge
by Hoffman's method. Although there are rapidly at smaller capillary numbers. Specifically,
substantial differences between the experiments the data for aqueous glycerol approach the static
in material properties and geometry, the data contact angle at a capillary number about two
correspond to some significant extent, particularly orders ofmagnitude smaller than for Santicizer 405.
180 , . . - - - . , . - - - . , . - - - . , . - -...... , - - - . , . . - - . ,
deg
150 -I----I----I---..,.....~.,.-....t-T.....
80 120 -I----I----I--..,.+-.......-~:?_-F-~
60
.-A.
-1----1----1----1----1---+----1
0.00001 0.001 0.1 10
u m/s
- 670 • 100 • 19 A. 4.2 - 1
Figure 3.16 Data of Blake plotted as 8D versus U for poly(ethyene terephthalate) tape vertically entering a
pool of aqueous glycerol. The numbers in the legend indicate viscosities in mPa s.
180 """--'T""--"""--"""--"""--"""--T'""'-""""
deg
150 +---+--+---+---+---+---~
eo 120 +---+--+---+---+---+-.....,
60 +---+--+--+---+---+---+--~
0.000001 0.0001 0.01
Ca
670 • 100 A 58 • 19 - lOt:.. 4.2 [J 1.5 - 1
Figure 3.17 Data of Blake plotted as eo versus Ca for poly(ethyene terephthalate) tape entering a pool of
aqueous glycerol. The numbers in the legend indicate viscosities in mPas.
180 .,....-----.....,~---~-..,. . .-"""":::__-.....,

••••
deg
I.
••
150 +------.....jH
eo 120 +-----....,
90+-~~
60 +-..----.....jf-------+--------t
0.01 0.1 10
shifted Ca
Figure 3.18 Data of Blake plotted as eo versus Ca for poly(ethyene terephthalate) tape entering a pool of
aqueous glycerol (light symbols); the data of Hoffman for a meniscus advancing in a capillary tube (dark symbols).
180 r------.,..------.....,----eJ----.
deg
8 0
150 +--------+- _+...L - - - - - ~
eo 120 +--------+-----......~------~
....ID--_+------~
.- ..
90 +--------+---:~
·0
o
60
• o
+--------+-------+--------f
0.00001 0.001 0.1 10
shifted Ca
• gly, 0.67 • gly, 0.10 0 San, 11.2 0 Adm,109
Figure 3.19 Data of Blake plotted as (JD versus Ca for poly(ethyene terephthalate) tape entering a pool of
aqueous glycerol (dark symbols), and data of Hoffman for liquid advancing in a capillary tube (light symbols).
All liquids have about the same static contact angle. The numbers in the legend give the viscosities in Pas.
The data for the lowest viscosities in Fig. 3.17 the more irregular behavior of the data for water
show rapid changes in slope that could imply a shown in Figs 3.17 and 3.20. Blake (1993)
change in mechanism. In particular, the data for postulated a change in wetting kinetics to account
water (viscosity 1mPa s) show a region where for the apparent break in the slope of the data
the contact angle falls as speed increases. The and suggested that molecular-kinetic theory could
effect is even more evident for water on gelatin- therefore be applied separately to the data above
subbed (having a thin subbing layer of dried and below the break. Both high-speed and low-
gelatin) poly(ethylene terephthalate) tape, as shown speed fits of molecular-kinetic theory are shown
in Fig. 3.20. Observational evidence supporting in Fig. 3.21. The values of the parameters for
a change in mechanism is that the wetting line each curve indicate that the surface of the solid
exhibits 'stick-slip' behavior in the region where contains a low density of adsorption sites that
the contact angle falls: the wetting line appears interact strongly with the liquid, together with
to advance irregularly and to catch and release a higher density of more weakly interacting sites.
in a way that is suggestive offluctuations between At low wetting speeds, the strong interactions
two states. dominate the wetting process, but despite long
The data for 10mPas aqueous glycerol on relaxation times for each molecular displacement,
poly(ethylene terephthalate) tape are plotted a more or less equilibrium solid-liquid interface
again in Fig. 3.21. Neither hydrodynamic theory is expected to remain after the passage of the
nor molecular-kinetic theory can describe the wetting line. On the other hand, at high wetting
entire range of the data. Also shown is a curve speeds, the strongly interacting molecules are
for Petrov's combined theory. The combined unable to relax within the available time interval,
theory gives a smooth curve that can reasonably and a nonequilibrium absorbed layer is entrained.
fit the data in Fig. 3.21, but cannot cope with The wetting process is then controlled by the
180
deg
•
•
160
•
•
140
eo ••
•
•
120
III
•
•• ~
•
100
• • •• jill •• ,
80 I
0.00001 0.001 0.1 10
u m/ s
Figure 3.20 Data of Blake plotted as eD versus U for gelatin-subbed poly(ethylene terephthalate) tape vertically
entering a pool of water.
180,....------"'T------....,------.....,
deg
150 +-------+-------+----1\:01--__1
e{) 120 +--------+-------i~-7I_---__1
90 +-------I----"""""=::r'I:J"'!J;il!f;.,.--------i
60 +------~I-------+-----~
0.0001 0.01 100
mN/m
e
Figure 3.21 Data of Blake plotted as D versus J.l.U for poly(ethyene terephthalate) tape entering a pool of
aqueous glycerol. Viscosity, 10 mPa s; surface tension, 65 mNjm; static contact angle, 64S. - - - - molecular-kinetic
theory for low speeds with N = 4.7 X 10 13 cm -2 and K s = 9.0 X 106 S-I; - - - molecular-kinetic theory for
highspeedswithN = 2.9 x 10 14 cm- 2 andK s = 3.4 x 10 1 0s- 1 ,-----hydrodynamictheorywithln(LjLs) = 6.46;
- - combined theory with In(LjLs ) = 7.5, N = 2.0 x 1013 cm -2 and K s = 1.3 X 10 5 S-I.
weaker interactions, and the locus of molecular (1979) compared Hoffman's data for a capillary
displacement effectively shifts from the surface tube with some data of Schwartz and Tejada for
of the solid to the plane immediately above the a wire entering a pool and found reasonable
absorbed layer, where the influence of the solid agreement for cases where inertial effects were
is reduced. At intermediate speeds, the unsteady negligible and gravitational effects modest. They
transition between these two planes gives rise to suggested that the contact angle may be inde-
the observed stick-slip behavior. pendent ofmacroscopic geometry under restricted
The notion that there is more than one circumstances. There are other observations and
mechanism for the effect of wetting speed on the results that support this hypothesis. The above
dynamic contact angle is well rooted in the comparison between Hoffman's data and Blake's
literature. Schwartz and Tejada (1972) found an suggests some degree of invariance to flow
apparently discontinuous change in slope with geometry. Burley and Jolly (1984) found in
speed in their data and concluded that there were plunging-tape experiments that the dynamic
at least two mechanisms in operation. The lower contact angle was not affected by the angle with
speed data could be represented by molecular- respect to the vertical that the tape entered the
kinetic theory, whereas the higher speed data pool. Bracke, De Voeght and Joos (1989) applied
could be fitted to an expression of the form their contact-angle data for plunging strips to
tan(B o) proportional to Ca raised to a power. predict successfully the rate of spreading of drops
Similarly, Seebergh and Berg (1992) offered and the unsteady advance of the wetting line
different correlations for data at capillary numbers following the initial contact of a strip with the
below about 10- 5 and above 10- 3 . Hoffman surface of a pool. On the other hand, Ngan and
(1975) suggested that changes in interfacial forces Dussan V. (1989) measured contact angles for
accounted for any low-speed variations in the low-capillary-number displacement between par-
contact angle that lay outside his correlation. allel plates spaced 0.1 to 1.2 mm apart and found
A small degree of surface roughness, up to at that the apparent contact angle depended on the
least 1 ~m, appears to have little effect on the gap between the plates.
dynamic angle (Schwartz and Tejada 1972). Virtually no dynamic contact angle data are
However, Seebergh and Berg (1992) associated available for non-Newtonian liquids, despite
degrees of surface roughness so light as to be their prevalence in industrial coating processes.
unmeasurable by mechanical profilometry with The effect of viscosity for Newtonian liquids and
unsteady, stick-slip motion of the wetting line. the expected high shear rates near the wetting
Cain et al. (1983) also reported that a surface line suggest the importance of shear-thinning
roughed to a fraction of a micron caused the behavior. Elastic properties need to be considered
wetting line to be serpentine at rates of advance as well.
of about 10 ~m/s. Jansons (1985, 1986) modeled Although the origins of the dynamic contact
the effect of random roughness on the moving angle are not clear, this review of theory and
wetting line at small capillary number by experimental results allows a few tentative
considering successive pinning and release events conclusions to be drawn. At low capillary and
at individual asperities. According to the theory, Weber numbers, the dynamic contact angle has
the motion of the wetting line can appear uniform some degree of invariance to macroscopic
at a distance from the wetting line which is large geometry. The mechanism or mechanisms that
compared with the scale of the roughness, and cause the static and dynamic contact angles to
the dynamic contact angle is predicted to be differ operate locally at submicroscopic distances
larger than the microscopic contact angle. from the wetting line. Some portion of the
The degree ofsensitivity ofthe dynamic contact difference between the dynamic and static contact
angle to flow geometry is fundamentally and angles is likely to be due to hydrodynamic
practically important. Jiang, Oh and Slattery bending of the interface on the submicroscopic
scale, particularly at capillary numbers above 3.4.1 EXPERIMENTAL OBSERVATIONS
about 0.001. Hydrodynamic theory is well
developed, and certain experimental data appear The development of dynamic wetting failure for
to correlate with the capillary number. Although the well studied case of a tape entering a pool is
there may be systems for which hydrodynamic typical. The wetting line becomes unsteady and
bending is the only mechanism, there clearly are breaks up into straight-line segments, and an air
other systems whose behavior at low capillary film forms. Each segment of the wetting line is
numbers cannot be explained by hydrodynamic inclined from the horizontal at an angle that
bending alone. For these systems, the submicro- increases steadily with speed. At any instant, the
scopic contact angle appears to depend upon wetting line has the appearance of sawteeth
speed, and the variation is not always smooth. (Perry 1967; Burley and Kennedy 1976; Blake
Thermodynamic equilibrium does not prevail at and Ruschak 1979; Burley 1992), and air bubbles
the wetting line, and molecular-kinetic theory may be formed at the downstream vertices of the
provides a nonhydrodynamic mechanism for the sawteeth and carried into the liquid. The scale
difference between the dynamic and static contact of the sawteeth depends upon fluid properties
angles. In general both mechanisms may operate and flow geometry and can be large enough to
together, but so far there is no definitive way to be readily visible or small enough to require
sort out the relative contributions. magnified viewing.
All considered, it is still necessary to conduct The case where only one sawtooth, having one
experiments to obtain the dynamic contact angle vertex, forms is special but particularly instructive.
for particular material systems and geometries. The photograph reproduced in Fig. 3.22 shows
Some invariance to geometry is likely at small such a case for a tape plunging into aqueous
capillary and Weber numbers and moderate glycerol. Fig. 3.23 is a schematic drawing of the
Bond numbers, but the degree of invariance and air film. The wetting line consists of two slanted
the precise parameter ranges are not established. but steady straight-line segments in the shape of
In practical coating flows, the capillary and a ·V.' The vertex may be several centimeters
Weber numbers are commonly not small, in downstream of the pool surface, and over this
wl,ich case the macroscopic hydrodynamics will distance the liquid is separated from the substrate
affect the dynamic contact angle. Under these by a thin, triangular-shaped layer ofair contained
conditions, the dynamic contact angle cannot be within the 'V.' As speed increases, the angle of
expected to be invariant to flow geometry. inclination of the segments with respect to the
horizontal increases. For polyester tape plunging
into a pool of l00mPas aqueous glycerol, Blake
and Ruschak (1979) showed that the angle of
3.4 DYNAMIC WETTING FAILURE
inclination increases such that the component of
In the context of coating, the most important substrate speed normal to each segment remains
phenomenon associated with the dynamic contact constant. They termed this speed the maximum
angle is dynamic wetting failure. At sufficiently speed of wetting. Initially, no visible air bubbles
high speeds, the wetting line segments, and a thin are produced, but as the speed of the tape and
film of air intrudes between the liquid and the the inclination of the wetting line segments
solid. The flow becomes unsteady and three- increase, air bubbles begin to be entrained from
dimensional, and air bubbles may be entrained the vertex. At still higher speeds, an air tube may
into the liquid. Dynamic wetting failure is observed form and persist for a significant distance
to occur when the dynamic contact angle downstream before breaking up.
approaches 180°. The term 'air entrainment' is Perry (1967) was the first to examine air
most appropriate when visible air bubbles are entrainment in liquid/solid systems in detail. He
produced. found that he could influence the air film by
Dynamic wetting failure 87
air
film -+
wening
line -+
possible
air bubbles 0
I '\.0 A
o
Figure 3.23 Schematic of a sawtooth wetting line and

its associated air film for a tape vertically entering a pool.
cellulose tape and inspected the dried coatings

by microscopy. For conditions where air bubbles
appeared in the dried coating, Miyamoto estimated
Figure 3.22 Photograph of gelatin-subbed, 35 mm from the number and size of the bubbles an
wide poly(ethylene therephthalate) tape entering a average air-film thickness in the range of 0.05 to
pool of aqueous glycerol (nominal viscosity, 50 mPa s; 0.8 J.lm, with thickness increasing with speed.
surface tension, 62 mN/m). An air film, triangular in Below a certain speed no air bubbles were found,
shape, separates the liquid and the tape. The segments
of the wetting line are inclined at about 51 from
0 and the estimate is zero. However, Miyamoto
horizontal. The tape is transparent, so wetting lines did find craters in the dried coating at the lower
on both sides are visible. The tape speed is 0.585 mis, speeds, which he postulated were created by air
and the dynamic contact angle reaches 180· at 0.39 m/s. bubbles that had dissolved. He established
experimentally that micron-sized bubbles dissolve
making changes in the macroscopic geometry or in seconds and concluded that some air is
flow field. For example, he could change or entrained below the speed at which air bubbles
eliminate the air film by placing a solid object appear in the dried coating. However, it is not
in proximity or by jetting the liquid toward the clear how this interpretation can be reconciled
air film. The application of a slight suction to with observations that the onset of both a
the three-phase zone in bead coating increases sawtooth wetting line and the entrainment of air
the air entrainment speed (Beguin 1954; Gutoff bubbles appear to be sudden.
and Kendrick 1987). Similarly, application of an Mues, Hens and Boiy (1989) applied laser-
electric field attracts the liquid to the solid and Doppler velocimetry to bead coating at a speed
increases the speed at which dynamic wetting about two-thirds of that for the obvious onset
failure occurs (Miller and Wheeler 1962). of air entrainment. They inferred that a thin air
Using an optical method, Perry estimated the film extends a distance on the order of tOO J.lm
thickness of the visible air film as about 0.9 J.lm. downstream of the apparent wetting line. They
Miyamoto (1991) used curtain coating to coat did not determine whether bubbles were created
122 mPa s gelatin onto gelatin-subbed triacetyl- in the coating.
Some correlations for the onset of dynamic and polypropylene:
wetting failure in experiments are available from
a)O.77 a
the literature. The speed at which the dynamic U = 0.395 ( ~ 0.2 < - < 3.2m/s (3.31)
contact angle reaches 180·, that for the onset of fJ.
the sawtooth wetting line, or that for visible air For a 180· dynamic contact angle, Gutoff and
entrainment may be reported. In the following Kendrick (1987) found that the expression
correlations with dimensional parameters, speed
5.1
is given in mis, viscosity in mPa s, and surface U =- - 1< fJ. < 1000, 22 < a < 72
fJ. 0.67
tension in mN/m. Perry (1967) found that the
air-entrainment speed for magnetic recording (3.32)
tape plunging into a pool of aqueous glycerol is
inversely proportional to viscosity raised to a holds for various liquids on gelatin-subbed
power: polyester tape. Similarly, the data of Blake (1993)
for poly(ethylene terephthalate) tape entering a
3.5 pool of aqueous glycerol can be represented as
U = -0 20 < fJ. < 300, fJ A = 70· (3.29)
fJ. .8
10.7
U = li'827 1 < fJ. < 670, a ~ 65 (3.33)
fJ.'
u=~
fJ. 0 . 73
20 < fJ. < 300, fJ A = 20· (3.30)
The data and the fitted correlation are shown in
The tape with a 20· advancing contact angle Fig. 3.24. Virtually identical results, not shown,
entrained air at a higher speed than that with a were obtained on gelatin-subbed poly(ethylene
70· contact angle. Increased surface tension also terephthalate).
raised the air-entrainment speed. Burley and Bracke, De Voeght and Joos (1989) obtained
Jolly (1984) determined the following relationship from the empirical equation (3.27)
for plunging tapes ofcellophane, Melinex polyester, Ca = 0.25 (3.34)
100...- -----.....,,....-----.,
m/s
10 . .~----_+---------it-------I
0.1 +------_+---------it----.;:::IIioolll::"""---I
0.01 +------_+---------it-------f
10 100 1000
J1 mPa·s
Figure 3.24 Data of Blake plotted as U versus J.l for the onset of air entrainment for poly(ethyene terephthalate)
tape entering a pool of aqueous glycerol. The solid line is equation (3.33).
However, constant capillary number is clearly fibrous materials could lead to the formation of
not a precise criterion. a composite liquid/solid-liquid/air interface. Small
Finally, Wilkinson (1975) obtained data for pockets ofair might not be detected. Furthermore,
the onset ofair entrainment for a scraped cylinder they might reduce the net drag ofthe solid surface
rotating into a pool; presumably the cylinder and so could increase the speed for the onset of
surface was damp as it entered the pool. The visible air-entrainment. A related effect has been
liquids were aqueous glycerol and mixtures of observed experimentally by Menchaca-Rocha
hydrocarbon oils. The data can be represented by (1992), who found that the mobility of mercury
drops sliding across roughened glass plates
(J
Ca = 0.64J,l°·13 or U = 0.64---0-87 increased with roughness due to the formation of
J,l' a composite mercury/glass-mercurylair interface
29 < J,l < 288, 30 < (J < 74 (3.35) that reduced solid/liquid contact.
Overall, it does not seem possible to make
So, different experimenters using different generalized predictions on the effects of roughness.
materials arrived at similar results. The inverse Once again, experimentation is necessary.
dependence ofspeed on viscosity is well established; Buonopane, Gutoffand Rimore (1986) measured
a direct dependence on surface tension is probable the effects ofsurface properties on air-entrainment
but less firm. If the above expressions are evaluated speed in a plunging-tape experiment. The prop-
for a viscosity of 10 mPa s and a surface tension erties of the solid changed the air-entrainment
of 65 mN/m, the results are 0.55 m/s from equation speed by about a factor of 10. They concluded
(3.29); 1.1, (3.30); 1.7, (3.31); 1.1, (3.32); 1.6, (3.33); that speed increased with root-mean-square
1.6, (3.34) and 5.6, (3.35). The speed from the roughness, but that surface wettability had
scraped cylinder data stands out as high; otherwise, relatively little effect. These conclusions have yet
there is a variation of about a factor of three to be confirmed, and our own experience leads
that might be attributable to the range of materials us to doubt their generality.
used. In practice, liquids are frequently non-New-
At first glance, it may appear that surface tonian. Shear thinning is likely to result in
roughness should always decrease the air en- reduced viscosity near the wetting line and
trainment speed, as the wetting line must move increased speed. We have found, for example,
a greater distance across a rough surface than that the onset of air entrainment for poly(ethylene
across a flat one of equivalent macroscopic terephthalate) tape plunging into 21 mPa s aqueous
dimensions. With a rough surface, the local speed gelatin, a modestly shear-thinning solution with
of the wetting line must exceed the overall a surface tension of 48 mN/m, occurs at 2.8 m/s.
wetting speed for at least part of the time, hence For a Newtonian liquid, equation (3.31) predicts
air entrainment may be prematurely triggered. 0.75 mis, (3.32) 0.66, and (3.33) 0.86; all of these
Theoretical studies of wetting such as those of are well below the actual value.
Jansons (1985, 1986) and Joanny and Robbins A basic limitation of plunging tape and fiber
(1990) indicate that the unsteady pinning and experiments is that hydrodynamic forces are
release of the wetting line as it moves across a never more than comparable with capillary effects.
rough or otherwise heterogeneous surface leads Kistler (1983) and Blake, Clarke and Ruschak
to energy dissipation in the vicinity ofthe wetting (1994) investigated stronger hydrodynamic effects
line above that expected on a smooth surface. on dynamic wetting failure using curtain coating
As a result, the dynamic contact angle is expected (Chapter llc). Kistler reported experiments for
to increase and the speed of dynamic wetting aqueous glycerol on a scraped cylinder. He found
failure decrease. that the air entrainment speed at first increases
On the other hand, trapping of air within with flow rate, reaches a maximum, and then
surface cavities and hollows of very rough or decreases. That is, there is an optimum flow rate
that maximizes speed. Thus, air entrainment can as speed is increased can be substantially higher
depend significantly upon a flow parameter. than the speed at which it ceases as speed is
Blake, Clarke and Ruschak coated aqueous decreased. The effect can be as large as a few
solutions of gelatin and other polymers onto dry meters per second. Flow visualization shows that
base and found the same qualitative trend but air-film formation in this situation is accompanied
much higher speeds for a given low-shear viscosity. by major changes in the macroscopic flow field.
They achieved conditions where the Weber To illustrate the effects of macroscopic hy-
number based on final film thickness was of order drodynamics on air entrainment, data from a
unity or larger and viscous and inertial effects coating experiment on a scraped cylinder are
were of comparable importance. They found that shown in Fig. 3.25. Aqueous glycerol solutions
coating speed at the onset of air entrainment with viscosities 23 and 60 mPa s were delivered
varied by over two orders of magnitude as flow through a vertical pipette spaced 0.064 cm above
rate, impingement angle, and impingement speed the surface of the cylinder. The outside diameter
were changed. Flow visualization established of the nozzle was 0.29 cm and the inside diameter
that the speed maximized over flow rate corre- 0.13 cm. The experiment was performed by fixing
sponds to a wetting line position directly beneath the flow rate and increasing speed until an air
the curtain. For aqueous gelatin the maximum film formed. Speed was then decreased until the
speed can be represented as air film disappeared. A hysteresis loop was found.
Within the loop, an air film might or might not
812Vo. 81
U =. be observed, depending upon how the point was
n (3.36)
0 19
J1. . approached. The measured speeds were also
inconstant, and the data points plotted are
in which J1. is the low-shear viscosity and v" the averages of three or four observations. When an
impingement speed normal to the solid in m/s. air film formed, the liquid formed drops around
The value ofthe exponent of viscosity in equation the tip of the pipette. The speed for air-film
(3.36) is much less than the 0.7-0.8 found in formation was markedly affected by the flow rate,
plunging tape experiments, but the hydrodynamic with speed increasing with flow rate. Speed also
regime differs. Also, the shear-thinning nature of increased as viscosity decreased. Wilkinson's
aqueous gelatin probably accounts for at least correlation, equation (3.35), predicts air entrain-
part of the difference. Because of high shear rates ment at 2.7 mls for 23 mPa s viscosity and 1.2 mls
near the wetting line, the apparent viscosity is for 60 mPa s viscosity. These values are roughly
significantly reduced below its low-shear value. the results obtained at the highest flow rates.
We have found that the air-entrainment speed
for 63 mPa s aqueous gelatin in a plunging tape
3.4.2 THEORIES FOR DYNAMIC WETTING
experiment is about 1.8 m/s. For an impingement
FAILURE
speed normal to the same tape of 2 mls in curtain
coating, the maximum air entrainment speed is Dynamic wetting failure can be considered from
about 6.5 m/s. In this latter case, the Weber two perspectives: how an air film forms between
number is close to 10, and the Bond number the liquid and the solid as speed increases, and,
based on final film thickness is much less than once formed, how the film vanishes as speed
unity. These conditions lie well beyond those decreases. If there is hysteresis, then the speed
achievable in the traditional flows by which for film formation is higher than the speed for
dynamic wetting has been studied. film disappearance.
Blake, Clarke and Ruschak (1994) also found Perry (1967, p. 103) applied the film-formation
that for flow rates exceeding the optimum, air theory of Landau and Levich, and Deryagin (see
entrainment may be subject to hysteresis. In such Ruschak 1985 for a review) to an air film in the
cases, the speed at which air entrainment begins plunging-tape geometry. This gives, to lowest
..... u
•
o. 0
•
0.75
~-
-
p • 0
•
0.5
- J •
0
• c
• •
0.25
~: c
0
•
ooeg . 0 •
I
d}jef1 ~
o
o 0.5 1.5 2 2.5 3 3.5
U m/s
• 60, In 0 60, out • 23, in 0 23, out
Figure 3.25 Data of Ruschak plotted as volumetric flow rate versus U for aqueous glycerol coated onto a
scraped cylinder from a pipette. Dark symbols represent the speed at which air entrainment began as speed
increased, and light symbols the speed at which air entrainment ceased as speed decreased. The numbers in
the legend indicate viscosities in mPa s.
order, is a liquid displacing air from a narrow slit. Away

from the walls, the meniscus is a cylindrical cap.
He: = l.34R (Ji:
U)2/3 (3.37) They examined what happens as capillary number
based on air viscosity decreases. At high capillary
number, the film is so thick that disjoining
where He: is the thickness of the air film, JiA the pressure is negligible. On the other hand, for a
viscosity of air, and R = LaJ.J2 the radius of sufficiently thin air film, disjoining pressure is
curvature at the solid surface ofthe static meniscus expected to be negative and increasing toward
that approximately describes the surface of the zero with increasing film thickness. Hence,
pool for a dynamic contact angle of 180° (see disjoining pressure promotes thinning of the air
equations (3.6) and (3.18)). An evident shortcoming film, and there is no residual film below a critical
of equation (3.37) is that it predicts an air-film capillary number. The critical capillary number
thickness regardless ofhow low the speed. Another based on the viscosity of air for a slit of order
weakness is that the prediction does not depend 1 mm in width is about 10- 7 , which implies a
upon the viscosity of the liquid. Perry compared minimum speed oforder 0.001 mls for the existence
his experimental estimate for the thickness of the of an air film. When the film is thick, the circular
air film, 0.9 Jim, with the molecular mean free meniscus appears tangent to the walls, but for
path for air, 0.09 Jim, and concluded that the air small film thicknesses the meniscus appears to
film is in slip flow. However, he did not take intersect the walls; that is, there is a dynamic
explicit account of this in his theory. contact angle of less than 180°. Bending of the
Teletzke, Davis and Scriven (1988) included meniscus by disjoining pressure occurs at sub-
the disjoining pressure of the air film and the microscopic distances from the walls.
viscosity of the liquid in their model. The flow To account for the possible existence of an air
film at speeds far below that for the onset of be used to predict dynamic wetting failure. Cox
visible air films in experiments, Teletzke, Davis (1986) applied his hydrodynamic theory to predict
and Scriven suggested that air is first entrained air-film formation. There are no solutions to his
as a film too thin to be seen by visible light. hydrodynamic problem beyond the speed at
Disjoining pressure causes such a thin film to be which a dynamic angle of 180° is reached, and
unstable; relatively thin regions tend to grow he postulated air-film formation at this point. A
thinner, and the film breaks up into microbubbles 180° contact angle corresponds to the following
that quickly dissolve in the liquid. The experi- condition:
mental results of Miyamoto (1991) and Mues,
Hens and Boiy (1989) qualitatively support the ~ln(_4Ji._) - X(Ow)
theory. Nevertheless, it is not clear that the 6 3nJi.A
Ca = ------'-:-...,...,,--:-:--,------ (3.38)
theory is in accord with experimental results. The In (LIL s )
thinnest, visible air film is on the order of 1 Ji.m
thick; for a slit about 1mm wide, the corresponding Accordingly, the critical capillary number increases
capillary number based on the viscosity of air is slowly with the viscosity of the liquid, although
on the order of 10- 5_10- 4 , which implies speeds dimensional speed falls. If all else remains constant,
on the order of 0.1 m/s. This estimated speed the critical capillary number also increases as the
seems much too low in light of the experimental submicroscopic contact angle decreases. Applying
data presented previously. At speeds far below this expression to Blake's (1993) data for a tape
that required for a visible air film, the dynamic plunging into aqueous glycerol gives the result
contact angle is well below 180°, so the analysis shown in Fig. 3.26. The value 10.8 was used for
of Teletzke, Scriven and Davis cannot apply at In(LIL s )' The theory also fits reasonably well
these speeds. Additionally, the theory predicts a Wilkinson's (1975) data for a scraped cylinder,
weak effect of viscosity, whereas experiments but with In(LILs ) set to 3.45. This smaller value
show a major effect. could reflect the presence ofa liquid film remaining
Theories ofthe dynamic contact angle can also on the solid surface. Cox's theory accounts for
0.5
04 ./
V
0.3 ~
~
/
Ca
0.2
0.1
V
0.0
10 100 1000
J.l mPa·s
Figure 3.26 Data of Blake plotted as Ca versus Ji. for the onset of air entrainment for poly(ethyene terephthalate)
tape entering a pool of aqueous glycerol. The solid line is equation (3.38).
the strong effect of the viscosity of the liquid but equations as -A/H~, where H A is air-film
does not directly address the air film. thickness and A is a positive constant taken to
Molecular-kinetic theory predicts dynamic be 10 - 21 J in the sample calculations. The speed
wetting failure when the dynamic contact angle at the liquid/air interface is simply considered to
reaches 180 because the wetting line can advance
0
be a constant V, expected to be greater than zero
no faster. The sawtooth wetting line consisting but less than U. If the pressure in the vicinity of
of slanted straight-line segments that advance the upstream circular section is PL' then the
normal to themselves at the maximum speed is radius of curvature ofthis section is approximately
a consistent outcome. However, the air film is R = (J/P L .
not addressed, and speed predictions require that The origin of the coordinate system is on the
the parameters K s' N and v be known in advance. solid surface where the upstream circular section
makes its closest approach (Fig. 3.27). The flow
in the air film is modeled by the lubrication
3.4.3 MODEL PROBLEM AND ESTIMATES FOR
approximation to the full set ofequations; basically,
AN ENTRAINED AIR FILM
hydrodynamic forces are strongest where the air
A model problem similar to that of Teletsky, film is thinnest and the streamlines are nearly
Davis and Scriven (1988) is helpful in explaining parallel. The lubrication approximation enables
and estimating the factors likely to affect an air the thickness of the air film beneath the upstream
film. In Fig. 3.27, a thin air film separates a solid circular section to be approximated by a parabola,
surface moving at speed U and a layer of liquid H A = H'J + X 2 /2R, which is accurate where the
of average depth L D . A gage pressure PH above film is thinnest. In the lubrication approximation,
the liquid presses on the air film. The ends of the X-component of the Navier-Stokes equation
the meniscus are most simply considered as reduces to ap/ax = J1.A02UA/Oy2 and the Y-
circular sections; this is justifiable if the capillary component to oP/oY = 0; therefore P does not
and Weber numbers are small. These circular vary across the film and the velocity profile is
sections are tangent to a connecting straight line parabolic:
running parallel to the substrate at a distance
UA 3
= [ -;;2(U + 6Q] Y 2
V) - -;;J
equal to the final air-film thickness, H'J. The air HA HA
_
film can be so thin that disjoining pressure affects
it. In the manner of Teletzke, Davis and Scriven + [6~ 4U _ 2VJY+ U (3.39)
(1988), disjoining pressure is included in the flow HA HA HA
Ps
~
~ L
JP
y
~
WA
!........ l
...L..
I
-'+X
u •
Figure 3.27 Model problem for an air film.
Here UA is the X-component of velocity in the gage pressure over the liquid and the hydrostatic
air, and Q is the volumetric flow rate per unit pressure in the liquid as
width of the air. From the velocity profile, the
pressure gradient can be expressed as
P L = PH + pgLo (3.44)
Equations (3.42), (3.43) and (3.44) can be solved
8P = 12J-lA[H A(U+ V)-Q] (3.40) iteratively for H'J as a function of U.
ax H~ 2 Below a specific substrate speed there is no
In the fully developed air film, the pressure solution for H'J, as Teletzke, Davis and Scriven
gradient is negligible and the velocity profile is (1988) found. An air film exists only if the
nearly linear, so Q = H'J (U + V)/2. Equation capillary number based on the viscosity of air is
(3.40) can be integrated with P ---'> 0, X ---'> - 00, large enough that the term under the radical
and the result evaluated at X = 0 to give the exceeds zero:
pressure developed in the air film:
J-lA U 16 PL fA (3.45)
3nJ-lAU(1 + m) - A --;;- > 3n(1 + m) -; './ 2-;;:
PL = 8(H'J)3 /2 .j2R - (H'J)3 (3.41)
The air-film thickness at this critical capillary
Here, m = V/U, and the term on the far right number is given by
represents the adjustment for disjoining pressure.
Lubrication pressure tends to thicken the air film, HOC> > _A )1/3 (3.46)
A (P
but disjoining pressure tends to thin it. A large L
radius of curvature increases the lubrication For air films much thicker than this, the disjoining
pressure. The pressure changes by just a factor pressure term is negligible, and a result similar
of two as the speed V at the meniscus ranges to Perry's is obtained:
between 0 and U. Solving equation (3.41) for H'J
/
gives
H'J ~ ;L[3.j2n~1 + m)(J-l:U)T 3 (3.47)
HOC> _ [31t.j~J-lAU(l + m)
A - 16P{12 These concise results are qualitatively consistent
with the complete calculations.
+ J( 3n.j2(JJ-lA U (l
16P{12
+ m))2 _ ~]2/3
PL
(3.42)
Sample calculations are shown in Fig. 3.28 for
viscosity 10 mPa s, depth 0.2 cm (roughly repre-
sentative of that in the plunging tape experiment),
For the case L » L o ' the lubrication approxi- and surface tension 65 mN/m. The curves for
mation can also be applied to the liquid layer. increasing pressure loads P L of 0, 100 and
The velocity profile is again parabolic, and a 1000 Pa show large speed increases. In plunging
pressure gradient 3J-lV/L~ exists so that the net tape experiments, the speed for visible air
flow rate is zero. Because H'J/L o « 1, the velocity entrainment is in the range of 0.55-1.7 mis,
profile in the air film is nearly linear, and depending upon the materials. The model predicts
equating shear stresses at the interface gives a very thin air film at speeds two or three orders
1 of magnitude below this range, where the
m = (3.43) experimental dynamic contact angle is far below
1 + 3H'J !!:... 180·. Moreover, the predicted speeds for an easily
L o J-lA visible air film 1 J-lm thick are still too low.
The pressure gradient in the upper liquid layer Accounting for Perry's observation that the air
causes the upper interface to be sloped, but in film is in slip flow would reduce the hydrodynamic
the special case where the layer thickness does pressure and increase speeds. However, the weak
not vary much, PL can be estimated from the effect of the viscosity of the liquid, just a factor
Concluding remarks 95
10 oy---T""""--T""""----,r---.....,~-....., an existing air film away from its edges, but the
existence of the film depends upon what takes
m place at its edges.
A complication is that the maximum speed of
wetting depends in general upon the local flow
field. The dynamics of the slanted segments of a
wetting line are determined by the maximum
speed of wetting there, the speed corresponding
0.' +---M;&--+-+-I--7'9I-----4
to a local dynamic contact angle of 180' (hence
the strong viscosity dependence). If the flow field
at the slanted wetting line is similar to the
0.01
original two-dimensional flow, then the maximum
0.0001 0.01
speeds may be essentially the same and there
U m/s may be little hysteresis. Such seems to be the
case for a solid entering a pool. If the flow field
Figure 3.28 Plots of H A versus U showing the effect at the slanted wetting line is markedly different
of three pressure loads P B on the thickness of an air from the original two-dimensional flow, then the
film as predicted by equations (3.42), (3.43) and (3.44). maximum speeds may be much different; the
onset of air entrainment may be catastrophic,
and there may be large hysteresis. Such seems
of two in speed, suggests a basic shortcoming in to be the case for curtain coating. Because the
the model for the air film. The model does not existence of the air film is linked to the maximum
explain dynamic wetting failure either qualitatively speed of wetting and hence the dynamic contact
or quantitatively. angle, the strong effect of the viscosity of the
What appears to be wrong is viewing the liquid found in experiments follows.
appearance or disappearance of the air film as Finally, because the dynamic contact angle
a gradual thickening or thinning with changing and maximum wetting speed generally depend
speed. Although solutions for two-dimensional on the macroscopic flow, the broad range of
air films can be obtained for speeds well below speeds at which a given liquid can undergo
the observed onset of dynamic wetting failure, dynamic wetting failure also follows.
they are evidently unstable and are not realized
in experiments. Observationally, dynamic wetting
3.5 CONCLUDING REMARKS
failure occurs at the maximum wetting speed
where the dynamic contact angle is 180'. The Many opportunities remain for research on
wetting line adopts a sawtooth shape so that the dynamic wetting and air entrainment. Although
component ofthe speed ofthe solid perpendicular the subject has received much attention, no
to the wetting line does not exceed the maximum theory is as yet experimentally verified, generally
speed. The lengthened wetting line continues to accepted, and complete enough for general use
wet the solid, except possibly at the downstream in coating applications. On the experimental
vertices. There is no visible air entrainment at side, there is a clear need to study flows in
the slanted wetting line segments bordering the addition to the plunging of solids into liquids
air film. The air film grows or shrinks from its and the displacement of air by liquid in capillary
edges. A thick, visible air film can appear or tubes. Air entrainment in other flows can occur
disappear suddenly. So viewed, the air film is at speeds much larger or smaller than that which
inherently a three-dimensional phenomenon not would be expected from data on these flows. The
entirely explicable by two-dimensional models. effects ofsurface properties, including topography
The two-dimensional models probably describe and electrical potential, need detailed examination.
Finally, non-Newtonian fluids are widespread in A. W. 1983. Dynamic contact angles on smooth and
practice and need study. rough surfaces. J. Colloid Interf Sci. 94: 123 -130.
The physics of dynamic wetting is key to Cherry, B. W. and Holmes, C. M. 1969. Kinetics of
wetting of surfaces by polymers. J. Colloid Interface
high-speed coating processes. Because of the Sci. 29: 174.
incomplete state of scientific study, most high- Cox, R. G. 1986. The dynamics of the spreading of
speed wetting problems have to be approached liquids on a solid surface. Part 1. Viscous flow. J.
experimentally. Nonetheless, the qualitative and Fluid Mech. 168: 169-194.
quantitative features of dynamic wetting and air de Gennes, P. G. 1985. Wetting: statics and dynamics.
entrainment reviewed in this chapter are useful Rev. Modern Physics. 57(3): 827-863.
in the design and interpretation of experiments. de Gennes, P. G., Hua, X. and Levinson, P. 1990.
Dynamics of wetting: local contact angles. J. Fluid
M echo 212: 55-63.
REFERENCES Dussan V., E. B. 1976. The moving contact line: the
slip boundary condition. J. Fluid Mech. 77: 665-684.
Ablett, R. 1923. An investigation ofthe angle of contact Dussan V., E. B. 1979. On the spreading ofliquids on
between paraffin wax and water. Phil. Mag. 46: solid surfaces: static and dynamic contact lines. Ann.
244-256. Rev. Fluid Mech. 11: 371-400.
Adamson, A. W. 1982. Physical Chemistry ofSurfaces. Dussan V., E. B. and Davis, S. H. 1974. On the motion
4th edn. New York: Wiley. of a fluid-fluid interface along a solid surface. J.
Beguin, A. E. 1954. Method of coating strip material. Fluid Mech. 65: 71-95.
US Patent 2,681,234. Dussan V., E. B., Rame, E. and Garoff, S. 1991. On
Berg, J. C. ed. 1993. Wettability. Surfactant Science identifying the appropriate boundary conditions at
Series, Volume 49. New York: Marcel Dekker. a moving contact line: an experimental investigation.
Blake, T. D. 1993. Dynamic contact angles and wetting J . Fluid M echo 230: 97- 116.
kinetics. In Wettability, Surfactant Science Series, Elliott, G. E. P. and Riddiford, A. C. 1967. Dynamic
Volume 49. ed. J. Berg. New York: Marcel Dekker. contact angles. I. The effect of impressed motion.
Blake, T. D., Clarke, A. and Ruschak, K. 1. 1994. J. Colloid Interface Sci. 23: 389-398.
Hydrodynamic assist of dynamic wetting. AIChE J. Glasstone, S., Laidler, K. J. and Eyring, H. J. 1941.
40: 229-242. The Theory ofRate Processes. New York: McGraw-
Blake, T. D. and Haynes, J. M. 1969. Kinetics of Hill.
liquidjliquid displacement. J. Colloid Interface Sci. Gutoff, E. B. and Kendrick, C. E. 1987. The flow limits
30(3): 421-423. ofcoatability on a slide coater. AIChE J. 33: 141-145.
Blake, T. D. and Haynes, J. M. 1973. Contact-angle Hansen, R. J. and Toong, T. Y. 1971. Dynamic contact
hysteresis. Progress in Surface and Membrane Science. angle and its relationship to forces of hydrodynamic
Volume 6. New York: Academic Press. origin. J. Colloid Interface Sci. 37: 196-207.
Blake, T. D. and Ruschak, K. J. 1979. A maximum Hocking, L. M. and Rivers, A. D. 1982. The spreading
speed of wetting. Nature. 282: 489-491. of a drop by capillary action. J. Fluid Mech. 121:
Bracke, M., De Voeght, F. and Joos, P. 1989. The 425-442.
kinetics of wetting: the dynamic contact angle. Prog. Hoffman, R. L. 1975. A study of the advancing
Colloid Polym. Sci. 79: 142-149. interface. I. Interface shape in liquid-gas systems.
Buonopane, R. A., Gutoff, E. B. and Rimore, M. M. T. J. Colloid Interface Sci. 50(2): 228-241.
1986. Effect of plunging tape surface properties on Huh, C. and Mason, S. G. 1977. The steady movement
air entrainment velocity. AIChE J. 32: 682-683. of a liquid meniscus in a capillary tube. J. Fluid
Burley, R. 1992. Air entrainment and the limits of Mech. 81: 401-419.
coatability. JOCCA. 5: 192-202. Huh, C. and Scriven, L. E. 1971. Hydrodynamic model
Burley, R. and Jolly, R. P. S. 1984. Entrainment of air of steady movement of a solidjliquidjfluid contact
into liquids by a high-speed continuous solid surface. line. J. Colloid Interface Sci. 35(1): 85 -101.
Chem. Eng. Sci. 39(9): 1357-1372. Inverarity, G. 1969a. The wetting of glass fibres by
Burley, R. and Kennedy, B. S. 1976. An experimental liquids and polymers. PhD thesis, Univ. Manchester.
study ofair entrainment at a solid/liquidjgas interface. Inverarity, G. 1969b. Dynamic wetting of glass fibre
Chem. Eng. Sci. 31: 901-911. and polymer fibre. Brit. Polymer J. 1: 245-251.
Cain,J. B., Francis, D. W., Venter, R. D. and Neumann, Jansons, K. M. 1985. Moving contact lines on a
References 97
two-dimensional rough surface. J. Fluid Mech. 154: Ngan, C. G. and Dussan V., E. B. 1989. On the
1-28. dynamics of liquid spreading on solid surfaces. J.
Jansons, K. M. 1986. Moving contact lines at non-zero Fluid Mech. 209: 191-226.
capillary number. J. Fluid Mech. 167: 393-407. Oliver, J. F., Huh, C. and Mason, S. G. 1977. Resistance
Jiang, T., Oh, S. and Slattery, 1. C. 1979. Correlation to spreading of liquids by sharp edges. J. Colloid
for dynamic contact angle. J. Colloid Interface Sci. Interface Sci. 59(3): 568-581.
69(1): 74-77. Perry, R. T. 1967. Fluid mechanics of entrainment
Joanny, J. F. and Robbins, M. O. 1990. Motion of a through liquid-liquid and liquid-solid junctures.
contact line on a heterogeneous surface. J. Chern. PhD thesis, University of Minnesota, Minneapolis.
Phys. 92: 3206-3212. Petrov, P. G. and Petrov, 1. G. 1992. A combined
Johnson, R. E. and Dettre, R. H. 1993. Wetting of molecular-hydrodynamic approach to wetting
low-energy surfaces. In Wettability. Surfactant Science kinetics. Langmuir 8: 1762-1767.
Series. Volume 49, ed. J. Berg. New York: Marcel Ruschak, Kenneth J. 1976. Limiting flow in a pre-
Dekker. metered coating device. Chern. Eng. Sci. 31: 1057-1060.
Johnson, R. E., Dettre, R. H. and Brandreth, D. A. Ruschak, Kenneth J. 1985. Coating flows. Ann. Rev.
1977. Dynamic contact angles and contact angle Fluid Mech. 17: 65-89.
hysteresis. J. Colloid Interface Sci. 62(2): 205-212. Schrader, M. E. and Loeb, G. L. eds. 1992. Modern
Joos, F. M. 1992. Leveling of a liquid layer over a Approaches to Wettability -Theory and Applications.
sorbing surface. AIChE Spring National Meeting, New York: Plenum Press.
New Orleans, LA, pap. 49d. Schwartz, A. M. and Tejada, S. B. 1972. Studies of
Kalliadasis, S. and Chang, H. 1994. Apparent dynamic dynamic contact angles on solids. J. Colloid Interface
contact angle of an advancing gas-liquid meniscus. Sci. 38: 359-375.
Phys. Fluids. 6: 12-23. Seebergh, J. E. and Berg, J. C. 1992. Dynamic wetting
Kistler, S. F. 1983. The fluid mechanics of curtain in the low capillary number regime. Chern. Eng. Sci.
coating and related viscous free surface flows with 47: 4455-4464.
contact lines. PhD thesis. Univ. Minnesota, Min- Tanner, L. H. 1979. The spreading of silicone oil drops
neapolis. on horizontal surfaces. J. Phys. D: Appl. Phys. 12:
Levich, V. G. 1962. Physicochemical Hydrodynamics. 1473-1484.
Englewood Cliffs, N.J.: Prentice Hall. Teletzke, G. F., Davis, H. T. and Scriven, L. E. 1988.
Menchaca-Rocha, A. 1992. The mobility of mercury Wetting hydrodynamics. Revue Phys. Appl. 23:
drops on rough glass surfaces. J. Colloid Interface 989-1007.
Sci. 149: 472-480. Tilton, J. N. 1988. The steady motion of an interface
Miller, Clarence, A. and Neogi, P. 1985. Interfacial between two viscous liquids in a capillary tube.
Phenomena. New York: Marcel Dekker. Chern. Eng. Sci. 43: 1371-1384.
Miller, F. D. and Wheeler, J. 1. 1962. Coating high Voinov, O. V. 1976. Hydrodynamics of wetting. Fluid
viscosity liquids. US Patent 3,206,323. Dynamics. 11: 714-721.
Miyamoto, Kimiaki. 1991. On the mechanism of air Washburn, E. W. 1921. The dynamics ofcapillary flow.
entrainment. Industrial Coating Research. 1: 71-88. Phys. Rev. 17: 273-283.
Mues, W., Hens, J. and Boiy, L. 1989. Observation of Wilkinson, W. L. 1975. Entrainment of air by a solid
a dynamic wetting process using laser-Doppler surface entering a liquid/air interface. Chern. Eng.
velocimetry. AIChE J. 35(9): 1521-1526. Sci. 30: 1227-1230.
SURFACTANTS: STATIC AND DYNAMIC 4
SURFACE TENSION
Yves-M. Tricot
4.1 INTRODUCTION
(N/m). Unlike water, surfactant molecules require
Surfactants - an acronym for surface-active little energy to reside at the interface, because
agents - are versatile and ubiquitous chemicals. they are not well stabilized in the bulk. The
They are found in industrial areas related to hydrophilic part will orient towards the aqueous
detergents (Hancock 1984; Thayer 1993), phar- phase, while the hydrophobic part will orient
maceuticals (Tadros 1984; Attwood and Florence towards the apolar phase. In the adsorption
1983), cosmetics (Ainsworth 1993), paints (Nylen process, surfactant molecules replace high-
and Sunderland 1965), pesticides (Wada et al. energy surface water molecules, which reduces
1983; Tann, Berger and Berger 1992) and oil interfacial tension and decreases the free energy
recovery (Neustalder 1984), to mention a few. of the system. During the coating process, liquid
Mother Nature used them, long before humans, films have to spread over solid surfaces or other
as major building blocks of biological membranes, liquid surfaces, an action generally called wetting.
taking advantage of the so-called hydrophobic This will occur spontaneously if the new surface
effect (Tanford 1980). This chapter focuses on has a lower surface tension than the surface it
applications of surfactants in liquid film coating covers. By reducing surface tension, surfactants
processes, and in particular in coatings involving therefore promote wetting.
aqueous solutions, as encountered in the photo- Another critical parameter for liquid film
graphic industry. coating is time. Coating is a dynamic process.
The action of surfactants is a result of their The process of reduction of surface tension by
partly hydrophilic and partly hydrophobic nature, surfactant adsorption is not instantaneous, but
which causes them to adsorb at interfaces is generally diffusion-controlled. After sufficient
separating media of different polarity. Water is adsorption time, the static or equilibrium surface
particularly important, because of the strong tension will be reached. Determination of static
hydrogen bonding which stabilizes water molecules surface tension is fairly easy (Section 4.4).
in the bulk. Half of the stabilization energy is Unfortunately, static surface tension is seldom
lost for the water molecules that must be at the reached during the most critical parts of a fast
interface. Additional energy is associated with coating process and is of little use in predicting
the interface, which is observed as an interfacial surfactant efficiency during coating. The dynamic
tension. It is the energy required to create one or nonequilibrium surface tension is the relevant
unit of interfacial area (J/m 2 ) and is equivalent surface tension acting during fast processes. It is
to a force acting along a boundary wetting line a recognized fact, and in coating practice surfactant
London. ISBN 0 41206481 2.
100 Surfactants: static and dynamic surface tension
optimization is not made according to static effect of surface expansion are presented in
surface tension data. However, the dynamic Chapter lId.
surface tension is not easy to measure and a The purpose of this chapter is to explain how
workable wetting recipe is often empirically surfactants can be better understood using a
determined from practical coating trials. While scientific experimental approach, as opposed to
coating tests cannot be completely eliminated, a the commonly used empirical methods which
more fundamental understanding of the dynamic consider surfactants as mysterious and unpredict-
wetting behavior reduces the risk of coating able additives. The data presented here, measured
failure due to poor surfactant performance and using industrial (not purified) surfactants in
could widen the coating window in particular aqueous and gelatin solutions, have not been
towards higher speed. previously published and were chosen because
The term 'dynamic surface tension' is not the media are of general interest. For practical
self-explanatory and requires further definition. applications, the most interesting results are
For a particular interface, there is only one value obtained directly from actual coating solutions.
for the static surface tension. It is a thermodynamic An example of data taken from a real product
property which characterizes the equilibrium is given, but with composition details withheld
state of that interface at rest. However, the for proprietary reasons.
dynamic surface tension of the same interface, Dynamic surface tension methods can be
when placed under non-equilibrium conditions, applied to many types of liquids other than
can take a whole range of values. The highest photographic coating solutions. However, one
value is generally that of the solvent (water), and should not believe that they provide a miracle
the lowest will be equal to, and in some cases cure to solve all coating problems. Other import-
even lower than the static surface tension. All ant factors, such as bulk and surface rheology
values in between are possible, depending on the and flow characteristics, may cause coating
time from creation of a fresh surface (surfactant failure. Contaminants may cause local coating
adsorption time), or the rate of expansion or defects, but have no effect on the measured
compression of the surface. The term dynamic dynamic surface tension, because it is not a
surface tension for a particular solution of local value. A systematic surfactant know-
surfactant has only a clear meaning if one ledge must be interactively combined with
specifies under which dynamic process. practical coating trials. Guidelines on how to use
Methods for measuring dynamic surface tension dynamic surface tension to determine wetting
abound but few of them are able to characterize recipes with good chances of success are pre-
surfaces at a time scale of relevance for coating sented.
processes (down to milliseconds or less). Some Unlike many other chapters in this book, this
methods are designed to study surfaces' stretching chapter is not about flow dynamics. Flow
or compression effects, others to study surfactant dynamics is treated in connection with surfact-
adsorption after fast surface creation with no or ants in Chapter ltd, using literature values for
little subsequent area change. A typical geometry parameters such as diffusion coefficient (diffus-
for the first case is an overflowing funnel and for ivity) and adsorption density to predict the effect
the second case a flow coming out of a slit down of surfactants on the flow field. Using the
an inclined plane. Both types are more complex methods described in this chapter, the re-
than static surface tension measurement quired parameters could be experiment-
methods. Section 4.5 presents experimental results ally determined in the solutions whose flow is
obtained with methods ofthe second type, chosen simulated, and the calculated surface ten-
because they have been found to be the most sions could be compared with measured values.
reliable and practical to use. Examples of The two approaches are therefore comple-
results obtained with a technique studying the mentary.
Surfactant properties and applications 101
4.2 SURFACTANT PROPERTIES AND • cleaning (through solubilizing and/or emul-

APPLICAnONS sifying);
• lubricating (solid surfaces).
4.2.1 BASIC SURFACTANT PROPERTIES
Surfactants exist in great variety. They all possess Several of these functions overlap. For liquid
an hydrophobic (apolar) part and an hydrophilic film coating applications, the first three are the
(polar) one, but they can be distinguished according most important. Further general information
to the chemical structure of each of their parts. about surfactants can be found in several extensive
Based on the polar headgroup, one can find the reviews and textbooks (Lucassen- Reynders 1981;
following types of surfactants: Schick 1967; Rosen 1978; Ross and Morrison 1988).
Adsorption of surfactants at the air-water
interface occurs because it lowers the free energy
• anionic (permanent negative charge);
of the system. The adsorbed layer is in equilibrium
• cationic (permanent positive charge);
with monomers in solution. However, when the
• non-ionic (only dipoles but no permanent
surface becomes saturated, and more surfactant
charge);
is added, it becomes more favorable to form
• zwitterionic (simultaneous positive and negative
surfactant aggregates in solution rather than to
charges);
increase the adsorption density (Kahlweit, Busse
• amphoteric (pH-dependent charge).
and Jen 1991). These aggregates are called micelles
and the concentration limit at which they start
and the hydrophobic part can be:
to form is called the critical micelle concentration
(emc). Micelles are more or less spherical structures
• either hydrocarbon or fluorocarbon chain(s); with an aggregation number of the order of 100.
• single, double, multiple chains; The surfactant polar groups in a micelle are in
• linear or branched chain structure. direct contact with water, and form a shell that
surrounds the core made ofthe hydrophobic part
The degree of hydrophilic or hydrophobic of the surfactants. The formation of micelles
character of surfactants has been described by manifests itself by an abrupt change in a number
the so-called HLB number (hydrophile-lipophile of bulk parameters as the surfactant concentration
balance) (Griffin 1949). The HLB number is a is varied (conductivity, osmotic pressure, diffusivity,
predictor for the degree ofsolubility ofa surfactant viscosity, etc.) and also by a plateau of surface
in a particular solvent or indicates its suitability tension (Ottewil 1984; Heimenz 1977). In a first
for a particular type ofapplication. HLB numbers approximation, the surfactant activity remains
lower than 10 indicate water-insoluble and above constant above the cmc, leading to a constant
12 water-soluble surfactants (Ross and Morrison adsorption density and bulk monomer concen-
1988, Chapter IVA). Large collections of HLB tration. The excess of surfactants is absorbed in
numbers have been compiled by Becher (1985) the form of micelles, which do not contribute
and Becher and Griffin (1985). significantly to the solution activity. The cmc
Surfactants can also be characterized by their and the size of the micelles depend strongly on
functions, which can be one or several of the the molecular properties of the surfactant, and
following (non-exhaustive) list: on external parameters (ionic strength, pH,
temperature, other solutes). Cmcs can vary by
• wetting and spreading (on solid or liquid orders of magnitude. Many aggregates other
surfaces); than micelles can exist, depending on the molecular
• emulsifying (insoluble solid, liquid or gaseous geometry of the surfactants and their concentration
secondary phases); (lsraeiachvili, Marceija and Horn 1980). Some
• solubilizing (insoluble molecules); typical structures are shown in Fig. 4.1.
t02 Surfactants: static and dynamic surface tension
monomer spherical globular or

micelle cylindrical micelle
nnmnnr
llHHllHH
rrrnmrrrr
HHHHHH
Lamellar bilayers Hexagonal packing
ot cylinders
Figure 4.1 The most common surfactant aggregate structures.
Micelles are highly dynamic aggregates, with a 4.2.2 SURFACTANT ACflONS IN LIQUID FILM
half-life for micellar formation or breakdown of COATING
the order of 10- 3 -1 s, while the residence time of
monomers within micelles is of the order of 10- 7 s During the process ofliquid film coating, surfaces
(Aniansson et at. 1976; Okubo et at. 1979). This are subjected to several disturbances. Surfactants
has important consequences for the interpretation respond to these disturbances by trying to restore
of diffusion processes of hydrophobic solutes adsorption equilibrium, through diffusion, con-
stabilized in micelles. A discussion of the various vection and micellization processes. Figure 4.2
mathematical models which have been developed illustrates schematically how micelles, monomers
to describe micelles is presented in Chapter ltd. and surface monolayers of surfactants interact.
AIR
Compressed At equilibrium Fresh or expanded

Surface
monolayer JJ-,W..UU J-l-!JJJ-l-1- J--1--J--l--
t
Monomers
'""""'"Q •
----0 ----0
d
Micelles
~~ ~~ ~~
WATE R
Figure 4.2 Dynamic response to a disturbed interface of surfactants at concentrations above the critical
micelle concentration.
Surfactant properties and applications 103
The micelles act as a reservoir of surfactant receive the liquid film. If placed within a few
molecules which diffuses to or from the surface. hundred microns of the substrate, one speaks of
Surface tension gradients induce the so-called slide or cascade or meniscus coating (as is the
Marangoni flows (Marangoni 1878; Ross and case in Fig. 4.3), and if the liquid film is left to
Morrison 1988, Chapter IVB). In this chapter, fall several centimeters before reaching the
we will be concerned mostly with diffusion and substrate, the technique is called curtain coating.
adsorption processes. Theoretical concepts will Chapter t t gives detailed descriptions. In the
be presented in Section 4.3. meniscus coating, the age that the back surface
When surfaces are created, stretched or com- will reach can be very short (fractions of
pressed, the effective surface age, or residence milliseconds) if the meniscus is held at or pinned
time of a particular element of the flow at the to the edge of the coating device. When this is
liquid-air interface, can vary widely. Many coating so, surfactants will usually have little impact
processes use flow metering and guiding devices because time is too short for adsorption to occur
that produce laminar, multilayer flow exiting significantly. However, if the meniscus produces
from parallel slits on an inclined plane and then a heel or vortex (recirculation), surface ages can
transfer the liquid film on a moving web (see become rather long and surfactant effects will be
Chapter 11). Such devices are usually called important. Surfactants affect also edge quality
hopper slides or cascades. Figure 4.3 illustrates and coating homogeneity after deposition on the
a simplified coating situation and indicates the substrate, until the layers have solidified and/or
range of surface ages corresponding to various dried.
parts of the coating process. The residence time
of the layer surface on the slide (up to a few
4.2.3 OTHER SURFACTANT ROLES
seconds) is long enough for the surfactants to
play an active and important role. Proper Using surfactants implies that they are present
surfactant response will ensure stable interlayer in the coating solutions before and after the
wetting, proper wetting of the slide itself, and coating process. Even during the process, sur-
can guard against the formation of waves (Chapter factants can help control other defects than those
t td). During operation, the coating devices are related to macroscopic surface tension. Unfor-
positioned near a moving substrate that will tunately, surfactants can also create unwanted
After coating: up to ca. 10 seconds (until solid)
Moving web
/ , . - - - - Flowing on the slide:
0.1 \0 ca. 2 seconds
~~~IIII
Pinned meniscus:
-3 -4
10 to 10 second
Dy"Om', wett',g U", ~ tt Coating device edge
Static wetting line (unpinned)

Free meniscus (vortex):
up to several seconds
Figure 4.3 Surface age time scales in slide coating. Dark grey flow only: near optimum flow field, very short
transit time at the back of the meniscus; dark + light grey flows together: free static wetting line and meniscus
with recirculation or vortex (long transit times).
104 Suifactants: static and dynamic surface tension
effects. These considerations are divided below the adjoining liquids. Additionally, changes of
into three categories. temperature, composition or dynamic effects can
all induce surface tension gradient-driven flows,
generally called Marangoni flows (Marangoni
4.2.3.1 Stability of coating solutions 1878; Ross and Morrison 1988, Chapter IVB).
A common problem with surfactants is the Surfactants should be able to 'outwet' these
formation and stability offoam. While foam may hydrophobic spots. Unfortunately, as many
be desirable in certain applications (cosmetic industrial surfactants are of technical grade, they
products, for instance), it is a nuisance for liquid may bring oily impurities with them and produce
film coating. Foam stability is high when the film hydrophobic spots which require other surfactants
surface elasticity, or excess surface tension due to fight them.
to non-equilibrium surfactant adsorption, is Surfactants can affect the response of coated
maintained for long periods of time (Ross and films on ambient pressure fluctuations. It has
Morrison 1988, Chapter IVB; Abe and Matsumura been shown that some degree of surface elasticity
1983). The concept ofsurface elasticity is discussed is necessary to dampen waves produced by
further in Section 4.3. Fast-diffusing surfactants external disturbances (Ruschak 1987). To a degree,
will quickly restore adsorption equilibrium and retaining some elasticity is incompatible with the
static surface tension, suppressing elastic energy fast diffusion required to fight potential repellency
and thus achieving low foam behavior. Some problems and avoid foam. A compromise may
long-chain alcohols such as n-octanol have a have to be found based on practical trials.
similar effect and are used as effective defoaming
agents. Even more powerful are non-volatile,
very low surface tension substances made of 4.2.3.3 Interaction with other properties of the
perfluorinated hydrocarbons or silane polymers. product
A completely different kind of problem can be The surface properties of finished products can
caused by the solubilizing effect of surfactants be strongly affected by surfactants. Some surfac-
on molecules adsorbed on suspended colloids. tants (often non-ionics) produce hydrophilic
An example of this phenomenon is known in the surfaces, which are easy to wet by processing
photographic industry as 'dye-stripping', which liquids (like photographic developer) or by another
is the removal of adsorbed sensitizing dyes from coating solution when the product is manufactured
silver crystals (Diaz Garcia and Sanz-Medel in several steps. However, a very hydrophilic
1986). This interaction later affects the properties surface is also easy to contaminate by hydrophobic
of the finished product. substances and may feel sticky. On the other
hand, other surfactants (often anionics) tend to
produce hydrophobic surfaces, which are more
4.2.3.2 Control of coating defects
difficult to wet or re-wet, but are also more
Surfactants must be able to avoid local defects difficult to contaminate. The above trend is not
caused by inhomogeneities in solution. These can necessarily general but was found for photographic
be undissolved particles, hydrophobic oil droplets, products where the main component of the
insoluble molecules or even microscopic air coating is gelatin (unpublished results).
bubbles. The solid substrate may be contaminated In a liquid film coating process, the main role
by hydrophobic impurities. All these defects can of surfactants is or at least should be to control
cause local de-wetting or 'repellency' spots (Gutoff the wetting behaviour of coating solutions. This
1992). The term 'repellency' refers to the phe- is of course a matter of opinion, and there are
nomenon ofliquid retraction from an hydrophobic situations where surfactants are also deliberately
zone. The driving force is the existence of surface used to modify properties of the final products,
tension gradients, which act as shear forces on not just to improve its manufacturability. However,
Theoretical treatment 105
assigning too many roles to an additive such as itself is rather trivial for anyone experienced in
a surfactant detracts from the robustness of the liquid film coating, but obtaining the relevant
manufacturing process, and should be avoided dynamic surface tension information is not.
whenever possible. Optimization of surfactant Demonstration ofthis concept has been reported
recipes is described in Section 4.5.6 based on (Valentini et al. 1991). The use of dynamic surface
their role of controlling dynamic surface tension. tension in optimizing surfactant recipe is further
Surfactant optimization under different criteria discussed in Section 4.7. Most methods for
has been reported elsewhere (Sturge 1977; Knox measuring dynamic surface tension are no longer
1970). patentable (although some patents had indeed
covered such methods), because their use has
been the object of scientific publications (see
4.2.4 SURFACTANT-RELATED PATENTS Section 4.4).
A very large number of patents have been issued
to describe the use and properties of many
4.3 THEORETICAL TREATMENT
different types of surfactants. Virtually every
surfactant on the market has been the object of
4.3.1 WETTING, SPREADING, ADHESION AND
one or several patents, some of them being still
COHESION
in effect. It is not within the scope of this chapter
to review them. However, from reading patents Surface tension is a property defined by two
about surfactants, it appears that it is difficult to immiscible phases. Wetting is a phenomenon
assess the scientific reality of the advantages that (static or dynamic) involving three generally
certain types of surfactants are claimed to bring. immiscible phases, usually a solid, a liquid and
In principle, patents exist for economic, not a gas. In all cases of interest in this chapter,
scientific reasons. It is probable that some claimed wetting is the displacement of air by a liquid
advantages are based on true and reproducible phase on the surface of a solid or another liquid
facts, but in other cases these advantages are phase. The boundary between the three phases
probably exaggerated. In my opinion, there are is called the wetting line. Wetting is a complex
no 'miracle surfactants', but many surfactants subject which is treated in detail in Chapter 3.
can be optimized to yield good results, based on Below is a short review of the basic concepts
simple physical principles describing surface related to wetting and spreading.
tension and wetting, principles which have been When a drop of liquid is placed on a solid
known for nearly two centuries. Interestingly, a surface, it can either spread completely (perfect
patent application has been issued, claiming the wetting) or only partially, in which case the
use of dynamic surface tension as one of the three-phase boundary defines a contact angle ()
objects of the patent (Ishiwata et al. 1990). It was (Fig. 4.4). The relation between the three interfacial
later withdrawn after publication of the search
report. Such a claim was unjustified because
there is nothing new in the desire to use dynamic (J gas
surface tension as a critical parameter for LG \,
optimizing coating. The idea goes back to Young's
equation (Young 1805, see also Section 4.3),
which quantifies how surface tension affects
(J SO
...
wetting. The only additional concept is that the
relevant surface tension for the fast coating
(JSG TIs
process is not the static surface tension, but the
actual surface tension under the dynamic condi- Figure 4.4 Vectorial representation of Young's
tions, i.e. the dynamic surface tension. The concept equation (equation 4.1). See text for definition ofsymbols.
tensions and the contact angle is a force balance the top layer will occur (positive S in equation
which has been proposed by Young (1805): (4.4» if its surface tension is lower than the surface
tension of the layer below. Further requirements
O"LG cos e = O"SG - O"SL (4.1) for multilayer coating stability are discussed in
and Section 4.5.6.
Important derived parameters are the work of
(4.2)
adhesion and the work of cohesion (Blake 1984).
where 0" is the interfacial tension and the indices The work of cohesion We of a liquid phase is the
S, Land G stand for solid, liquid and gas, energy required to create two new interfaces of
respectively. O"so is the intrinsic interfacial tension unit area:
of the pure solid and 1I:s is the two-dimensional (4.5)
film pressure exerted by the adsorbed gas on the
solid. The contact angle can take in principle any Equation (4.5) is used in the ring detachment
value between O· (perfect wetting) and 180· (no method for measuring static surface tension
wetting). The film pressure 1I:s is usually negligible (Section 4.4). Similarly, the work of adhesion WA
on apolar surfaces (hydrocarbons, PTFE). It is is the work required to separate one unit area
important for metals, which have high surface ofa solid-liquid (or liquid-liquid) interface. For
energies, causing strong water vapor adsorption. a solid-liquid interface, it is defined as:
When the liquid drop spreads completely, the WA = O"SG + O"LG (4.6)
contact angle e is equal to 0·, cos e = 1 and
- O"SL
equation (4.1) becomes: and measures the attraction between the solid
and the liquid phases. For two aqueous liquid
(4.3) phases, the interfacial tension is zero and WA
The difference between the two terms of equation reduces to the sum of their surface tensions.
(4.3) is the driving force of spreading, which is (4.7)
called the spreading coefficient, S:
combined with equation (4. 1), equation (4.6) gives:
S = O"SG - O"SL - O"LG (4.4)
WA = O"LG(l + cos e) (4.8)
S must be positive for spontaneous spreading.
Equation (4.8) means that the condition for
For spreading on solid surfaces, S is difficult to
calculate because the solid-gas and solid-liquid complete wetting (cos e = 1) is We = 20"LG ~ WA ,
or that the work of adhesion between the solid
interfacial tensions O"SG and O"SL are difficult to
and liquid must be equal or superior to the work
measure. However, if spreading of the liquid
of cohesion of the liquid. If not, a finite contact
occurs on a second liquid, O"SG in equation (4.4)
angle will be formed. The difference between the
should be substituted by O"L2G' the second liquid
work of cohesion and the work of adhesion is
surface tension and O"SL by O"LtL2 the liquid-liquid
another definition of the spreading coefficient
interfacial tension, which are both directly
(equation (4.4».
accessible. An interesting example is benzene on
water, for which S is initially positive and benzene
spreads, when both liquids are pure, but becomes 4.3.2 SURFACE TENSION COMPONENTS
negative about a minute later, after each liquid
On the molecular level, the attractive energy
saturates the other, causing retraction into droplets
which needs to be overcome to create an interface
(Gutoff 1992, 129- 130).
can be of different types. The interfacial tension
In a multilayer coating process, the top liquid
is decomposed into several components (Blake
layer must be able to spread over the layer just
1984; Hiemenz 1977):
below. Aqueous layers cannot sustain interfacial
tensions. Therefore, spontaneous spreading of 0" = O"d + O"h + O"m + O"X + O"i = O"d + O"SP (4.9)
where the superscripts refer to dispersion forces adsorption is a combination of thermodynamic

(d), hydrogen bonds (h), metallic bonds (m), relations, restricted to ideal systems, and empirical
electron interactions (n) and ionic interactions rules found to extend their application to practical
(i). Only the dispersion forces (d) are common to cases. Detailed derivations and analyses can be
all molecules. They arise from the interactions found in several standard textbooks (see for
offluctuating dipoles which exist even in non-polar instance Adamson 1967; Hiemenz 1977; Ross
molecules (London forces). All the other inter- and Morrison 1988). Following is a short summary
actions can be referred to as specific forces (sp). presenting the essential concepts relating surface
The dispersion forces are also the only ones to tension and surfactant adsorption. Based on
operate across phase boundaries, as they are thermodynamic arguments, for an ideal, very
long-range and non-specific. Thus they are dilute non-ionic surfactant, the variation ofsurface
responsible for the work ofadhesion WA (equation tension is related to surfactant adsorption
(4.6)). From the studies of many non-polar according to the Gibbs equation:
substances, for which O'LG = O'd, it has been found
experimentally that:
WA12 = 2(O't 0'~)1/2 (4.1 0)
(~)
v(ln C) T,P
= -RTr (4.13)
where WAI2 is the work of adhesion between where 0' is the surface tension, C the bulk
liquids 1 and 2, and O'~ and O'~ are their respective concentration of the solute (surfactant), T the
dispersion component of surface tension. In absolute temperature, P the pressure, R the
water, the surface tension O'w is composed of universal gas constant, and r the surface con-
hydrogen bonds and dispersion forces, or: centration (or Gibbs's excess surface concentration)
of the adsorbed surfactant. Equation (4.13)
(4.11)
expresses the fact that surfactant adsorption
By measuring the interfacial tension O'WH of water reduces surface tension. The reduction of surface
against pure hydrocarbon liquids, for which the tension from the solvent value to the surface
surface tension O'H is equal to its dispersion tension in the presence of adsorbed surfactant is
component O'~, it is possible, by combining called the spreading pressure n:
equation (4.6) with equation (4.10), to determine
the dispersion component of the water surface (4.14)
tension O'~:
where O's is the surface tension of the solvent. At
O'WH = O'w + O'H - 2(0'~O'H)1/2 (4.12) sufficiently low concentrations, the spreading
All quantities except O'~ are directly measurable, pressure n becomes proportional to the surfactant
therefore equation (4.12) provides a means to concentration, or in other terms a surface tension
evaluate this quantity. For water at 20·C, isotherm would be linear (ideal behavior). For
O'w = 72.5 mN/m and O'~ = 21.8 mN/m. A similar very surface-active solutes, the dilution required
procedure can be applied to complex aqueous to reach ideality may be extreme and of little
solutions, containing surfactants and/or colloids, practical interest. Nevertheless, under such
to gain insight about the dispersion component conditions an ideal surface equation of state can
in practical solutions, and about the surface free be derived from equations (4.13) and (4.14) (Ross
energy of solids (Janczuk et al. 1989). and Morrison 1988):
4.3.3 STATIC SURFACE TENSION AND n = rRT or nA = RT (4.15)

SURFACTANT ADSORPTION
where A = l/r is the molar surface area. For
The theoretical description of the relations between practical surfactant concentrations, equation (4.15)
surface tension (static and dynamic) and surfactant cannot be applied, but an empirical equation
(von Szyszkowski 1908), is often used: molecule is calculated from the adsorption density
= as - a = RT roo In(1 + Cia) (4.16)
r (mol/m 2 ). Most surfactants occupy an area
between 20 and 100 A2 per molecule at saturation
TC
where r 00 is the maximum excess surface concen- adsorption (when r = roo)' This area can vary
tration (or maximum adsorption density) and a considerably with the presence of other solutes
an empirical constant. Below the critical micelle that compete for the surface. The parameters roo,
concentration (cmc), experimental plots of surface a and as are experimentally determined by a fit
tension versus surfactant concentration can be of the static surface tension versus surfactant
very well fitted with equation (4.16) (see Section concentration with equation (4. 16)(see Section 4.5.3).
4.5.3). Above the cmc, surface tension usually stays
constant. Increase of surface tension is sometimes 4.3.4 DIFFUSION AND DYNAMIC SURFACE
found above the cmc in the presence of impurities. TENSION
The cmc is not always clearly defined, but can be
This section describes the action of diffusion as
seen sometimes as a broad transition region.
a mechanism trying to restore surfactant adsorp-
An adsorption isotherm can be derived from
tion equilibrium after surface generation. Under
equations (4.13) and (4.16), as shown below.
conditions of diffusion-controlled adsorption
Equation (4.13) is rewritten as:
(absence of adsorption barrier), the evolution of
RTr adsorption r with time is ruled by the Ward and
(4.17)
C Tordai equation (1946):
r = 2Co(~tY/2 - 2(~Y/2 III2 Cs(t - r)dr 1 / 2
(4.18)
(4.21 )
Equating the right-hand terms in equations (4.17) where Co is the bulk concentration (far from the
and (4.18) and rearranging, gives: surface), Cs the subsurface concentration (con-
centration in equilibrium with the surface), D the
r = r ooC or r c CIa diffusion coefficient (m 2 Is), t the age ofthe surface
a +C roo a +C 1 + CIa (time), TC the circle number TC (not the surface
(4.19) pressure) and r an auxiliary variable (s). The first
term on the right-hand side of equation (4.21)
Equation (4.19) is known as the Langmuir represents the diffusion from the bulk to the
adsorption isotherm, and the empirical constant subsurface. At the time of surface creation, the
a as the Langmuir constant. Because of its close subsurface concentration and the adsorption
relationship with equation (4.19), equation (4.16) density are nearly equal to zero. At very short
is also called the Langmuir-von Szyszkowski surface ages, adsorption is still negligible and no
equation. significant back-diffusion takes place. A simple
By eliminating the surfactant concentration estimation of the diffusion coefficient (diffusivity)
between equations (4.16) and (4.19), Frumkin can be made from the initial stage of adsorption:
(1925) obtained a new surface equation of state, a plot of r versus t 1/ 2 will be a straight line with
relating the surface tension with the surfactant a slope equal to 2Co (DITC)1/2. When adsorption is
adsorption: significant, diffusion from the subsurface to the
as - a = -RTr ooln(l - :(0) (4.20)

bulk (back-diffusion), represented by the second
term (integral) of equation (4.21) cannot be
neglected anymore. Evaluation of the integral
Equation (4.20) applies for both the static and term is complex and is usually approximated by
dynamic surface tension. The area per surfactant iterative methods (van den Bogaert and Joos 1980).
Equation (4.21) is a rigorous description of Morrison 1988). The equilibrium adsorption

truly diffusion-controlled adsorption. However, density r 0 is determined for each surfactant
it has often been found easier and more convenient concentration by substituting r with r o and (J
to use various approximations (van den Bogaert with (Jeq in equation (4.20) (Frumkin 1925). As
and Joos 1979). For surfactants following equation for equation (4.22), values of the diffusion
(4.16) (von Szyzskowski 1908) and for diffusion- coefficient D are varied until a good fit with the
controlled kinetics, Hansen (1960) has proposed experimental dynamic surface tension values is
approximations of equation (4.21) for short and obtained.
long times. The short-time approximation is: Figure 4.5 shows the type of curves obtained
( 4un) n 6zin) from equations (4.22) and (4.23). Using two

u=z+ 1 - - z2 + ( 1 -2- - - + -2 z3 independent equations to simulate the short- and
3u 2u long-time domains of the experimental data
(4.22) offers more flexibility than equation (4.21). The
best fit may lead to the determination of two
where different diffusion coefficients: one for the short-
time and one for the long-time domain. Often
U= Cs; it = Co and z = (2Co) (Dt)1/2 with real surfactants, a concentration-dependent
a a roo n
diffusion coefficient is found. Most industrial
and where Cs' Co, nand D have the same surfactants have indeed a complex diffusion
meaning as in equation (4.21). The parameters a behavior, as they are often mixtures of several
and roo are taken from the fit of the static surface surface-active compounds (see Section 4.5.1), and
tension data with equation (4.16). An initial value also because of the presence of micelles above
of the diffusion coefficient D is chosen and the cmc. An extensive review of mathematical
introduced in equation (4.22). The resulting values models and experimental data of surfactant
of the subsurface concentration Cs' as a function adsorption and diffusion has been given by
of the surface age t, are substituted to C in Chang and Franses (1994).
equation (4.16) (von Szyszkowski 1908) and yield Discussion of other mechanisms affecting surface
a calculated curve of dynamic surface tension as equilibrium, such as convection and surface area
a function of surface age. The calculated curve changes, can be found in Chapter lid. These
is then compared with the experimental data and effects are of less significance for the dynamic
the process is repeated with another value of the
diffusion coefficient until a good fit is obtained.
70
A diffusion coefficient for the short-time domain
~
at a particular surfactant concentration is thus -. E 60
determined. The procedure is repeated for all z
measured surfactant concentrations. c:
.s
0 50
The long-time approximation is: 'iii
c: '.
2 40
RTr~ ( n )1/2 QJ
kDt (4.23) U
(J = (J eq + ~ <1l
't:
:>
(/)
30 ._- ....
where (J is the dynamic surface tension, (Jeq the
20
equilibrium surface tension, r o the equilibrium 0.0001 0.001 0.01 0.1 10 100
adsorption density (different from roo) and k is
Surface age [5]
a constant equal to 1 for anionic surfactants at
low ionic strength, and equal to 4 for anionic Figure 4.5 Combination of the short-time (solid line)
surfactants at high ionic strength or for non-ionic and long-time (dashed line) Hansen approximations
surfactants at any ionic strength (Ross and for simulation of dynamic surface tension.
surface tension methods used to produce the to mixtures of surfactants or other complex
results presented in Section 4.5. solutions where static surface tension data do
not follow equation (4.16).
4.3.5 PARAMETRIC SIMULATION OF
DYNAMIC SURFACE TENSION 4.3.6 SURFACE ELASTICITY AND SURFACE
It is possible to simulate the experimental dynamic VISCOSITY
surface tension values without any knowledge of Many dynamic surface phenomena encountered
the static surface tension, using a four-parameter in coating flows are due to the viscoelastic
equation (Hua and Rosen 1988): properties of surfaces. The surface elasticity is
a o - aM defined by (Gibbs 1961; Lucassen and Giles 1975;
a = aM + 1 + (t/t*t (4.24) Lucassen-Reynders 1981):
da da
where aM is the surface tension near to equilibrium E = - - = A- (4.25)
(meso-equilibrium), a o the surface tension ex- dInA dA
trapolated at zero surface age (not to be confused where E is the dilational surface elasticity modulus,
with as' the surface tension of the solvent), t the da the surface tension increment corresponding
surface age, t* and n two characteristic constants. to a relative molecular area change dInA or
A typical curve produced by equation (4.24) is dA/A. An equivalent definition relates E to the
shown in Fig. 4.6. Its shape is very similar to adsorption density:
that of the combined Hansen approximations,
and excellent fits can easily be obtained (see da da
E = --- = -r- (4.26)
Section 4.5.4). Regression is made using a linearized dlnr dr
form of equation (4.24) (Hua and Rosen 1988). Surface elasticity, measured isothermally from
No molecular information can be derived from equilibrium surface pressure-molecular area
such a simulation. However, it is useful because curves, is called Gibbs elasticity. It is a ther-
static surface tension data are unnecessary and modynamic property of the interface. Calculated
prediction ofdynamic surface tension as a function values for the Gibbs surface elasticity can also
of surface age and surfactant concentration can be obtained as a function of parameters obtained
still be made. Equation (4.24) can also be applied from static surface tension data, as shown below.
Differentiating equation (4.20) (Frumkin) with
60 ---- respect to the adsorption density r gives:
[ ( roor)J
Ii
E
Z
E- 50 ° 0
d(a - as) = da = RTr

d In 1 - -
c
0 I
l\ dr dr 00 dr
'iii
c
2 40 (4.27)
<IJ
()
co The logarithmic derivative term is rearranged as:
't:
~
C/) 30 Om
0.001 0,01 0.1 10 100
Surlace age [sl
Figure 4.6 Typical curve of dynamic surface tension

obtained using equation (4.24). The parameters are (4.28)
defined in the text.
Overview of surface tension measurement methods 111
Combining equations (4.26), (4.27) and (4.28) yields: fJ.(Pa s = N s/m 2 ). The surface dilational viscosity
can only be determined from the surface under
r expansion or compression, unlike the Gibbs
E = RTr oor _r (4.29)
00 surface elasticity which is measurable from
equilibrium experiments.
Finally, it is easily shown using equation (4.19)
In oscillation experiments, the elasticity modulus
that 1/(r 00 - 1) = C/a, and the elasticity is finally
is small at low frequencies, when the surface
expressed as:
viscosity is dominant and dissipates most elastic
E = RTr 00 (C/a) (4.30) energy, but at high frequencies the elasticity
reaches a maximum limiting value which is
Equation (4.30) has been used to describe the generally higher than the Gibbs elasticity. At
effect of elasticity in the stabilization of waves sufficiently high frequencies, diffusion can no
on a disturbed planar flow (Ruschak 1987). longer contribute to energy dissipation and the
Surface elasticities measured under dynamic surface becomes purely elastic. Wave dampening
conditions are called Marangoni elasticities. As is optimum for a particular combination of
the surfaces are no longer isothermal under viscous and elastic behavior, depending on the
dynamic conditions, Marangoni elasticities are frequency of the perturbation (Ruschak 1987).
usually larger than Gibbs elasticities (Ross and Mathematical details of surface elasticity phe-
Morrison 1988). In fact, there is a certain confusion nomena, and a recently developed measurement
in the literature, because what is called Marangoni method based on an oscillating bubble, have
elasticity is often a surface viscoelasticity. For been described by Lunkenheimer et al. (1990).
instance, the dynamic surface tension response
obtained after a jump of surface area of a
surfactant solution, is an initial elastic response 4.4 OVERVIEW OF SURFACE TENSION
followed by a 'viscous' recovery to equilibrium MEASUREMENT METHODS
surface tension. In this case, the mechanism of
viscous dissipation ofenergy is surfactant diffusion 4.4.1 THE WILHELMY PLATE
and adsorption. Purely elastic behavior is Proposed by Wilhelmy (1864), this is one of the
characterized by a linear response to stress with simplest and most reliable methods used to
no phase shift, as expressed in equation (4.25). measure surface tension. A clean, thin plate made
However, under sinusoidal surface compres- of platinum, glass or even filter paper is placed
sion-dilatation cycles, the dynamic surface tension vertically at the surface of the measured liquid.
response was found to exhibit a phase shift with As the liquid wets the plate, the force F exerted
respect to the applied stress, which could reach by the meniscus on the plate is measured by a
up to 90° in the presence of micelles (Lucassen balance or a force transducer, and is expressed by:
1975). The 90° out-of-phase component of the
surface tension response to surface area change F = p(1 cos () (4.32)
is due to the surface dilational viscosity, defined
where p is the perimeter ofthe plate, (1 the surface
as (Bergink-Martens et al. 1990):
tension and () the contact angle of the liquid on
(1 - (1cq
the plate. If the plate is effectively clean, the
(4.31) contact angle is zero and the cosine term is equal
Jls = d(ln A)/dt
to 1.
where (1 is the surface tension under surface To avoid any buoyancy correction, the plate
expansion and (1eq the surface tension of the is positioned with its lower edge exactly at the
surface at rest (equilibrium surface tension). The average liquid level. The method can be calibrated
parameter fJ. s (N s/m) is the two-dimensional with standard weights, independently ofa reference
equivalent of the three-dimensional bulk viscosity liquid. It measures true static surface tension,
without any liquid motion. The reading is where AP is the difference of pressure across a
independent of the density of the liquid. In order spherical interface of radius of curvature R. If
to ensure zero contact angle, the plate must be the contact angle is smaller than 90·, the liquid
thoroughly cleaned (platinum or glass) or replaced will rise until the hydrostatic pressure just balances
(filter paper) after each measurement. the Laplace pressure. At equilibrium:
The Wilhelmy plate technique is also applicable
to dynamic surface tension measurements by AP = pgh = 2acos(}/R (4.35)
placing the plate along a flowing liquid. A
where p is the liquid density, 9 the gravity
practical set-up is based on an inclined plane
constant, h the meniscus height or capillary rise,
(van den Bogaert and Joos 1979). It will be
a the surface tension, () the contact angle and R
described in more detail in Section 4.5.2.3.
the radius of the glass tube. The method is used
preferably when the contact angle is zero
4.4.2 THE DV NOVY RING (cos () = 1) because if a finite contact angle exists,
its value is generally not known. The method is
Named after du Nouy (1919), this method uses limited to liquids of low viscosity and the radius
an horizontal ring instead of a vertical plate. The of the capillary tube must be precisely constant.
ring is normally made of platinum, to ensure Because of these limitations, this technique is
good wetting, and easy cleaning through heating used mostly for rough estimates ofsurface tension.
in a flame. Initially, the method consisted of It is not appropriate to measure even slow
measuring the force necessary to detach the ring variations of surface tension.
from the liquid. The force to overcome is the
cohesion of the liquid, equal to 2a, as given in
equation (4.5). The maximum force measured, 4.4.4 THE OSCILLATING JET
equal to the weight of the meniscus held by the The oscillating or vibrating jet has been originally
ring, is expressed as: described mathematically by Lord Rayleigh (Strutt
(4.33) 1897). It is applicable to the measurement offresh
F = a4rrRK
surfaces, aged between ca. 3 to 50 ms. The
where R is the ring radius and K is a correction measured liquid is ejected horizontally from an
factor given by Harkins and Jordan (1930), which elliptical orifice. The differences of curvature in
takes into account the small volume of liquid the elliptical section of the jet produce an internal
that remains attached to the ring and the difference pressure difference according to the Laplace
between the ring radius and the actual radius of equation:
the meniscus in the rupture plane. The method
AP = 2a(I/R 1 - I/R 2) (4.36)
is not as truly static as the Wilhelmy plate, as
the meniscus is stretched - although as slowly where a is the surface tension, R 1 is the small
as desired - during the measurement. Further and R 2 the large radius of curvature in the
refinements have been made to avoid the need elliptical section of the jet. The resulting flow
for ring detachment and make the measurement causes the jet to oscillate between two extreme
automatic (Sucker et al. 1964). Specific details elliptical sections oriented at 90· to each other.
are given in Section 4.5.2.1. When the flow is constant, this oscillation can
be seen as a stationary wave. The dynamic
surface tension can be calculated according to
4.4.3 THE CAPILLARY RISE
(Defay and Petre 1971):
The capillary rise is based on the Laplace equation:
4pV*2
a = 6aA 2
(1 ++
1
(37/24)(b 2/a 2))
(5/3)(rrR 2/A 2) 'P (4.37)
AP = 2a/R (4.34)
where 'II = 1 + 2(JlA/PV*)3/2 + 3(JlA/p v*f; Jl is given by (Brown 1961):

the liquid viscosity, p the density, V* the volumetric
flow rate, R the average radius of the capillary, (j = pH V2 sin2 (J. (4.38)
a = .J(R 1 R 2 ) the radius of a circle of same area 2
as the cross-section of the nozzle orifice, R 1 and
where p is the liquid density, H is the film
R 2 are as defined in equation (4.36),
thickness and V is the velocity of the film at the
b = (R 2 - R 1 )/2 is the oscillation amplitude and
point of rupture. For a symmetric curtain, the
A the oscillation wavelength. For solutions of
surface tension (j can be easily calculated from
viscosity Jl < 20 mPa s, the term 'II is approxi-
directly measurable quantities knowing that:
mately unity.
As the liquid moves away from the orifice, the
surface ages and the surface tension is reduced Q=HoVo=HV or H=~ (4.39)
by adsorption of surfactants, which causes an
increase of the wavelength of the oscillation. The where Q is the flow rate per unit width, H 0 is
age t of the surface can be calculated from the the initial film thickness (slit width), Vo is the
horizontal surface velocity V. and the distance initial velocity, while H and V are the same
X from the nozzle orifice (t = X/Ys). The surface parameters at some distance X of the slit.
velocity is estimated from the flow rate According to Brown (1961), the velocity can be
(V. = V*/(na 2 )) or measured from the jet free-fall expressed as:
trajectory (see Section 4.5.2.2).
V 2 = V 2 + 2g[X - 0.5(Jl/p)(2/3)] (4.40)
0
where Jl is the liquid viscosity. Combining

equations (4.38), (4.39) and (4.40), the dynamic
4.4.5 THE FALLING CURTAIN
surface tension is expressed as:
The falling curtain technique is based on a thin
sheet of liquid extruded from a slit and falling Qp . 2
(j = -sm (J.
between two vertical guide wires. There are two 2
basic possibilities: a symmetric curtain, where (4.41)
both sides of the curtain emerge vertically from
the slit at the same time, or an asymmetric curtain Equation (4.41) can be used as a rough
where one surface is generated on an inclined approximation also for the asymmetrical curtain.
plane before producing a falling curtain. The However, Balbaert et al. (1987) have found that
symmetric curtain has the advantage ofsimplicity, the Mach angle method applied to an asymmetrical
as both sides of the curtain have the same surface curtain seems to measure only the lowest surface
age. The asymmetric curtain is a better model tension, and not the average between both sides
for the curtain coating technique (see Chapter of the curtain. Instead of the Mach angle, they
11c), but is more difficult to interpret, since both used a technique based on generating, with a
sides of the curtain do not have the same surface vibrator, capillary waves on the curtain and
age. Detailed mathematical descriptions of both measuring the resulting wavelength by optical
methods have been reported (van Havenberg and means. The method works, but is more delicate
Joos 1983; Balbaert, Bleys and Joos 1987). than the Mach angle technique. The mathematical
The determination of the surface tension in treatment of the asymmetrical curtain is also
the curtain is generally made by introducing a more complex.
perturbation in the flow, causing the rupture of Another problem with the falling curtain
the film and the formation of a V-shaped edge, technique is to relate the dynamic surface tension
also called a Mach wave. The relation between with a surface age. It is not trivial, as the surface
the Mach angle (J. and the surface tension (j is is under expansion as the curtain is accelerated
by gravity. The detailed treatment ofthis problem The problem of determination of the true
can be found in the above reference. surface age may have been recently solved through
the design of a new maximum bubble pressure
instrument, developed by an Ukrainian research
4.4.6 THE MAXIMUM BUBBLE PRESSURE group (Fainerman 1992; Fainerman et al. 1994;
When a gas is blown through a capillary tube Fainerman and Miller 1995) and available from
into a liquid, bubbles are formed and escape the company Lauda. This instrument uses the
when a critical pressure is reached. This 'maximum transition between continuous jet and discrete
bubble pressure' is determined by the Laplace bubble, as the flow rate is reduced, to determine
equation and is related to surface tension by the dead time in each experiment, combined with
(Garrett and Ward 1989): an acoustic or electric detection of the bubble
detachment. The instrument was tested in our
AP - pgh) laboratory and found to work very well for
(1=
( R (4.42) aqueous surfactant solutions. Dynamic surface
2
tension could be measured easily down to 5 ms
where AP is the pressure difference, p the liquid or less of surface age. Some difficulties remained,
density, g the gravity constant, h the hydrostatic however, in the presence of dispersed particles,
height of liquid at the capillary tip and R the caused by contamination of the small capillary
radius of the capillary. The method appears to walls. Further improvements in the quality of
work regardless of the orientation ofthe capillary, the capillary may be able to minimize the
either pointing upwards (Garrett and Ward contamination tendency.
1989), downwards (Bendure 1971; Hua and Rosen
1988) or sideways (Mysels 1989).
One advantage ofthis technique is its application 4.4.7 THE OVERFLOWING FUNNEL
to surface tension determination in remote places, The overflowing funnel is an apparently very
for instance inside pipes along a production line. simple and convenient method for measuring
It can in principle measure dynamic effects over dynamic surface tension caused by a surface
a wide range of surface ages, by simple variation dilatation effect. The term 'apparently' is used
of the bubble rate. The technique has been because the interpretation of the technique is not
developed with automatic, computer-controlled simple and it shows an effect which cannot be
operation and is highly suitable for routine related to a defined surface age. The method
analysis. consists of feeding a liquid from underneath a
However, problems arise when one is interested vertical funnel and leaving it to overflow over
in the details of the surface ages. The process is the rim. The surface tension is measured with a
deceptively simple. In fact, a complex sequence Wilhelmy plate (or a du Nouy ring) usually at
of events is taking place during the formation the center of the funnel (loos and de Keyser 1980;
and escape of the bubbles. The bubble growth Bergink-Martens et al. 1990). The surface is
period is a 'dead time', during which pressure subjected to a relative expansion rate A::
drops. Pressure builds up while a fresh meniscus
ages inside the capillary. When the pressure * 1 dA d(1n A)
A ------ (4.43)
reaches the maximum Laplace pressure, the R - A dt - dt
bubble starts to grow rapidly and irreversibly.
According to Ward and Garrett (1989), the dead where A is the surface area of the emerging liquid,
time is dependent of the measured dynamic equal to the area of the funnel exit. The increase
surface tension as well as the capillary radius, of surface tension A(1 can be related to physico-
thus complicating the estimation of true surface chemical parameters through a relation proposed
age. by van Voorst Vader et al. (1964)(equation(4.44)).
Au = RTr In ( 2(1 + Cola) ) (4.44)

00 1 + Cola - K + J(l - Cola + K)2 + 4Cola
where Co is the surfactant concentration at analysis was confirmed by direct measurements
equilibrium, a is the Langmuir constant and the of the radial surface velocity. Therefore, due care
parameter K is: must be taken in interpreting the results obtained
with the overflowing funnel. The type of over-
K = roo
a
(1tA:)1/2
2D (4.45)
flowing funnel described in Chapter lId has a
non-ideal geometry, resulting in a very complex
The surface expansion rate A; must be
flow pattern which cannot be treated using the
theory developed by Bergink-Martens et al. (1990).
calculated independently of the change of surface
tension, if one wants to determine the diffusion
coefficient D by comparing the experimental data
with the calculated values. The problem is complex, 4.4.8 THE SESSILE OR PENDANT DROP
as elements of liquid are forming new surface The profile of a sessile ('sitting') drop resting on
areas not just from the center but from the entire a plane or pendant ('hanging') drop is the result
funnel, with a defined velocity profile of the of the balance between surface tension and
incoming liquid. The presence ofsurfactant causes gravity forces. The shape of the drop has to be
surface tension gradients which can further precisely measured and analyzed. Computer
complicate the analysis. Nevertheless, Joos and programs can now perform this task very
de Keyser (1980) have proposed a simple express- accurately, and with ease in contrast to the use
ion for the average surface expansion rate based of the original tables from Bashforth and Adams
on the volumetric flow rate d l1dt and the rise (Rotenberg et al. 1983; Lopez de Ramos, Redner
AH of the overflowing surface as compared to and Cerro 1993). However, the method is still
its position when the liquid is at rest: complex and tedious compared with the Wilhelmy
plate or the du Nouy ring. It may be worthwhile
dV dA
Q= dt = AH dt (4.46) when the greatest accuracy of surface tension is
required. Interfacial tension can be measured and
and combining equation (4.46) with (4.43) gives: automatic operation of the technique is possible.
Even highly viscous liquids can be measured.
A* -.-5L
AAH (4.47) Slow dynamic (aging) effects can be studied.
R - Pendant drops can be generated more quickly
than sessile drops, so that faster dynamic effects
Joos and de Keyser (1980) found that values of can be followed. Pulsating techniques have been
surface tension recalculated using equation (4.44) reported (I"unkenheimer et al. 1984). A general
agreed well with their experimental data, taking description of the method has been given by
independently determined values for roo' a and Patterson and Ross (1979) and Ambwani and
D. On the other hand, Bergink-Martens et al. Fort (1979).
(1990) have analyzed in depth the problem of
calculating the relative expansion rate din Aldt
and found more complex expressions, even with
4.4.9 OTHER METHODS
an optimized geometry of the funnel. The reader
is referred to the above paper for more details. Many other methods have been developed to
The geometry is optimum when the conical part measure static andlor dynamic surface tension,
of the funnel has a small slope (1:10) and the axis or surface expansion. They are, for example, the
of the funnel is exactly parallel to the gravity maximum drop weight or volume, the maximum
vector. The validity of Bergink-Martens et al. pull on a rod, the spinning drop, the plunging
tape, the wave propagation, the flowing bell, the or potential interest, and is intended to be neither
flowing sheet, the vertical jet. The description of exhaustive nor even representative of the very
these techniques can be found in several reviews large number of available industrial surfactants.
(Ross and Morrison 1988; Defay and Petre 1971; The names and main characteristics ofthe selected
Defay and Hommelen 1958). surfactants are listed in Table 4.1. The measure-
ments were done in water at 40·C (deionized for
dynamic studies, ultrapure, chromatographic grade
for static ones) and in 4% gelatin at 40·C
4.5 EXAMPLES OF PRACTICAL (inert cow bone type, medium viscosity and
APPLICATIONS deionized).
Further information on the active chemicals
4.5.1 MEASURED INDUSTRIAL is shown below. The molecular weight, when
SURFACTANTS known, is an approximate value, due to the
Static and dynamic surface tension data of a existence of a chain length distribution.
number of industrial surfactants have been
measured in our laboratory with the methods Triton 770 (70% anionic, 30% non-ionic):
described in the next section. Our data have not
been previously published. These products contain
impurities to various degrees, but were chosen 70% mol. WI. 617
because they represent what is or could be used (average mol. WI. 586)
in reality. The detailed analysis of some of them
30% mol. WI. 515
is complex and will not be given here. The choice
of the surfactants was based on their known use
Table 4.1 List and main characteristics of various industrial surfactants
Approx. conc.t
Surfactant Type Manufacturer (%) Solvent
Triton 770 Anionic * Rohm & Haas 30 33% i-prop.tt

66% H 2 O
Texapon K-12 Anionic Henkel 100
Nekal BX Anionic BASF 100
AerosolOT Anionic American Cyanamid 100
Empimin LSM30 Anionic Albright & Wilson 30 H 2O
Fenopon CO-433 Anionic GAF 35 H 2O
Niaproof #4 Anionic Niacet Co. 27 25% EDGt
75% H 2 O
Triton X-200 Anionic Rohm & Haas 28 H 2O
Olin lOG Non-ionic Olin Co. 50 H 2O
Emcol5130 Non-ionic Witco 100
Igepal CA-720 Non-ionic GAF 100
or Triton X-102 Non-ionic Rohm & Haas 100
Lodyne S-103 Fluoro** Ciba-Geigy 45 30% HEGH
70% H 2O
F1uorad FX-1002 Fluoro** 3M 25 H 2O
*contains also ca. 30% non-ionics ** also anionic

t'I00' means no solvent, not 100% pure tt isopropanol
t diethylene glycol monoethylether U hexylene glycol
Examples of practical applications 117
Texapon K-12: Na-dodecyl sulfate (SDS), mol. Emcol 5130 (non-ionic):
weight 288.38.
o
II
Nekal BX (anionic): CH 3 -(CH 2)n - C- N(CH 2CH 20H)2 45 malo/. mol. wi. ca. 315
n = 6.8.10~.~.16,18
" glycerine 15 malo/.

- diethanolamine 4Omol%
- fatty acid mono- and diglycerides (traces)
- free fatty acids (traces)
60% <1% 40%

Igepal CA-720 or Triton X-102 (non-ionic):
mol. wt. 342
Aerosol OT (anionic): n = 11
CSH17-COfH-OS03"Na+ mol. wI. 42S
CsH17 -COCH 2 Lodyne S-103 (ftuoro-anionic):

'Fluoroalkyl sodium sulfonate', estimated mol. wt.
ca. 700
Empimin LSM30 (anionic): (exact formula is proprietary information).
Fluorad FX-1002 (ftuoro-anionic):

(average and approximate formula; exact formula
Fenopon CO-433 (anionic): is proprietary information).
C9H19-@-(OCH2CH2)4-0S03"Na+ mol. wI. 498 o

1/
CF3 - (CF2 )6 - C- O' NH 4 + mol. WI. ca. 431
Niaproof #4 (anionic):
H H
I I
C4 H9 -y-C 2H4 -y-CH2CH(CH3)2 mol. wt. 316
C 2 HS S04Na 4.5.2 APPLIED MEASUREMENT METHODS
4.5.2.1 The du Nou)' ring (static surface tension)

Triton X-200 (anionic):
A Lauda TEIC automatic tensiometer, fitted
with a platinum du Nouy ring attached to a force
mol. wt. 661
transducer (Hottinger Baldwin Messtechnik, range
± 5 g), was used. The instrument takes advantage
of a special technique designed to avoid film
Olin lOG (non-ionic): rupture during measurement (Sucker et al. 1964).
Consequently, the surface tension reading could
be repeated several times until a constant value
was reached, without need for ring cleaning. The
C9H19-@-O-(CH2iHCH20ho-H mol. wt. 961
accuracy of the surface tension values was better
OH than 0.1 mN/m.
4.5.2.2 The oscillating jet (dynamic surface Precision elliptical nozzles were obtained from
tension) Ronda SA, Lausen, Switzerland, and were made
by an electrocorrosion technique. The nozzle
The oscillating jet technique was used to measure radii were R 2 = 0.695 mm and R 1 = 0.515 mm
the dynamic surface tension of industrial surfac- (low ellipticity, used for surfactants in deionized
tants in aqueous and gelatin solutions on the water) and R 2 = 1.010mm and R 1 = 0.325mm
millisecond time scale. Jets were formed in front (high ellipticity, used for surfactants in 4%
of a millimeter scale. A photograph of the entire gelatin solutions). The high ellipticity, producing
continuous jet was taken. A bright lamp was larger initial oscillation amplitudes, was necessary
placed above the jet, in such a way that each for gelatin solutions because of the dampening
convex part of the jet would reflect the image of effect of viscosity.
the lamp. This method was slightly less accurate The liquid jet was supplied by a pressurized
than the optical front illumination method usually container. The jet was free of vibrations or other
reported (Defay and Petre 1971; Abe and instabilities that could have been produced by a
Matsumura 1983), but was very simple and pump. The flow rate was not measured, but the
convenient. Typical photos (enlargements of the horizontal velocity was determined directly on
initial part of the jets) are shown in Fig. 4.7. the photograph of the jet by fitting its trajectory
A) WATER at40·C
0.515 x 0.695 mm
nozzle radii
B) 4% GELATINE at 40·C
0.325 x 1.010 mm
nozzle radii
Figure 4.7 Typical photos (ca. 1/30 s) of the initial part of oscillating jets in the absence of surfactant. In the
presence of surfactant (not shown), the oscillation length increases along the jet as surface tension decreases:
(a) persistent oscillations due to the low viscosity of water at 40·C; (b) fast viscous dampening of the oscillations,
despite stronger initial amplitude as in case (a). The numbers on the scale are cm; the horizontal velocity of
the jets was ca. 2 m/s.
with the parabolic free-fall equation: tension values (up to 90 mN/m). The cause of the
error may lie in the strong distortion caused to
Y = a + bX + eX 2 (4.48) the jet by the high ellipticity, distortion which is
assumed small in equation (4.50). In that case, a
where Y and X are the vertical and horizontal
nozzle constant had to be determined in order
coordinates, respectively, and a, band c are the
to reconcile the surface tension of pure gelatin
variable coefficients to find from a fit with the
solutions with results from the inclined plane
measured trajectory. The only interesting one is
method described below (ca. 65 mN/m for short
c, which relates to the horizontal velocity Vx as
surface ages at 40°C). The nozzle constant value
follows:
was, after adjustment, 1.15 x to- 9 m 3 .
(4.49)
4.5.2.3 The inclined plane (dynamic surface
where g is the gravity constant. The uncertainty tension)
of Vx is estimated to be less than 1%.
Equation (4.37) (Section 4.4.4) can be modified The inclined plane method is an application of
by expressing the volumetric flow rate V* as a the Wilhelmy plate method to a flowing liquid,
function of the horizontal velocity with the plate oriented in the flow direction. It
Vx = V*/(na 2) = V*/(nR 2R 1 ), the parameters a has the same advantages of simplicity, ease of
and b as a function of R 2 and R 1 , and by operation and interpretation, and robustness as
r
separating the variable flow parameters: the Wilhelmy plate method. The inclined plane
had a slope of ca. 14° versus horizontal, was 1
(J = p(i ~n2(JR2Rl)3 meter long, 10 cm wide, and was attached to a
small two-slit liquid delivery device. The whole
+ (37/24)(R 2 - R 1 f/(4R 2 R 1»)
system was made of stainless steel and was
x (1 (4.50) temperature-controlled. The measured solution
1 + (5/3)(n2R2RdA.2) was circulated by a peristaltic pump, with a flow
or rate of ca. 3cm 3/s. Recently, the maximum flow
rate was increased up to ca. 16cm 3/s. The set-up
is represented in Fig. 4.8. Only the lower delivery
(4.51 ) slit was used. The Wilhemy plate used for most
of the reported results was 5 mm long, giving a
where K contains all the geometric terms of time resolution of 30 to 50 ms for the 3 cm 3/ S
equation (4.50), plus one term in r 2 • However, flow rate. The time resolution was estimated as
the dependence of K on A is weak. Variation of the time necessary for the flowing liquid to travel
surface tension from 70 to 30 mN/m corresponds one plate length (see calculation ofsurface velocity
to an increase of the wavelength A by ca. 50% below). The time resolution was recently improved
and a variation of K of less than 5% for low to ca. 6 to 8 ms by using a shorter 2 mm Wilhelmy
viscosity aqueous solutions. K is in essence a plate in conjunction with the higher flow rate.
'nozzle constant'. The plates, made of platinum, were cleaned by
With aqueous solutions and the low ellipticity briefly heating to glowing red in a flame. Either
nozzle, direct calculations of the surface tension of the plates was attached rigidly to a force
with equation (4.50) gave satisfactory values, transducer of ± 5 g capacity (Hottinger Baldwin
within a few percents for pure water. The Messtechnik) with the plate axis normal to the
corresponding nozzle constant K was liquid surface. The slight drag force of the liquid
1.27 x to- 9 m 3 . However, with data obtained had no influence on the 5 mm plate, and possibly
measuring gelatin solutions with the high ellipticity a small effect on the 2 mm plate, which was taken
nozzle, equation (4.50) yielded too high surface into account by a correction factor. The accuracy
Force transducer (moveable along the plane)
SIGNAL -----, I Wilhelmy Plate (5 or 2 mm long)

CONVERTER \ ~ ~
Measured solution (recirculated)
Figure 4.8 Experimental set-up for the inclined plane method of measuring dynamic surface tension.
ofthe dynamic surface tension measurement was where V. starts at zero. More accurate models
ca. 0.5 to 1.0 mN/m for the 5 mm plate and ca. have been described (Rillaerts and Joos 1982;
1.0-1.5 mN/m for the 2 mm plate. Calibration Ziller, Miller and Kretzschmar 1985; Wu and
was done by placing the transducer in a vertical Whitaker 1986), but were not deemed necessary
position and suspending a standard weight to for the comparative purpose of our experiments.
the plate, as for a static instrument. Viscosity was measured using a Haake CV-100
The surface age t at different positions of the viscometer, at shear rates between 0 and 1000 s- 1.
plane was calculated from the surface velocity Vs All measured solutions were Newtonian.
and the distance d from the slit exit (van den
Bogaert and Joos 1979):
4.5.3 STATIC SURFACE TENSION RESULTS
t= d/Vs (4.52)
The static surface tension of solutions of all
The surface velocity was measured experimentally surfactants listed in Table 4.1 have been measured
by following surface tracers and compared with in deionized water and 4% gelatin at 40°C using
the theoretical surface velocity (Bird, Stewart and the du Nouy ring technique. The data obtained
Lightfoot 1960): below the critical micelle concentration were
I f '_
t',
3
-_ [Q pg sin IXJ 1/3
2
(4.53)
fitted with equation (4.16) (von Szyszkowski
1908) to determine the values of r co and a. The
s 2 3J1.
results are summarized in Table 4.2.
where Q is the flow rate per unit width, p the The surface tension results are very different
liquid density, g the gravity constant, IX the in water and in gelatin solutions. The main
inclined plane angle (14°) and J1. the liquid viscosity. reason is that gelatin is itselfalready surface-active.
The experimental and theoretical values usually The static surface tension of 4% gelatin with no
agreed to within 10%. For the sake of simplicity, surfactant was 42-45 mN/m, while the static
we have assumed in reporting our results (see surface tension of water (at 40°C) was 69.6 mN/m.
Section 4.5.4) that the surface velocity V. was Therefore, the range of accessible surface tension
constant along the entire inclined plane. This is was much reduced in the gelatin solutions.
not strictly correct, in particular near the slit exit Gelatin is known to form complexes with some
Table 4.2 Parameters derived from the static surface tension data of the measured surfactants in pure water
and in 4% gelatin at 40·C
Max. ads. density Area per moleculett Langmuir constant 104 x CMC Min. surface
106 x r", (mollm 2 ) (A"2) 106 x a (molldm 3 ) (molldm 3 ) tension (mN1m)
Surfactant Water 4% gel. Water 4% gel. Water 4% gel. Water 4% gel. Water 4% gel.
Triton 770* 5.57 2.35 30 70 2.22 16.3 0.50 2.83 26.2 27.2
Texapon K-12* 5.11 0.96 32 173 152 13.1 47.1 106 25.1 29.8
3.98t 41t 297t 63.1H
Nekal BX* 2.63 0.71 63 235 50.6 10.8 83.8 9.55 32.0 36.5
AerosolOT* 2.45 1.41 67 117 11.0 6.97 58.6 4.33 26.5 30.6
1.30§ 126§ 5.0§ 31.6§
Empimin
LSM30* 3.16 0.93 53 177 16.2 35.0 14.2 9.74 32.5 35.1
Fenopon
CO-433* 6.87 1.25 24 132 3.55 4.47 0.30 2.11 29.0 33.1
Niaproof #4* 3.28 0.87 51 191 119 40.2 226 1160 26.1 26.0
Triton X-200* 2.82 3.21 59 52 0.16 103.0 4.04 5.99 27.5 27.6
Olin IOG** 7.84 1.90 21 87 1.92 10.1 0.13 1.39 28.8 30.0
Emcol 5130** 11.40 2.72 15 61 3.68 11.3 0.14 0.99 26.4 26.9
Igepal CA-720** 5.17 1.79 32 92 4.67 15.8 0.44 1.90 30.5 31.1
Lodyne S-103t 28.40 1.64 6 101 7.81 0.49 0.08 2.32 18.7 19.0
F1uorad
FX-IOO2t 7.78 3.72 21 45 2960 3510 336 358 19.0 19.5
* anionic (Triton 770 30% non-ionic) t from Van den Bogaert and Joos (1980), no specified temperature (probably
** non-ionic ambient)
t fluoro (also anionic) H from Knox and Parshall (1970), no specified temperature (probably ambient)
tt at maximum adsorption density § from Chang and Franses (1992), data taken at 25·C
surfactants, in particular anionic ones. Knox and highly branched hydrocarbon group (see Section
Parshall have studied the interaction of gelatin 4.5.1), while SDS is normally a straight C 12 chain.
with Na-dodecylsulfate or SDS (1970) and Aerosol This structural difference has a strong effect on
OT (t972). Interestingly, they found no interaction the micelle formation tendency and probably
with Triton X-tOO, a non-ionic surfactant (pre- also on the interaction with gelatin. A specific
sumably analogous to Triton X-t02). The SDS interaction between Niaproof # 4 and gelatin
or Aerosol OT surfactant-gelatin complexes cannot be demonstrated from Fig. 4.9. The only
appeared more surface-active than the monomeric visible effect is a compression of the surface
surfactants. The structure of complexes between tension range due to the reduced surface tension
SDS and other types of polymers has been of the gelatin solution. On the other hand, Emcol
studied using low-angle neutron scattering by 5130 (Fig. 4.10), a non-ionic surfactant, seems to
Cabane and Duplessix (1982, 1987). For other have a stronger interaction with gelatin than the
and more recent SDS-gelatin studies, see Griffith anionic Niaproof # 4, based on the distortion of
et al. (1996). Figures 4.9, 4.10 and 4.11 show the static surface tension curve. One should be
examples oflog (concentration) - surface tension cautious, therefore, in generalizing the finding of
curves for Niaproof #4 (anionic), Emcol 5130 Knox and Parshall (1970) stating that non-ionic
(non-ionic) and Lodyne S-103 (f1uoro-anionic surfactants do not interact with gelatin. In the
surfactant) in water and 4% gelatin solutions. case of Emcol 5130, some interaction may also
Niaproof # 4 is a Na-tetradecylsulfate, chemi- involve secondary components of the product
cally similar to Na-dodecylsulfate except for a (see Section 4.5.1).
70
E
.sZc 60
o
.~ 50
2
Ql
(.)
co
'§ 40
CJl
.2
co
(jj 30
20 L-..L.....J'-J....L.J..U..LL-..L.....J'-J....L.J..U..LL-..J....-L..J...l...I..U.IL-..J....-L..J...l...l..U.lcl....-...J......J
1~ 1~ 1~ 1~ 1~
Niaproof #4 concentration (mol/dm 3)
Figure 4.9 Static surface tension versus Niaproof # 4 concentration. (0): in water at 40·C and (e): in 4%
gelatin at 40·C. Lines: equation (4.16).
70
E
Z 60
.s
c
o
.~ 50
2
Ql
(.)
co
'§ 40
CJl
(.)
,\
~ \
Cii 30 \
\
\_----------~~~~
1~ 1~ 1~ 1~
Emcol5130 concentration (mol/dm 3 )
Figure 4.10 Static surface tension versus Emcol 5130 concentration. (0): in water at 40·C and (e): in 4%
gelatin at 4O.C. Lines: equation (4.16).
In the last example (Fig. 4.11), Texapon K-12 curve in the presence of gelatin as compared with
(SDS), an anionic surfactant, exhibits a very the curve in water, which is explained by the
strong distortion of the static surface tension formation of complexes with gelatin. The
--E.sz 60
c
o
.~ 50
2
Ql
al
'§ 40
rJ)
.2
ro
(jj 30
10-5 10-4 10-3 10-2 10-1

Texapon K-12 concentration (mol/dm 3 )
Figure 4.11 Static surface tension versus Texapon K-12 concentration. (0): in water at 40°C and (e): in 4%
gelatin at 40°C. Lines: equation (4.16).
presence of a mInImum in the static surface roo, explained by the presence of gelatin at the
tension curve indicates the presence of impur- interface.
ities (possibly dodecanol). In Table 4.2, the cmc An increase of surface tension above the cmc
were determined from the fit of the initial part region is a sign of surface active impurities. The
of the curves, at low surfactant concentration, surfactants other than Texapon K-12 which
with equation (4.16), using the regions of the appear to contain such impurities are Fenopon
curves that could give a precise fit. Then the CO-433 and Empimin LSM30, and, to a lesser
curve was extrapolated down to the lowest degree, Triton X-200 and Igepal CA-720. Only
measured surface tension. The concentration at Nekal BX has a sharp cmc transition followed
which the fitted curve was equal to the lowest by a constant surface tension at higher concen-
surface tension was taken as the cmc. Another trations. Triton 770, Triton X-200, Olin lOG,
choice could be to take the concentration at Emcol 5130 and Lodyne S-103 have a poorly
which the lowest surface tension is effectively defined cmc region, indicating a broad distribution
reached. Some literature values of surface tension of chain length or a complex mixture of com-
parameters were found for Na-decylsulfate pounds. Fluorad FX-I002, although showing a
(equivalent to Texapon K-12) and Aerosol OT rather well defined cmc transition in water,
and are indicated in Table 4.2. The values are exhibits more complex behavior in gelatin,
within 50% of each other, which is reasonable indicating an interaction of the surfactant with
considering the differences of measurement the gelatin molecules. Niaproof # 4 hardly reaches
temperatures and of sample purity. Because of a surface tension plateau up to 10- 1 moljdm 3 ,
the possibility of complex formation, the par- indicating a lack of micelle formation, which
ameters found in the presence of gelatin may could be the consequence ofits bulky hydrocarbon
only be apparent values and have to be interpreted group. Triton 770 reveals a simpler behavior in
with caution. A probably real effect is, however, gelatin than in water, but this could also indicate
a lowering of the maximum adsorption density some interaction with the gelatin molecules.
4.5.4 DYNAMIC SURFACE TENSION RESULTS surface tension becomes equal to the static
surface tension at long surface ages.
All the surfactants listed in Table 4.1 have been Without using any equation to fit the data,
measured at several concentrations in water and the experimental results already describe well the
4% gelatin, both at 40·C, with the oscillating jet dynamic behavior of the surfactant (in the case
and the inclined plane methods. The concentra- of Fig. 4.12, Olin lOG). They also demonstrate
tions were chosen to include what are considered that the gelatin solution without surfactant
practical concentrations in coating operations. maintains a dynamic surface tension higher than
An example of experimental results (Olin lOG 60 mN/m for several seconds, even though its
in 4% gelatin) is shown in Fig. 4.12. The oscillating static surface tension is 42 to 45 mN/m. Even
jet data points are in the range 5 to 30 ms of more remarkable is the fact that all measured
surface age, and the inclined plane data points Olin lOG concentrations produce nearly the
in the range 60 ms to 4 s of surface age. These same static surface tension (limit of the curves
measurements were made before the inclined in Fig. 4.12 at long surface ages). This proves
plane was modified to permit measurement at clearly, at least in the case of Olin lOG, that
less than 10 ms of surface age (examples of newer measuring only static surface tension would miss
data are given in Section 4.5.6). The combination most of the information and could not predict
of the two methods, oscillating jet and inclined the dynamic behavior of a solution containing
plane, is useful because each technique can be this surfactant.
used to confirm the other. The data in Fig. 4.12 A lot more can be extracted from the data if
show that the results of both methods are very they can be fitted with suitable models. In Fig.
consistent, and therefore reliable. A further check 4.12, Olin lOG data in 4% gelatin follow rather
can be made by verifying that the dynamic well the models of equations (4.22) and (4.23),
70
E
-..
z 60
.s
c:
0
'(i)
c: 50
2
Q)
()
eu 40
't:
::J
Ul
()
'E 30
eu
c:
>-
0
20
0.0001 0.001 0.01 0.1 1 10 100
Surface age [s]
Figure 4.12 Dynamic surface tension of Olin lOG in 4% gelatin at 40°C. Lines: equations (4.22) and (4.23).
(*): no surfactant; (0): 2.00 x 1O- 4 moljdm 3 ; ( . ) : 3.67 x 1O- 4 mol/dm 3 ; (6): 5.24 x 1O- 4 mol/dm 3 ; (e):
1.05 x 1O- 3 moljdm 3 ; (0): 2.10 x 1O- 3 mol/dm 3 ; (A): 5.24 x 1O- 3 moljdm 3 .
which are based on diffusion-controlled adsorp- surface ages within reasonable limits of extrapo-
tion. The diffusion coefficient of the surfactant lation, without further measurement.
can be extracted from these fits (see Figs 4.14 There are cases when equations (4.22) and
and 4.15), and the good quality of the fits (4.23) cannot fit the data in a satisfactory manner.
indicates that the surfactant does not form This is shown, for instance, by the dashed lines
complexes with gelatin and has no adsorption in Fig. 4. t 3, for Empimin LSM30 in gelatin
barrier. From a more practical point of view, the solution. The flexibility of the fits with equations
successful use of equations allows one to predict (4.22) and (4.23) is limited, because the surface
the dynamic surface tension as a function of tension at zero surface age is determined by the
surface age for intermediate concentrations of solvent (69.6 mN/m for water at 40'C or 65 mN/m
surfactant, using linear interpolation, for instance. as the short-time dynamic surface tension of
To some extent, it allows one also to extrapolate gelatin at 40'C without surfactant). In addition,
the dynamic surface tension at shorter surface the parameters derived from static surface tension
ages (also at longer surface ages, but it is less data (r 00 and a), which are used in equations
useful). The surface age scale in Fig. 4.12 extends (4.22) and (4.23), cannot be varied. The applicability
to 0.1 ms, a surface age found in the pinned of these parameters is doubtful in the presence
meniscus in slide coating (Fig. 4.3), which is not of gelatin, as gelatin is much more surface-active
experimentally accessible. However, the extrapo- at long surface ages (static situation) than at short
lated dynamic surface tension curves are plausible, surface ages. Therefore, equations (4.22) and
considering the general quality of the fits. Based (4.23) are not always able to fit the dynamic
on the data in Fig. 4.12, it is possible to predict surface tension data, and even if they do, the
the dynamic surface tension of the measured diffusion coefficients derived in the presence of
solution at any surfactant concentrations and gelatin may only have apparent values. In the
70
'E
.-..
'z 60
'E
,~
c:
o
"en
c:
2 50
Q)
~
"t:
:J
l/l
u 40
'"E
C1l
c:
>-
10
30 L....J.--I....Wu..wJ..----'-J....l...I.J.u;l-..J...J..~JJ_L...L.L.......,.,,--'--'---'-UUllL___'_'..LilUlJ
0.0001 0.001 0.01 0.1 10 100

Surface a~e JSJ
Figure 4.13 Dynamic surface tension of Empimin LSM30 in 4% gelatin at 40°C. Dashed lines: equation
(4.22) and (4.23); solid lines: equation (4.24). (.): 5.79 x 10- 4 mOl/dm 3; (6): 1.45 x 10- 3 mOl/dm 3; (.):
2.90 x 10- 3 mol/dm 3; (0): 1.45 x 10 - 2 mOl/dm 3 .
case of Empimin LSM30, surfactant-gelatin 'poor' (even 'very poor' for Texapon K-12) and
complexes are probably formed, as is known for can be found in Table 4.3. 'Good' means that
SDS and Aerosol OT, and/or its adsorption is the derived parameters can accurately predict the
not entirely diffusion-controlled. Chang and data (although diffusion coefficients in gelatin
Franses (1992) have developed models taking should all be considered as apparent values),
into account concentration-dependent adsorp- while 'poor' means that the given values should
tion/desorption barriers. Surfactant-gelatin com- be taken with a pinch of salt. In general, it was
plexes would require new and more elaborate easier to fit the data with the parametric equation
models. (4.24) than with equations (4.22) and (4.23), for
More pragmatically, the parametric equation the reasons explained above.
(4.24) can be used to obtain a better fit of the A comparison of the short-time diffusion coef-
data (solid lines in Fig. 4.13). It does not give ficients of the above surfactants is shown in Fig.
any molecular information about the surfactant, 4.14, at three typical concentrations. The fits obtained
neither does it describe the mode of adsorption, with equations (4.22) and (4.23) for Texapon K-12
but it still allows prediction of surface tension were so poor, even in water, that no diffusion
over a wide range of surface ages and surfactant coefficient could be reasonably estimated. In water,
concentrations (see Section 4.5.6). The extrapolated Na-dodecylsulfate or SDS (active substance of
dynamic surface tensions at very small surface Texapon K-12) has been found to exhibit an
ages are, however, no longer realistic, because adsorption barrier (Chang and Franses 1992), and
the parameter (To is freely adjusted to obtain the SDS is known to interact strongly with gelatin
best fit. (Knox and Parshall 1970 and 1972). In fact, it is
However, equation (4.24) did not always give unfortunate that this surfactant has been so popular
very good fits of the data, probably because some as 'standard' anionic surfactant (see also van den
surfactants produced more experimental errors Bogaert and Joos 1980; Weinheimer et al. 1981;
than others. As an approximative guide, the fit Hempt et al. 1985; Chari and Hossain 1991)
quality of all measured surfactants with both because it is one of the most peculiar as far as
types of equation has been rated 'good', 'fair' or surface properties are concerned.
Table 4.3 Fit quality of the dynamic surface tension data with equations (4.22) and (4.23) (Hansen's
approximations, diffusion model) and with equation (4.24) (parametric equation) of the measured surfactants
Equations (4.22) and (4.23) Equation (4.24)
Surfactant Water 4O·C 4% gel.40·C Water 4O·C 4% gel.40·C
Triton 770 good good good good

Texapon K-12 very poor very poor good fair
Nekal BX fair good fair good
AerosolOT good fair fair good
Empimin LSM30 poor poor fair good
Fenopon CO-433 poor good good good
Niaproof #4 fair fair fair good
Triton X-200 fair poor good good
Olin lOG good good good good
Emcol5130 good good good good
Igepal CA-720 good fair good good
Lodyne S-103 good good good good
Fluorad FX-lOO2 fair poor fair fair
It can be seen in Fig. 4.14 that the diffusion sion measurements using equations (4.22) and
coefficients are not concentration-independent. (4.23).
It is not surprising, considering that the surfactants As mentioned previously, the presence ofgelatin
are not pure substances, and that several of them has a profound effect on the dynamic surface
form micelles in the range of concentrations tension, either through the formation of surfac-
indicated in Fig. 4.14. For similar reasons, the tant-gelatin complexes, or simply through an
diffusion coefficient found for the long-time increase of viscosity. The ratios of the diffusion
approximation was often different from the one coefficients (apparent values) found in gelatin
found for the short-time approximation. However, and in water are reported in Fig. 4.15. The
the long-time domain is less critical for coating anionic surfactants, in particular Niaproof # 4,
operations and detailed results are not given Nekal BX and Aerosol OT, exhibit larger values
here. The surfactants have been ranked according in gelatin solutions than in pure water. Non-ionic
to the value ofthe short-time diffusion coefficient surfactants show the opposite effect. Igepal CA-
found at 10- 3 moljdm 3 . Few diffusion coefficients 720, the fastest of the group at 10- 3 moljdm 3 in
for the above surfactants have been reported. water, is much slower in gelatin solutions. The
The diffusion coefficient ofa straight-chain isomer fluorosurfactants, although anionic, behave like
of Aerosol OT has been reported to be non-ionic surfactants towards gelatin.
9.6 x 10- 10 m 2 Is from conductivity measurements The effects seen on the short-time diffusion
and 2.7 x 10- 10 m 2/sfrom dynamic surface tension coefficients are confirmed by the variations of
methods at 7.5 x 10- 4 moljdm 3 (Serrien and the parameter t* used in equation (4.24) (surface
Joos 1990). These values are reasonably close to age at the inflexion point of the curves). Figure
the value of 12 x 1O- 10 m 2/s shown in Fig. 4.14 4.16 shows the t* values of the series of surfactants
at 6 x 1O- 4 moljdm 3 Aerosol OT. The diffusion in water and Fig. 4.17 the ratios of the values in
coefficient of pure SDS is usually taken as water and in gelatin. The ratios were taken in
4 x 10- 10 m 2 Is (see above references), but we the other direction, so that as for Fig. 4.15, a
cannot determine it from dynamic surface ten- ratio larger than one would mean aJaster action
60 . - - - - - - - - - - - - , - - - - - - - - - - - ,
50
3
40
30 2
20
Figure 4.14 Short-time diffusion coefficients of surfactants in water at 40°C. (III): 6 x 1O- 4 mol/dm 3 ; (III):
10- 3 mol/dm 3 ;(III): 3 x 10- 3 mol/dm 3 . The surfactants are ranked according to the value found at 10- 3 mol/dm 3 .
2.0,--------------------,
1.5
1.0
0.5
·0.5
Figure 4.15 Ratios of the short-time diffusion coefficients of surfactants in 4% gelatin and in water. Positive
log values mean faster diffusion in gelatin solution than in water. (.): 6 x 10- 4 mol/dm 3 ; (.): 10- 3 mol/dm 3 ;
(.): 3 x 10- 3 moljdm 3 •
in gelatin than in water. The trends are the same t. the speed of action of the surfactant as a
as for the diffusion coefficients, but the absolute function of its concentration, which can be
values of the ratios are larger. The parameter t* judged from the diffusion coefficients D and
is expected to decrease at higher surfactant the characteristic times t*;
concentration as it measures the direct effectiveness 2. the lowest attainable surface tension, defining
of the solution to reduce surface tension and is the wetting limit of the surfactant, its ability
not a theoretically concentration-independent to spread on surfaces and 'outwet' hydrophobic
parameter as the diffusion coefficient should be. particles, which can be seen from the static data;
The parameter t* effectively decreases with 3. the range ofconcentrations producing a change
increasing surfactant concentrations in all cases ofstatic surface tension, determining the degree
except for Texapon K-12 (SDS) in gelatin (not of control of the static surface tension by
directly shown; can be seen from the combined adjustment of the surfactant concentration,
Figs 4.16 and 4.17). Again, SDS exhibits a also seen in the static data;
peculiar behavior in the form of a very strong 4. 'the solubilizing power for hydrophobic mol-
concentration dependence in the parameter t*. ecules, which will determine the ability of the
surfactant to stabilize insoluble substances in
solution, and which will be efficient only
4.5.5 SURFACTANT SELECTION CRITERIA above the cmc (static data).
The basic properties of surfactants and their de-
sired functions have already been discussed in Sec- The static surface tension data are therefore
tion 4.2. However, they appear under a new light important, as three out of the four above criteria
when actual results of dynamic surface tension can be judged from them. However, the speed
are known. Based on the presented experimental criteria is essential for fast coating operations
results, it is possible to select which type of and for this kind of application has at least as
surfactant would be appropriate for a particular much weight as the three others combined. It
purpose. Some criteria to be considered are: would be rather arbitrary to establish a definitive
10000
0
Ul
oS
1000 • 0
0
,g
Ql
• • 0 0
0 0
<>
100 l:>
• • • • 0
~
.~
U 10
l:>
l:> l:> •A
-
0
0 0
~
ell
.J::.
l:>
l:>
D-
l:>
• • •
<> l:>
A
~
l:>
l:>
I I I
Figure 4.16 Comparison of the characteristic surface ages t* of surfactants in water at 40°C. (0):
6 x 10 - 4 mol/dm 3; (.): 10- 3 mol/dm 3; (.6.): 3 x 10 - 3 mol/dm 3.
2.5 r-------------------,
2
1.5
0.5
o
-0.5
-1
Figure 4.17 Ratios of the characteristic times t* of surfactants in water and in 4% gelatin. Positive log values
mean smal1er t* (faster surfactant) in gelatin solution than in water. (.): 6 x 10- 4 mOl/dm 3 ; ( . ) : 10- 3 mol/dm 3 ;
(.): 3 x 10- 3 mol/dm 3 .
ranking of the surfactants whose experimental addition, the group of surfactants chosen in this
data are presented in the previous sections. Based chapter is very limited and finding a 'best one'
on the present data, it should be fairly easy for among them would not mean a lot. If direct
the reader to make his or her own judgment. In experimental data cannot be obtained from other
surfactants, some analogies could be made using ofthe surfactant concentration for different bubble
the molecular structures shown in Section 4.5.1. rates. As explained in Section 4.4.6, the bubble
It is unlikely that one single surfactant will rate in the maximum bubble pressure apparatus
perform best for all the required criteria under does not give precise information of the effective
one set ofconditions. There is no 'perfect' surfactant. surface age at which the dynamic surface tension
Therefore, usually two or more surfactants are is measured. The bubble rate is only a qualitative
used. Some of the best properties can be combined parameter which cannot be easily related to the
to match the criteria. However, using mixtures physics of the process. The recent maximum
complicates the optimization process and the bubble pressure instrument developed by Lauda
number ofsurfactants should be kept to a minimum. (see Section 4.4.6) has the capability to determine
If a surfactant is needed for testing flow the effective surface age, but it was not available
simulation studies by numerical methods (see to Valentini et at. (1991) and may still have
Chapter lid), in particular when the effects of difficulties measuring real coating solutions
surfactant convection and diffusion on surface containing large amounts of dispersed colloids.
tension are to be taken into account, the best Dynamic surface tension measurements have
surfactant to use should be one that follows been done in our lab with actual photographic
simple diffusion models, such as Olin lOG, or solutions, using the inclined plane method in
another non-ionic surfactant. It appears from the which the shortest accessible surface age has been
present data that several anionic surfactants recently reduced to ca. 10ms. As an example, a
behave in a more complex manner, in particular three-layer package has been measured over a
Na-dodecylsulfate (SDS). wide range of surfactant concentrations. This
package, designed to be coated with the curtain-
coating technique (see Chapter llc), was plagued
4.5.6 SURFACTANT MULTILAYER RECIPE
by curtain instabilities which were attributed to
OPTIMIZATION
wrongly matched interlayer wetting properties.
The transition from surface tension measurements For proprietary reasons, details of the coating
to a successful coating formula is not an easy solutions and of the surfactants used are not
task. The reality of coating involves many given, but this information is not necessary for
interacting parameters and the applicability of the purpose of the demonstration.
the dynamic surface tension data must be proven The aim ofthe work was to find a combination
by practical tests. A good illustration of this of surfactant concentrations for each layer (a
concept has been given by Valentini et at. (1991). surfactant recipe) which would result in stable
These authors measured dynamic surface tensions interlayer wetting, in particular at the surface age
using a maximum bubble pressure apparatus and time scalefound in the curtain itself (ca. 10-100 ms).
correlated the data with the accessible coating The general principle on which the optimization
window. The main message of their work is based is directly derived from Young's equation
confirmed that it is the dynamic, not the static (equation (4.1»: a liquid layer must have a surface
surface tension which is the relevant parameter tension as low or lower than the surface tension
to measure. Similar conclusions have been reached ofthe layer it is supposed to spread on. The point
by Hirt et at. (1990). In addition, Valentini et at. is that it must be true already when the surfaces
(1991) found that surface elasticity was stabilizing are very fresh, down to the millisecond time scale.
the flowing film against waves, confirming the It means that when both layers are suddenly
work of Ruschak (1987). exposed to air, having both an initial surface
Improvements from the work of Valentini et tension close to the solvent value, the one that
at. (1991) could be made in the area of dynamic is supposed to spread on the other must be able
surface tension measurements. These authors to reduce its surface tension faster than the other.
reported dynamic surface tension as a function The top layer of a multilayer coating package is
always exposed to air, and to possible external will be numbered from 1 to 3 starting at the back
disturbances. It is the most vulnerable layer. of the curtain. The necessary gradient of surface
Therefore, this layer must be able to reduce its tension required to ensure stable wetting has not
surface tension very shortly after it emerges from a definite value, but a few mN/m is generally
the slit of the coating device. The requirement considered sufficient (Gutoff 1992). On the other
for the surface tension or potential surface tension hand, a wrong gradient of the same order of
reduction speed of internal layers is not as magnitude can easily promote dewetting.
critical, because these layers are generally not Knowing what surface tension relation is
exposed to air and they do not sustain interfacial desired between the layers of the coating package,
tension, provided they are miscible. However, a the optimization procedure using surface tension
too large difference in surface tension between data is done as follows:
internal layers can have undesirable edge effects
(Gutoff 1992). In practice, it is customary to 1. The static surface tension ofthe three solutions
maintain a similar or slightly increasing wetting is measured as a function of the surfactant
tendency from the lower to the top layer in order concentration (data not shown). This should
to facilitate at start-up the formation of the match the long-time dynamic surface tension
multilayer package on a slide coater. and gives a general picture of the accessible
In the curtain coating operation, the back of range of static surface tensions.
the curtain is exposed to air during a significant 2. The dynamic surface tension of each solution
time (ca. 50 to 100 ms). Therefore, the layer at is measured on the inclined plane as a function
the back of the curtain must be considered, at of surface age for several surfactant concen-
least temporarily, as a surface layer. In our trations. Actual data for layer 1 are shown in
example, there is only one internal layer (total Fig. 4.18. The accessible surface age range was
three layers), and this layer must be easy to wet 7 ms to 3 s, using the 2 mm-Iong Wilhelmy
by the two others. For convenience, the layers plate and a flow rate of ca. 16 cm 3 Is.
60,------------------,
E
--
z
.s
55 ====- u--.r-.- ""- ~-----
(/)
Q)
~
(ij
c 50 -_ >
o ---~ .... ....
.2
'00 .....'2'1 iii
c
2 45 Ci5
Q) )(
()
<ll 40 -s.-. -0..____ 0
o
.
't:
-.-
~
(/)
() 35
'E -....
<ll
C
>..
o
30 .-.-.-. ----
-.--.- -----.
25 L..--'--'--'-'-'.LU.L--'-,-"'-L.1..1.uL--'---'--'-.<..U-L'-'---------'--''-'-'-.L..U..l.l----J
0.001 0.01 0.1 10

Surface age [5]
Figure 4.18 Dynamic surface tension for several surfactant concentrations of layer 1 in a three-layer
package (inclined plane). (*): 0; (0): 2; (0): 5; (L0.): 10; (.): 20; (.): 50 (arbitrary units). The lines are fits with
equation (4.24).
3. The experimental data are fitted with equation

(4.24) and provide a set of the four parameters ~ 50
of equation (4.24) at each surfactant concen- .sc
o
tration. Then by using a simple computer 'iii
c
45
worksheet, it is possible to determine these 2
parameters as a function of the surfactant ~
ell
't:
40
concentration only, anywhere within or near :::3
----------------------- .::::::::;;;;;~.-...
(/)
the measured concentration range. A simple o

'E 35
and safe method consists of using linear ell
c
>.
interpolation. A graph showing the dynamic o 0.001 0.01 0.1 10
surface tension versus surface ages according
Surface age [s)
to equation (4.24) can be generated from the
four parameters and modified automatically
every time a new surfactant concentration is Figure 4.19 Recalculated dynamic surface tension
entered. (equation (4.24)) corresponding to an inadequate
4. A similar automatically generated curve can surfactant recipe: (--): layer 1 with conc. 10; (-----):
layer 2 with conc. 8; C············): layer 3 with conc. 30
be established for each coating solution and (arbitrary units).
combined in the same graph as the others.
The combined graph will show automatically
the relations between the dynamic surface coating and could be susceptible to repellency
tension versus surface ages of the three coating holes if hydrophobic particles reach its surface.
solutions together, and can be modified just After trying a few other surfactant recipes, one
by entering a new surfactant recipe. The finally arrives at the situation shown in Fig. 4.20.
surfactant concentrations are then empirically The internal layer (layer 2) has no more surfactant,
changed until the required criteria for the causing its surface tension to remain above the
relative dynamic surface tensions of the three surface tension of both outside layers of the
coated layers are satisfied. curtain (layers 1and 3). The amount of surfactant
The dynamic surface tensions curves corre-

sponding to the current surfactant recipe, which 45 , - - - - - - - - - - - - - - - - - - - - - - - ,
was causing curtain instabilities, is shown in Fig.
~
4.19. It can be seen that the required wetting .sc ---------------------------------------------------------_.
stability criteria were violated on both sides of o 40
'iii
the curtain. In particular, the layer 1 has a higher c
.2!
dynamic surface tension than layer 2 (internal
layer) up to ca. 1 s of surface age. This situation ~ :::3
35
could cause the internal layer to repel layer 1 if (/)
o
any particle could cause a temporary rupture of "E
ell
c: 30
layer 1. The curtain layers are flowing along edge o
>. '--~~""_____~...LW.'_U.L.____'_~.LU..<L__'___'__'_.L.U..U.L_J
guides on each lateral end of the curtain. Layer 0.001 0.01 0.1 10
2 has more tendency to wet the edge guide than Surface age [s)
layer 1, because its surface tension is lower, and
could have a tendency to repel layer 1 from the
Figure 4.20 Recalculated dynamic surface tension
edge guide (Gutoff 1992). There is also a wrong (equation (4.24)) corresponding to an improved
gradient between layer 3 and layer 2, at least surfactant recipe. (-): layer 1 with conc.18; (- •. _): layer
during the first second of surface aging. As a 2 with conc. 0; ( ): layer with conc. 100 (arbitrary
result, layer 3 may retract from the edge of the units).
Summary and conclusions 133
in layer 1 and especially in layer 3 has been at very short surface ages and others are better
increased to ensure quick surface tension reduction. at controlling near-equilibrium conditions.
The final surface tension for layer 3 is not very Characterization of surfactants through their
low, even with a relatively high surfactant static and dynamic surface tension brings a
concentration. If a lower surface tension is found detailed understanding of their behaviour in real
necessary after practical test coatings, another processes. Dynamic surface tensions can be
surfactant should be used. Experiments on pilot estimated at surface ages shorter than can actually
coating machines have shown that the curtain be measured - down to the submillisecond time
stability responded as expected to the relative scale - by simulating experimental data with
surface tension gradients between its layers. appropriate models. Practical coating tests and
The lower dynamic surface tension of layer 1 sometimes empirical corrections are still necessary,
versus layer 2 could produce some difficulty for but fewer are needed and they can be made with
layer 2 to wet layer 1 in the slide flow region greater confidence when the fundamental surfac-
proceeding the curtain. However, the main wetting tant behavior is known. Surfactants can be
requirement for the slide flow stability is that the qualified based on their effective static and
top layer (in this case layer 3) is able to wet the dynamic properties and not on arbitrary char-
next layer (in this case layer 2), a requirement acteristics.
which is fulfilled in the above example. The This chapter has not directly considered
dynamic surface tension relationship between dynamic flow situations, but focused on diffusion
internal layers is less critical, as discussed above. and adsorption processes near the interface.
Therefore, the use of dimensionless numbers,
very common in flow characterization, was not
found useful. The experimental methods used,
4.6 SUMMARY AND CONCLUSIONS
the oscillating jet and the inclined plane, have
This chapter has reviewed the basic aspects of very simple flow characteristics which were not
surfactants, their applications to liquid film coating taken into account in discussing the results. The
processes, and general concepts related to purpose of the work was to demonstrate how
surfactants and wetting. The main focus of this static and dynamic surface tension can be measured
work is the determination and use of static and and used to optimize the dynamic wetting behavior
dynamic surface tension. An overview of the of multiple layers in liquid film coating. When
major experimental methods able to determine diffusion coefficients could be determined, their
static and/or dynamic surface tension have been value can be used as input parameter in numerical
given. flow studies which consider surfactants and the
Two experimental techniques, the oscillating influence of their diffusion (Chapter lId; Barthod
jet and the inclined plane methods, were used to 1993).
characterize a group of industrial surfactants by Despite the detailed information already gained
their static and dynamic surface tensions. The from surface tension data, and the fact that it is
measurements were made under conditions where a 'mature' science, one should not believe that
the surfaces ofthe solutions were instantaneously there is nothing more to learn. In fact, we are
created and not subjected to significant expansion only beginning to tackle more real and complex
afterwards. Industrial surfactants were chosen, situations. For instance, a lot of current research
despite their rather undefined purity, because activity is focused on surfactant interactions with
they represent what is actually used by manu- other surfactants (Siddiqui and Franses, 1996),
facturers involved with aqueous liquid films with latex particles (Bolze, Horner and Ballauff,
coating. It is apparent from the reported results 1996), silica particles (Goloub et al. 1996), polymers
that the studied surfactants differ very widely (Holmberg and SundelOf, 1996; De Oliveira,
from each other. For instance, some are active Tiera and Neumann, 1996) or lipsomes (de la
Maza and Parra, 1996). There is a need to Becher, P. ed. 1985. Encyclopedia ofEmulsion Technology,
understand the behavior of surfactants in more Volume 2, Applications. New York: Marcel Dekker.
and more complex systems. However, the kind Becher, P. and Griffin, W. C. 1985. In McCutcheon's
Emulsifiers and Detergents. McCutcheon Division,
of complexity encountered in actual coating
Glen Rock, NJ: MC Publishing.
solutions, such as in the photographic industry, Bendure, R. L. 1971. Dynamic surface tension deter-
will require direct and pragmatic experimentation mination with the maximum bubble pressure method.
for a long time still. Fundamental research J. Coll. Interf Sci. 35: 238-248.
should be balanced and completed with more Bergink-Martens, D. J. M., Bos, H. J., Prins, A. and
applied investigations. Schulte, B. C. 1990. Surface dilation and fluid-
dynamical behavior of Newtonian liquids in an
overflowing cylinder. I. Pure liquids. J. Coll. Interf
ACKNOWLEDGMENTS Sci. 138: 1-9.
The author gratefully thanks Professor P. Joos, Bird, R B., Stewart, W. E. and Lightfoot, E. N. 1960.
Transport Phenomena. New York: Wiley.
University of Antwerpen, Belgium, Professor A. Blake, T. D. 1984. Wetting. In Surfactants, ed. Th.F.
Prins, University ofWageningen, The Netherlands, Tadros, pp. 221-275. London: Academic Press.
and Dr R. Traber, Ciba-Geigy, Basel, Switzerland, Van den Bogaert, Rand Joos, P. 1979. Dynamic
for stimulating discussions and advice on selecting surface tension of sodium myristate solutions. J.
dynamic surface tension measuring methods; Ms Phys. Chem. 83: 2244-2248.
S. Buser, Mr R. Carfagno, Mr M. Pasquier and Van den Bogaert, Rand Joos, P. 1980. Diffusion-
particularly Mr L. Casarico, Ilford AG, for their controlled adsorption kinetics for a mixture of
surface active agents at the solution-air interface.
essential contributions to the development of the J. Phys. Chem. 84: 190-194.
methods and for obtaining experimental results; Bolze, J., Horner, K. D. and Ballauff, M. 1996.
and Dr E. Kramp, Ilford AG, for initiating this Adsorption ofthe nonionic surfactant Triton X-405
work. on polystyrene latex particles as monitored by
small-angle X-ray scattering. Langmuir. 12:
2906-2912.
REFERENCES
Brown, D. R 1961. J. Fluid Mech. 10: 297.
Abe, Y. and Matsumura, S. 1983. The relationship Chang, C. H. and Franses, E. I. 1992. Modified
between defoaming power and dynamic surface Langmuir-Hinshelwood kinetics for dynamic ad-
tension ofsodium dodecylbenzene sulfonate solutions sorption of surfactants at the air/water interface.
with alcohols. Tenside Detergents, 218-224. Coll. and Surf 69: 189-201.
Adamson, A. W. 1967. Physical Chemistry ofSurfaces, Chang, C. H. and Franses, E. I. 1994. Adsorption
2nd edn. New York: Wiley. dynamics of surfactants at the air/water interface:
Ainsworth, S. J. 1993. Cosmetics, Product Report. a critical review of mathematical models, data and
Chem. & Eng. News. 26 April: 36-48. mechanisms. Coll. and Surf 100: 1-45.
Ambwani, D. S. and Fort, Jr. T. 1979. Pendent drop Chari, K. and Hossain, T. Z. 1991. Adsorption at the
technique for measuring liquid boundary tensions. air/water interface from an aqueous solution of
Surf Colloid Sci. 11: 93-119. poly(vinylpyrolidone) and sodium dodecyl sulfate.
Aniansson, E. A. G., Wall, S. N., Almgren, M., J. Phys. Chem. 95: 3302-3305.
Hoffmann, H., Kielmann, I., Ulbricht, W., Zana, R, Defay, Rand Hommelen, J. 1958. Ind. Chim Beige.
Lang, J. and Tondre, C. 1976. J. Phys. Chem. 80: 905. 23: 597.
Attwood, D. and Florence, A. T. 1983. Surfactant Defay, R and Petre, G. 1971. Dynamic surface tension.
Systems, Their Chemistry, Pharmacy, and Biology. In Surf Coll. Sci. 3, ed. E. Matijevic, pp. 27-81.
New York: Chapman and Hall. New York: Marcel Dekker.
Balbaert, I., Bleys, G. and Joos, P. 1987. Measurement Diaz Garcia, M. E. and Sanz-Medel, A. 1986. Dye-
of the dynamic surface tension in an asymmetric surfactant interactions: A review. Talanta. 33: 255-264.
free-falling film. J. Coll. Interf Sci. 115: 362-371. Fainerman, V. B. 1992. Adsorption kinetics from
Barthod, D. 1993. Contribution aI'etude de I'ecoulement concentrated micellar solutions of ionic surfactants
d'un rideau symetrique (avec ou sans tensio-actifs). at the water-air interface. Coll. and Surf 62: 333-347.
PhD thesis, Ecole Centrale de Lyon, France. Fainerman, V. B., Miller, Rand Joos, P. 1994. The
References 135
measurement of dynamic surface tension by the factant-polymer interaction in solution. The

maximum bubble pressure method. Coil. Polym. Sci. EHECjSDS/water system. Langmuir. 12: 883-889.
272: 731. Hua, X. Y. and Rosen, M. J. 1988. Dynamic surface
Fainerman, V. B. and Miller, R. 1995. Dynamic surface tension of aqueous surfactant solutions I. Basic
tension measurements in the sub-millisecond range. parameters. J. Coil. Inter! Sci. 124: 652-659.
J. Coil. Inter! Sci. 175: 118-121. Ishiwata, M., Kawame, N., Sakurai, T., Nakano, M.,
Frumkin, A. 1925. Die Kapillarkurve der hoheren Tamazawa, K., Kimura, N., Saito, T. and Hada, G.
Fettsiiuren und die Zustandsgleichung der Ober- 1990. Process for producing photographic materials.
fliichenschicht. Z. Phys. Chem. Stoechiom. Verwan- EU patent application 0383347.
dschaftsl. 116: 466-484. Israelachvili,1. N., Marcelja, S. and Horn, R. G. 1980.
Garrett, P. R. and Ward, D. R. 1989. A reexamination Physical principles ofmembrane organization. Quart.
of the measurement of dynamic surface tension Rev. Biophys. 13: 121-200.
using the maximum bubble pressure method. J. Janczuk, B., Bialopiotrowicz, T. and Wojcik, W. 1989.
Coil. Inter! Sci. 132: 475-490. The components of surface tension of liquids and
Goloub, T. P., Koopal, L. K. and Bijsterbosch, B. H. their usefulness in determination of surface free
1996. Adsorption of cationic surfactants on silica. energy of solids. J. Coil. Inter! Sci. 127: 59-66.
Surface charge effects. Langmuir. 12: 3188-3194. Joos, P. and de Keyser, P. 1980. The overflowing
Griffin, W. C. 1949. Classification of HLB values of funnel as a method for measuring surface dilational
non-ionic surfactants. J. Soc. Cosmet. Chem. 1: properties. Levich Birthday Conference, Madrid.
311-326. Kalhweit, M., Busse, G. and Jen, J. 1991. Adsorption
Griffith, P. c., Stilbs, P., Howe, A. M. and Cosgrove, of amphiphiles at water/air interfaces. J. Phys.
T. 1996. A self-diffusion study ofthe complex formed Chem. 95: 5580-5586.
by sodium dodecyl sulfate and gelatin in aqueous Knox, W. J. 1970. Photographic surfactant composi-
solutions. Langmuir. 12: 2884-2893. tions. US patent 3,514,213.
Gutoff, E. B. 1992. Premetered coating. In Modern Knox, W. 1. and Parshall, T. O. 1970. The interaction
Coating and Drying Technology, eds E. Cohen and ofsodium dodecyl sulfate with gelatin. J. Coil. Inter!
E. B. Gutoff, pp. 117-167. New York: VCH Sci. 33: 16-23.
Publishers. Knox, W. 1. and Parshall, T. O. 1972. The interaction
Hancock, R. I. 1984. Macromolecular surfactants. In of Aerosol OT with gelatin. J. Coil. Inter! Sci. 40:
Surfactants, ed. ThF. Tadros, pp. 287 - 321. London: 290-296.
Academic Press. Lopez de Ramos, A., Redner, R. A. and Cerro, R. L.
Hansen, R. S. 1960. The theory of diffusion-controlled 1993. Surface tension from pendant drop curvature.
adsorption kinetics with accompanying evaporation. Langmuir. 9: 3691-3694.
J. Phys. Chem. 64: 637. Lucassen, J. 1975. Adsorption kinetics in micellar
Harkins, W. D. and Jordan, H. F. 1930. J. Am. Chem. systems. Disc. Faraday. 59: 76-87.
Soc. 52: 1571. Lucassen, J. and Giles, D. 1975. Dynamic surface
van Havenberg, J. and Joos, P. 1983. The dynamic properties of nonionic surfactant solutions. J. Chem.
surface tension in a free-falling film. J. Coil. Inter! Soc. Faraday Trans. I. 71: 217-232.
Sci. 95: 172-182. Lucassen-Reynders, E. H. ed. 1981. Anionic Surfactants.
Hempt, C., Lunkenheimer, K. and Miller, R. 1985. On New York: Marcel Dekker.
the experimental determination of the dilational Lunkenheimer, K., Hartenstein, c., Miller, R. and
elasticity and the exchange of matter of mixed Wantke, K. -D. 1984. Investigations on the method
gelatin surfactant adsorption layers. Z. Phys. Chemie of radially oscillating bubble. Coli. and Sur! 8:
(Leipzig). 266: 713-720. 271-288.
Hiemenz, P. C. 1977. Principles ofColloid and Surface Lunkenheimer, K., Serrien, G. and Joos, P. 1990. The
Chemistry. New York: Dekker. adsorption kinetics of octanol at the air/solution
Hirt, D. E., Prud'homme, R. K., Miller, B. and interface measured with the oscillating bubble and
Rebenfeld, L. 1990. Dynamic surface tension of oscillating jet methods. J. Coil. Inter! Sci. 134:
hydrocarbon and fluorocarbon surfactant solutions 407-411.
using the maximum bubble pressure method. Coil. Marangoni, C. 1878. Difesa della teoria dell'elasticita
and Sur! 44: 101-117. superficiale dei liquidi: Plasticita superficiale. Nuovo
Holmberg, C. and SundelOf, L.-O. 1996. Temperature Cimento. 3(3): 50-68,97-123 and 193-211.
dependence of hydrodynamic properties and sur- De la Maza, A. and Parra, 1. L. 1996. Interaction of
equimolecular mixtures of nonionic/anionic surfac- and Reprography. Materials, Processes and Systems.
tants with Iiposomes. Langmuir. 12: 3393-3398. 7th edn. New York: van Nostrand Rheinhold.
Mysels, K.1989. Some limitations in the interpretations Sucker, Ch., Meskat, W. and Heusch, R. 1964. German
of the time dependence of surface tension measured patent 1211821, 10.10.64.
by the maximum bubble pressure method. Langmuir. von Szyszkowski, B. 1908. Experimental studies of the
5: 442-447. capillary properties of aqueous solutions of fatty
Neustalder, E. L. 1984. Surfactants in enhanced oil acids. Z. Phys. Chem., Stoechiom. Verwandschaftsl.
recovery. In Surfactants, ed. Th.F. Tadros, pp. 64: 385-414.
277-285. London: Academic Press. Tadros, Th.F. 1984. Some applications of surfactants
Du Nouy, P. 1919. Journal General Physiology. 1: 521. in biological systems. In Surfactants, ed. Th.F.
DIN-Blatt 53914. Tadros, pp. 323-335, London: Academic Press.
Nylen, P. and Sunderland, E. 1965. Modern Surface Tanford, C. 1980. The Hydrophobic Effect: Formation
Coatings, pp. 650-660. London: John Wiley. ofMicelles and Biological Membranes, 2nd edn. New
Okubo, T., Kitano, H., Ishiwatari, T. and Isem, N. York: Wiley.
1979. Proc. Roy. Soc. A366: 81. Tann, R. S., Berger, P. D. and Berger, C. H. 1992.
De Oliveira, V. A., Tiera, M. 1. and Neumann, M. G. Applications ofDynamic Surface Tension to Adjuvants
1996. Interaction of cationic surfactants with acrylic and Emulsion Systems (Conference on Pesticide
acid-ethyl methacrylate polymers. Langmuir. 12: Formulation and Application Systems, San Antonio,
607-612. TX USA, November 1991). ASTM Special Technical
Ottewil, R. H. 1984. Introduction. In Surfactants, ed. Publication 1112, pp. 158-172. Philadelphia, USA:
Th.F. Tadros, pp. 1-18. London: Academic Press. ASTM.
Patterson, R. E. and Ross, S. 1979. The pendent-drop Thayer, A. M. 1993. Soaps and detergents, Product
method to determine surface or interfacial tension. Report. Chem and Eng. News. 25 January: 26-47.
Sur! Sci. 81: 451-463. Valentini, 1. E., Thomas, W. R., Sevenhuysen, P., Jiang,
Rillaerts, E. and Joos, P. 1982. Rate of demicellization T. S., Lee, H. 0., Liu, Y. and Yen, S. C. 1991. Role
from the dynamic surface tension ofmicellar solution. of dynamic surface tension in slide coating. Ind. Eng.
J. Phys. Chem. 86: 3471-3478. Chem. Res. 30: 453-461.
Rosen, M. 1. 1978. Surfactants and Interfacial Phenomena. Van Voorst Vader, F., Erkelenz, Th.F. and van den
New York: Wiley. Tempel, M. 1964. Trans. Farad. Soc. 60: 1170.
Ross, S. and Morrison, I. D. 1988. Colloidal Systems Wada, Y., Nakahara, T., Orrii, T., Okano, Y., Aya,
and Interfaces. New York: Wiley. M., Kasui, K., Kamochi, A., Yamada, Y., Katsumata,
Rotenberg, Y., Boruvka, L. and Neumann, A. W. 1983. 0., Sakawa, S. and Kurahashi, Y. 1983. In Proc.
Determination of surface tension and contact angle 5th Int. Congress Pesticide Chem. Volume 4. Oxford:
from the shapes of axisymmetric fluid interfaces. J. Pergamon Press.
Coli. Inter! Sci. 93: 169-183. Ward, A. F. H. and Tordai, L. 1946. Time-dependency
Ruschak, K. J. 1987. Flow of a thin liquid layer due of boundary tensions of solutions. I. The role of
to ambient disturbance. AIChE Journal. 801-807. diffusion in time effects. J. Phys. Chem. 14: 453-461.
Schick, M. J. ed. 1967. Nonionic Surfactants. New Wilhelmy, L. 1864. Ann. Physik. 1: 521.
York: Marcel Dekker. Wu, A. and Whitaker, S. 1986. The recirculation zone
Serrien, G. and Joos, P. 1990. Dynamic surface at the entrance of a falling liquid film: consequences
properties of aqueous sodium dioctyl sulfosuccinate for the surfactant adsorption problem. J. Coli.
solutions. J. Coli. Inter! Sci. 139: 149-159. Inter! Sci. 110: 389-397.
Siddiqui, F. A. and Franses, E. I. 1996. Equilibrium Young, T. 1805. An essay on the cohesion of fluids.
adsorption and tension ofbinary surfactant mixtures Phil. Trans. Royal Soc. (London). 95: 65-87.
at the air/water interface. Langmuir. 12: 354-362. Ziller, M., Miller, R. and Kretzschmar, G. 1985. A
Strutt, J. W. (Lord Rayleigh) 1897. Proc. Roy. Soc. model of the adsorption kinetics for a flowing liquid
29: 71. layer of a surfactant solution. Z. Phys. Chemie
Sturge, J. M.1977. Neblette's Handbook ofPhotography (Leipzig). 4, 721-730.
COATING RHEOLOGY: COMPONENT 5
INFLUENCE ON THE RHEOLOGICAL
RESPONSE AND PERFORMANCE OF
WATER-BORNE COATINGS IN ROLL
APPLICATIONS
J. Edward Glass and Robert K. Prud'homme
5.1 INTRODUCTION
the result of major formulation components
Even though much ofthe published coating-flow necessary to achieve the ultimate performance
literature - including many chapters in this book of the coated film, and can give rise to serious
- focuses on Newtonian liquids, most coating difficulties in applying uniform layers at acceptable
liquids in commercial applications are strongly speeds. A primary case in point is that of
non-Newtonian: the viscosity is not simply a adhesives, for which the desired 'tackiness' in the
constant, but depends on the deformation rates final product is often associated with substantial
the liquid experiences; viscoelasticity gives rise viscoelasticity during processing.
to additional stresses that can significantly alter The purpose of applied rheology in general is
the local force balances in the flow; and, at to understand and predict the response of non-
sufficiently low deformation rates, coating liquids Newtonian liquids to the complex flows in
can exhibit solid-like yield behavior preventing industrial processes from their response in much
flow altogether. In some instances, non-Newtonian simpler flows that subject the test liquids to
characteristics of the formulation are adjusted well-defined deformations and are amenable to
for the specific purpose of enhancing coating laboratory instrumentation. The use of rheometry
process performance. In the pre-metered processes to determine material functions such as viscosity,
discussed in Chapters lla-llc, for instance, normal stress coefficients, or dynamic moduli, is
small amounts ofpolymeric additives can enhance well documented in numerous texts (Walters
the stability and hence the operating latitude of 1980; Whorlow 1980; Schoff 1988). Also well
the process. Another example comes from paints established is the broader field of rheology which
and varnishes, for which much research has been encompasses not only rheometric measurements,
invested to achieve ease of flow during application but also the development of so called constitutive
yet resistance to sagging and other defects once equations that mathematically describe the
the coating is applied. In many instances, on the material functions and associated non-Newtonian
other hand, the non-Newtonian behavior is flow behavior; the explanation of these equations
London. ISBN 041206481 2.
138 Coating rheology: component influence on water-borne coatings in roll applications
in terms of the miscrostructure or molecular beyond their limits the most advanced computa-
structure of the liquid; and the application ofcon- tional algorithms available to date (see Chapter
stitutive models to predict the flow in complex geo- 9). In fact, at least for viscoelastic liquids, there
metries and flow fields (e.g., Tanner (1985); Bird, are virtually no publications that successfully
Armstrong and Hassager (1987); Larson (1988)). predict coating flows and their stability from
The specialized field of coating rheology, or hydrodynamic principles and constitutive models.
the science ofhow non-Newtonian liquid behavior Given the rather modest state-of-the-art of
affects coating flow performance, is not as well coating rheology in general, the present chapter
developed, in spite of its industrial significance. does not attempt a broad review of the entire
A major difficulty arises from the extremely high field but limits itselfto reviewing basic rheological
and rapidly varying deformation rates found in concepts and then describing in detail three
most industrial coating processes (e.g., Schunk particular studies for which some progress has
and Scriven (1990)). These deformations are been made in quantifying the interplay between
difficult if not impossible to replicate in laboratory formulation components, rheology, and ultimate
measurements with conventional rheometric performance during coating. The first is the
equipment. Thus, the application of data from application of architectural paints by means of
such measurements often requires considerable a small porous mat roll at low speeds to wallboard,
amounts of interpretation and speculation. Even wood and stucco surfaces. House paints in which
if an apparent correlation between coating the lay-person is involved are typical examples.
behavior observed in experiments and material The second case is the coating of paper, ranging
functions measured on rheometers can be estab- from low-cost newspaper to highly glossy cor-
lished, it does not unequivocally prove a cause- porate reports. Here, the substrate consists of
and-effect relationship. Constitutive theories and flocculated cellulose fibers and is highly porous.
non-Newtonian fluid mechanics can in principle Modern paper coaters rely on blades to apply
provide such a proof, but still face immense the coating liquid and smooth the coated surface,
difficulties for realistic coating flows. They tax as illustrated in Fig. 5.1. They can operate at
Figure 5.1 Schematic of a blade applicator for high-speed paper coating.

Formulation concepts 139
speeds in excess of 25 mis, or 5000 ftlmin (see in the literature. Section 5.3 concludes with a
Chapter 12d). The third case, finally, is that of brief review of some simple but useful rheological
protective coatings on metal substrates, used by models (for an in-depth treatise of constitutive
a wide range oforiginal equipment manufacturers models, which constitute an entire field of study
for refrigerators, washing machines, automobiles, by themselves, see Larson 1988). Section 5.4
etc., and henceforth referred to as OEM coatings. attempts to illuminate how the coating formulation
Methods for applying OEM coatings include roll and resulting rheology influence the success ofthe
coating, as exemplified by the so-called Giordano application step and the fate of the liquid film
coater shown in Fig. 5.2. OEM coatings on high- after application. The main focus is on architectural
energy metal surfaces can exhibit unique defects coatings, for which sufficient data is available to
different from those on other surfaces (Bierwagen draw a fairly unified picture of the interplay of
1975). components, rheology, and performance during
Illuminating the interplay of chemical compo- and after application. For formulations used in
nents, rheological material properties, and per- paper and OEM coatings, systematic rheological
formance during and after application of typical studies are not as common in the published
coating formulations is the primary focus in this literature, but the difficulties in the application of
chapter. Section 5.2 provides a brief introduction these coatings are nevertheless discussed in terms
ofthe components and processes used to formulate of the knowledge gained from the architectural
coatings of the sort addressed in this chapter. area. Hot melts and adhesive coatings are not
After a brief review of basic rheological concepts, discussed in this chapter because the component
such as measurement techniques and material influences on rheology and flow behavior in these
functions, Section 5.3 focuses on recounting systems are poorly identified in the published
various kinds of rheological behavior that are literature. Section 5.5 finally recapitulates the
characteristic of common types of liquids in most significant formulation effects on the rheology
general and coating solutions in particular, and of water-borne coating formulations.
shows how some of that behavior can be
interpreted in terms of the composition and
5.2 FORMULAnON CONCEPTS
microstructure of the coating system. Section 5.3
also selectively cites component influences on There are two key ingredients in most coating
melt polymer rheology that have been identified formulations of the sort considered in this chapter:
un_.., Enamel pan
Figure 5.2 Schematic of 'Giordano' reverse-roll coater configuration (after Matsuda and Brendley, 1979).
the polymer that gives desired flow properties Acrylic esters lower the glass transition tempera-
during application and mechanical integrity of ture of the latex particle and thereby facilitate
the coated layer once it is solidified; and the particle coalescence in the film forming sequence.
pigment that provides color or opacity to the In coating formulations for less demanding interior
applied film, or merely serves as a cheap filler. applications, vinyl acetate copolymers with a low
Most architectural and paper coatings are latex percentage of butyl acrylate are used.
based, i.e., the polymer is high in molecular Styrene/butadiene latices were developed during
weight and forms a dispersion of insoluble solid World War II as a natural rubber substitute.
particles in water. The final film forms from the Styrene is still used in many coating formulations,
coalescence or fusion of the latex particles rather expecially in the paper industry, but generally is
than chemical cross-linking. The films in OEM no longer a primary monomer in architectural
applications, in contrast, are cross-linked from coatings because of its poor durability.
dissolved or dispersed oligomers or resins to The stability of latices is well described by the
achieve hard and wear-resistant surfaces. Derjaguin, Landau, Verwey and Overbeck
(DLVO) theory, as is documented in numerous
textbooks (e.g., Napper (1983); or Sato and Ruch
5.2.1 LATICES IN ARCHITECTURAL AND
(1980)). For latices with small particles in the
PAPER COATINGS
range of 100 nm or so in aqueous phases containing
Latices are prepared by an emulsion polymeriz- low salinity levels, electrostatic stabilization by
ation process that begins with an emulsion of anionic surfactants is not sufficient (Ottewill
monomer droplets of limited solubility in the 1982). To provide high and low temperature
aqueous phase (Odian 1993). A free-radical stability as well as mechanical stability to such
polymerization is started by a water-soluble latices, 1-4 wt.% methacrylic acid is included in
initiator, and occurs in the presence of surfactant. the latex synthesis. With a proper balance of
The surfactant molecules aggregate into groups hydrophilicity and glass transition temperature
of generally spherical shapes called micelles (cf of the acrylate ester, the methacrylic acid units
Chapter 4). In the polymerization of styrene, the can be positioned at the methacrylate par-
'stung' monomer enters the micelle to propagate ticle/water interface in a two stage emulsion
the polymerization, which is relatively immune synthesis (Karunasena and Glass 1989). In alkaline
to termination as compared with solution media, these surface segments stabilize the particle.
polymerization. Vinyl acetate, a more hydrophilic Methacrylic surface acid stabilization is used in
monomer, on the other hand, does not participate vinyl acetate latices for paper coatings, but not
significantly in micellar propagation. It polymer- in vinyl acetate latices for architectural coatings.
izes primarily in the aqueous phase and the In the latter, water sensitivity of the final film is
polymer chains aggregate to form the latex important (Ma et al. 1993), and stability of large
particles. This sort of emulsion polymerization latex particles is achieved by chemically grafting
process is referred to as homonuclear. Methyl water-soluble polymer to the growing chain,
methacrylate, which is ofintermediate hydrophilic- which places nonionic surface segments on the
ity, polymerizes by both micellar and homonuclear latex surface (Craig 1986). With the high chain
propagation. In the final latex dispersion, sur- transfer rate and reactivity of the hydrophilic
factants play an important role in stabilizing the vinyl acetate monomer, grafting of the polymer
polymer particles. Typically, there is an excess segments leads to a broad particle size distribution
of surfactant in the aqueous phase, particularly that is beneficial in lowering the viscosity at lower
after the synthesis ofsmalliatex particles (100 nm). shear rates. Section 5.4.2 discusses in detail the
For architectural coatings, methyl methacrylate, influence on rheology ofgrafted polymer segments
with its proven durability for exterior applications, for a 600 nm vinyl acetate latex coating stabilized
is generally copolymerized with acrylic esters. by hydrorxyethyl cellulose (HEC) segments, and
Formulation concepts 141
a 100 nm methyl methacrylate latex with meth- paper coatings typically include a kaolinite clay
acrylic acid surface segments. and calcium carbonate, with particle sizes of
In architectural coatings, there is an even about 100 nm and 400-600 nm respectively (Garey
approach to using both polymer and pigment 1977). The content of latex binder and Ti0 2
particles. The percentage of the latex binder is pigment is often very small to minimize
varied to produce a level of film quality that fits cost (e.g., Lundberg et al. (1990)). The latex
a given market need and cost point. The 'hiding' has traditionally been a styrene/butadiene
pigment that provides opacity is titanium dioxide, copolymer, but vinyl acetate latices have captured
which is optimal based on its refractive index part of the market over the last decade. Small
difference with air and the polymer binder. Its particle latices (100 nm) are used almost exlusively
concentration in reference to the latex is close to in paper coatings to maximize the amount of
21% by volume, which is the minimum amount clay that can be used (Schaller 1968). In some
needed in an architectural coating for optimum paper coating formulations, the latex and/or the
opacity. Titanium dioxide pigment is seldom used Ti0 2 pigment are omitted altogether.
in higher amounts because it is one of the most
costly components in the formulation. However,
5.2.2 DISPERSE POLYMER SYSTEMS IN OEM
other low-cost 'extender' pigments such as calcium
COATINGS
carbonate, silicas and clays are often added to
paint formulations because a number of coating In conventional high-solids OEM coatings, an
properties (i.e., tendency for blistering, corrosion oligomer is applied as a 60-80% solution (by
resistance, opacity) improve with increasing pigment weight) in organic solvents and is cross-linked
volume concentration. Clays and other inorganic after application to achieve the desired mechanical
disperse phases can also be added - or substituted properties. Environmental factors and political
for more costly latex particles in poorer quality legislation, however, are forcing the transition of
coatings - to achieve higher formulation viscosities OEM coatings from the high-solids approach in
without the need for additional, water soluble organic solvents to water-borne systems.
thickener polymers. One approach to drastically reducing organic
Past the so-called critical pigment volume solvent emissions is to employ so-called 'water-
concentration (Cpvq, the amount of polymer reducible' resins whose solutions in a water-
binder is insufficient to form a continuous film. miscible organic solvent can be diluted with
As a result, numerous coating properties exhibit water to form acceptably stable dispersions of
abrupt changes as the particle concentration is polymer aggregates swollen by solvent and water.
increased past the CPVC, as was identified early There is indeed a close analogy between a
in the development of latex coatings (e.g., 'water-reducible' acrylic and an acrylic latex, the
Bierwagen and Hay (1975)). It was also realized main difference being that hydroxyethyl meth-
early on that higher CPVCs could be obtained acrylate is used in the synthesis of the OEM resin
in lower cost architectural paints if a small to provide cross-linking sites with multi-functional
particle size latex was used (Schaller 1968). In melamine or isocyanate groups after application.
coatings with high particle volume concentrations The low molecular weight resin is first prepared
(pvq, a small 100 nm latex is typically employed in a water-miscible solvent, and is diluted with
in an attempt to maintain a continuous film. water before being applied (Lambourne 1987).
Paper coatings are low-cost formulations that In OEM acrylic resins, methacrylic or acrylic
serve market needs entirely different from those acids are often included to provide electrosteric
of architectural coatings. They are applied at stabilization to the newly formed disperse phase
very high speeds (sometimes in excess of 25 mis, when the organic, water-miscible solvent is diluted
or 5000 ft/min) on highly porous cellulose-fiber with water. In both the latex and in the OEM
substrates. The dominant disperse phases in dispersions, the amount ofoligometric acid buried
within the polymer particle is ca. 25% by weight the test, or time since inception of flow, as
(Richards 1977; Ma 1992), and in both binder explained in more detail below. Material functions
systems the acids are incorporated into the chain provide invaluable information about the structure
by a chain-growth mechanism. In a chain-growth and flow-induced changes in the structure of
polymerization, several methacrylic acid mono- non-Newtonian liquids. They are useful to classify
mers may add to a propagating radical. This liquids. They are furthermore important to
permits the acid groups to be incorporated in construct or select appropriate mathematical
sequence runs that, on neutralization, will provide models, called 'constitutive equations', that
greater stability to the particle than an individual describe relationships between imposed defor-
acid unit. mation and resulting stresses (e.g., Larson (1988)),
In the synthesis of water-reducible polyester and to determine adjustable constants in such
and polyurethane dispersions, acid groups are models.
incorporated into the chain by a step-growth The definition ofstress requires the specification
mechanism (Odian 1993) prior to water dilution of the direction of the force and the orientation of
of the organic solvent- resin solution (Lambourne the surface upon which the force acts. Similarly,
1987). The acid groups, incorporated by the the rate of deformation, or velocity gradient,
reaction of the terminal hydroxyl in dimethylol- requires specification of the direction of the
propionic acid with the terminal carboxylic acid velocity and the direction in which the velocity
units of a diacid, can enter the polyester chains varies. Figure 5.3 illustrates five flow fields that
only as individual units. These isolated units are extensively used in rheological measurements,
cannot sequence in the manner observed for namely steady shear, transient shear, uniaxial
acrylic resins prepared by a chain growth extension, biaxial extension, and oscillatory shear.
mechanism, and the percent of the individual
acid units buried in the disperse has not been
determined. Section 5.4.4 examines the influence 5.3.1.1 Steady shear flows
of individual components on rheology and
In the flow illustrated in Fig. 5.3a, a fluid between
performance during application for 'water- two plates is sheared as the top plate moves with
reducible' acrylic and polyurethane dispersions.
velocity U x in the x-direction while the bottom
In OEM coatings, pigments are typically not plate is at rest. The velocity gradient, or shear
used. The disperse polymer phase is not a
rate, Yyx = dvx/dy == Y, is given by U x/<> where <>
classical latex but is rather unstable, making is the distance separating the plates. The shearing
pigmentation with Ti0 2 or other extender fillers
action of the flow generates stresses that act
difficult. parallel to the direction ofshear, i.e., shear stresses
5.3 RHEOLOGICAL CONCEPTS (5.1)
5.3.1 STRESS AND STRAIN IN RHEaMETRIC and stresses that act perpendicular to the direction
FLOWS of shear, i.e., normal stresses. The latter include
the stress induced by the fluid motion as well as
As indicated in the Introduction (Section 5.1), the isotropic hydrostatic pressure. It is customary
most rheological measurements are performed to eliminate the isotropic part by taking the
in so-called rheometric flows that subject the test difference between normal stresses, and these
liquid to a very simple deformation, often uniform differences,
in space. The purpose is to isolate specific
'material functions' of the tested liquid, such as (5.2)
the viscosity, normal stress coefficients, or dynamic
moduli as a function of shear rate, frequency of (5.3)
Rheological concepts 143
(a) Vx= yy
Steady Shear Vy= Vz • 0
y = Shear Rate
(b) y Vx =(Ymax cos rot) y

Vy = Vz = 0
Dynamic
Oscillatory Ymax = Maximum Shear Rate
Shear
(c) t < 0:
Step
Shear ~lob'>O
t5 ~
t
Vx
~o:
Vx
= Vy = Vz = 0
= YoY
o Tome
x
Vy = Vz = 0
Yo= Shear Rate
z
(d)
Uniaxial Vx = -~ x
Extension
vy = -ty
x
Vz =£ z
£ = Extension Rate
(e) z
Biaxial
Extension Vx =- £ z
Vy =0
Vz =£Z
£ = Extension Rate
Figure 5.3 Common flow fields used to define material functions in rheological measurements.
are in fact measured in experiments. N 1 and N 2 define the material functions for steady shear
are referred to as the primary and secondary flow, namely the viscosity '1:
normal stress differences. The relationships
between the stresses and the velocity gradient, y, (5.4)
144 Coating rheology: component irifluence on water-borne coatings in roll applications
and the primary and secondary normal stress -dvz/dz, whereas in uniaxial extension
coefficients, '1'1 and '1'2: dVy/dy = -1/2(dv z/dz) and dvx/dx = -1/2
(dvz/dz).
(5.5)
(5.6) 5.3.1.3 Dynamic oscillatory flows
The material functions '1, '1'1 and '1'2' generally In oscillatory shear measurements, often also
vary with shear rate. The normal stress coefficients referred to as dynamic mechanical measurements,
are defined in terms of the square of the velocity a sinusoidally varying shear field is imposed on
gradient because the stress difference must be an a fluid and the amplitude of the resulting shear
even power of shear rate; that is, changing the stress and the phase angle between the imposed
direction of the shear, which makes Y negative, shear and the measured stress is detected. The
does not change the direction or sign of the test is particularly simple to interpret and is said
normal stress, whereas changing the direction of to be in the 'linear viscoelastic' regime if the stress
the velocity gradient does change the direction is linearly proportional to the imposed strain
of the shear stress. and the stress response is also sinusoidal and
involves only the first harmonic and no higher
5.3.1.2 Extensional and compressive flows harmonics in frequency (in experimental work,
the validity of these conditions must, of course,
Fig. 5.3d depicts a flow that either converges or
be verified).
diverges with respect to the z-axis. Such flows
In the test shown in Fig. 5.3b, the imposed
are produced by uniaxial stretching of a filament
oscillatory velocity field is
as in the case of fiber spinning (extension), or by
biaxial stretching of a sheet as in the case of Vx = [Ymax cos(wt)]y (5.10)
inflating a balloon (compression in the z-direction). where w is the frequency, Ymax is the maximum
The measurable material responses are the tensile velocity gradient, and the maximum value of the
normal stresses. As for the normal stresses in strain is given by Ymax = Ymax/w. The resulting
shear flow, the difference in stress is used in order stress also oscillates and goes through a maximum
to eliminate isotropic pressure terms, viz., value, <max' with some phase shift, </J, from the
(5.7) imposed shear.
The uniaxial elongational viscosity, '1., is defined <yx = <max cos(wt - </J) (5.11)
as the proportionality factor between this normal This oscillatory stress response can be decomposed
stress difference and the velocity gradient, into two terms, one in-phase with the velocity
E = dvz/dz: and one 90· out-of-phase. These, in tum, can be
(5.8) expressed in terms of the maximum velocity
gradient.
In general, '1. is a function of E, the extension rate.
In planar extension, illustrated in Fig. 5.3e, the <yx =- '1' Ymax cos (wt) - '1" Ymax sin (wt) (5.12)
material is being stretched in the z-direction. This where
deformation arises when a sheet of material is
pulled in one direction. The planar elongation <max cos </J == '1' Ymax
viscosity is, as above, defined via the tensile stress and <max sin</J == '1" Ymax (5.13)
difference:
These expressions define the two dynamic viscosity
(5.9) coefficients, '1' and '1". A complex viscosity '1*
can be defined as
As in (5.8), the velocity gradient, or extension
rate, is E = dvz/dz, but dVy/dy = 0 and dvx/dx = '1 * = '1' - i '1" (5.14)
which is the ratio of shear stress to shear rate, to be sensitive probes of the molecular structure
viz., ,,*= .yAt)IYyx(t). At low frequencies, ,,' of polymer solutions and gels, and a large body
approaches the viscosity measured in steady of literature exists on this subject (Ferry 1980).
shear as the shear rate approaches zero, and is As an illustrative example, Fig. 5.4 shows the
usually given the name 'dynamic viscosity'. dynamic moduli of a polymer solution during
Alternately, coefficients can be defined in terms gelation (Clark, Wellinghoff and Miller 1983).
of the maximum strain instead of the strain rate. The material begins as a liquid solution in Fig.
.yx = - G' Ymax cos(wt) - G" Ymax sin(wt) (5.15) 5.4a and ends as a solid gel in Fig. 5.4d. For a
polymer solution at low frequency, elastic stresses
where relax sufficiently rapidly and viscous stresses
.max cos4J == G' Ymax dominate. As a result, the loss modulus, G", is
higher than the storage modulus, G'. Both decrease
and .max sin 4J == G" Ymax (5.16) with decreasing frequency, but G' decreases more
This defines the two material functions G' and quickly. For a gel, in contrast, the stress cannot
G", referred to as the storage and loss moduli, relax and G' therefore is independent offrequency.
respectively. It is customary to write Also, because the gel is highly elastic, the storage
modulus is higher than the loss modulus.
G* = G' + iG" (5.17)
where G* is called the complex modulus. The
term 'complex modulus' is sometimes also used 5.3.1.4 Transient flows
to refer to the magnitude of G*, i.e., Fig. 5.3c shows an experiment in which the shear
rate is stepped instantaneously from one value
(5.18)
to another and the response in shear stress is
This is the case in the early literature on coating monitored with time. The resulting material
rheology where, in a somewhat arbitrary choice, function is then the 'stress growth function':
IG*I was selected to project the importance of
elastic effects and, at the same time, minimize .Xy(t) = ij+(yo.j,t)yo (5.21)
data redundancy. This choice is adopted for which describes the growth from an initial shear
much of Section 5.4 in this chapter. The modulus rate YO.i to the final shear rate Yo' Most commonly,
and viscosity coefficients are related via the initial shear rate is zero.
G' = w,," (5.19) In an alternative transient test, the step strain
experiment, an instantaneous strain is applied to
G" = w,,' (5.20) the material and the decay in the stress is
G', which is proportional to the stress in phase monitored with time. This defines the shear
with strain, provides information about the modulus, G(t), for an applied shear strain of
recoverable elasticity of a material. An ideal magnitude Yo:
elastic rubber band subjected to an oscillatory
(5.22)
test, for instance, would have all of its stress in
phase with strain or displacement. G", in contrast, or the Young's modulus, E(t) for an elongational
is proportional to stress that is out of phase with strain of magnitude t: o
displacement but in phase with rate-of-displace-
(5.23)
ment or shear-rate. It accounts for the dissipative
losses in the material. For a purely viscous liquid, A more complicated transient deformation test
of course, all of the stress would be out of phase that provides qualitative information about
with displacement. time-dependent fluid rheology ramps the shear
The linear viscoelastic dynamic moduli, G' and rate continuously from zero to a higher value
G", are functions of frequency. They have proven over a prescribed time period, and monitors the
.-..
cf ro
10
-
CL
(J)
::::l
10
::::l
"0
o
(A) ~
(C)
-12°C -18°C
0.1 ........................u.w...-'-........._ _...........~
1 10 100 0.1 1 10 100
Frequency (rad/s) Frequency (rad/s)
G'
-~ ro
(J)
::::l
10
-
CL
(J)
.2
10
::::l ::::l
"0
"0
o o
~ ~ (0)
(8)
-15°C -22°C
O. 1 ................................---'-............UllL.~~""
0.1 1 10 100 1 10 100
Frequency (rad/s) Frequency (rad/s)
Figure 5.4 Dynamic moduli versus frequency during gelation upon cooling of polystyrene in carbon disulfide
(after Clark, Wellinghoff and Miller, 1983).
resulting shear stress. This sort of test is sensitive tests can be performed in a transient mode.
to the kinetics of structure evolution, which can Indeed, because of the difficulties involved in
be important in aggregated colloidal dispersions. realizing purely extensional steady-state flows in
If the structures in the material are broken apart laboratory tests, transient extensional flows are
by shear and cannot reform during the time of very important in rheometric practice. For the
the shear rate ramp, the stresses during the sudden start-up ofextensional flow, the transient
decreasing shear rate ramp are lower than the stress is described by the elongational stress
stresses during the increasing leg. The region growth function ij:, in the coating rheology
between the stress and shear rate curves in the literature often referred to as dynamic uniaxial
increasing and decreasing ramps is known as the extensional viscosity, or DUEV.
'thixotropic loop'. It is typically difficult to
interpret the measured data quantitatively because
5.3.1.5 Other flow configurations used in
the response is a complex convolution of the
rheometry
kinetics of shear induced structure breakdown
and the kinetics of aggregation. Nevertheless, Although the material functions are strictly defined
thixotropic loops often provide useful 'fingerprints' for the simple flows just described, it is often
of complicated coating materials such as paints most convenient to measure material properties
and other pigmented dispersions. in alternate experimental configurations that
Not only shear, but also extensional rheological approximate the ideal flow geometries. Table 5.1
Table 5.1 Experimental geometries for measuring fluid rheology
Experimental geometry Measured quantities Material function determination

3
. 'R [ d In (Q/1tR )]
Flow in a tube (capillary viscometer) Q = volume rate of flow 17(YR)=Q/1tR 33 + dlnr
t1P = pressure drop through tube R
r"'\ '\ Shearing surface
R = tube radius . I d(r;Q/1tR 3 )
L = tube length Y --
R - r~ d rR
t1PR
rR = 2L
\ / ~,' Line of shear and

partIcle path line
3T
Torsional flow between cone and disk R =
radius of circular plate 17(Y) = 21tR 3 y
00 = angle between cone and plate
Wo = angular velocity of cone 2F
T = torque on plate 'P l(Y) = 1tR 2 f
F = force required to keep tip of
Wo cone in contact with plate WO
Y=-
00
0<1
- >9 Shearing surface
~
Line of shear lind
pllrticle path line
Table 5.1 Continued
Experimental geometry Measured quantities Material function determination
3 3
. T/2rtR [ d In (T/2rtR )]
Torsional flow between parallel plates R = radius of disks '1(l'R) = -.-2- 3+ dI .
h = separation of disks l'R n l'R
Wo = angular velocity of upper 'I't{Ji R) - 'I' 2(YR)
disk 2 2
T = torque required to rotate = F/rtR [2 + dln(F/rtR )]
upper disk Y; dlnYR

F = force required to keep
. WoR
separation of disks constant l'R=-h-
Th
-L Shearing
surface
Line of shear and

particle path line
T(R 2 - R t )
Torsional flow between concentric cylinders (Couette geometry) Rt • R 2 = radii of inner and outer '1(y) = 2rtR:hW
cylinders
H = height of cylinders
W = angular velocity of inner
cylinder
T = torque on inner cylinder
Shearing
surface
Line of shear
and particle
path line
gives several examples of configurations from 1985), exactly like the polymeric counterparts;
which material functions can be determined for but they exhibit virtually no normal stresses.
low viscosity fluids (see also Bird, Armstrong and For dispersions with either Newtonian or
Hassager (1987)). polymeric continuous phases, viscosity increases
In industry, unfortunately, many other 'vis- with increasing volume fraction up to a critical
cometers' are in use that involve very complicated volume fraction above which the viscosity diverges
flow fields and hence cannot produce absolute to extremely large values and the material ceases
viscosity data or other material functions needed to flow. This critical volume fraction is about
for research and development purposes. In the 63% for noninteracting, monodisperse spheres,
US paint industry, the most notorious among and increases to about 80% for a broad distribu-
such devices is the Stormer viscometer in which tion ofsphere sizes. It is markedly lower, however,
a shaft with two rectangular paddles is immersed for particles with long aspect ratios, such as
in the paint and rotated at 200 rpm (for details, chopped glass fibers. The limiting values for
see ASTM Test Method D 562-81; and also Van volume fraction are, of course, intimately linked
Wazer (1963)). The force required to maintain to the closest packing densities inherent to the
this rate of rotation is measured with weights particles. Figure 5.5 illustrates the rise of the
added to a platform that hangs from the end of viscosity at high shear rates with increasing
a chord which, via a pulley and simple gear, particle volume fraction for polymer latex particles
drives the paddle. That force, in turn, is converted of different sizes (Krieger 1972). The viscosity
into so-called 'Krebs Units' (KUs) with an arbitrary plotted is a relative viscosity made dimensionless
conversion scale. Even though the value of this with the viscosity '10 of the suspending liquid.
sort of measurement is quite dubious, the Stormer Figure 5.5 shows that, when the volume fraction
instrument is entrenched in the paint industry of the particles is corrected to account for the
and very commonly used to test flow properties increase in particle volume due to the adsorbed
of paints. It is mentioned here because much of surfactant layer, the data points for the different
the literature cited below makes reference to data particle sizes fall on the same master curve. This
in Krebs Units (see Section 5.4). example makes plain that caution is in order
when interpreting rheological data for chemically
complex systems, a point reiterated in Section
5.3.2 EXAMPLES OF NON-NEWTONIAN
5.4 below.
MATERIAL BEHAVIOR
For particles that are strongly interacting, the
As rheology is frequently used as a probe of critical volume fraction at which the viscosity
molecular structure and material interactions, it rises sharply is also markedly lower than that
is helpful to have a general idea of what the expected from simple volume calculations. This
material functions look like for commonly is typically the case for particles smaller than
encountered fluids. 1000nm for which surface forces can become
Low molecular weight fluids and resins generally dominant. The effect ofparticle surface interactions
are Newtonian, that is, they have a constant is illustrated for a polystyrene calcium carbonate
viscosity, independent of shear rate. They also system in Fig. 5.6. Decreasing particle size from
generate no elastic normal stresses. The viscosity 1700 to 70 nm at the same volume fraction of
of polymer melts, filled polymers, polymer particles (30%) results in a ten-fold increase in
solutions and dispersions, on the other hand, is viscosity (Suetsugu and White 1983). Additional
not constant but decreases with increasing shear information on filled melt rheology and particle
rate, and the elastic normal stresses can be orientation for non-spherical particles can be
considerable. Dispersions of solid particles in found in a review by Khan and Prud'homme
nonpolymeric media may also exhibit shear- (1987).
thinning viscosities (Wildemuth and Williams Polymer melts and other viscoelastic fluids
180 r-----------'i"""----.....,.-----~-----~----...,
o
160 - - - -- - - - .-- - ------- ----..+-------''--------'----t---I
I
140· ... -.- - - _. ~ - --- - -- --.----- .. - - - - - - - - - - - , - - - - 1 - - - 1
I
... I
.... 120 --.--- .• - . - . - - - --- -.- --.' ------.---.------------i-----I---~
~
~u 100 . _ •.• - - _.-l. __. -,- +-_-1-_ _--1
ltl
> 80 - -.- -- _._---
~
i .-------. -
~ 60 -O-~--------I
A
40 . - - .. - .'. .- .---.-.---~---r-------___f
o~
, 0
20 . -.- --- _.- . - 0 - -0--- --!:-O-:o-.-::..::.-----------I
A I
0 .....- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -....
35 40 45 50 55 60 65
Particle Volume Fraction rio]
Figure 5.5 High-shear viscosity of latex dispersions versus nominal polymer volume fraction (open symbols)
and effective particle volume fraction with adsorbed surfactant layer (closed symbols): • 0.2 ~m particle
diameter; • 0.66 ~m; • 1.1 ~m (adopted from Krieger, 1972).
display elongational ViscosItIes that strongly with a large aspect ratio have also been observed
depend on strain rates. Figure 5.7 shows repre- to impart a dramatic increase in the dynamic
sentative data of time-dependent or dynamic uniaxial elongational viscosity, as illustrated in
uniaxial extensional viscosity (DUEV) for fibers Fig. 5.8 (Weinberger and Goddard 1974).
of two types of polystyrene melt at several Steady-state elongational viscosity measure-
constant elongation rates (Meissner 1985). The ments free of significant shear viscosity contri-
elongational viscosity increases with time initially, butions are difficult to obtain, and hence are only
may reach a constant asymptotic value that is qualitative in nature.
equal to three times the zero shear viscosity, and The relationship between the elongational
then may increase until the fiber breaks. The viscosity of a liquid and its shear viscosity
onset of the final rise in viscosity depends on the depends on the mechanisms by which the flow-
elongation rate, occurring earlier at higher rates. induced microstructure transmits stresses. For
This so-called 'strain hardening' is a key mech- Newtonian fluids, there are no extra flow-induced
anism in stabilizing the stretching of polymer mechanisms, and the elongational viscosity is
fibers and is therefore very desirable in polymer simply three times the shear viscosity. The rigid
melt spinning. On the other hand, it can be very rods reported in Fig. 5.8, on the other hand, tend
detrimental to successful application of coating to align with the extensional flow and then
formulations, as discussed in Section 5.4.2 of this transmit stress in the axial direction more efficiently
chapter. In dilute aqueous solutions, added fibers than in the transverse direction, causing the
elongational viscosity to become much higher

than the shear viscosity. For very high molecular
weight linear polymer molecules in solution, or
melts of the sort characterized in Fig. 5.7, the
molecular chains undergo a conformational
change at high elongational strain from random
coils to highly stretched configurations. It is this
change that results in the rise in elongational
viscosity which, in turn, correlates with spatter
and filament growth discussed for paint systems
in Section 5.4.2. For colloidal or dispersed phase
systems such as emulsions, the elongational
viscosity is roughly equal to three times the shear
viscosity over the entire shear rate range and,
thus, shows thinning with increasing deformation
rate just as the shear viscosity does (see, for
example, Anklam, Warr and Prud'homme (1994)).
Dynamic oscillatory measurements are sensitive
Shear Rate y [S-I]
probes of molecular structure and interactions
in melts and solutions. Figures 5.9 and 5.10
Figure 5.6 Viscosity versus shear rate for polystyrene display the frequency dependence of the storage
melts with CaC0 3 fillers of various particle sizes and loss moduli, G' and G" respectively, for
labeled in the figure (after Suetsugu and White, 1983). several types of polymeric materials (adapted
The filler loading is constant at 30% by volume. from Ferry 1980). The materials are designated
108
E[S·1]
d 0.0034
...
,
0.016
0.051 :w-
, 0.02 ~
& 0.04 - 10 7 l='
fL1 I
~
'00
o()
~
nlit I ~
~~~~O~g6cPO'U6tc__--=
3~o (/l
:>
Cii
coc: 106 c:
o
/
o ~ 60.001 '00
'00 t1 0.003 c:
c: PS II 0- 0.009
Q)
X
Q)
q, 0.03 X
T=403 K 9 0.09 UJ
UJ
~ I I I pO.2
1()4 L..- I
-'---- ----'-- --'...,....- _ _0_.3_ _-' 105
-'--_-o
10,1 100 101 102 103 104
Time [s]
Figure 5.7 Elongational viscosity versus time for polystyrene melts at various constant elongation rates
shown on the figure (after Meissner, 1985). The samples had a narrow molecular weight distribution, with
average molecular weights of M w = 7.4 X 104 (PS I) and M w = 3.9 X 104 (PS II).
(;'
~ 7 Ge IV
Ci1
0-
-
(f)
(f)
0
~
~
Cl
.2
5
1
j
(J) 3
"- 10
( /)
co 3
'x
«
solution -10 -5 0
Constant + log co
Figure 5.9 Storage modulus G' versus frequency (j)

for several classes of materials (after Ferry, 1980). The
1 10 materials are identified in the text.
Extension Rate (8 0')
Figure 5.8 Axial stress versus extension rate for

200 j.lm long glass fibers with an aspect ratio of 1/57 9
suspended at l.3vol.% in 'Indopol' (Weinberger and
Goddard, 1974). The lower solid line is the calculated (;'
axial stress for a pure 'Indopol' solution based on ~
7
g
;1
shear viscosity measurements.
in the figure by Roman numerals (complete

".2
Cl 5
details on the materials can be found in the

original citation): 3
• Material I is representative of dilute polymer -5 a 5 10

solutions. The data are for polystyrene with Constant + log co
narrow molecular weight distribution around
860000 M w and dissolved in a chlorinated Figure 5.10 Loss modulus Goo versus frequency (j) for
biphenyl at 25°C (the data are taken at several several classes of materials (after Ferry, 1980). The
temperatures and then normalized to 25°C materials are identified in the text.
via time/temperature superposition as described
by Ferry (1980)). • Material IV is a lightly cross-linked amorphous
• Material II is prototypical for a low molecular polymer. The data are for a lightly vulcanized
weight, amorphous polymer. The data are for Hevea rubber with an equilibrium modulus
a 10500 M w poly(vinyl acetate) at 75°C. of E = 7.10 5 Pa at 25°C.
• Material III is a typical high molecular weight,
amorphous polymer. The data are for poly- For both the dilute polymer solution (material I)
styrene at 100°C, with 600000 M w and a and the low molecular weight polymer (material
narrow molecular weight distribution. II), the stresses decrease, or relax, with decreasing
frequency or increasing time. The elastic stresses, viscosity characteristic of that shear rate. At high
measured by G', relax more quickly than the shear rates, however, the viscosity can overshoot
viscous stresses, indicated by G". In the limit of the equilibrium viscosity, as Fig. 5.11 illustrates
low frequencies, the loss modulus G" plotted for a 2 wt.% solution polyisobutylene in primol
against frequency attains a constant slope of + I (Huppler et al. 1987). The maximum in the stress
on a log-log plot. In this region, G"/w = '10' i.e., value occurs at increasingly shorter times as the
the zero-shear viscosity '10 is given by the loss shear rate increases. Indeed, experiments suggest
modulus divided by frequency(cj equation (5.20)). that the maximum occurs at a constant value of
For the high molecular weight polymer (material strain (i.e., shear rate times the time that rate is
III), the storage modulus exhibits a plateau applied). The demarcation between shear rates
toward high frequencies. The plateau is commonly that are mild enough so as to allow monotonic
referred to as the 'glassy' regime because the behavior and shear rates that produce overshoot
frequency imposed on the material is faster than is given by the dimensionless Weissenberg number
the natural relaxation processes of the polymer (see, for example, Bird, Armstrong and Hassager
chains and the material behaves like an amor- (1987)):
phous, or glass-like, solid. Another plateau in
Wi = A,'y (5.24)
the storage modulus, G~, appears at a lower
frequency. It corresponds to a 'pseudo network' Here, A, is the characteristic relaxation time of
made up of temporary chain entanglements that the molecular structures within the liquid, and
do not have sufficient time to relax during the y is the shear rate imparted externally on the
time of an oscillation at frequency w. In this liquid. For Wi> I, the imposed deformation is
intermediate frequency range, the material behaves faster than the material-internal relaxation, and
like a cross-linked rubber. Finally, toward lower non-linear effects such as stress overshoot are
frequencies, the moduli drop because the stresses observed. Similarly, the onset of shear thinning
can relax during the time scale of the imposed, in steady shear viscosity is governed by the
slow oscillation. The zero shear viscosity is Weissenberg number, i.e., shear thinning is
'10 = G"/w in this regime. observed for Wi > 1. The concept ofcharacteristic
For the cross-linked rubber (material IV), the relaxation times versus flow time scales is essential,
moduli approach the same glassy plateau at high and arises repeatedly in rheology.
frequencies as for polymers without cross-links.
At these frequencies, the mechanical properties
of the polymer are not affected by the constraints 3.0 I I I I I
of the chemical cross-links because the spacing Yo = 17.0 5"
between cross-links is longer than the lengths of
the polymer chain that can relax over the time- .;2 2.0 Yo = 1.75. 1
-
scale of an oscillation. In contrast to materials II ~
.;2
and III, however, the 'rubbery' plateau for material NO=0.17S·'
I~
----
1.0
IV at Ge extends over the entire low frequency \.
range because the cross-links prevent elastic ~o ~ 0.017 5"
stresses from relaxing no matter how slow a 0
I I I
0 50 100 150 200 250
deformation is applied. The loss modulus, G",
decreases with decreasing frequency, and is two t(s)
orders of magnitude below G'.
Figure 5.11 Shear stress growth function ij+(t;yo),
When a steady shear rate is imposed on a normalized by the equilibrium viscosity ,,(Yo) at the
sample that is initially at rest (i.e., a step shear given shear rate Yo' versus time t (after Huppler et aI.,
rate experiment), the stresses ofpolymeric materials 1987). The data are for a 2.0% by weight polyisobutylene
rise with time and ultimately reach a steady shear solution in primol.
5.3.3 CONSTITUTIVE MODELS OF the magnitude of the rate of deformation tensor
RHEOLOGICAL BEHAVIOR and is given by:
The simplest rheological equation of state is that
of a Newtonian fluid for which only one material !~~y
2 L.:- L.:-
... y, .. =~I11
'J J' 2
(5.30)
function - the viscosity 11- is needed to characterize , J
the liquid in any rheometric flow, and predict For shear flow, the magnitude of the rate of
the flow behavior in any other laminar flow. This deformation tensor is just equal to the velocity
equation is gradient or shear rate.
t = -111 (5.25) Generalized Newtonian fluid models work
well for describing steady shear flows of polymeric
where Yis the rate-of-deformation tensor defined by fluids, but not transient or elongational flows.
1 = Vv + (VV)T (5.26) They are also inadequate for shear flows when
elastic normal stresses are to be predicted (see,
where Vv, in turn, is the velocity gradient tensor for instance, Chapter 10).
and the superscript T denotes its transpose (see One ofthe most popular generalized Newtonian
also Chapter 2). The components of yare fluid models is the power-law model:
, oV i oVj
Yij = ox· + ox.
J ,
(5.27) _
11 -
K,(n-l) _ _ L xy
Y xy - •
Yxy
(5.31)
where Vi are the Cartesian components of the

which has two adjustable parameters, nand K.
velocity vector v and Xi are the components of
This model describes the viscosity versus shear
the position vector x. For the shear flow given
rate curve on a log-log plot as a straight line.
in Fig. 5.3a, for instance, the shear stress is
It is attractive because it is manageable in the
. oVx direct mathematical solution of many simple
L xy = -11Y xy = -11 oy (5.28) flow problems. Also it adequately describes the
high-shear-rate viscosity behavior of polymer
For non-Newtonian liquids, the Newtonian model solutions under realistic process conditions and
(5.25) may be generalized by allowing the viscosity requires only two parameters. The power-law
to become a function of shear rate. In such model fails, however, to capture the constant,
generalized Newtonian fluid models, the viscosity low-shear-rate viscosity observed at low flow
is made a function of the square root of the rates (discussed in the preceding Section 5.2), or
second invariant of the rate of deformation the limiting high-shear viscosity often seen at
tensor because continuum mechanics arguments high processing speeds. Other models with more
show that the viscosity can depend only on parameters must be chosen to account for these
certain combinations of the rate of deformation viscosity limits (see, for instance, Bird, Armstrong
tensor called invariants. Ofthe three independent and Hassager (1987». Very successful among
combinations that can be formed, the second these is the Carreau model
invariant is chosen because the first invariant is
equal to zero for an incompressible fluid and the
third invariant is equal to zero in a shear flow.
The second invariant is defined by
Here, 110 and 1100 are the low- and high-shear
11= LL YijYji
i j
(5.29) viscosities respectively, A. is a time constant that
measures the characteristic shear rate at which
Instead of II directly, the viscosity is made a the viscosity begins to drop off, and n is the
function of the square root of 1/2 II, which is power-law index as in (5.31).
Very concentrated filled melts, emulsions, foams, integral form

dispersions and gels display an additional
non-Newtonian behavior toward low deformation
rates, namely solid-like yield. A true yield stress
implies the material will not flow until some -00
minimum stress is achieved. In many liquid where the quantity in brackets is the relaxation
systems there may be very slow flow at low shear modulus, G(t), for the Maxwell fluid. The integral
stresses; however, over the time scale of practical form of the Maxwell model can, furthermore, be
importance, the material can be treated as if it written in terms of strain, yxy' rather than rate
does not yield. This can be called an 'apparent' of strain, YXy,
yield behavior. Models describing the shear
stress versus the shear rate for materials with
apparent yield stress have been reviewed by
Prud'homme et al. (1983), and also Bird, Armstrong
-00
and Hassager (1987). Two popular and common
models among these are the Bingham model, where the term in brackets is called the memory
function, m, for the Maxwell fluid.
'xy = '0 + rrp Yxy for 'xy> '0 (5.33a) The Maxwell model, with its single time constant
and single viscosity parameter (or modulus
YXY = 0 for 'Xy::::; '0 (5.33b) parameter) can qualitatively reproduce the creep
and the Casson Model, behavior of 'ideal' viscoelastic fluids; however, it
cannot qualitatively reproduce the linear vis-
.../'Xy = .../'0 + .../rrp.../yXY for 'xy > '0 (5.34a) coelastic behavior for real materials which, because
of their composition, most often have a range of
YXy = 0 for 'xy ::::; '0 (5.34b) relaxation times and associated modulus par-
ameters. To account for this, the Maxwell model
In both, '0 is the apparent yield stress, and rrp is can be generalized by replacing the relaxation
the plastic viscosity. modulus by a sum of contributions:
For liquids that exhibit viscoelastic behavior,
polymer rheologists have constructed an impres-
sive array of constitutive models. Most of these
models have their origin in polymer melt processing -00
where elastic effects often are quite large. To
describe the flow ofa viscoelastic fluid, constitutive Fig. 5.12 shows an example of the fitting of
models must account for both viscous and elastic modulus data for a low density polyethylene melt
behavior. The simplest model is the Maxwell with a generalized Maxwell model with ten time
model, which results from the superposition of constants, A. i , and ten modulus constants, Gi
the equations for a Newtonian fluid and a (Laun 1978). The agreement between the experi-
Hookean solid. For a shear flow, the Maxwell mental data with the fitted moduli is excellent.
model can simply be written as: However, even with the complexity that arises
from multiple relaxation modes, the Maxwell
model is still insufficient to predict shear thinning
(5.35) that is characteristic of most viscoelastic liquids,
or stress overshoot as illustrated in Fig. 5.11
Here, rr is the viscosity as above and G is the above. To account for these and many other
elastic shear modulus. The ratio rrlG is a time aspects of viscoelastic flow behavior, more
constant, often called the 'relaxation time', A.. sophisticated constitutive models are needed (see
Alternatively, equation (5.35) can be written in Larson (1988), for an in-depth treatise).
105
104 a)
-a...co 103
CD 102
10 1
10°
10- 3 10-2 10- 1 10° 10 1 102 103 104
Reduced Frequency (racl/s)
106
105
-co
a... 10
'-'
4
CD 103
102
10 1
10- 3 10- 2 10- 1 10° 10 1 102 10 3 104
Reduced Frequency (rad/s)
6
10
105 C)
-co
--
Cl..
( J)
:J
104
103
:J
\J
0
~
10 1 "'----:-~~
L.......olI:::'O"'U'WAl................u..w.IL.....J...............J.WL...-L...L..u..w'"---'.............u.w...----'----'.........
10- 3 10- 2 10- 1 10° 10 1 102 103 104

Reduced Frequency (racl/s)
Figure 5.12 Fitting the modulus of a low-density polyethylene melt with a generalized Maxwell model with
10 time constants, Aj , and 10 modulus constants, G j (after Laun, 1978): (a) and (b) individual G' and G"
contributions from the terms comprising the Maxwell elements; (c) comparison offitted model to experimental data.
Interplay among components, rheology and application performance 157
5.4 INTERPLAY AMONG COMPONENTS, rates and radial or extensional flows from cavity
RHEOLOGY AND APPLICATION centers; and (4) a filamentation region where the
PERFORMANCE lamellar liquid sheets progress into filaments
extended in air between the two roll surfaces,
5.4.1 ROLL COATING APPLICATIONS - A with high localized shear at their roots and highly
BRIEF OVERVIEW extensional flow normal to the roll applicator
elsewhere. They dissipate into little droplets
For many if not most commercial paints and causing what is commonly referred to as 'misting'.
protective coatings, application by roll is the Building on these early investigations, a wealth
method of choice. For that reason, this section of experimental information has been amassed
focuses primarily on roll applications. The on roll coating processes, and numerous rigorous
principles involved in the application of coatings and comprehensive hydrodynamic analyses have
by brushes, blades, or spray share some similarities been performed to explain the experimental
with those of roll application, but are not discussed findings (see Chapters 12a and 12c for a complete
in detail here. review). In fact, roll coating is one of the most
For Newtonian liquids, the laminar flow in thoroughly investigated and best understood
the nip region of a roll applicator generates coating application processes. Still, from the
pressure variations that are largely responsible point of view ofcoating rheology, many unresolved
for the amount of coating material that flows issues remain. In particular, much of the academic
through the narrow nip passage, and also for the and published work on roll coating has been
stability of the film split where the coated film confined to Newtonian fluids, yet small amounts
separates from the material that remains on the of elasticity can drastically affect ribbing and
roll. Numerous early investigators, interested filamentation. In addition, hydrodynamic stability
primarily in printing ink applications, recognized analyses, as difficult and sophisticated as they
the importance of the pressure variations in may be, predict only the critical parameters at
controlling coating thickness (e.g., Gaskell (1950), the onset of ribbing but, in their most commonly
or Miller and Myers (1958)). They also identified used, linearized form cannot describe the fate of
various regimes of film-split instabilities, viewing ribs at flow conditions past the initial onset of
the coating process from the interior of a coating the instability or their break-up into filaments
roll. In particular, they noted that, when the and ensuing misting.
product of velocity and viscosity exceeded a Even though ribbing can be interpreted as a
critical value, a regime of regularly spaced air form of cavitation, as suggested in the early
cavities that are separated by 'ribs' is superseded literature cited above, its initial onset is triggered
by a regime in which the ribs break up into by a purely hydrodynamic instability of the
filaments (Miller and Myers 1958; Myers and two-dimensional, translationally symmetric film
Hoffman 1961). In total, four characteristic flow split meniscus between the substrate and the
regions were identified early on in roll application applicator roll to three-dimensional disturbances,
processes: (1) a recirculating bead ofliquid, often as explained in detail in Chapter 12a. In fact,
referred to as 'bank', at the inlet to the nip ribbing is very similar to the phenomenon of
between the two rolls, with widely fluctuating viscous fingering (Saffman and Taylor 1958),
shear rates, sometimes turbulent flow, and often which occurs when a low-viscosity fluid, such as
ejection of material; (2) a nip region with laminar air, displaces a high-viscosity liquid from a
lubrication-type flow and varying shear rates; (3) confined flow channel between parallel walls (in
a film-split region where ribs form due to an roll coating, an observer sitting on the substrate
instability mechanism akin to ventilated cavitation or moving with the applicator roll sees air
in hydrodynamic lubrication and often grow into penetrate into the liquid coating bead in the nip
very thin, lamellar liquid sheets with low shear region). Both the ribbing instability between
coating rolls and the viscous fingering instability into one large 'rib', the three-dimensional flow
between parallel plates are driven by the pressure state in roll coating is stabilized by the diverging
gradients generated by viscous flow, and are geometry of the flow channel, allowing perfectly
stabilized by surface tension (see Chapter 12a). regularly spaced, large-amplitude ribs to exist
For Newtonian liquids, the onset of these over some parameter ranges (see Fig. 5.13).
instabilities is therefore governed by the capillary Eventually, these three-dimensional flow states
number, viz., Ca = "U /(1, where U is the appli- become unstable themselves, giving way to
cation speed and (1 is surface tension. The ribbing transient, semi-chaotic flow states (e.g. Rabaud,
instability is further stabilized by the divergence Michelland and Couder (1990)). These phenomena
of the flow channel between the applicator roll are presumably strongly affected by elastic effects,
and the substrate (e.g., Ruschak (1985); Hakim but, unfortunately, no data exists in the published
et al. (1988)). Thus, the ratio Ho/R between the literature for well-characterized systems.
closest gap between the roll and substrate, Hej, For elastic liquids, the ribs often also break
and the roll radius, R, becomes an additional up into fibers spanning the air space between the
parameter controlling the onset of ribbing. Mill applicator roll and the substrate: see Fig. 5.13.
and South (1967); see also Mill (1967), for instance,
found
Ca = 10.3(Ho/R)0.75 (5.39)
to provide a good fit for their own data for

printing inks. Figures 12a.ll in Chapter 12a
summarizes many other data for mostly Newto-
nian liquids. For non-Newtonian liquids, even
small amounts of elasticity can drastically reduce
the critical capillary number at the onset of
ribbing. Bauman, Sullivan and Middleman (1982),
for example, found that 10ppm of a high-
molecular-weight polyacrylamide added to a
90/10% mixture of glycerin/water reduced the
critical capillary number by a factor of two to
five while no normal stress was detectable. At
100 ppm, normal stresses could be measured and
the reduction in capillary number was ten to
thirty fold. For all practical purposes, roll
applications for viscoelastic liquids - and even
Newtonian liquids - operate well past the onset
of ribbing.
As the capillary number is raised past the
initial onset of ribbing, the ribs grow rapidly in
amplitude to form lamellar liquid sheets, and
their number per unit length increases. Mill and
South (1967) noted that, when "U/(1 exceeds
about 2.4, the rib frequency becomes independent
of the capillary number and dependent solely on
H 0/R. In contrast to the viscous fingering between Figure 5.13 Comparison of ribbing in two industrial
parallel plates, where the unstable air pockets OEM coatings: (a) little misting; (b) significant misting
grow into irregular shapes and eventually combine (Soules, Fernando and Glass, 1988).
Raleigh instabilities, in turn, promote the break- example of very poor roll coating performance
up of the fibers into spatter droplets, often with extreme filamentation is shown in Fig. 5.14
referred to as 'misting'. Elastic effects can delay for an architectural latex paint that was purchased
the surface tension induced break-up and, in fact, at a chain store (Glass 1978a). Massouda (1985)
aggravate misting problems. Soules, Fernando found calculated viscoelastic axial stresses to
and Glass (1988), for example, investigated the correlate well with spatter severity for eight
performance of two commercial OEM acrylic commercial latex paints, and thereby provided
dispersions in a Giordano roll coater of the sort additional evidence that elongational effects are
described by Matsuda and Brendley (1979). key in controlling filamentation and misting.
Although their study was not adequate in detail All coating formulations (architectural, paper,
and the coating compositions they used were and OEM) are applied at high deformation rates,
proprietary, their data provide sufficient indication which tend to break up aggregates, followed by
that the differences in ribbing and misting at high 'flow-out' or leveling of surface imperfections at
roll velocities are related to the dynamic uniaxial very low deformation rates. To simulate this in
elongational viscosities, as might be expected. An rheometric equipment, a steady high shear rate
Figure 5.14 Example of ribbing, filamentation and fiber growth in mat-roll application of a commercial
architectural coating at about 37 ft/s (Glass, 1978a).
is applied and then dropped rapidly to a low not change with time. The stress overshoot of
deformation rate, typically in an oscillatory formulations that did vary with time in oscillatory
mode, for instance in a cone-and-plate configur- studies were less than that of the formulation
ation. The cone angle in such experiments must reported in Fig. 5.15. Clearly, stress overshoot of
be small to ensure a low deformation response coating formulations in particular, and dynamic
during oscillation. This limits the shear rate that rheological studies attempting to simulate rapidly
can be achieved under steady shear to about varying deformation histories in coating application
3000 s - 1, which is much less than the shear rate steps in general, need to be investigated more
in real applications. Nonetheless, valuable infor- systematically.
mation can be extracted from such 'simulated'
coating rheology measurements. Figure 5,15 gives
5.4.2 ARCHITECTURAL LATEX COATINGS
an example. It shows the dynamic viscosity
response of a formulation thickened with a Among the coatings considered in this chapter,
maleic acid/methyl vinyl ether copolymer to a water-borne latex coatings for architectural
shear rate of 2750 s - 1 imposed from an initial applications have seen the greatest change in
rest state (Glass 1975). The behavior is similar component compositions over the last three
to that illustrated in Fig. 5.10 yet, to our knowledge, decades, and resulting improvements in coatings
the data in Fig. 5.15 are the only data of this quality have been considerable. At first, the changes
type reported for complicated non-Newtonian focused on the latex itself. The goal was to use
coating systems. The formulation chosen in Fig. small latex particles for better binding of pigments
5.15 exhibited the greatest stress overshoot of all and fillers (Schaller 1968). The viscosities of
those studied by Glass (1975); curiously, however, small-particle latex dispersions, however, were
the oscillatory response of this formulation did high at low shear rates and prevented applied
30
'G(/)i'
'0
25
~ 0 0
•
0 .
e:.
0
0 0 0 0
20
';
(/)
It)
N
..... '5
N
. .
1ii
~
'iii
0
10 .
0
(/)
:>
0.5
0,0
-2 2 6 '0 14 '8 22 26 30
Time [51
Figure 5.15 Transient viscosity response of interior coating formulations to a sudden shear rate of 2750 S-1
(after Glass, 1975). The formulations are thickened with a methyl vinyl ether/maleic acid copolymer to 81
Krebs Units (triangles) and 92 Krebs Units (diamonds) as measured with a Stormer instrument (cf Section 5.3.1).
coatings from leveling out evenly. The high coating formulation. Higher viscosities provide a
viscosities were attributed to inter-particle bridging thicker film during application, which indirectly
via thickener molecules (Kreider 1964), but leads to improved leveling, or 'flow-out', as
comprehensive attempts to quantify such mech- leveling is commonly referred to in paint-industry
anisms were not made. The size of a hydroxyethyl jargon. The effect of film thickness on leveling
cellulose (HEC) thickener molecule of 410000 can be inferred from the modified Orchard
molecular weight is about 120 nm and thus equation (Smith, Orchard and Rhuid-Tutt 1961;
comparable to a l00nm latex particle. However, Orchard 1962; see also Chapter 6 in this volume):
the number of thickener molecules in a paint with
90 Krebs Units (see Section 5.3.1.5 for a discussion (5.33)
of this empirical viscosity measure) is too low to Here, t 1{2 is the halflife for an amplitude decrease
account for bridging of a significant number of in surface irregularity, r,o is the low-shear-rate
latex particles. More recently, the industry moved viscosity of the fluid, (j is the surface tension, and
toward bimodal particle size distributions, often H is the applied film thickness. In equation (5.33),
achieved by intentional blending of latices with only the variables that can be influenced by
different particle sizes, which permitted formulation formulation components are listed. In architectural
at higher volume percent solids (Chong, Chris- latex formulations, there is an excess ofsurfactant
tiansen and Baer 1971). The industry also moved from the emulsion polymerization process (Section
to the synthesis of latices with larger particle sizes 5.2) and, therefore, surface tension is not a key
that provide multimodal distributions when the variable in influencing coating 'flow-out' (it is,
water-miscibility of the monomer is high as, for however, in high-solid solvent formulations applied
instance, for vinyl acetate. without surfactants to high energy surfaces).
The influence ofindividual components on the rhe- The median particle size and the particle size
ology and application performance ofwater-bome distribution of the disperse phases do not have
latex coatings is summarized in Table 5.2, and is much of a direct influence on high-shear-rate
discussed in more detail in the subsections that follow. viscosity. However, increased viscosities at high
shear rates can arise from nonionic water-soluble
polymers that are grafted to the polymer chains
5.4.2.1 Influence of components on viscosities at
during the emulsion polymerization process in
high shear rates
order to stabilize larger latex particles (Craig 1986).
Achieving high viscosities at high shear rates Adding water-soluble thickeners to a latex
(viz., y > 104 S -1) is one ofthe goals in architectural formulation also raises the viscosity at high shear
Table 5.2 Overview of effect of major formulation components on the rheology and application performance
of water-borne latex coatings
Viscosities at high shear Viscosities at low shear

rates rates
Component y~104(s-1) y < lO(s-l) Spatter
Conventional thickener:
Chemical structure no yes yes
Molecular weight yes no yes
Latex median particle size:
Monodisperse no yes no
Bimodal no yes no
Protective colloid surface segments yes
rates. However, even to someone familiar with constant. Differences in chemical structure of the
the structure-rheology relationships in polymer thickener can lead to different degrees of asso-
solutions, the changes imparted by thickener ciation at low shear rates, as is discussed below.
additions to latex paints may seem difficult to At high shear rates, however, such associations
understand. A major difficulty centers around are disrupted and, thus, thickeners affect high-shear
the Stormer instrument, which is deeply entrenched viscosities not by their chemical structure. but
in the paint industry yet, at best, reflects a primarily by their molecular weight which sets
viscosity at an intermediate shear rate of about their hydrodynamic volume (Arney and Glass
50 to 100 S -1. With this 'intermediate viscosity' 1976).
in Krebs Units fixed and dictated, the formulator
has to live with whatever the components produce
5.4.2.2 Influence of components on viscosities at
for viscosities at high or low shear rates. For
low shear rates
example, Fig. 5.16 shows the effect on steady-shear
viscosity at two shear rates (1800 and 14000 s - 1) In the traditional coatings literature, the viscosity
of varying the total amount of hydroxyethyl at low shear rates is discussed more frequently
cellulose thickener while holding different for- than the viscosity at high shear rates. This bias
mulations at the same 'intermediate viscosity' (89 stems from paint applications on vertical surfaces
Krebs Units) by blending thickeners of different where it is important that the applied coating
molecular weight (Glass 1975). Evidently, the use does not sag under gravitational forces yet has
oflower molecular weight thickeners can provide a viscosity low enough to permit 'flow-out' of
higher viscosities at high shear rates than are surface irregularities that arise during application.
observed with higher molecular weight thickeners Equation (5.33) above highlights the importance
but, in Fig. 5.16 the latter are, of course, used in of the viscosity at low shear rates, '10' for 'flow-out'
smaller amounts to keep the Stormer 'viscosity' or leveling of a coating.
4.8
4.4
Q)
<J) 4.0
'0
e:. 3.6
Z:'
'iii 32
0
u
<J) 28
:>
ca
(1)
2.4
.c 2.0
C/)
>.
-g 1.6
•
(1)
./"
ii5 1.2
.8
,42 .4 .6 .8 1.0 1.2 1.4 1.6 1.8 2.0
Total Hydroxyethyl-Cellulose Concentration [wt. %]
Figure 5.16 Steady-state viscosity of architectural latex coatings as a function of hydroxyethyl cellulose
thickener content at a fixed intermediate 'viscosity' of 89 Krebs Units (Glass, 1975). The values of M w marked
by the arrows in the figure give the average molecular weight of the particular thickener used in blending
with the highest molecular weight thickener.
In contrast to the viscosities at high shear cellulose (HEC), of course, also contribute to the
rates, viscosities at low rates are affected by the viscosity at low and moderate shear rate. They
latex size: they increase with decreasing median do so through both their size and their amount.
particle sizes below 150 nm. Figure 5.17 illustrates The volume fraction of the thickener must be
this effect for several model latices based on a added to the total volume fraction of solids, and
variety of hard (i.e., high glass-transition tem- thereby again moves the viscosity to higher
perature ~) and soft (i.e., low ~) methacrylate, values and closer to the sharp rise in Fig. 5.5.
acrylate and styrene compositions (Karunasena, Table 5.3 summarizes the calculated relative
Brown and Glass 1989). The synthesis of small volume fractions for a comparative set of coating
latices requires an excess of surfactant. This formulations investigated by Lundberg et al.
excess plus oligometric surface acid salts stabilize (1991). The size of the hydroxyethyl thickener
the latex and, at the same time, prevent adsorption was calculated from the root mean square end-
of cellulose-ether thickeners onto the latex. The to-end distance of the random polymer coil, R e •
surface stabilizers themselves, however, form Figure 5.18 shows how the corresponding,
hydration layers which can signficantly increase measured viscosities rise with increasing amounts
the effective volume fraction of disperse phases ofthickener added. Clearly, the rise is more sharp
(Krieger 1972, Nielsen 1977), and thereby drive and the viscosities are higher for the small latex
the viscosity of smaller latices closer to the sharp dispersions.
rise seen in Fig. 5.5 (Karunasena and Glass 1992). Figure 5.19 illustrates thickener effects on
The size effect is due to specific material interactions viscosity for another formulation study (Glass
in real-world dispersions, and cannot be explained 1978c) in which an existing vinyl acetate/butyl
with the standard model approach for normalizing acrylate latex dispersion was split into different
viscosities ofdifferent particle size latices (Krieger batches, and hydroxyethyl cellulose (HEC)
1972). thickeners of different molecular weights as well
Water-soluble thickeners, such as hydroxyethyl as two other water-soluble polymers, namely
30
Ii)
til
!!:.
<l>
(ij
a: 20
iiiQ)
J::.
(f)
ill
(\J
;a 10 o
~
·iii Doo
0
u
CIl B cP 0
:>
0
50 100 150 200 250
Particle Size [nm]
Figure 5.17 Effect of median particle size on viscosity at low shear rate (2 s -1) for latex paint at total volume
fraction of Ti0 2 pigment and latex of 0.32, and a volume fraction of pigment as compared to latex particles
of 0.21 (Karunasena, Brown and Glass, 1989; for complete descriptions of the formulations, refer to the original
paper).
Table 5.3 Contributions of hydration layer and hydroxyethyl cellulose diameter to total effective volume
fraction of latex dispersions at a nominal volume fraction of 0.32 (after Lundberg et al. (1991))
Effective volume Wt.% HEC in Estimated HEC

Latex size fraction with aqueous volume fraction Total effective
(nm) hydration layer phase (p = 23nm) volume fraction
100 0.42 0.50 0.15 0.57

0.90 0.27 0.69
1.41 0.42 0.84
220 0.36 0.51 0.15 0.51
0.89 0.26 0.62
1.33 0.39 0.75
30
Vi' 100 nm Latex
cO
~
'"
1ii
II:
(;j 20
'"
.t:
(/)
it>
C\l
1ii
J!;- 10 220 nm Latex
'iii
0
()
(I)
:>
0
0.0 0.5 1.0 1.5 2.0 2.5
Thickener Concentration in Aqueous Phase [wI. %]
Figure 5.18 Effect of median particle size on viscosity at low shear rate (2 S-l) as a function of hydroxyethyl
cellulose concentration for model latex dispersions (Lundberg et at., 1991).
poly(ethylene oxide) and an acrylic acid/acrylate thickeners. On the other hand, when the highest
ester copolymer, were added to each batch to and lowest molecular weight hydroxyethyl cel-
achieve the same elevated viscosity as measured lulose thickeners are employed in a realistic
by a Stormer viscometer (127 Krebs Units). Table coating formulation of lower viscosity (about 90
5.4 lists the amounts of thickener added. Figure Krebs Units), the latex dispersions not only have
5.19 shows that, toward low shear rates, the the same nominal 'formulation viscosity' (around
viscosity measured with a 'Rheometries' cone/plate 50 to 100 s - 1 is the shear rate that may
rheometer is higher for the dispersion containing approximate the deformation rate in the Stormer
a lower amount of higher molecular weight HEe, viscometer), but the viscosities match remarkably
and that dispersion is more shear thinning than well over the entire range of low shear rates, as
the dispersions containing higher amounts of the shown in Fig. 5.20. Viscosity measurements in
lower molecular weight hydroxyethyl cellulose similar studies of a simplified formulation
101
Shear Rate [s·11
Figure 5.19 Steady-state viscosity as a function of shear rate for aqueous vinyl/acrylic latex dispersions (0.32
volume fraction) thickened to 127 Krebs Units with water-soluble polymers (Glass, 1978c). Filled squares:
poly(ethylene oxide); closed circles: acrylic acid/ethyl acrylate copolymer; open circle: hydroxyethyl cellulose
(HEC), Mw = 800 000; open inverted triangles: HEC, M w = 650000; open triangles: HEC, Mw = 300 000;
open diamonds: HEC, Mw = 70000.
containing only thickener and Ti0 2 pigment cellulose without surfactant (Glass 1975). As a
indicate that the pigment is flocculated (Lundberg result of this synthesis, the latex had a broad
and Glass 1992a). The formulations examined in particle size distribution, and was stabilized with
Figs 5.19 and 5.20 contain many more ingredients grafted hydroxyethyl cellulose fragments. The
such as calcium carbonate, silicas, etc. but the only surfactant present in the formulations came
independence of the low-shear-rate viscosity on with the subsequent addition of Ti0 2 pigment.
the molecular weight of the thickener is probably Poly(vinyl alcohols) with different residual acetate
also related to aggregate structures. contents were then used to thicken the formula-
The chemical structure of water-soluble polymer tions. Poly(vinyl alcohols) with 99% alcohol
thickeners can also have a dramatic influence on units are known to be difficult to solubilize in
viscosities at low shear rates. This effect is water at ambient temperatures, and solutions
attributed to associations that arise from hydrogen obtained at higher temperatures will readily
bonding, which can lead to crystallinity or precipitate under shear at lower temperatures.
macromolecular superaggregation (e.g. Klenin et With increasing acetate content, up to 25 mol. %,
al. (1973)). Figure 5.21 documents a formulation the copolymers are easier to solubilize and
study with ca. 700 nm vinyl acetate latex that remain soluble in aqueous solution under defor-
was synthesized in the presence of hydroxyethyl mation. The acetate units serve to inhibit inter-
Table 5.4 Solution properties ofaqueous vinyl acetate/acrylic latices with different thickeners (after Glass (1978c))
Thickener Stormer Power law

Latex viscosity of coefficient n
Cone. cone. mixture between 2.5 and
Type Mw (wt.%) (wt.%) (KU) 25 S-l
Hydroxyethyl cellulose 80 x 104 1.9 2.6 127 0.30

Hydroxyethyl cellulose 7x 104 14.0 2.6 126 0.97
Poly(ethylene oxide) 50 x 104 6.9 2.6 126 0.64
Acrylic acid/acrylate
ester copolymer unkwn. 23.3 2.9 127 0.93
103
•
A
A
102
• •
A
A
Q)
til
'0
eo.
A
A •
~
A
• •
'iii A
0
u A
til
:>
10'
A
A •
10° '-------~-------'------------'
10-' 102
Shear Rate [5·']
Figure 5.20 Steady-state viscosity as a function of shear rate for interior architectural coatings thickened to
comparable intermediate viscosities with hydroxyethyl cellulose of high and low molecular weight (data from
Glass, 1978b). Triangles: M w = 15 x lOs, 94 Krebs Units; circles: M w = 3 x lOs, 89 Krebs Units.
and intra-hydrogen bonding that facilitate super- the thickener on the response of the coating
molecular order in poly(vinyl alcohols) with high formulation after shearing as measured by the
hydroxyl content (Klenin et al. 1973). Figure 5.21 transient rise of the complex modulus, jG*1 (see
shows the influence of the chemical structure of equations (5.17) and (5.18) and associated dis-
400 the structure formation right after the coating is

applied. Indeed, in one ofthe earliest examinations
350 of coating behavior with oscillatory rheometry
N
E (Glass 1975), the second dynamic viscosity
--
<..>
c
>.
300 coefficient, 1'1", and the storage modulus, G', were
noted to be primary variables that increased after
.
~
250 coating application. An earlier theoretical pub-
<.9 lication predicted that the poor leveling charac-
<J)
200 teristics of coated films ought to be related to
::::l
:; the elasticity of the coating formulation (Bierman
"0
0 150 1968).
~ Another example illustrating the build-up of
)( 89%
<ll
100 elastic rheological properties from structure
C.
E formation is provided by the formulation study
0
(.) 50 discussed already in reference to Fig. 5.16, which
compared latex paints with the same Stormer
0 'viscosity' but with hydroxyethyl thickeners of
0 20 40 60 80 100 120 140 different molecular weight (Glass 1975). The data
Time [s] in Fig. 5.22 show how the complex viscosity 1'1*1,
the dynamic viscosity '1', and the storage modulus
Figure 5.21 Time-dependent recovery of complex G' increase with time after shear application for
modulus IG*I (at 28lJlIl amplitude of the oscillatory three formulations. Two were formulated with
measurement after steady-state shearing) of interior different molecular weights and concentrations
latex coatings thickened with alcohol/vinyl acetate
of hydroxyethyl cellulose thickener to achieve
copolymers with different degree of hydroxyl content
(Glass, 1975). the same intermediate formulation viscosity (90
Krebs Units as measured on a Stormer instrument).
Despite substantial formulation differences, the
cussion nearby). Clearly, with more hydrophobic elastic rheological responses of these two formu-
groups, the IG*I values are lower and, as a result, lations are close, within experimental error -
much improved leveling of the coating was consistent with the findings from another study
observed (see Glass 1975, for further details). A with hydroxyethyl cellulose reported in Fig. 5.20.
similar change in rheological response has been The intermediate data set in Fig. 5.22 is for a
observed for cellulose ether in both aqueous formulation with a lower amount of high-
solutions and complete coating formulations as molecular-weight hydroxyethyl cellulose, and its
the ratio of hydrophobic to hydrophilic groups G'(t) is markedly lower. Its viscosity in Krebs
(viz., methoxyl, oxyethylene and carboxyl) was Units is also lower.
varied (Glass 1976). The elastic rheological response of coating
formulations quite often proves to be very 'system
specific', as is exemplified by a formulation study
5.4.2.3 Influence ofcomponents on elastic respome
in which two different latices were thickened by
and leveling
three different types of thickeners (Glass 1978d).
As the example in Fig. 5.21 illustrates, oscillatory In Figs 5.23 and 5.24, the open symbols denote
measurements of elastic rheological properties coating formulations with a 560 nm vinyl acetate
have proven to be sensitive probes ofthe structure latex, which is representative of a modern latex
of polymer solutions and gels in general, and of with moderate viscosity at low shear rates; and
coatings in particular. For the latter, the ability the closed symbols denote coatings containing
to 'flow out', or self-level, strongly depends on a 117 nm acrylic latex, which is representative of
180
120 WI. % KU ~i~-
-7 ~F-~-
I{;()
100 .28 90
~8.0"O5
140
80 HEC' MW
120
Q) {;()
(/)
~~ C\J
-
r
'0 100
E
a..
........ 83 __ -0\)',--0-0·0--
()
.23 c
-~ {;() .0-- -
.iJ-D-'.o C:' '" >.
:£
-... 100
~ 40 0 CD
<0 ••
100
1.50 90 140
80
o
..Lr
D-,--
.
...c~ '" -
HEC: Mw = 6.8 x 10
4
120
{;()
0 0 0 '"
~
100
0 0 0 0
40 ::'
0 12 24 3.6 48 {;()
Time [5]
Figure 5.22 Time-dependent build-up of complex viscosity 1'1*1 (circles), dynamic viscosity '1' (hexagons) and
storage modulus G' (triangles) of latex coatings thickened with hydroxyethyl cellulose, The molecular weight
of the thickeners, the amount of thickener and the resulting viscosity in Krebs Units (KU) as measured by a
Stormer instrument are given in the figure (Glass, 1975).
early, small particle latices which tended to dominates the rheological response. For the
exhibit high viscosities at low shear rates. The HEC thickened paints, in contrast, IG*I and N 1
thickeners include a modified acrylic acid exhibit a time and shear rate dependence, with
copolymer, hydroxyethyl cellulose (HEC) and an values of IG*I that are notably higher for the
acrylic acid/ethyl acrylate copolymer (the effect smaller 117 nm latex formulation than for the
on viscosity versus shear rate of the latter two 560 nm latex. Considerably lower IG*I values are
thickeners in a latex dispersion is illustrated in obtained for either latex with the low molecular
Fig. 5.19 above). With the modified acrylic acid weight acrylic acid/ethyl acrylate copolymer
copolymer thickener, which is highly cross-linked, thickener, which is more hydrophylic than HEC.
both latex formulations exhibit a high complex The IG*I values are particularly low for the
modulus, IG*I (Fig. 5.23), and the normal stress 560 nm latex. These findings suggest that, in the
difference, N l' of the small acrylic latex also is HEC formulation, both the thickener and the
high (Fig. 5.24). Neither IG*! or N l' however, latex contribute to rheology, whereas in the
exhibit a dependence on time or shear rate linear acid copolymer, the rheological reponse is
respectively. This suggests that the thicktmer dominated by the latex. Figure 5.25 shows that
the amount of leveling of the six formulations 1978d). The coatings were applied with a rake-
investigated in Figs 5.23 and 5.24 correlates quite like 'draw-down bar' and the profiles of the dried
closely with the measured IG*I values (Glass films were measured with a surface profilometer
(Dodge 1972); the S-values on the charts denote
1000 , . . . - - - - - - - - - - - - - . . . . . , sag ratings that are inversely related to the ability
of a non-uniform coating to level out after
900 application.
800
5.4.2.4 Influence of components on elongational
~ viscosity and roll spatter
l/l
300
-. .
:J
"S
. .
"8 _ - - A - - - - A - - - .... Contrary to what might be expected, dynamic
--...-
::E
><
<I>
a.
200
~ ....
-4/('". -cr--- t : : r - - - - --l::r- measurements of shear elastic formulation prop-
erties are not good indicators of the susceptibility
E 100 of coating formulations to misting or spattering
o
()
() in roll applications. The tendency for spattering
o O~~2!i;;O~2!402:0:::::::::::!i:60t::::::::;80==IOi:::O=::12:L:O=:J140 appears to be more closely correlated with
measurements of dynamic uniaxial extensional
Time[s)
viscosity.
Figure 5.26 illustrates the dynamics of the film
Figure 5.23 Time dependent recovery of complex split in roll applications for the 117 nm all-acrylic
modulus IG*I at ±28 11m strain amplitude for large- latex coating formulations discussed in connection
particle latex with bimodal size distribution (open
with Figs 5.23 and 5.24 above (Glass 1978c). The
symbols) and latex with small monodisperse particles
(closed symbols), both thickened to 90 Krebs Units formulation with the acrylic acid/ethyl acrylate
with an acrylic acid/ethyl acrylate copolymer (circles), copolymer thickener (Fig. 5.26b) exhibits signifi-
hydroxyethyl cellulose (triangles), or a modified acrylic cant fiber elongation and resultant spatter, even
acid copolymer (squares) (Glass, 1978d). though its IG*I and N 1 responses are rather low.
N
E
~ 10',.,....---------------------,
.
>.
:£
. •
-1
Shear Rate [s )
Figure 5.24 Shear-rate dependence of first normal stress difference for the same latex paints as in Fig. 5.23
(Glass, 1978d).
==- -_. l-fI- _.
- =-1- -h1
I-
-- - ,~~iirf- H I:+
__ -- 1-1
'1'1 -t- ./
--
::'1- - - .- 141
W -
~ ...
i-
- -- 1-1-
-1
-i -
ftf f-'- I- f----
r---
••
--I- I-
I- • -..1''11 '1:•
- I-
I-f-'-I-- -
.. 1 - 1 -
- - '~. _ ,_ _ _ i--
=-~
.. -. - - --I-~"
~~-- - -
,--'-"-
-,
- f-'--+.-1-_-i-_+ ,III
- I--I--IH--I
,- Jj _f-~ A,
- ,-
1-- '-= - ,._ 1- , ' -
- 1-
1-
-
= I":: 1- il
1- - I- -
'- --
-I'-
- - - - I-
,- 1-' .- 1--
i- 'I't-I'-
-- ..- - - I-- !It - -- --,

~
- j I- 11\
l---j ~ ,- \LJ . ) - '-H h m .,.. -"J.- ,i ~i=

!-
.- i -
Figure 5,25 Profiles of dried films after leveling of the latex paints characterized in Figs 5.23 and 5.24 (Glass,
1978d). Left column ('BMD'): large-particle latex with bimodal size distribution; right column ('SMD'): latex
with small monodisperse particles; first row ('PAAC'): acrylic acid/ethyl acrylate copolymer thickener; second
row ('MAP'): modified acrylic acid copolymer; third row ('HEX-M'): hydroxyethyl cellulose (the label
1-1212-90 refers to a particular paint formulation standard).
On the other hand, its dynamic uniaxial elonga- low IG*I and N 1 values are consistent, however,
tional viscosity as measured by the spinning-fiber with the good 'flow-out' of surface striations after
technique is very high, as Fig. 5.27 shows. The applications seen for that formulation (cf Fig. 5.25
Figure 5.26 Instantaneous film-split patterns of interior paints formulated at 90 Krebs Units, and thickened
with hydroxyethyl cellulose of high (a) and low molecular weight (c), acrylic acid/ethyl acrylate copolymer
(b), and modified acrylic acid copolymer (d) thickeners (Glass, 1978c).
IZ,O , - - - - - - - - - - - . - - - - - - - - - - - , proportional to the molecular weight of the

thickener (Soules, Fernando and Glass 1988).
Consistent with this finding, the low-molecular-
11.5 weight HEC system exhibited substantially less
filamentation and spatter than the high-molecular-
weight system (compare Fig. 5.26c with 5.26a).
11.0 The relationship between decreasing molecular
weight of the thickener and decreasing dynamic
><
uniaxial elongational viscosity of a latex paint
~ 10.5 appears to be universal for a wide range of
'0
~ formulations. Figure 5.28 illustrates elongational
~
.;;;
o
~ 10.0
':>
"iii
c:
o
';;;
c:
<l>
;;: 40
W
C
~
.'t'"
0-
3,0
Z,5
;;;- 30
o
><
<l>
40 50 eo 90 100
'"
'0
~
Extension Rate (s"] ~
'Vi
8
'"
~ 20
Figure 5.27 Apparent uniaxial elongational viscosity c:
o
as a function of extension rate for vinyl/acrylic latex 'Vic:
paints at 120 Krebs Units containing different thickeners: ;;:<l>
o acrylic acid/ethyl acrylate copolymer of unknown W
c
molecular weight; £':" acrylamide/acrylic acid copolymer, <l>
M w = 6 X 10 ; [J poly(ethylene oxide), M w = 4 X 10 li
6 6
(Glass, 1978c). .'t

1.0
above). The formulation with the modified acrylic

acid copolymer follows exactly opposite trends:
it did not support fiber growth at all and its 50 60 70 eo 90
elongational viscosity was too low to even be Extension Rate (5"]
quantified (Glass 1978c, d); and, as expected from
the high IG*I and N 1 values noted in Figs 5.23
Figure 5.28 Apparent uniaxial elongational viscosity
and 5.24, the coating showed little evidence of
as a function of extension rate for vinyl/acrylic latex
'flow-out' of surface striations (Fig. 5.25). The paints at 120 Krebs Units with acrylamide/acrylic acid
dynamic uniaxial elongational viscosities of copolymer thickeners of different molecular weights
hydroxyethyl cellulose (HEC) solutions are (after Glass, 1978c).
viscosities for another case study with acrylam- 5.4.2.5 Recent trends in surfactant modified,
ide/acrylic acid copolymers (Glass 1978c). In this water-soluble polymers
case also, high molecular weights led to stable
fiber growth (Glass 1978c). The dependence of Cellulose ethers have dominated the water-borne
filamentation and spatter on molecular weight thickener market for architectural coatings from
is even more dramatic for poly(ethylene oxide) their first commercialization in the 1960s until
thickeners, as Fig. 5.29 shows. The sequence of the early 1980s. During the last decade, however,
high-speed photographs in Fig. 5.30 reveals that, surfactant modification of water-soluble polymers
for the high-molecular-weight poly(ethylene oxide) has produced a new generation of thickeners
system, large spatter drops result from the breakage (Glass 1984). Hydrophobic modification of alkali
at two points and subsequent 'snap-back' of a swellable emulsions and hydroxyethyl cellulose
filament. have produced thickeners that have gained
acceptance due, in part, to the lower spatter and
misting they provide for certain market segments.
The most dominant of these new thickeners has
become the hydrophobically-modified, ethoxy-
lated urethane (HEUR) family which, to date,
has proven to be the only class of thickeners
whose viscosity is enhanced by nonionic surfac-
tants at low polymer concentrations (Lundberg,
Glass and Eley 1989). These water-soluble
polymers follow general surfactant behavior.
With increasing size of the hydrophobic group,
the thickening efficiency of the polymer increases
(Brown and Glass 1989), and the solutions exhibit
an increasing storage modulus, G' .
In the HEUR family, where structural features
can be controlled, the optimum spacing between
terminal hydrophobic groups is approximately
500 oxyethylene units (Kaczmarski and Glass
1993). The size of the alkyl units ofthe diisocyanate
is an important complement to the terminal
hydrophobe in promoting an 'effective terminal
hydrophobe size' for viscosity enhancement, but
in internal placements (i.e., as coupling units for
6000 molecular weight oxyethylene chains) these
units do not contribute to solution viscosities
(Kaczmarski and Glass 1992). Broadening of
molecular weight distributions, as observed in
classical step-growth polymerizations, diminishes
the viscosifying properties of HEUR thickeners
(May, Kaczmarski and Glass 1993). In coating
formulations, the associative thickener may also
interact with the latex and pigment but competes
Figure 5.29 Effect ofmolecular weigth ofpoly(ethylene in this interaction with a variety of surfactants
oxide) thickeners on filamentation in roller application present.
at 2 ft/s ofa latex paint at 120 Krebs Units (Glass, 1978c). If the surface of Ti0 2 is pretreated with an
Figure 5.30 High-speed flow visualization at 2000 framesls of film split in roller application at 37 ft/s of latex
paint thickened with poly(ethylene oxide) at M w = 40 X 10 5 (Glass, 1978c). The time intervals given in the
figure are in milliseconds.
oligometric polyacid that contains hydrophobic bution is difficult to quantify because of the many
entities and acts as a dispersant, HEUR thickeners equilibria with the different types of surfactants
will 'adsorb' on the pigment through hydrophobic that are present in coating formulations and
interactions with the dispersant (Lundberg and because of the variability in surface stabilizers
Glass 1992b), even in the presence of other and the median size ofthe latex. For that reason,
surfactants in the formulation (Kaczmarski, Glass a comprehensive study of such interactions has
and Buchacek 1993). The impact ofthis interaction not been reported to date. Associative thickeners
on the viscosity of the coating formulation is in the HEUR family have been shown to contribute
minimal, but the interaction stabilizes the pigment to the viscosity at high shear rates in much the
and tends to promote higher gloss in the applied same manner as observed with hydroxyethyl
film. The interaction of associative thickeners cellulose thickeners: the more thickener added
with latices could provide a higher dispersion the higher the viscosity. For HEUR thickeners,
viscosity via an increased effective volume fraction this effect arises from a reduction in the number,
of particles in the system. However, this contri- spacing (Kaczmarski and Glass 1993) and size
of terminal hydrophobes (Howard et al. 1991; In practical paper coating applications, a

Brown and Glass 1989). The high-shear viscosity primary problem arises from the need to increase
obtained by blending different HEUR thickeners the solids content of the formulation in order to
is not linear with concentration, but blends of cope with dryer capacity limitations when
high-molecular-weight hydrophobe-modified production rates are raised. An increase in solids
hydroxyethyl cellulose and an HEUR thickener content increases the viscosity which, in turn,
provide linear relationships (Anwari and Schwab can give rise to considerable difficulties in the
1989). blade application step, such as 'weeping', spraying
or blade 'whispering' (see Chapter 12d for an in-
depth discussion).
5.4.3 PAPER COATINGS Low-molecular-weight water-soluble polymers,
Even though the latex content in paper coatings such as starch or carboxymethyl cellulose, are
is typically quite small (cf Section 5.2.1), the latex usually added to paper coating formulations.
appears to have a significant influence on the They decrease the rate of drying immediately
formulation's 'runnability' (Barber and Bartel after application while at the same time raising
1976), a loosely defined term that reflects a the viscosity, which retards penetration of the
shear-thickening behavior under deformation, coating into the porous cellulose fiber substrate.
sometimes also referred to as dilatancy. The The addition of low-molecular weight thickener
dilatant behavior of monodisperse, well charac- polymers contributes minimally to the elongational
terized latices without steric stabilizers has been viscosity of the formulation. However, the elastic
described by Hoffman (1972, 1974). In paper response of paper formulations that results from
coatings, the latex is stabilized by oligomeric the interaction of clays with starch or car-
methacrylic surface segments, and both types of boxymethyl cellulose could be considerable and
latices used in paper coatings, styrene/butadiene warrants further, detailed study (e.g., Preston,
copolymers and vinyl acetates, have been described McGenity and Husband (1993». Hydrophobically-
as superior in runnability. However, neither latex modified, alkali-swellable emulsion (HASE)
has been studied by state-of-the-art rheometers thickeners have achieved some success in paper
with acceptable geometries, alone or conjunction coatings (Roper et al. 1995). These thickeners
with the clays used in paper coatings to probe bring about viscosity increases without a notable
for specific latex-clay interactions. increase in the elastic component of the solution
The clay is usually the primary component in (Tarng and Glass 1992). They may also reduce
paper coatings and finely dispersed clays such the electro-viscous response of the clays, in part
as kaolinite, montmorillonite or attapulgite are by adsorbing onto the clays and thereby stabilizing
notably shear thinning and often also give rise them from flocculation into a 'house-of-cards'
to substantial elastic responses (Lundberg et al. arrangement (Heinle, Shah and Glass 1986). This
1990). This is illustrated in Fig. 5.31 by the latter effect is under further investigation (Tarng
low-frequency oscillatory response of a latex 1995).
coating containing attapulgite (Glass 1975). Prior
to and within 30 s of the oscillatory test, the
5.4.4 OEM COATINGS
formulation was subjected to an intermediate
shear rate to break up loose aggregates and reach Even though the composition of water-reducible
an equilibrium viscosity. The deformation was acrylic resins is similar to that of latices used in
then dropped into a low-amplitude oscillatory paper or architectural coatings (cf Section 5.2.2),
mode. The immediate, substantial low frequency their response to dilution with water is unusual.
response of the storage modulus reveals the 'Water-reducible' dispersions are water-swollen
dominance ofelasticity in formulations with high during dilution, giving rise to an abnormal
clay content. viscosity maximum as illustrated in Fig. 5.32
g
.
0
0
o 0
'"
..,'"
o 0
o 0
o
'"
.,
0
'" '"
o 0
IG*I
..
o
..
0
.,
o
\
\ / 0
'"
\
o 0
o \ 0
o \ 0
~o 0
C\I
E
..
\ o '" a>
o \ Ul
'0
"Cg 0
>. .
"0 0
\. o
0
0
9::-
~:::
Ul ~J ~ \ 0
~
~
::l 'iii
::;0
"00
(7']' '\ 0
0
0
o . '\ 0
0
Ul
~Z '\
0
~
:>
"-
"-
"- ,
o
o 0
0
o
to 0
.,
G'-- " ...... .....
o 0
o 0
.,
o ,
0
o <:>
o 0
9J+.-00---'-20-.O-O------,~Or-.-O-O- -6"'-0-.o-o------,eor-.-oo--I.,...O-O.-00--'12-0-.0 -0--+.~ni. 00
Time[s]
Figure 5.31 Time dependence, after shearing, of complex modulus, loss modulus and storage modulus of an
architectural coating thickened with attapulgite clay at 2.6% by weight (Glass, 1975).
(Richards 1977). Such a maximum is not evident thickened with nonadsorbing hydroxyethyl cel-
upon dilution of latex dispersions, or dilution of lulose. This so-called syneresis increases with
solvent-based resins used in coatings with oligomer increasing particle size. The latex size effect
synthesis. Figure 5.33 shows that the molecular suggests depletion flocculation of a poorly
weight of the resin and its acrylic acid content stabilized dispersion as a mechanism. When
influence the magnitude of the maximum (Richards aqueous polyurethane dispersions are thickened
1977). In the vicinity of the maximum the with hydrophobically-modified, ethoxylated
dispersions are shear thinning, but detailed urethane (HEUR) thickeners, on the other hand,
rheological studies of such systems have not been phase separation is not observed. Commercial
reported. The molecular weight of the disperse HEUR thickeners that exhibit high loss modulus
phase influences the misting behavior during roll responses with conventional latices (Lundberg et
application to metal substrates but, again, there al. 1991) exhibit high storage modulus responses
are no comprehensive rheological studies pub- with aqueous polyurethane dispersions: see Fig.
lished, for either clear or pigmented formulations. 5.34 (Kaczmarski, Fernando and Glass 1993). To
Polyurethane-based aqueous dispersions, or elucidate the mechanism responsible for this
PURADs, exhibit phase separation when difference, model systems without the high
Concluding remarks 177
10.000 SO/50 acrylic acidlbulyl methacrylate

10.000 copolymer ratio
Wa er
1000
VI t-Butyl
cil alcohol
c..
.s
~ 100
'in
0
u
Ul
:>
/
/
/
/
10 /
/ 100
/
/
10 20 30 4() SO 60
Polymer Concentration [%1 10 L-_--'-_ _--..1_ _--..1_ _--'-_--.J

10 20 30 40
Polymer Concentration [%J
Figure 5.32 Viscosity dependence on polymer con-
centration for a latex and for water and t-butyl alcohol
solutions of a 10/90 acrylic acid/butyl methacrylate Figure 5.33 Viscosity dependence on polymer con-
copolymer with 75% effective neutralization with centration for dilutions in water of acrylic acidjbutyl
dimethylaminoethanol (Richards, 1977). methacrylate with different copolymer ratios and 75%
effective neutralization with dimethylaminoethanol
(Richards, 1977).
cosolvent contents typical of commercial poly-
urethane dispersions are under investigation
(Chen 1997). simpler model flows that mimic essential features
ofthe application flow of interest. As this chapter
also makes plain, the connections between
5.5 CONCLUDING REMARKS
formulation, rheology and flow behavior can be
As the case studies in this chapter illustrate, very subtle and material specific and, hence, it
establishing the interplay among formulation is very difficult if not impossible to extract very
components, rheological behavior and ultimate general and broad conclusions from most inves-
coating application performance requires judicious tigations in coating rheology. Nevertheless, with
selection of the chemical constituents and detailed sufficient diligence and effort invested into a
study of possible associations between the particular case study, conclusions and insights
constituents; careful selection of the appropriate of some general validity typically emerge, as is
material functions and comprehensive and some- the case for the water-borne latex coatings featured
times tedious rheological measurements of these in detail in this chapter.
material functions; and systematic application In the rheology of water-borne latex coatings,
testing in either actual coating configurations or the median particle size of the latex plays an
100
80
60
e::.
(Q
l/l
;;)
"5
"0
~ 40
20
0.1 10
Frequency [S-l]
Figure 5.34 Storage modulus G' (circles) and loss modulus G" (triangles) as a function of frequency for 20 wt%
polyurethane-based aqueous dispersions (PURAD) with 0.6 wt% hydrophobe-modified ethoxylated urethane
(HEUR) thickeners (after Kaczmarski, Fernando and Glass, 1993). Closed symbols: HEUR thickener with
dicyclohexylmethane disocyanate end groups; open symbols: HEUR thickener with hexamethylene diisocyanate
end groups.
important role, similar to the influence of the but the amount of thickeners used to achieve an
size of the disperse phase on the rheological 'intermediate formulation viscosity' can have a
response ofpolymer melts. The viscosity increases dramatic effect at high shear rates. Higher
with decreasing size of the disperse phase and viscosities at high shear rates are achieved by
influences phenomena associated with shear using lower-molecular-weight conventional
deformations (i.e. leveling and thickness of applied thickeners which allow greater amounts to be
film). Water-soluble polymer thickeners may added to the formulation. Among a new class of
playa subordinate role in the rheological behavior surfactant-modified thickeners, higher viscosities
of coating formulations containing larger particle can be achieved with a thickener of a given
size latices, yet the thickener may dominate the molecular weight by decreasing the size of the
rheology if there are chemical associations among terminal hydrophobic group. Surfactant-modified
its macromolecular chains. Chemical associations polymers thicken through hydrophobic associ-
predominantly influence the viscosities at low ations, and can also be used to lower the viscosity
shear rate and thus affect the leveling or 'flow-out' of coating formulations containing small-particle
behavior of applied coatings. Raising the viscosity latices, because the latter contains an excess
at high shear rates (> 104 S - 1) increases the surfactant.
coated film thickness and, hence, the hiding In roll coating applications, the onset ofribbing
power of the applied film. Through their influence and the ensuing spatter and misting behavior are
on film thickness, high-shear-viscosities also affect influenced by the dynamic uniaxial elongational
the leveling of the coating after application. The viscosity (DUEV) of the coating formulation, but
size and particle size distribution of the latex do not the elastic functions associated with shear
not have much of an effect at high shear rates, deformations. This rheological response is deter-
References 179
mined primarily by the chemical composItIon viscosity model to the onset ofviscosity enhancement
and molecular weight ofthe thickener. Surfactant- in solutions of linear HM W-SP's. Proc. Am. Chem.
modified, water-soluble polymers are an effective Soc. Div. Polym. Mater.: Sci. Eng. 61: 634-638.
Chen, M. 1997. Synthesis and adsorption behavior of
way to lower the DUEV of coating formulations
model pigment dispersants. PhD thesis (in prepara-
and thereby minimize spatter and misting during tion), North Dakota State University, Fargo, North
roll application. The viscosities at low and high Dakota.
shear rates and the DUEV of surfactant-modified, Chong, J. S., Christiansen, E. B. and Baer, A. D. 1971.
water-soluble polymers are influenced by the Rheology of concentrated suspensions. J. Appl.
amount of surfactant present in the formulation. Polym. Sci. 15: 2007-2021.
Surfactant-modified, water-soluble polymers are Clark, 1., Wellinghoff, S. T. and Miller, W. G. 1983.
Rheological properties ofpolystyrene-carbon disulfide
also observed to inhibit the depletion flocculation
gels. ACS Polym. Preprints. 24: 86-87.
of step-growth aqueous dispersions. Craig, D. H. 1986. Grafting reactions of (hy-
droxyethyl)cellulose during emulsion polymerization
of vinyl monomers. Chapter 18 in Water-Soluble
REFERENCES
Polymers: Beauty with Performance, ed. J. E. Glass,
Anklam, M. R., Warr, G. G. and Prud'homme, R. K. Advances in Chemistry Series, Volume 213, pp.
1994. The use of opposed nozzles configuration in 351-368. Washington, DC: American Chemical
the measurements of the extensional rheological Society.
properties of emulsions. J. Rheology. 38: 797-810. Dodge, J. S. 1972. Quantitative measures of leveling.
Anwari, F. M. and Schwab, F. G. 1989. Optimizing J. Paint Technol. 44(564): 72-78.
latex paint rheology with associative thickeners. Ferry, J. D. 1980. Viscoelastic Properties of Polymers,
Chapter 23 in Polymers in Aqueous Media: Perform- 3rd edn. New York: Wiley.
ance through Association, ed. J. E. Glass, Advances Garey, C. L. 1977. Physical Chemistry of Pigments in
in Chemistry Series, Volume 223, pp. 527-542. Paper Coating. Atlanta, GA: TAPPI Press.
Washington DC: American Chemical Society. Gaskell, R. E. 1950. The calendering ofplastic materials.
Arney, W. C. and Glass, 1. E. 1976. The role played J. Appl. Mech. (Trans. ASME). 17: 334-336.
by water-soluble polymers in paint performance. Glass, J. E. 1975. The role played by water-soluble
Part III: molecular weight and concentration effects polymers in paint performance. J. Oil Col. Chem.
in mixed thickener studies. J. Oil Col. Chem. Assoc. Assoc. 58: 169-177.
59: 372-378. Glass, J. E. 1976. The role played by water-soluble
Barber, E. J. and Bartel, D. G. 1976. Soluble polymers polymers in paint performance. Part II: Chemical
in air knife coating. Tappi. 59(6): 123-126. modeling studies. J. Oil Col. Chem. Assoc. 59: 86-94.
Bauman, T., Sullivan, T. and Middleman, S. 1982. Glass, 1. E. 1978a. Dynamics of roll spatter and
Ribbing instability in coating flows: effect ofpolymer tracking. Part I: Commercial latex trade paints. J.
additives. Chem. Eng. Commun. 14: 35-46. Coatings Technol. 50(640): 53-60.
Bierman, M. 1968. Das Fliessen ebener Schichten Glass, J. E. 1978b. Dynamics of roll spatter and
inkompressibler viskoelastischer Fliissigkeiten mit tracking. Part II: Formulation effects in experimental
einer in Ablaufrichtung infinitesimal gestorten paints. J. Coatings Technol. 50(640): 61-68.
Oberftiiche. Rheol. Acta. 7: 138-165. Glass, J. E. 1978c. Dynamics of roll spatter and
Bierwagen, G. P. 1975. Surface dynamics of defect tracking. Part III: Importance of extensional
formation in paint films. Prog. Org. Coatings. 3: viscosities. J. Coatings Technol. 50(641): 56-71.
101-113. Glass, J. E. 1978d. Dynamics of roll spatter and
Bierwagen, G. P. and Hay, T. K. 1975. The reduced tracking. Part IV: Importance of G* recovery and
pigment volume concentration as an important N 1 in tracking. J. Coatings Technol. 50(641): 72-78.
parameter in interpreting and predicting the prop- Glass, J. E. 1984. Perspectives in thicker development
erties of organic coatings. Prog. Org. Coatings. 3: for water-borne coatings. Amer. Paint J. August 6th
281-303. issue, pp. 45-47.
Bird, R. B., Armstrong, R. C. and Hassager, O. 1987. Greener, 1., Sullivan, T., Turner, B. and Middleman,
Dynamics of Polymeric liquids - Volume 1: Fluid S. 1980. Ribbing instability of a two-roll coater:
Mechanics 2nd edn. New York: Wiley-Interscience. Newtonian fluids. Chem. Eng. Commun. 5: 73-83.
Brown, R. G. and Glass, 1. E. 1989. Application of a Hakim, V., Rabaud, M., Thome, H. and Couder, Y.
1988. Directional growth in viscous fingering. In Media: Performance through Association, ed. 1. E.
New Trends in Nonlinear Dynamics and Pattern Glass, Advances in Chemistry Series, Volume 223,
Forming Phenomena: The Geometry ofN onequilibrium, pp. 495-525. Washington, DC: American Chemical
Proceedings of the NATO Advanced Research Society.
Workshop, Cargese, eds P. Coullet and P. Huerre. Karunasena, A. and Glass, J. E. 1989. Interactions in
Heinle, S. A., Shah, S. and Glass, 1. E. 1986. Influence associative thickenerj220nm methyl methacrylate
of water-soluble polymers on the filtration control latex dispersion. Prog. Org. Coatings. 17: 301-320.
of bentonite muds. Chapter II in Water-Soluble Karunasena, A. and Glass, 1. E. 1992. The effect of
Polymers: Beauty with Performance, ed. 1. E. Glass, latex particle size on the rheology of thickened
Advances in Chemistry Series, Volume 213, pp. dispersions. Prog. Org. Coatings. 21: 53-67.
183-196. Washington DC: American Chemical Khan, S. A. and Prud'homme, R. K. 1987. Melt
Society. rheology of filled thermoplastics. Rev. Chem. Eng.
Hoffman, R. L. 1972. Discontinuous behavior in 4: 205-270.
concentrated suspensions. Trans. Soc. Rheol. 16: Klenin, V. J., Klenina, O. V., Kolnibolotchuk, N. K.
155-173. and Frenkel, S. Va. 1973. Structural levels of
Hoffman, R. L. 1974. Discontinuous and dilatant supermolecular organization in aqueous solutions
behavior in concentrated suspensions. II. Theory ofpoly(vinyl alcohol). J. Polym. Sci.: Polym. Sympos.
and experimental tests. J. Colloid Interface Sci. 46: 42: 931-942.
491-506. Kreider, R. W. 1964. Factors controlling latex paint
Howard, P. R., Leasure, E. L., Rosier, S. T. and rheology and their measurement. Official Digest -
Schaller, E. 1. 1991. Systems approach to rheology J. Paint Techno/. Eng. (Federation of the Societies
control. Chapter 12 in Polymers as Rheology for Paint Technology). 36(478): 1244-1260.
Modifiers, eds 1. E. Glass and D. N. Schulz, ACS Krieger, I. M. 1972. Rheology of monodisperse latices.
Symposium Series, Volume 462, pp. 207-221. Adv. Colloid Interface Sci. 3: 111-136.
Washington, DC: American Chemical Society. Lambourne, R. 1987. Paint and Surface Coatings:
Huppler, J. D., Macdonald, I. F., Ashare, E., Spriggs, Theory and Practice. Chichester, England: Ellis
T. W., Bird, R. B. and Holmes, L. A.1987. Rheological Horwood.
properties of three solutions. Part II. Relaxation Larson, R. G. 1988. Constitutive Equationsfor Polymer
and growth of shear and normal stresses. Trans. Melts and Solutions. Boston: Butterworths.
Soc. Rheol. 11: 181-204. Laun, H. M. 1978. Description of the non-linear shear
Kaczmarski, 1. P., Fernando, R. H. and Glass, 1. E. behavior of a low density polyethylene melt by
1993. Interactions of HEUR associative thickeners means of an experimentally determined strain
with waterborne polyurethanes. J. Coatings Technol. dependent memory function. Rheol. Acta. 17: 1-15.
6(818): 39-46. Lundberg, D. J., Alahapperuna, K., Fernando, R. H.
Kaczmarski, J. P. and Glass, J. E. 1992. Structure and Glass, J. E. 1990. Viscoelastic behavior in paper
property relationships in HEUR associative coating formulations containing associative
thickeners: influence of the diisocyanate coupler on thickeners. Proc. TAPPI Coatings Cmiference(Boston,
aqueous solution rheology. Proc. Am. Chem. Soc. May 1990), pp. 479-489. Atlanta, GA: TAPPI Press.
Div. Polym. Mater.: Sci. Eng. 67: 282-283. Lundberg, D. 1. and Glass, 1. E. 1992a. Pigment
Kaczmarski, J. P. and Glass, 1. E. 1993. Synthesis and stabilization through mixed associative thickener
solution properties of hydrophobically modified interactions. J. Coatings Technol. 64(807): 53-61.
ethoxylated urethanes with variable oxyethylene Lundberg, D. 1. and Glass, 1. E. 1992b. Titanium
spacer lengths. Macromolecules. 26: 5149-5156. dioxide 'association' with HEUR thickeners through
Kaczmarski, J. P., Glass, 1. E. and Buchacek, R. 1. model dispersant interactions. Proc. Am. Chem. Soc.
1993. Development of SEC technique for analysis Div. Polym. Mater: Sci. Eng. 66: 21-22.
of competitive adsorption of surfactant, HEUR Lundberg, D. 1., Glass, 1. E. and Eley, R. R. 1989.
associative thickeners and oligometric carboxylate Model-HEURjsurfactantjmodellatex matrix rheol-
dispersants on titanium dioxide. Proc. Am. Chem. ogy studies. Proc. Am. Chem. Soc. Div. Polym.
Soc. Div. Polym. Mater.: Sci. Eng. 69: 199-200. Mater.: Sci. Eng. 61: 533-538.
Karunasena, A., Brown, R. G. and Glass, J. E. 1989. Lundberg, D. 1., Ma, Z., Alahapperuna, K. and Glass,
Hydrophobically modified ethoxylated urethane 1. E. 1991. Surfactant influences on hydrophobically
architecture - importance for aqueous- and dispersed- modified thickener rheology. Chapter 14 in Polymers
phase properties. Chapter 26 in Polymers in Aqueous as Rheology Modifiers, eds 1. E. Glass and D. N.
References 181
Schulz, ACS Symposium Series, Volume 462, pp. formation of polyacrylamide/chromium (III) gels.
234-253. Washington DC: American Chemical Soc. Petrol. Eng. J. 23(5): 804-808.
Society. Rabaud, M., Michelland, S. and Couder, Y. 1990.
Ma, Z. 1992. Associative thickener disperse systems. Dynamical regimes of directional viscous fingering:
PhD thesis, North Dakota State University, Fargo, spatiotemporal chaos and wave propagation. Phys.
North Dakota. Rev. 1£lt. 64: 184-187.
Ma, Z., Lundberg, D. J., Roberts, S. and Glass, 1. E. Richards, B. M. 1977. Viscosity of water reducible
1993. Phase behaviors and film properties of acrylic copolymers. MS thesis, North Dakota State
dispersions and coatings containing associative and University, Fargo, North Dakota.
conventional thickeners. J. Appl. Polym. Sci. 49: Roper, J., Chonde, Y., Salminen, P. and Pollok, M.
1509-1527. 1995. Determining the rheological and water retention
Massouda, D. F. 1985. Analysis and prediction of properties of various thickeners for paper coatings.
roll-spatter from latex paints. J. Coatings Technol. Proc. Am. Chem. Soc. Div. Polym. Mater.: Sci. Eng.
57(722): 27-36. 73: 52-53.
Matsuda, T. and Brendley, W. H. 1979. Transfer and Ruschak, K. J. 1985. Coating flows. Ann. Rev. Fluid
leveling processes and rheological design in water- Mech. 17: 65-89.
borne reverse-roll coating systems. J. Coatings Saffman, P. G. and Taylor, G. I. 1958. The penetration
Technol. 51(658): 46-60. of a fluid into a porous medium or Hele-Shaw cell
May, R., Kaczmarski, 1. P. and Glass, J. E. 1993. containing a more viscous liquid. Proc. Royal Soc.
Influence of molecular weight distributions of Lond. A. 245: 312-329.
associative thickener on solution rheology. Macro- Sato, T. and Ruch, R. 1980. Stabilization of Colloidal
molecules. 29: 4745-4753. Dispersions by Polymer Adsorption. New York and
Meissner, 1. 1985. Experimental aspects in polymer Basel: Marcel Dekker.
melt elongational rheometry. Chem. Eng. Commun. Schaller, E. 1. 1968. Critical pigment volume concen-
33: 159-180. tration of emulsion based paints. J. Paint Technol.
Mill, C. C. 1967. The behavior of liquids on rotating 40(525): 433-438.
rollers. J. Oil Col. Chem. Assoc. 50: 396-406. Schoff, C. K. 1988. Rheological measurements. In
Mill, C. C. and South, G. R. 1967. Formation of ribs Encyclopedia of Polymer Science and Engineering,
on rotating rollers. J. Fluid Mech. 28: 523-529. Volume 14, 2nd edn., eds H. F. Mark et aI., pp.
Miller, J. C. and Myers, R. R. 1958. A photographic 454-451. New York: Wiley.
study of liquid flow in a roll nip. Trans. Soc. Rheol. Schunk, P. R. and Scriven, L. E. 1990. Constitutive
2: 77-93. equation for modeling mixed extension and shear
Myers, R. R. and Hoffman, R. R. 1961. The distribution in polymer solution processing. J. Rheol. 34:
of pressure in the roll application of Newtonian 1085-1119.
fluids. Trans. Soc. Rheol. 5: 317-328. Smith, N. D. P., Orchard, S. E. and Rhind-Tutt, A. J.
Napper, D. N. 1983. Polymer Stabilization and Dispersion 1961. The physics of brushmarks. J. Oil Col. Chem.
Stability. New York: Academic Press. Assoc. 44: 618-633.
Nielsen, L. E. 1977. Polymer Rheology. New York: Soules, D. A., Fernando, R. H. and Glass, J. E. 1988.
Marcel Dekker. Dynamic uniaxial extensional viscosity (DUEV)
Odian, G. 1993. Principles of Polymerization. New effects in roll application. I. Rib and web growth in
York: John Wiley & Sons. commercial coatings. J. Rheol. 32: 181-198.
Orchard, S. E. 1962. On surface leveling in viscous Suetsugu, Y. and White, J. L. 1983. The influence of
liquids and gels. Appl. Sci. Res. All: 451-464. particle size and surface coating ofcalcium carbonate
Ottewill, R. H. 1982. The stability and instability of on the rheological properties of its suspensions in
polymer latices. Chapter I in Emulsion Polymerization, molten polystyrene. J. Appl. Polym. Sci. 28:
ed. I. Piirma, pp. 1-49. New York: Academic Press. 1481-1501.
Preston, D. 1., McGenity, P. M. and Husband, 1. C. Tanner, R. I. 1985. Engineering Rheology. Oxford:
1993. The use of rheological techniques in the study Clarendon Press.
ofinteractions between components ofpaper coating Tarng, M.-R. 1995. Interactions of water-soluble
colours. In Thin Film Coating, ed. H. Benkreira, pp. polymers with surfactants and disperse systems.
23- 33. Cambridge: The Royal Society of Chemistry. PhD thesis, North Dakota State University, Fargo,
Prud'homme, R. K., Uhl, 1. T., Poinsatte, 1. P. and North Dakota.
Halverson, F. 1983. Rheological monitoring of the Tarng, M.-R. and Glass, 1. E. 1992. The role of
surfactant in RASE thickener rheology. Proc. Am. suspensions in a spinning motion. Int. J. Multiphase
Chem. Soc. Div. Polym. Mater.: Sci. Eng. 67: Flow. 1: 465-486.
280-281. Whorlow, R. W. 1980. Rheological Techniques. New
Van Wazer,J.R.1963. Viscosity and Flow Measurements. York: Wiley.
New York: Inter Science Publishers. Wildemuth, C. R. and Williams, M. C. 1985. A new
Walters, K. 1980. Rheometry. New York: Wiley. interpretation of viscosity and yield stress in dense
Weinberger, C. B. and Goddard, J. D. 1974. Extensional slurries: coal and other irregular particles. Rheo/.
flow behavior of polymer solutions and particle Acta. 24: 75-91.
THE FATE OF THIN LIQUID FILMS 6
AFTER eOATING
H aroon S. Kheshgi
6.1 INTRODUCTION
The fate of liquid films is determined by their and to dry patch formation, if the liquid does
state following application, and their environment. not wet the underlying substrate. Films containing
In most coating operations the region of film solvents must be dried. Drying can lead to
application is spatially separated from that
following application. Exceptions do exist in
operations such as spin coating where application,
leveling, and drying all occur simultaneously as
is described in Chapter 14. Phenomena of Film received from
coating operation
importance after application are caused by a
largely different set of mechanisms than do those
of importance in coating application, which is
the primary emphasis of this book. During Leveling
application, the liquid is often severely deformed J1U~t§~~
I
through a complex geometry and the resulting
fluid mechanics often requires sophisticated
computational methods to analyze. After appli-
Sagging
cation liquid films are usually, and often desirably,
of uniform thickness - a simple geometry; the Drying
film, however, is composed of material of
!
increasingly complex properties as the film is set
and dried. ~ Orange peel
eee:.,
Stress
During the transition of a recently applied film buildUp ~
to a 'dry' solid film, the film may undergo a
~~/
wealth of experiences as illustrated in Fig. 6.1.
Irregularities in film thickness originated during
application or by ambient disturbances can heal ............
by capillary leveling, if there is time before the ~~
film becomes too viscous or undergoes transition Cracking Curling
to a solid. Alternatively, large amplitude irregu- Figure 6.1 Phenomena that films undergo after coating
larities in film thickness can lead to film rupture application.
London. ISBN 0 412 06481 2.
184 The fate of thin liquid films after coating
composition variation with depth into the film. which occur simultaneously in coatings after
Drying also leads to liquid/solid transition. Once application.
solid, the film may still continue to dry, and This chapter first addresses leveling of films
shrink, leading to stress build-up within the film of uneven thickness by capillary pressure - the
which is the cause of a variety of coating defects. normal resultant of surface tension in a curved
Experiences are governed by a diverse group interface. Surface leveling is an essential and
of physical mechanisms. Fluid mechanics is crucial beneficial phenomenon in, for example, most
in healing of nonuniform films or breaking of paint coating operations where uneven films are
unstable ones. Diffusion and phase behavior in applied. The leveling of nonuniformities in simple
complex media, chemical kinetics in reacting Newtonian liquid films is a well understood
films, and boundary layer theory define phenom- phenomenon; however, in practice the effects of
ena during drying and setting. Mechanical gravity, film rheology, curved solid substrate
properties as a function of evolving composition shape and variations in surface tension brings
and temperature are needed to analyze stress build- us to the current state of research. In the following
up and relaxation in solidifying films, in order section the phenomenon of film dewetting is
to predict the final state of stress in the film considered. Local thin or dry regions sometimes
which, if severe enough, leads to modes of film form in otherwise uniform films. These defects
failure such as peeling, cracking or curling (see go by the names of pinholes or craters and can
Fig. 6.1). Understanding all of these phenomena limit how thin one can coat. A connection is
becomes a challenge when we realize that the sought between the long-range intermolecular
evolving film is a complex media, e.g., a polymeric forces which constitute the effect of disjoining
film or a film of suspended solids undergoing pressure (see Chapter 2) and this phenomenon.
liquid to solid transition. Next, the phenomena important in film drying
Many defects arrive after coating application are considered. Drying technologies (see, for
as the ultimate result of coating formulation and example, Cohen (1992a and b)) speed drying by
the way in which the coated film is treated after application of heat and mixing of the gas overlying
coating. While it is beyond the scope of this the film. However, diffusion limitation or unwanted
chapter to examine the entire catalogue ofspecific phase separation in the film, or film disturbance
defects, many of which are still not understood by gas flow or by surface tension gradients can
or even well defined, the phenomena which are present operability limits on the maximum rate
thought to cause the majority of defects will be of drying. Finally, once the film has passed
addressed. It is hoped that applying understanding through the sometimes-ambiguous transition from
of these phenomena to specific defect problems liquid to solid, it can still shrink by further drying.
will lead to ways of removing them by optimal Shrinkage in a solid film leads to the build-up of
choice ofcoating formulation and post-application stress which can result in the defects shown in
environment. Formulation could be altered for Fig. 6.1. And while the effects of film stress are
example by addition of solvent, plasticizer reduced in practice by altering film formulation
(non-volatile solvent), surfactants or treatment or drying/curing conditions, the theory for doing
of the coating substrate. Environment could be so is still in its infancy.
altered by changing temperature, residence time,
overlying gas flow or solvent vapor pressure.
This wide range of choices offers both the
6.2 SURFACE LEVELING: HEALING OF FILM
opportunity to arrive at favorable coating
IRREGULARITIES
properties, or those that are not so favorable,
such as defects. It remains the challenge of the Coating operations often leave films of irregular
coating engineer to make these choices with thickness as illustrated in Fig. 6.2. Instabilities
proper consideration of the many phenomena in the coating operation such as brushmarks in
Surface leveling: healing of film irregularities 185
less, is often a critical aspect of film quality. To
achieve a uniform film thickness, the surface
smoothing effect of capillarity may be required
to heal film irregularities. There is, however,
limited time between coating application and the
point at which the film is immobilized by, for
(a) example, drying; therefore, the leveling rate of
an irregular surface is crucial. While films often
have complex rheological behavior and may be
nonhomogeneous, it is instructive to first consider
the ideal case ofleveling ofa Newtonian liquid film.
Horizontal films ofliquid that wet the underlying
solid are absolutely stable: any disturbance decays
to leave ultimately a quiescent flat film. There
are two mechanisms that can act. Capillarity
(b) straightens a curved free surface. Gravity levels
an inclined free surface. Either or both set
disturbed liquid films in motion. Once in motion,
inertia tends to maintain the momentum ofliquid
in motion while viscous stresses tend to distribute
the momentum uniformly over the entire mass
and to transmit it to the rigid horizontal surface
on which the film sits. How disturbances evolve
depends on which of these forces dominates the
(c) situation.
Disturbances of infinitesimal amplitude in
Figure 6.2 Sources of film irregularities: (a) coating
Newtonian liquid decay exponentially to leave
applicator instability: ribbing; (b) local disturbances: a flat, quiescent film provided the viscosity is
impact ofsurfactant aerosol particle; and (c) irregularities high, the film thickness h is small or the horizontal
inherent to the coating method: powder coating. length scale of the disturbance is great compared
with the film thickness. The evolution of a
disturbance of infinitesimal amplitude, i.e.
painted films (Smith, Orchard and Rhind-Tutt I> == a/h o--+ 0 where a is the amplitude and ho is
1961), or ribbing instabilities in roll (Pearson the undisturbed film thickness, which is chosen
1960; Pitts and Greiller 1961; Coyle 1984) or slot as the unit of film thickness measure, is identical
(Bixler 1982) coated films lead to nearly-sinusoidal to the evolution of the sum of the Fourier
two-dimensional variations in film thickness. components of the disturbance. It has long been
Defects caused by scratching, edge effects, bubbles, known (e.g. Lamb (1945)) that the exponential
particulates, electrostatic forces, or surface tension decay of a sinusoidal disturbance is exact in the
gradients can leave localized irregularities in film asymptotic limit of infinitesimal amplitude (I> --+ 0)
thickness oflarge amplitude (Hansen and Pierce and small Reynolds number (Re--+O). The
1974; Gutoff 1993). The inherent nature of the Reynolds number Re == ph6a/3J1.2J3 where
coating operation produces a film of irregular 3J1./4/ah~ is the unit of time measure to, P is
film thickness in, for example, powder coating density and J1. is viscosity, 2nl is the wavelength
(Spitz 1973; Anshus 1973; de Lange 1984), or of the sinusoidal disturbance as seen in Fig. 6.3,
gravure coating (Munter 1990). and I is chosen as the unit of film length measure.
Uniformity ofthe final coated product, nonethe- The evolution of two-dimensional disturbances
sinusoidally oscillates without loss of amplitude:
it executes a standing wave. Liquids, however,
are not inviscid fluids; damped standing waves
are observed (cf Lamb 1945; Batchelor 1967).
The theory for decay ofinfinitesimal disturbances
for intermediate values of Reynolds number was
given by Wehausen and Laitone (1960; see also
Kheshgi 1989). Kheshgi (1989) shows that inertial
effects are negligible for aRe « 1, which is true
for essentially all coated films.
In horizontal films the leveling effect of
x capillarity dominates that ofgravity when Bo « a2
which is the case for irregularities of length less
Figure 6.3 Sinusoidal film irregularity. than the capillary length - i.e. 1« (U/pg)1 /2. In
this instance (6.1) can be approximated by
considered here is (Orchard 1962):
H(X,T) = 1 +esin(X)
H(X,T) = 1 + esin(X)
3 tanh(a) - asech 2 (a)
exp [ - -3
J
exp[_~(tanh(a) -
2
a SeCh (a)) (1 + Bo/(
2
)TJ 2a 1 + a 2 sech 2 (a)
T (63)
.
2 1 + a 2 sech 2 (a) a3
In the limit of long irregularity length compared
(6.1) with film thickness, a « 1, (6.3) can be further
simplified to
where the wavenumber a == ho/I, and the Bond
number Bo == h~gp/u is the ratio of gravitational H(X, T) = 1 + esin(X)exp( - T) (6.4)
force to capillary force, u is the surface tension
Although this equation gives only the evolution
of the film free surface, 9 is the acceleration of
of a sinusoidal disturbance, the decay of a small
gravity, H == h/h o' X == x/I, and T == t/t o' Anshus
amplitude irregularity of arbitrary shape can be
(1973) has since extended (6.1) to predict the
represented as the linear superposition of a series
evolution ofthree-dimensional, doubly-sinusoidal
of Fourier components. Since the decay time to
disturbances.
ofthe Fourier components is proportional to the
When the effects of liquid inertia become
fourth power of the component wavelength (when
sizable in relation to viscous stresses, the evolution
a is small), the longest wavelength component is
of small disturbances changes. Not always does
the most persistent.
a disturbance decay exponentially: it may instead
To get a feel for the time required for healing
overshoot the flat rest state to eventually approach
of disturbances by surface leveling, consider the
periodically a flat film. In the asymptotic limit
following example. A typical (Cohen 1992a) slide
of infinitesimal amplitude (e -+ 0) and vanishing
coating operation could have a coating speed of
viscous stress (Re-+ 00), a solution has again
3 mis, viscosity p. = 50 cp, surface tension
long been known (Lamb 1945; Whithan 1974):
u = 50 dynes/cm, and film thickness ho = 1()() p.m.
H(X, T) = 1 + esin(X) The time scale for capillary-driven leveling of a
long-wavelength irregularity in film thickness is
tanh(a) 1 + Bo/a 2 ] 112 }
to == 3p./4/uh~. If the characteristic length of the
cos {[- - - T
a aRe disturbance I = 0.5 mm (a = 0.2) - which may be
(6.2) produced by a local disturbance such as a scratch
or crater - then the exponential decay time
In this asymptotic limit liquid disturbed to = 0.08 s; one decay time will expire after the
Surface leveling: healing of film irregularities 187
coating has translated only 0.2 m. If the charac-

teristic length of the disturbance I = 4 mm
(oc = 0.025) - which may be produced by flow
maldistribution in the coating head, edge effects,
or a long wavelength ribbing instability (Coyle
1992)- then the exponential decay time to = 768 s;
one decay time would require 2.3 kilometers of
travel before setting of the coating. Only short
wavelength disturbances can be healed by surface
leveling when coatings are rapidly dried, as is
often a requirement for high speed coating
operations.
Most detrimental to coated film quality are
irregularities of large amplitude (I> » 0) and such
irregularities are often oflong wavelengths (oc « 1).
Kheshgi and Scriven (1988) show that irregularities
level significantly slower than predicted by linear
theory, if the amplitude is sufficiently large. When
oc « 1, the evolution of a finite-amplitude two-
Initial Condition
dimensional irregularity can be predicted by the
film profile equation (Orchard 1962)
(6.5)
The evolution of initially sinusoidal (two-

dimensional) irregularities was calculated (Kheshgi
1984; Kheshgi and Scriven 1988) by numerical
~!~~~~~]!!!~
solution of (6.5), and examples are shown in Fig.
6.4. Evident is the distortion of the irregularity N.B. '<~~ I
for the larger amplitude disturbance in Fig. 6.4b.
The shallow region of the film rises much slower
o -rt/2
(b) Distance, X
than the thickest region falls. Soon after leveling
proceeds there is even a slight thinning of the Figure 6.4 Capillary-driven surface leveling of a
film at its shallowest point and an increase in sinusoidally shaped disturbance. Free surface profiles
surface inclination. predicted by finite difference solution of film profile
The rate at which the shallowest regions of equation (6.5) shown at time increments of 0.2. (a)
film are flooded is, for instance, of prime e = 0.3, (b) e = 0.8 (Kheshgi and Scriven, 1988).
importance for smoothing brushmarks, because
thin paint films often do not optically hide the
underlying surface. Figure 6.5 (Kheshgi 1984; mation leads qualitatively to the right behavior
Kheshgi and Scriven 1988) shows the decay time as confirmed by finite element solution of the
T1/2 required for the shallowest region of film to full Navier-Stokes system (Kheshgi and Scriven
rise halfway to the mean film thickness, I, predicted 1988). Linear theory for long wavelength irregu-
by solution of (6.5). While brushmarks are larities given by (6.4) predicts T1/2 = In(2). As is
predominantly ofmoderate wavelength (Orchard evident in Fig. 6.5, the time required to raise the
1962), i.e. oc = 0(1), the long-wavelength approxi- shallowest portion of film increases as the
The effect of gravity on a horizontal film with
a solid substrate below is to speed leveling as is
N
.....
... 1.2 evident in (6.1). However, if the film is inclined
or inverted, then the effect ofgravity is to increase
CIl the size of disturbances. Formulators of coatings
E 1
i= have long been active in trying to formulate
>- coatings so that they will level without sagging
eu
CJ as is evidenced by the review by Quach (1973).
CIl 0.8
C Many types of non-Newtonian and/or time-
dependent behavior have been proposed to
0.6 1::.-_....1-_----'-_ _. . . L . . - _ - - ' - _ - - - l accomplish this: yield-stress liquids, viscoelastic
o 0.2 0.4 0.6 0.8 liquids, thixotropic liquids and viscosity increase
Amplitude, f. by evaporation of solvent. Although a theoretical
justification for these approaches has not been
Figure 6.5 Decay time required for thinnest portion developed, the goal is clear: to formulate a
of disturbed film to rise half way to the mean film non-Newtonian or time-dependent behavior that
thickness H = 1. allows surface leveling but does not permit other
detrimental effects. For example, a coating may
amplitude 8 becomes greater, and appears to be formulated so that it has a yield stress small
become unbounded as 8 ---+ 1 (a ---+ ho)' While the enough to allow capillary driven leveling of
initial rate of film leveling of irregularities is large-amplitude, short-wavelength disturbances,
sensitive to its amplitude, the final rate ofleveling but large enough to prevent gravity effects. And
will eventually approach that of linear theory. estimates, based on the theory ofcapillary leveling,
Therefore, if we consider the time required to of the magnitude ofstress during capillary leveling
level an irregularity to an amplitude of one fifth (Orchard 1962) can be used to guide design of
(instead of one halO its initial amplitude, as did such yield-stress fluids.
Orchard (1962), then the effect of amplitude on In practice, coated films are rarely homogeneous
leveling rate does not appear significant until the Newtonian liquids which solidify abruptly upon
amplitude of the irregularity becomes greater drying, but rather complex fluids exhibiting
than about 80% of the mean film thickness viscoelasticity, surface tension gradients, complex
(8) 0.8). geometry and depth dependent properties. These
While leveling of a coating (especially paints), complications have been included in analyses to
is a desired effect, unwanted are effects such as illustrate the phenomena which are possible,
capillary straightening of the free surface above however analysis of specific applications requires
a substrate of complex geometry or sagging of individual consideration.
wet films due to gravity. Keunings and Bousfield (1987) show that
When a uniform film is coated on an irregular viscoelastic effects can significantly retard capillary
surface, the effect of capillary leveling is to leveling. It is not clear, however, what rheology
squeeze liquid from the convex regions of the would result in enhanced leveling as opposed to
surface to the concave regions. This causes sagging as Quach (1973) suggests. Viscoelastic
thinning of the coating above local asperities effects may well retard gravity effects less than
common in sandpaper or edges common in capillary leveling.
painting, and thickening in depressions or corners Schwartz and Eley (1992a) found a wide variety
(see Kornum and Nielsen (1980)); all of these are of effects caused by surface tension gradients
usually detrimental effects to be avoided. The induced by variations in surface concentrations
spin coating of an uneven substrate is considered of surfactants. Lowering surface tension with a
in Chapter 14. surfactant will, by itself, lead to a slower rate of
Dewetting: nucleation and growth of dry regions 189
surface leveling as indicated by the unit of time

measure to == 3J1.l4/(Jh~. In addition, the elastic
effect of slowly diffusing surfactant can further
delay leveling; the corollary was shown (Ruschak :::l..
1987) that uniform films better resist disturbance ~

by fluctuations in the overlying ambient gas ~
u
pressure if surfactant is present. During drying -$
the presence of volatile surfactants can have ~
larger effects as is discussed later in this chapter. ~
Often coatings are not homogeneous. Joos and
Jacobmeyer (1992) analyze the leveling of an
initially two layer coating with a solvent rich ...J'--
~ =. ~
o
surface layer on top of a soluble polymer base
layer. In the situation they study, diffusion of the Low HIGH
polymer into the solvent rich film leads to a SURFACE TENSION, a

growing (or swelling) immobile or high viscosity
polymer base which limits leveling to a thinning Figure 6.6 Operability diagram for leveling ofpowder
solvent rich film. Bousfield (t 991) analyzes the coating which illustrates the conclusions of de Lange
leveling of a liquid/solid slurry layer on a porous (1984).
substrate, a common situation in the paper
coating industry. He finds that absorption of to a high contact angle with the substrate which
the liquid into the porous substrate has a often are consequences of high surface tension.
more rapid effect on raising film viscosity than
does drying in such situations. Removal of the
6.3 DEWETTING: NUCLEATION AND
liquid leads to solid behavior and termination
GROWTH OF DRY REGIONS
of leveling.
A good example of the analysis of the trade- It is common to see the edges of a nonwetting
offs between leveling effects and those of defect liquid film retract soon after the film is created
growth is given by de Lange's (1984) analysis of by coating a solid or by causing bulk liquid to
the leveling of powder coatings. Such coatings drain away from a solid. Retraction continues
are applied as a solid coating. Heating melts the until the liquid has drawn up into small beads
coating, allowing the particles to coalesce and or larger patches. At the contact line of these,
irregularities to consequently level. Thermosetting the apparent contact angle varies to some extent
leads back to a solid film. Low viscosity or high about a characteristic value that should be called
surface tension speed the leveling phenomenon. the recently receded contact angle (cf Huh and
And leveling must be complete (with some Scriven (1971». It is not uncommon to see such
tolerance) before solidification occurs, in order a liquid film also rupture spontaneously at places
to give an acceptable product. This leads to distant from its original edges, the rupture creating
operability bounds as shown in Fig. 6.6. Viscosity an apparently dry patch with a brand new edge,
and surface tension can be adjusted with additives. which retracts in the same fashion as that on the
However, low viscosity enhances capillary- outer limits of the film (cf Padday (1971». In
driven flow away from corners, edges or asperities these ways a nonwetting film tends to dewet.
and gravity-driven flow which may take the form Retraction and rupture/retraction can be seen
of sagging of the film. And high surface tension in a water film spread on a waxed automobile,
can lead to the formation of pinholes, which are in a water film left by draining of a sink basin,
discussed in the next section of this chapter, due in a paint film applied to an ill-prepared surface,
to either enhanced surface tension gradients or and in many other everyday situations. These
phenomena can also be seen on occasion in neously to cover ever more of a smooth solid
carefully coated liquid films that are meant to surface. The term nonwetting is applied to those
be not only continuous but also precisely uniform, cases in which the apparent contact angle is
as in the manufacture of photographic film, greater than zero. (Sometimes 'nonwetting' is
photoresist, optical coatings and similar products. reversed for cases where the angle is greater than
The phenomena are then greatly undesired, of some arbitrarily chosen nonzero value, but that
course. Here it is hypothesized that there are is not the usage in this chapter).
three stages which lead to and can limit the
formation of a new dry patch.
Dewetting of a solid surface covered by a film
6.3.1 FILM-THINNING DISTURBANCES
of nonwetting liquid either proceeds from a
pre-existing dry patch or edge, or initiates from Liquid films delivered in coating operations are
some film-thinning disturbance that continues to usually of average thickness great enough, e.g.
grow until the film ruptures. Local thinning can more than 1 Jlm, that the film rupturing effect of
be caused by evaporation; by electrostatic forces; conjoining force is not appreciable if the film is
by drainage due to gravity or capillary-driven uniform. For assumed models of conjoining
flow, especially from sharp protuberances; or by force, it has been estimated (Scheludko 1967;
surface tension gradients, such as are caused by Sharma and Ruckenstein 1986) that film rupture
surface-active substances delivered by particles by conjoining pressure proceeds at an appreciable
falling on or already in the film. Once a nonwetting rate only in films thinner than 0.1 Jlm. In careful
film has been sufficiently thinned, long-range experiments, Padday (1971), however, observed
intermolecular forces can accelerate thinning that films of average thickness, shown in Table
until rupture. Rupture can be followed by film 6.1, of about 500 Jlm to dewet. (Note that the
retraction and growth of a dry patch or to film contact angles and critical film thicknesses in
advance and healing of the rupture. Table 6.1 are significantly larger than those
A liquid film spread over a solid surface may found in most coatings). His observation can be
be a wetting film or a nonwetting one, depending explained: if a film that is on average stable to
on the nature of the long-range interactions infinitesimal amplitude disturbances is thinned
between molecules of the liquid and solid. The by some disturbance until its thickness is small
consequences of these interactions collectively enough, conjoining forces can rupture the film:
can be modeled by disjoining pressure in the see Fig. 6.7. For this explanation to be true, film
wetting case, and by its negative, conjoining thinning disturbances would have to be responsible
pressure, in the nonwetting case as reviewed in for thinning the film by more than a factor of ten.
Chapter 2. They are appreciable effects only in Film-thinning disturbances trigger the con-
submicroscopically thin liquid films. However, joining-driven rupture of nonwetting films.
sufficiently near a contact line where the surface Numerous are the external sources ofdisturbances.
of a liquid film, a liquid/gas interface, appears One is film thinning by gravity-driven and
to intersect a solid surface, the thickness of the capillarity-driven drainage, an obvious mechanism.
liquid diminishes to submicroscopic scale and A second is film-thinning by surface tension
these effects of stress anisotropy and solid-liquid gradients, which may be common and the most
interaction are active. So it is that the apparent dangerous source in coating operations.
contact angle at the contact line between liquid, From a solid substrate that is locally elevated,
gas and solid is an index to whether the liquid gravity drains and thins liquid films. From solids
is wetting or nonwetting (Frenkel 1946; Mohanty which are locally convex, capillary pressure
1981). When the apparent contact angle, measured squeezes thin liquid films and makes them ever
through the liquid, is zero, the liquid is said to thinner (the difficulty encountered when trying
wet the solid, for usually it will spread sponta- to paint a convex comer is a testament to this
Table 6.1 Observed thickness he for film rupture and asperity of a surface to be coated can cause
equilibrium contact angle 8 for several polar and dewetting; coating ofsandpaper provides a familiar
nonpolar solid/liquid systems example.
Contact Critical
The surface of a small gas bubble attached to
Solid/liquid angle thickness a solid substrate is also sharp, i.e. it has a small
system 8t he radius of curvature, and is a common source of
coating defects (Gutoff 1993). Kheshgi and Scriven
Teflonjbenzene 45" 160 Ilm * (1991) present photographs which illustrate
Teflon/l-decene 33° 260llm * bubbles' role in nucleation of dry patches. In
Teflon/l-tetradecene 45" 270llm * coating operations considerable care is often
Teflon/l-octadecene 50° 270llm *
obligatory to remove gas bubbles from a liquid
Parafin/water 65"tt 310 Ilm *
before it is coated; one of several reasons for this
Polymethylene metacrylate/
water 70-80° 3931l m ** is to avoid nucleation sites for dewetting.
Polyethylene/water 94° 420llm ** If surface tension is reduced locally by some
Teflon/water 110° 510Ilm * agency, the accompanying surface-tension gradi-
Teflon/ethylene glycol 92° 560llm * ents can flow that can thin a liquid film: see
Teflon/water 110° 7871lm ** Fig. 6.8. If film thinning continues until the effect
of conjoining forces become profound, thinning
t Contact angles estimated by Sharma and Ruckenstein (1989).
tt Receding contact angle. can continue ever more rapidly and the liquid
* Critical thickness measured by Padday (1971). film can rupture.
** Critical thickness measured by Doughman, Holly and Surface-tension gradients arise from sources
Dohlman (1971).
often present in coating operations. Locally high
temperature lowers surface tension in the vicinity
and results in surface-tension-gradient-driven
fact). However, film drainage from blunt surfaces thinning (shown vividly in the National
is so slow that it is rarely the sole film-thinning Committee for Fluid Mechanics Films movie
mechanism leading to a dewetting occurrence; coordinated by Trefethen 1972). Suractant con-
other film-thinning mechanisms intrude. But centration gradients can incite flow instabilities
from a sharp object, film drainage is rapid. For driven by surface tension gradients. This is a
example, the film covering a pinpoint rising up common problem as reviewed by Scriven and
out of a body of liquid drains quite quickly Sternling (1960) and recognized by, among others,
because of the high curvature at the tip. Thus, Hansen and Pierce (1974) in certain coatings
Flat Film
(a) (b) (c)
Figure 6.7 A flat coated film of nonwetting liquid: (a) usually greater than Illm thick can dewet if a disturbance
thins the film; (b) to the extent (ordinarily less than Illm) that the effect of conjoining force is to dewet; (c)
the solid substrate.
High Low High

Surface-Tension-Gradient
..
Surface Surface Surface
Tension Tension Tension Driven Flow
~
(a)
--~ ~~~!W4.~
Disjoining Pressure
Active
Figure 6.8 Surface-tension-gradient-driven flow draws liquid from regions of locally low tension (a-b). Film
thinning might continue until conjoining forces become appreciable (c), and rupture the film (d).
during drying processes where solvent concen- 6.3.2 FILM RUPTURE

tration, which affects surface tension, decreases
and does so nonuniformly. Surface tension can If a nonwetting film is sufficiently thinned, then
also be lowered locally by surfactant impurities conjoining pressure will rupture it; a dry patch
arriving at the free surface from within the film, is nucleated. For this to occur, a nonwetting
or from the gas above. liquid film must have been created on a solid
Dewetting occurrences triggered by the last of substrate. In order to resolve the mechanisms of
these causes of surface tension gradients, namely film rupture, conjoining forces are considered
airborne surfactant particles is ubiquitous to all here. These act on an invisible (submicron) scale
but the most carefully isolated films. Kheshgi to make a film nonwetting. Also considered are
and Scriven (1983) easily observed, with the aid the hydrodynamic forces of pressure gradient
of moire topography, local depressions (compare and viscous shear that are activated in the
with Hahn (1971» which formed when unseen process of dewetting. Computational solution of
particles lit on a wetting film. Such depressions the equations of film motion augmented with
can lead to the coating defect called craters, conjoining forces gives prediction of whether or
discussed by Hahn (1971), should the film set not an externally-generated, initial disturbance
before the depression heals. Particles of or leads to rupture; and if it does, how much time
containing surface-active material can dewet a elapsed before the film ruptures.
nonwetting film, as demonstrated in experiments The equilibrium chemical potential, or partial
by Kheshgi and Scriven (1991) and Gaver and molar (or mass) free energy, of material in a thin
Grotberg (1992) where surfactant was intentionally liquid film differs from that in bulk liquid. This
placed on a film. However, surfactant-bearing difference in chemical potential is referred to as
particles are more often unnoticed microscopic the disjoining potential or, as Deryagin (1955)
dust motes, droplets, or even dandruff or lint expressed it, as the disjoining pressure ll. When
that fall on the film. Precision coating facilities the difference II is negative, it is referred to here
are sometimes placed in clean rooms to, in part, as conjoining pressure, for it is conjoining force
reduce the likelihood of dewetting; there are which tends to make liquid conjoin into beads,
reports that this practice is becoming more or drops, and hence dewet the solid in which the
prevalent in industrial manufacturing. film lies.
The ongms of conjoining pressure are the c

long-range intermolecular forces that are always €::J d.Il
dh <0
present between neutral molecules (e.g., dispersion, (/)
(/)
and multipolar electrostatic forces) and are often Q)
stronger between charged molecules (i.e., ionic- 0:.01------=:::::..----...

Cl Film Thickness h
electrostatic forces which are quite important, for .5
r:::
instance, when electrostatic double layers are '0
'Vi
present), and short-range or structural interactions o
(e.g., hydrogen bonding and so-called solvation
(a) Wetting
forces): Derjagin (1989) and Teletzke, Scriven and
Davis (1987) give details. In a flat or nearly flat
film, i.e. a flat film altered by disturbances of
length I that is great in comparison with the mean c Adsorbed Layer
film thickness h o, conjoining (or disjoining)
pressure beneath the liquid - gas interface (the gas Unstable Film ~ >0
lying above) can be approximated as a function
of film thickness h (see, for example, Mohanty h
(1981); Teletzke, Scriven and Davis (1987»
00
ll(h) = L aJ(h)i (6.6) (b) Nonwetting

i= 1
The coefficients a i are determined by the particular Figure 6.9 Disjoining pressure profiles of(a) a wetting
forces that contribute to conjoining pressure; liquid film, and (b) a nonwetting liquid film.
these forces are specific to each solid-liquid-gas
combination. The disjoining (conjoining) pressure pressure does not alone determine the time-
profiles, i.e. the thickness dependencies, ofdifferent evolution of a disturbance; hydrodynamic forces,
contributing forces are listed by, among others, viz. capillary pressure, gravity, and viscous stresses,
Mohanty (1981) and Teletzke, Scriven and Davis influence the behavior.
(1987). Conjoining pressure depends sensitively To estimate the time evolution of translationally
on thickness and ordinarily affects films no symmetric, long disturbances of horizontal liquid
greater than 111m thick; if electrostatic contri- films with a gas above and a flat solid surface
butions are absent the conjoining pressure may beneath, it is appropriate to adopt the film profile
be inconsequential until the film thins to a equation employed by Teletzke (1983; see also
thickness of only a few score of nanometers. Teletzke, Scriven and Davis (1987); Williams and
Conjoining pressure influences the stability of Davis (1982); Sharma and Ruckenstein (1986);
thin films. When disjoining pressure decreases and Kheshgi and Scriven (1991):
monotonically as film thickness increases, i.e.
dll/dh < 0, it tends to level any disturbance to
aH a
a liquid film; a disjoining pressure profile like aT ax
that in Fig. 6.9a is of a wetting film. However,
when disjoining pressure is negative and so an { H
3 a[ a H Bo (/-l14)]}
2
ax - ax2 + ~ H - O'h~ II (6.7)

increasing function of film thickness, i.e.
dll/dh > 0, it tends to unlevel, i.e. to reinforce The x-coordinate runs along the solid-liquid
any disturbance and make unstable a flat liquid interface. Film viscosity is presumed not to
film; a conjoining pressure profile like that shown depend on film thickness; this presumption may
in Fig. 6.9b is of a nonwetting film. Conjoining not be warranted in films thinner than about
5-10 molecules, at which thickness the structural learnings notwithstanding, it remains a challenge
contribution to disjoining pressure (cf Fig. 6.9b) to predict the nucleation of a dry patch quanti-
might influence nonequilibrium momentum tatively, beginning from a uniform film.
transfer.
Solution of film profile equation (6.7) leads to
predictions (Williams and Davis 1982; Kheshgi
6.3.3 GROWTH OF HEALING OF DRY
and Scriven 1991; Yiantsios and Higgins 1991)
PATCHES
of the time required for film rupture, given an
initial irregularity in film thickness h(x) at time Once a rupture is formed, energy arguments of
t = O. Results show that the time required for the sort used in fluid statics determine whether
rupture is short if the film thickness is less than the hole will grow or shrink. Surface tension
a critical thickness which is somewhat less than perpendicular to the contact line - which tends
1 ~m depending on the model for conjoining to expand the size of the hole - competes with
pressure. tension parallel to the contact line. The rate of
Experimental confirmation of this mechanism growth, however, depends on film viscosity and
for film rupture is confounded by the difficulty this rate, like that of surface leveling, must be
in producing thin films, monitoring and modeling viewed relative to the time increment designed
film thinning disturbances, and modeling con- for film setting.
joining pressure and its effects. For example, Pinholes or dry patches in an otherwise
Jensen and Grotberg (1992) attempted to analyze continuous film may either grow in size with the
the observed dewetting initiated by surfactant, film retracting, or shrink with size and eventually
and continued by rupture dominated by conjoining heal should there be sufficient time. If the fluid
pressure; their theoretical analysis, however, led properties are not changing in time, the growth
to the disconcerting result that the surfactants or shrinkage of the dry patch can be determined
considered do not thin the film sufficiently to by fluid static considerations.
lead to rupture, contrary to observations. There This is a familiar observation in for example
has been recent improvement of measurement, a sink where dry patches in thick films or puddles
modeling and analysis of the effects of conjoining of water on a horizontal surface can be seen to
pressure. For example, ellipsometry has been either grow or heal. If the film is thick enough,
used to measure (Gupta and Sharma 1992a) and the effect of gravity (in thick horizontal films)
aid in developing models for disjoining/conjoining dominates the effect ofsurface tension to advance
pressure (Gupta and Sharma 1992b). However, the contact line and heal the dry patch. Lamb
even fundamental predictions like that of contact (1945) and then Taylor and Michael (1973) found
angle from a measured conjoining pressure profile that the critical film thickness where this occurs is
have not been made quantitatively. Recently
Hocking (1993) proposed a new approach to (J
hi = 2 ( pg
)1/2 sin (())
"2 (6.8)
calculating contact angle which includes the
effect of finite inclination of the free surface with
respect to the solid substrate (as opposed to just where (J is the surface tension, g the acceleration
the film thickness as is the case in (6.7)), although of gravity, () the equilibrium contact angle and
it has not been used actually to predict measured p is the liquid density. For films greater than this
contact angles. To complicate the situation further, critical film thickness, the continuous film wi1l
for some solid/liquid systems 'dewetting' will always have lower free energy than a film with
leave behind a thin film instead of a dry surface a hole of any size. In thick films, long-range
due to short-range interactions (see Brochard- intermolecular forces do not have an appreciable
Wyart et al. (1991)); and this complication may contribution to the energy of the film except at
be enhanced by the presence of surfactants. These the free surface and contact line. If the film is
thinner than the critical thickness hi' then dry Note that as (J approaches zero, he/r h also goes
patches will grow from a pre-existing hole in the to zero. So for a film of thickness h with an initial
film, if the hole is of sufficient size. In fact, the hole of radius r, if h/rh > he/r h then the whole
fate of a hole of radius r h (see Fig. 6.10) was will shrink and if h/rh < he/rh then the hole will
shown to be a function of static considerations, grow. The metastable equilibrium state is a
and properties such as advancing and receding maximum in free energy of the system, so to
contact angle, gravity, density, and surface tension, make a hole of radius rein an otherwise continuous
and not dynamical properties such as viscosity. film of thickness he' energy must be added.
For a hole of radius r h with contact angle (J, there Sharma and Ruckenstein (1990) argue that it is
is but some film thickness, he' which yields an more likely that holes significantly larger than
equilibrium solution to the Young-Laplace re, for a given film thickness h, form initially
equation for hydro-capillary statics. This equi- because it requires less energy to form the hole,
librium represents a balance between surface yet they do not address what are the mechanisms
tension perpendicular to the contact line, which for hole formation.
tends to expand the size of the hole, and the It is interesting to note that there is a correlation
effect of surface tension parallel to the contact between the thickness he at which holes form in
line plus gravity (which is negligible for thin liquid films, and the equilibrium contact angle
films), which tends to close the hole. An important shown in Table 6.1. Of course, contact angle is
imperfection of this theory comes from the fact one effect of conjoining pressure (Mohanty 1981)
that in real systems there is a difference between and so the magnitude of conjoining pressure is
the recently advanced and receded contact angle, also apparently related to the critical film thickness
which lead to a range in film thicknesses (or he; and so the frequency and size of film ruptures
alternatively, hole sizes) over which the hole will may be greater when the contact angle is greater.
be an equilibrium solution. The equilibrium hole According to the theory based on fluid statics,
radius r e (or range of radii if there is contact the unstable equilibrium film thickness he is
angle hysterisis) increases monotonically with related to contact angle by (6.9). Sharma and
film thickness until at h = hI' r e--+ 00. Sharma Ruckenstein (1989) observe that if initial hole
and Ruckenstein (1990) calculate the ratio of size is constant the measurements shown in Table
equilibrium film thickness to hole radius, he/rh , 6.1 can be explained by fluid statics. The initial
as a function of contact angle (J and the ratio of hole size in coating situations is, however, most
hole radius to capillary length, r == r h /(a/pg)1/2. often caused by dynamic effects, as is the growth
For most coatings the effects of gravity is small, rate of an unstable hole. In a coating, pinholes
which means that both r « 1 and h « hI' If the may be an important defect even if the static
gravity is neglected, Sharma and Ruckenstein contact angle is near zero if there is not enough
(1990) find that a good approximation for he/rh is time for a hole formed by a disturbance to heal.
Alternatively, significant dewetting may not occur
he in a nonwetting film if there is not time for either
-=(I-cos(J) (6.9)
rh disturbances to form holes or for holes to grow.
Adachi and Marukawa (1992) found the time t h
required for holes to become visible on a rotating
roll after a point where the wetted roll is scraped
by a knife coater to be
th = (1500) V- 1/ 2 h~/3 Jll/6 a-1/12 g-1/4 p -1112
(6.10)
Figure 6.10 Small rupture (hole) in a liquid film. where V is the roll speed. This is evidence that
dynamics is critical in the rupture and growth nearly always a solid film. Films are often
of dry patches in contrast to Sharma and solidified by drying, removal of solvent by
Ruckenstein's (1989) observation. vaporization at the gas/liquid interface or sorption
Once a large hole in a thin nonwetting film is of the solvent by the substrate. The rate of drying
formed, the contact line recedes at velocity, v, in coating operations can be limited by the mass
building up a thick rim ofliquid which originally transfer into the overlying gas, volatility of the
resided over the area of the hole which is evident solvent, contamination ofthe liquid/gas interface,
in photographs shown by Kheshgi and Scriven or diffusion through the liquid. Film dryers can
(1991). Redon, Brochard-Wyart and Rondelez be designed in response to these phenomena by
(1991) observed that the contact line of large providing heat, resistance time and enhanced
holes retracts at a constant velocity gas-side mass transfer. An overview of drying
devices can be found in Cohen (1992b).
a
v oc _()3 (6.11)
Jl
6.4.1 GAS-SIDE MASS TRANSFER RESISTANCE
and is independent of film thickness.
The mechanism for film thinning, rupture and The maximum evaporation rate in a vacuum is
contact line retraction is now clear. If a liquid given by the Langmuir-Knudsen equation:
made nonwetting by conjoining forces is thin M )1/2 (6.12)
enough, those forces can rupture the liquid, as G = Pe ( 2nR()
demonstrated by the theory of dewetting hydro-
dynamics. Continuously coated liquid films are where G is the mass flux, Pe is the equilibrium
ordinarily too thick for conjoining to break them, solvent vapor pressure, R is the gas constant, M
unless some disturbance thins the film to a is the molecular weight of the solvent and () is
thickness of less than 1 Jlm. If after a film is the surface temperature of the liquid film (Knudsen
formed it is thinned by evaporation or by some 1946). The kinetic theory ofgas diffusion originates
local film thinning disturbance, then conjoining from the same line of thought.
forces can rupture the film. Examples of film- At finite rates of vaporization, however, solvent
thinning disturbances were shown by Kheshgi molecules leaving the interface have a chance to
and Scriven (1991) to lead to dry patch formation collide with others, reducing the actual evaporation
in nonwetting films: drainage driven by gravity rate to (Palmer 1976)
and capillary pressure gradient thins liquid above
an underlying gas bubble, and film thin- M )1/2 (6.13)
G = BPe ( 2nR()
ning driven by surface tension gradients caused
by airborne surface-active particles impinging on In a vacuum the accommodation coefficient, B,
the film. To decrease the likelihood of dewetting, is somewhat less than one (Cong and Bird 1978;
sources of film-thinning disturbances should be Kaplon, Kawala and Skoczylas 1986).
removed in order to reduce the number of However, most relevant drying situations are
nucleation sites where dewetting can occur; in not operated in a vacuum, but at near atmospheric
addition, the solid or liquid can be treated or pressure. Moreover, in many cases condensation
altered to lessen the strength of conjoining of solvent vapor (e.g. water) in the overlying gas
pressure (contact angle) so that thinner films are can occur at a rate proportional to the ambient
stable. partial pressure, P., modifying the mass flux to
6.4 FILM DRYING M )1/2 (6.14)

G = B(Pe - P.) ( 2nR()
While coatings of concern in this book are
applied as liquids, th final coated product is Near atmospheric pressure the accommodation
Film drying 197
coefficient, B, in this formulation must account drying and must be accounted for when it is an
for diffusion through the overlying gas and is appreciable effect. Drying rate can be increased
much less than one. Although some have opted by heating the surrounding gas, or directly
to use this formulation for mass flux from the heating the film, thereby increasing the equilibrium
surface (e.g., Bierwagon (1979); Burelbach, vapor pressure of the solvent. While high
Bankoff and Davis (1988», the more prevalent vaporization rate leads to shorter required
formulation uses a mass transfer coefficient k residence time for drying, it also can make
which represents the mass flux by concentration gradients through the film more
severe. For this reason, it is beneficial in some
k cases to slow down drying.
G = - (p - P ) (6.15)
Po c a One source of disturbances to coatings is
agitation by motion of the overlying gas. These
where Po is a reference pressure. disturbances often take the form of Dryer Bands
Drying is often enhanced by agitating the which are stripes of film thickness or property
overlying gas and thereby increasing the mass variation aligned with the direction of coating
transfer coefficient. The mass transfer coefficient, translation, or Mottle which is an irregular
k, can be approximated by well-known correlations pattern of variation (see Cohen (1992b». Analysis
for analogous flows as seen in dryers (Bird, (Ruschak 1987) shows that the amplitude variation
Stewart and Lightfoot 1960). Often gas jets are is dependent on the pattern and intensity of gas
used to enhance mass transfer, and Martin (1977) motion, coating viscosity and surface tension,
gives both theoretical and empirical correlations gravity and the effects of surfactants (if present).
for common geometries. For example, the One obvious cure to these defects is to reduce
correlation for k that Martin finds for a square gas agitation in the initial stages of drying until
of hexagonal array of round jets impinging on the coating is sufficiently dried or cooled so that
a flat surface is given by its viscosity (or yield-stress) is high enough to
resist disturbances.
Here the Sherwood number Sh = kdn/D g where

6.4.2 DIFFUSION LIMITED TRANSPORT
dn is the diameter of the nozzle exit and Dg is
WITHIN THE FILM
the diffusivity of the solvent vapor through the
gas. The Schmidt number Sc == vg/D g where vg If removal of solvent at either the liquid/gas or
is the kinematic viscosity of the gas. The gas the liquid/substrate interface exceeds the rate of
Reynolds number Reg = wndn/v g where wn is the diffusion of the solvent through the liquid to the
mean exit gas velocity from the nozzle. The interface, then the interface will be depleted of
height of the nozzle exits above the surface is hn , solvent and concentration gradients within the
and the distance between neighboring nozzles is film will develop. In polymer solutions, especially,
In. The function F of nozzle geometry is given diffusivity decreases markedly with removal of
by Martin (1977). When details of the drying rate solvent. With continued removal of solvent,
dependence on gas flow are needed, the fluid transition to a solid state can lead to even smaller
mechanics and mass transfer in the gas phase coefficients of diffusion of the solvent through
can be accounted for as is done for a spin coating the solid and may lock solvent or impurities in
process by Bornside and Brown (1992). the core of the film. In extreme cases 'skinning'
Drying is not an isothermal process. As the can occur, where a solid forms at the interface
equilibrium solvent vapor pressure, Pc' is often while the film core remains liquid. At the interface
a sensitive function of temperature, film tempera- between the gas and the liquid film, surfactants
ture can change the rate of evaporation as given which may be contaminants can also form a solid
by (6.15). Evaporative cooling of the film slows 'skin' and limit diffusion of solvent from the
underlying liquid to the gas phase, however this of binary and ternary solutions containing polymer
phenomenon has yet to receive attention. have been studied by Yapel (1988).
Diffusive resistance in films has been analyzed In slurry coatings Whithan and Perera (1992);
by application of Fick's law of diffusion, conser- (see also Bierwagon (1979)) find that below a
vation of mass and constitutive equations for critical moisture content, ¢cril' the drying rate
diffusivity in the material being considered. For starts to fall rapidly. They find that this occurs
a multi-component mixture, conservation ofmass at roughly the liquid content consistent with the
dictates that the variation of mass concentration maximum packing fraction derived from rheologi-
of component i, Pi' with time and the one space cal studies. Drying rate can be explained with a
dimension z perpendicular to the film substrate model where diffusivity decreases significantly
is given by when the moisture content falls below ¢crit. A
detailed analysis of diffusion for this system,
however, remains to be carried out.
(6.17)
Surface tension gradients set up during drying
cause flow - and disturbance of film thickness -
where gi(Z, t) is the mass flux of component i in when there is a gradient of concentration or
the z-direction. If we assume constant partial temperature tangent to the film base. In cases
specific volume and no flux through the substrate where there is only a gradient normal to the film
located at z = 0, then the fluxes gi can be base there remains potential for the so-called
attributed solely to diffusion by Marangoni instability which manifests itself by
forming a hexagonal pattern of roll cells (Pearson
1958; Scriven and Sternling 1960; Anand and
gi = - L
n-1 ( 0)
Dij--.!!J. (6.18) Karam 1969). This instability is the cause for
j =1 OZ
well known defects that lead to pigment floc-
where Dijs are the diffusion coefficients for the culation, pigment segregation and lack of gloss
n-component system. The movement of the in paint coatings (Hansen and Pierce 1973, 1974).
gasfliquid interface due to shrinkage of the film The instability is expected (Pearson 1958; Scriven
can be accounted for by a coordinate transform- and Sternling 1960; Anand and Karam 1969;
ation as shown by Vrentas, Duda and Ling Kornum and Nielsen 1980) at Thompson or
(1985). At the gasfliquid interface the boundary Marangoni's number
condition is that the mass flux is equal to that
given by (6.15) 00 OU h 2
N =---
Diffusivity in polymer films can be modeled TH - OZ 00 KJl
by the free-volume theory of polymer solvent
diffusion laid out by Duda et al. (1982); (see also greater than a critical value. Here, 0 is temperature
Vrentas and Duda (1979)), although simpler (or concentration) and K is the thermal (or
empirical models abound (e.g., Okazaki et al. molecular) diffusion coefficient.
(1974); Blandin et al. (1987); Ng and Ross (1991)).
Vrentas, Duda and Ling (1985) found, in a study
6.4.3 PHASE TRANSITION
of ternary diffusion in polymer films, that a
formulation including a second less-volatile solvent Phase behavior of a multi-component film
can be used to promote removal of the first undergoing drying can impact product quality.
solvent, which may be deemed an impurity in Examples have been studied in films containing
the final product. This may be advantageous three components: polymer, solvent and a non-
compared with devolatilization by increasing solvent which may serve the purpose of a
temperature, since this may lead to degradation plasticizer.
of the polymer. Non-isothermal effects in drying In the 'phase inversion' method of membrane
Stress buildup 199
production, phase separation leads to useful containing plasticizer, solvent and polymer leads
properties of the membrane, while in a protective to variation in composition with depth in a
coating, for example, phase separation may lead drying film, as shown with respect to a phase
to a permeable coating which would be of poor diagram in Fig. 6.11. If the time-dependent
quality. Consider the ternary phase diagram concentration of the film were uniform with
shown in Fig. 6.11. This type of phase diagram depth - no diffusion limitation - the ratio of
can be generated from experimental measurements plasticizer to polymer would remain equal to the
or using a modified Flory-Huggins theory as initial ratio. However, with diffusion limitation,
discussed by Yilmaz and McHugh (1986). What counter-diffusion of the solvent and plasticizer
is seen in Fig. 6.11 is that there is a large two will, in most cases, lead to an increase in the
phase region bounded by the binodal curve for ratio within the film and a decrease at the free
high concentrations of non-solvent component. surface (see Adachi and Scriven 1992)). Significant
Between the binodal and spinodal curve the concentration gradients in the film may lead to
liquid is metastable and phase separation is slow conditions with the potential of phase separation
- time delayed phase separation is important to depicted in Fig. 6.11. While higher plasticizer
membrane quality (Tsay and McHugh 1990). In concentration near the solid substrate may be
the phase inversion process a film of sol- preferred, because the softening effect of the
vent/polymer is submerged in a bath of non- plasticizer may be of great use near the substrate
solvent in order to induce phase separation where stress build-up is greatest and may lead
which gives the membrane beneficial qualities. to delamination (peeling), there is also a danger
Phases separation can also lead, however, to of phase separation.
various defects. Diffusion during phase separation
has yet to receive serious attention with regards
6.5 STRESS BUILDUP
to coating.
Diffusion limitation during drying of films A liquid film once solidified can evolve to have
significant stress buildup. Stress, if beyond some
critical value, can lead to stress related defects
such as crazing, cracking, curling or peeling. An
Polymer
obvious mechanism for stress buildup is from
shrinking of a solid film due to removal (drying)
of solvent. Other mechanisms for stress buildup
such as phase transitions, phase separations,
reactions or temperature changes may well exist,
but have not been studied in detail in the context
of liquid film coatings.
The buildup of stress leading to defects occurs
nonuniformly through the film. Stress is concen-
trated at inhomogeneities such as cracks or
interfaces (at the substrate or within the film
between phases) as discussed by Tam, Stolarski
and Scriven (1992). Before delving into the
Solvent
complexities of stress buildup, it is instructive to
review the theory for homogeneous stress buildup
Figure 6.11 Illustrative ternary phase diagram showing
a single- and two-phase region. Also shown is the time in drying films.
evolution of film concentration from an initially Stress in ordinary (e.g. Newtonian) liquids is
uniform concentration to depth dependent concentra- nonzero (aside from the isotropic stress which is
tions. pressure) only during deformation (a nonzero
rate of strain). This is not the case in, for example, with idealized predictions based, for example, on
viscoelastic or yield-stress liquids, elastic or plastic packing of spheres.
solids, and suspensions of solids where stress can Once a transition to solid occurs, further
exist without concurrent strain. In these materials, drying causes film shrinkage and stress buildup.
stress buildup can be determined by past Consider a film drying slowly enough that there
deformation due, for instance, to drying. The is little diffusional resistance within the film, so
point at which a film goes through a transition that the volume fraction ofsolvent <P is independent
from liquid to solid state is often not well defined of position. With either type of liquid/solid
in practice, although it was a useful approximation transition mentioned above, there is a transition
in analyses of Croll (1979) and Bierwagon (1979) solvent volume fraction <PI. It is interesting to
which will now be recounted. note that for the liquid/glass transition the
Croll (1979) found that liquid polymer films transition solvent volume fraction <PI can be
(both polyisobutyl methacrylate and polystyrene) altered by changing (or programming) drying
underwent a glass transition to an amorphous temperature. Removal of solvent reducing the
solid as solvent (toluene) is removed by drying. volume fraction below <PI leads to stress buildup.
The transition was detected by measuring the The extent the solid film shrinks from its
glass transition temperature by a differential volume Jt; at the liquid/solid transition to its final
scanning calorimeter as a function of solvent volume Vr is related to the volume fraction of
volume fraction <p. Croll (1979) found the solvent solvent by
volume fraction to be very nearly proportional
to the solvent concentration. Once the glass
transition occurs, we should expect the physical
Vr=(~)
Jt; I - <Pr
(6.19)
properties of the solid to depend rather strongly

where <Pr is the final solvent volume fraction. In
on solvent concentration. In addition, diffusion
a film that adheres to the substrate, this film is
will likely limit drying in the solid due to low
fixed in the x-y plane tangent to the film and
values of diffusivity. Plasticizers (non-volatile
can only shrink in the direction, z, normal to the
solvents) are sometimes added to allow a final
film. And so the film is stretched in the x- y plane
uniform solvent concentration that is sufficient
and compressed in the z-direction giving the
to maintain flexibility in the product and can
strain tensor
allow drying of volatile impurities (Vrentas,
Duda and Ling 1985), e.g., another solvent.
YXX
Bierwagon (1979) considered drying of latex
paint films which are suspensions oflatex polymer Y= 0 (6.20)
[
particles in water. As water is removed the o
polymer coalesces leading to increased elasticity
and solid behavior. Bierwagon (1979) modeled where
this transition to occur when enough solvent was
removed so that the particles were close packed.
At this point diffusion of the solvent slows,
hindering drying of the film. Similar behavior
was seen in slurry coatings by Whithan and _1 =:: _ ~
3
(1 _Vr)
Jt;
= _~
3
(<PI -
1 - <Pr
<Pr)
Perera (1992) who measured rheological behavior
and drying rates, and compared these with (6.21)
idealized predictions of the solvent fractions at
which the particles would have their densest The approximation is made for a small extent
packing; they found that rheology was correlated of shrinkage.
with diffusional resistance but not well correlated If the material is approximated to be a Hookean
Summary 201
(elastic) solid, then the stress tangent to the film is stresses that build up, and the motion of the
particles in order to understand this defect.
Stress buildup can lead to a variety of defects
as shown in Fig. 6.1. Even if stress builds up
uniformly in an ideal film, stress will be concen-
(6.22) trated at imperfections or boundaries of a film.
The defects which occur relieve stress locally.
where E is the Young's modulus and v is the Bierwagon (1979) related stress in films to a
Poisson's ratio. And so if the mechanical properties critical stress for cracking in order to predict
in the film E and v can be measured, then (6.22) 'mudcracking' in paint films, and found the
gives an estimate of the stress in the film. It is critical stress does decrease with increasing film
interesting to note that the residual internal stress thickness. Croll (1979) noted that in drying of
does not depend on the thickness of the coating polymer films there was a critical stress at which
or on the initial concentration ofsolvent. Estimates point crazing occurred.
of homogeneous stress buildup in films, such as
that given by (6.22), can be used to form a criteria
6.6 SUMMARY
for the occurrence of stress related defects.
Diffusion limited drying leads to in- Phenomena which occur after coating application
homogeneous distributions ofsolvent and nonuni- often limit coating speed, restrict formulation
form stress buildup. Solidification of the surface and affect product quality.
prior to the core of the film is responsible, for Healing of irregularities in film thickness born
instance, for a coating defect where the solid skin during coating application is required in many
becomes wrinkled while the film core solidifies coating situations such as painting operations.
leading to nonuniform film thickness. The Capillary straightening of liquid surfaces or
mechanical properties and the diffusivity vary interfaces can make uniform films on flat
widely with depth even in the finished product. substrates. Healing of irregularities requires time
While this can degrade product quality, this and a film of not-too-high viscosity. While the
could alternatively be tailored, for instance, to effects of capillary pressure can straighten
give flexibility near the substrate to give better irregularities ofshort length quickly, irregularities
adhesion, while having a hardened film surface that span a long extent of coating heal slowly.
(low solvent fraction near the free surface). So healing of irregularities requires a time period
Drying of suspensions of solid particles leads after coating before the coating has set and film
to nonuniform stress buildup due to nonuniform viscosity or yield stress has risen. However, the
mechanical properties and, if drying is diffusion action ofgravity on inclined or inverted substrates,
limited, nonuniform diffusivity. Christodoulou or capillarity on curved substrates can lead to
and Lightfoot (1992) examined the consequences growth or formation of film irregularities in time.
of these phenomena in the context of the defect, And so formulation of coatings to allow healing
common in production of graphic arts photo- of nonuniformly applied films, yet resist subsequent
graphic films, known as 'starry night'. As shown disturbance, is often critical especially for paints.
in Fig. 6.1, the defect is caused by internal stresses Liquid films are vulnerable to rupture and film
built up in the upper layer of the coating which retraction, leading to holes in coatings and dry
push solid particles (matting agent) into the lower patches on coated substrates. The mechanism
coating containing silver halide particles leading for film thinning, rupture and contact line
to holes or thin spots in the silver halide coating retraction is now clear. If a liquid film made
(see, e.g., Gutoff (1993)). Christodoulou and nonwetting by conjoining forces is thin enough,
Lightfoot (1992) modeled the nonuniform drying those forces can rupture the liquid film, as
of the film around particles, the nonuniform demonstrated by the theory of dewetting
hydrodynamics. However, a quantitative predic- Spring National Meeting, New Orleans, March
tion of dewetting has not been successfully made 29-April 2.
from models of the physical mechanisms: film Adachi, K. and Marukawa, T. 1992. The critical film
thickness for formation of a pinhole and its growth
thinning, conjoining forces and film retraction. time. Paper No. 48b. AIChE, 1992 Spring National
Nevertheless, the qualititative description suggests Meeting, New Orleans, March 29-April 2.
actions. To decrease likelihood of dewetting, Anand, J. N. and Karam, H. 1. 1969. Surface deformation
sources of film-thinning disturbances should be of thin coatings caused by evaporative convection,
removed in order to reduce the number of III. J. Coli. and Interface Sci. 31: 208-215.
nucleation sites where dewetting can occur; in Anshus, B. E. 1973. The leveling process in polymer
addition, the solid or liquid can be treated or powder painting - a three-dimensional approach.
A. Chem. Soc. Div. Org. Coating Plastics Chem.
altered to lessen the strength ofconjoining pressure 33(2): 493-501.
(contact angle) so that thinner films are stable. Batchelor, G. K. 1967. An Introduction to Fluid Dynamics.
Curing and drying of coatings can be rate Cambridge: Cambridge University Press.
limiting and defects can occur during this process. Bierwagon, G. P. 1979. Film formation and mudcracking
Drying is limited by evaporation of solvent from in latex coatings. J. Coatings Technology. 51(658):
the surface of the film, and diffusion of solvent 163-170.
in the film. Evaporation is often decreased by Bird, R. B., Stewart, W. E. and Lightfoot, E. N. 1960.
Transport Phenomena. John Wiley: New York.
agitation of the overlying gas and heating of the
Bixler, N. E. 1982. Stability of coating flows. PhD
film. Excessive gas agitation can lead to disturb- thesis, University of Minnesota, Minneapolis.
ance of once uniform liquid films. Diffusion Blandin, H. P., David, 1. c., Vergnaud, 1. M., Illien,
limitation can lead to severe gradients of con- J. P. and Malizewicz, M. 1987. Modeling the drying
centation with depth into the coating. This can process of coatings with various layers. J. Coatings
lead to (Marangoni) instability driven by surface Technology. 59: 27-32.
tension gradients, or unwanted phase separation. Bornside, D. E. and Brown, R. A. 1992. The effects of
forced convection in the overlying gas on the rate
Control of the rate of drying through dryer of mass transfer from the thinning liquid film on a
conditions, and diffusivity of the components in spinning wafer during spin coating. Paper No. 39b.
the coating through coating formulation holds A IChE, 1992 Spring National Meeting, New
promise for increasing drying rate and reducing Orleans, March 29-April 2.
the occurrence of defects. Bousfield, D. W. 1991. A model to predict the leveling
Film rheology changes significantly as the of coating defects. Tappi J. 163-170.
coating is dried. The film usually changes from Brochard-Wyart, F., di Meglio, J.-M., Quere, D. and
de Gennes, P.-G. 1991. Spreading on nf>nvolatile
liquid to solid phase well before the coating is liquids in a continuous picture. Langmuir. 335-338.
dry. Buildup of stress in the coating can lead to Burelbach, J. P., Bankoff, S. G. and Davis, S. H. 1988.
defects such as cracking or peeling. Limiting Nonlinear stability ofevaporating/condensing liquid
stress buildup by changing coating formulation films. J. Fluid Mech. 195: 463-94.
or drying conditions, while not fully understood, Christodoulou, K. N. and Lightfoot, E. J. 1992. Stress
has long been practiced in the paint coating induced defects in the drying of thin films. Paper
industry. Translation of this know-how to No. 46c. AIChE, 1992 Spring National Meeting,
New Orleans, March 29-April 2.
relatively newer coating applications along with
Cohen, E. D. 1992a. Choosing the coating method. In
heightened theoretical understanding is an area Modern Coating and Drying Technology. eds E. D.
of future research. Cohen and E. B. Gutoff, pp. 1-22. VCH Publishers:
New York.
Cohen, E. D. 1992b. Thin film drying. In Modern
REFERENCES Coating and Drying Technology. eds E. D. Cohen
and E. B. Gutoff, pp. 266-298. VCH Publishers:
Adachi, A. and Scriven, L. E. 1992. Migration of New York.
solute during drying. Paper No. 46h. AIChE, 1992 Cong, T. T. and Bird, G. A. 1978. One-dimensional
References 203
outgassing problem. Physics of Fluids. 21: surfactant spreading on a thin viscous film: shock
327-333. evolution and film rupture. J. Fluid Mech. 240:
Coyle, D. J. 1984. The fluid mechanics of roll coating 259-288.
- steady flows, stability and rheology. PhD thesis, Joos, F. M. and Jacobmeyer, R. L. 1992. Levelling of
University of Minnesota, Minneapolis. a liquid layer over a sorbing surface. Paper No. 49d.
Croll, S. G. 1979. The origin ofresidual internal stress AIChE, 1992 Spring National Meeting, New
in solvent-cast thermoplastic coatings. J. Applied Orleans, March 29-April 2.
Polymer Science. 23: 847-58. Kaplon, 1., Kawala, Z. and Skoczylas, A. 1986.
de Lange, P. G. 1984. Film formation and rheology Evaporation rate of a liquid from the surface of a
ofpowder coatings. J. Coatings Technology. 56: 23-33. rotating disc in high vacuum. Chern. Eng. Sci. 41:
Deryagin, B. V. 1955. The definition and magnitude 519-522.
of disjoining pressure and its role in the statics and Keunings, R. and Bousfield, D. W. 1987. Analysis of
dynamics of thin fluid films. Kolloid Zhurnal. 17: surface tension driven leveling in viscoelastic films.
205-214. J. Non-Newt. Fluid Mech. 22: 219-33.
Deryagin, B. V. 1955. Theory and Stability of Colloids Kheshgi, H. S. 1984. The motion of viscous liquid
and Thin Films. New York: Plenum. films. PhD thesis, University of Minnesota, Min-
Doughman, D. J., Holly, F. J. and Dohlman, C. H. neapolis.
1971. Preliminary studies on the stability ofaqueous Kheshgi, H. S. 1989. Profile equations for film flows
films on solids. Presented at the ARVO Spring at moderate Reynolds numbers. AIChE J. 35:
Meeting, Sarasota, Florida. 1719-1727.
Duda, J. L., Vrentas, J. S., Ju, S. T. and Liu H. T. Kheshgi, H. S. and Scriven, L. E. 1983. Measurement
1982. Prediction ofdiffusion coefficients for polymer- of liquid film profiles by moire topography. Chern.
solvent systems. AIChE J. 28: 279-285. Eng. Sci. 38: 525-34.
Frenkel, J. 1946. Kinetic Theory of Liquids. Oxford: Kheshgi, H. S. and Scriven, L. E. 1988. The evolution
Oxford University Press. of disturbances in horizontal films. Chern. Eng. Sci.
Gaver, D. P. and Grotberg, J. B. 1992. Droplet 43: 793-801.
spreading on a thin viscous film. J. Fluid Mech. 235: Kheshgi, H. S. and Scriven, L. E. 1991. Dewetting:
399-414. nucleation and growth of dry regions. Chern. Eng.
Gupta, A. and Sharma, M. M. 1992a. Stability of films Sci. 46: 519-26.
on solid surfaces I. ellipsometric measurement of Knudsen, M. 1946. The Kinetic Theory of Gases.
disjoining isotherms for aqueous films bounded by Methuen, London.
glass and mercury. J. Colloid and Interface Sci. 149: Kornum, L. O. and Nielsen, H. K. R. 1980. Surface
392-406. defects in drying paint films. Progress in Organic
Gupta, A. and Sharma, M. M. 1992b. Stability of films Coatings. 8: 275-324.
on solid surfaces lI. a model for aqueous bounded Lamb, H. 1945. Hydrodynamics. Dover, New York.
by dissimilar charge regulating surfaces. J. Colloid Martin, H. 1977. Heat and mass transfer between
and Interface Sci. 149: 407-424. impinging gas jets and solid surfaces. In Advances
Gutoff, E. B. 1993. Avoid coating and drying defects. in Heat Transfer. eds J. P. Harnett and T. F. Irving
Chern. Eng. Prog. 89: 49-55. Jr. New York, Academic Press, 13: 1-60.
Hahn, F. J. 1971. Cratering and related problems. J. Mohanty, K.K. 1981. Fluids in porous media: two-
of Paint Tech. 43: 58-67. phase distribution and flow. PhD thesis, University
Hansen, C. M. and Pierce, P. E. 1973. Cellular of Minnesota, Minneapolis.
convection in polymer coatings - an assessment. Munter, 1. D. 1990. Attenuation of cell patterns for
Ind. Eng. Chern. Prod. Res. Develop. 12: 67-70. gravure coating. Paper No. 51 f. AIChE, 1990 Spring
Hansen, C. M. and Pierce, P. E. 1974. Surface effects National Meeting, Orlando, Florida, March 19-22.
in coatings processes. Ind. Eng. Chern. Prod. Res. Ng, K. C. and Ross, D. S. 1991. Diffusion in swelling
Develop. 13: 218-225. gelatin. J. of Imaging Science. 35: 356-361.
Hocking, L. M. 1993. The influence of intermolecular Okazaki, M., Shioda, K., Masuda, K. and Toei, R.
forces on thin fluid layers. Phys Fluids A. 5: 793-799. 1974. Drying mechanism of coated film of polymer
Huh, C. and Scriven, L. E. 1971. Hydrodynamic model solution. J. of Chern. Eng. Japan. 7: 99-105.
of steady movement of a solidjliquid/fluid contact Orchard, S. E. 1962. On surface levelling in viscous
line. J. Coll. Inter! Sci. 35: 85-101. liquids and gels. Appl. Sci. Res. 11: 451-464.
Jensen, O. E. and Grotberg, 1. B. 1992. Insoluble Padday, 1. F. 1971. Cohesive properties of thin films
of liquids adhering to a solid surface. In Special Paper No. 48f. AIChE, 1992 Spring National Meeting,
Discussion Faraday Soc. on 'T hin Liquid Films and New Orleans, March 29-April 2.
Boundary Layers', pp. 64-74. London: Academic Taylor, G. I. and Michael, D. H. 1973. On making
Press. holes in a sheet of fluid. J. Fluid Mech. 58: 625-639.
Palmer, H. J. 1976. The hydrodynamic stability of Teletzke, G. F. 1983. Thin liquid films: molecular
rapidly evaporating liquids at reduced pressure. J. theory and hydrodynamic implications. PhD thesis,
Fluid Mech. 75: 487-511. University of Minnesota, Minneapolis.
Pearson, 1. R. A. 1958. On convection cells induced Teletzke, G. F., Scriven, L. E. and Davis, H. T. 1987.
by surface tension. J. Fluid Mech. 4: 489-500. How liquids spread on solids. Chem. Eng. Commun.
Pearson, 1. R. A. 1960. The instability of uniform 55: 41-81.
viscous flow under rollers and spreaders. J. Fluid Trefethen, L. M. 1972. Surface tension in fluid mechanics.
Mech. 7: 481-500. Illustrated Experiments in Fluid Mechanics. ed. A.
Pitts, E. and Greiller, J. 1961. The flow of thin liquid H. Shapiro, MIT Press, Cambridge, MA.
films between rollers. J. Fluid Mech. 11: 33-50. Tsay, C. S. and McHugh, A. J. 1990. Mass transfer
Quach, A. 1973. Polymer coatings. Physics and modeling of asymmetric membrane formation by
mechanics of leveling. Ind. Eng. Chern. Prod. Res. phase inversion. J. Polymer Science. 828: 1327-1365.
Develop. 12(2): 110-116. Vrentas,1. S. and Duda, J. L. 1979. Molecular diffusion
Redon, c., Brochard-Wyart, F. and Rondelez, F. 1991. in polymer solutions. AIChE J. 25: 1-24.
Dynamics of dewetting. Phys. Rev. Letters. 66(6): Vrentas, J. S., Duda, J. L. and Ling, H.-C. 1985.
715-18. Enhancement of impurity removal from polymer
Ruschak, K. 1. 1987. Flow of a thin liquid layer due films. J. Applied Polymer Science. 30: 4499-4516.
to ambient disturbances. AIChE J. 33: 801-807. Wehausen, J. V. and Laitone, E. V. 1960. Surface
Schwartz, L. W. and Eley, R. R. 1992a. Numerical waves. Handbuch der Physik. ed. S. Flugge,9: 446- 778.
simulation of post-application flows, incuding flows Whitham, G. B. 1974. Linear and Nonlinear Waves.
arising from surface tension gradients. Paper No. John Wiley, New York.
48i. AIChE, 1992 Spring National Meeting, New Whitman, P. K. and Perera, W. 1992. Prediction of
Orleans, March 29-April 2. coating thickness and drying rate for forced air dried
Schwartz, L. W. and Eley, R. R. 1992b. Coating flows slurry coatings. Paper No. 46i. AIChE, 1992 Spring
on complex surfaces. Paper No. GI4. 1992 APS National Meeting, New Orleans, March 29-ApriI2.
Fluid Dynamics Division Meeting, Tallahassee, Williams, M. B. and Davis, S. H. 1982. Nonlinear
November 22-24. theory of film rupture. J. Call. Interf Sci. 90:
Scriven, L. E. and Stemling, C. V. 1960. The Marangoni 220-228.
effects. Nature. 187: 186-188. Yapel, A. 1988. A physical model of the drying of
Sharma, A. and Ruckenstein, E. 1986. An analytical coated films. MS thesis, University of Minnesota,
nonlinear theory of thin film rupture and its Minneapolis.
application to wetting films. J. Call. Inter! Sci. 113: Yiantsios, S. G. and Higgins, B. G. 1991. Rupture of
456-479. thin films: nonlinear stability analysis. J. Call. Inter!
Sharma, A. and Ruckenstein, E. 1989. Dewetting of Sci. 147: 341-350.
solids by the formation of holes in macroscopic Yilmaz, L. and McHugh, A. J. 1986. Analysis of
liquid films. J. Call. Inter! Sci. 133: 358-368. nonsolvent-solvent-polymer phase diagrams and
Sharma, A. and Ruckenstein, E. 1990. Energetic criteria their relevance to membrane formation modeling.
for the breakup of liquid films on nonwetting solid J. Appl. Polymer Sci. 31: 997-1018.
surfaces. J. Call. Inter! Sci. 137: 433-445.
Sheludko, A. 1967. Thin liquid films. Advances in
Colloid and Interface Science. 1: 391-464.
Smith, N. D. P., Orchard, S. E. and Rhind-Tutt, A. J. NOTATION
1961. The physics of brushmarks. J. Oil Colour a amplitude of irregularity
Chemist's Assoc. 44: 618-633.
aj coefficients in (6.6)
Spitz, G. T. 1973. Leveling in powder coatings. Amer.
Chem. Soc. Div. Org. Coating Plastics Chem. 33.2: B accommodation coefficient
502-507. Bo Bond number, h~pg/(J
Tam, S., Stolarski, H. K. and Scriven, L. E. 1992. How Dg binary diffusion coefficient of solvent vapor
to analyze defect-related stresses in drying films. through gas
N alation 205
Dij multicomponent diffusion coefficient for 10 unit of time measure, 3/3J.1/uh~

component pair i-j T l/2 dimensionless time for sinusoidal disturb-
dn nozzle diameter ance to level to half its initial amplitude
E Young's modulus u liquid velocity in x-direction
F function of nozzle geometry U roll speed
g acceleration of gravity v contact line speed
G mass flux V f final film volume
Gi mass flux of component i VI final volume at liquid/solid transition
h film thickness Wn mean exit gas velocity
H dimensionless film thickness, h/h o x,y spatial coordinate along and normal to the
he film thickness below which spontaneous solid/liquid interface
formation of holes is observed X dimensionless x-coordinate, x/x o
he equilibrium film thickness for a hole of z spatial coordinate perpendicular to the
radius rh solidjIiquid interface
hI film thickness above which gravity prevents
formation of holes
hn distance between nozzle exit and surface
Greek letters
ho mean film thickness and unit offilm thickness
measure wave number of irregularity, holt
k mass transfer coefficient y strain tensor
I characteristic length scale parallel to sub- r capillary length, rh /(u/pg)l/2
strate e dimensionless amplitude, a/h o
distance between neighboring nozzles () contact angle
molecular weight e temperature
Thompson number K thermal (or molecular) diffusion coefficient
ambient partial vapor pressure liquid viscosity
equilibrium solvent vapor pressure Poisson's ratio
reference pressure gas kinematic viscosity
gas constant disjoining potential
equilibrium hole radius for a film of thickness liquid density
h mass concentration of component i
hole radius surface tension
film leveling Reynolds number, ph6u/3J.1213 xx component of the stress tensor
gas Reynolds number, wndn/v g solvent volume fraction
Schmidt number, vg/D g cP erit critical moisture content for a suspension
Sherwood number, kdn/D g of particles
time final solvent volume fraction
dimensionless time, 1/10 solvent volume fraction at liquid/solid
time for hole to become visible transition
PART TWO
METHODS OF INVESTIGATING
COATING PROCESSES
EXPERIMENTAL METHODS 7
Peter M. Schweizer
7.1 INTRODUCTION - VALUE OF

photographic materials, metallic dispersions for
EXPERIMENTAL METHODS
magnetic recording media, polymers, surfactants
Experiments - operations carried out under and cross-linking agents. The manufacturing
controlled conditions in order to discover an process comprises sub-process or unit operations
unknown effect or law, to test or establish a like film formation, dynamic wetting, solidification,
hypothesis, or to illustrate a known law - as drying, etc. A product with desired and well
defined in Webster's dictionary, are the basis of specified properties can only be manufactured if
scientific knowledge and hence, their value is the performance of the relevant unit operations
paramount. In the beginning of science, there and the actions and interactions of the chemical
were only experiments. Later, experimental components among themselves and in conjunction
observations combined with human ingenuity with the unit operations are understood and
led to the formation oftheoretical models. Finally, controlled. Photographic films and papers, for
mathematical tools have been developed to solve example, consist of several layers each of which
the theoretical models for arbitrary and complex contains a variety of chemical components that
applications. Today, most branches of science affect photographic and physical aspects of the
are characterized by the fact that some aspects final product. In modern manufacturing operations
are well known and can accurately be modeled, these layers are coated simultaneously and it is
but other aspects need yet to be investigated and therefore not surprising that there exist no
clarified. This is also true for coating technology. theoretical models that can completely describe
It was recognized long ago that science advances the performance of such complex systems, much
most effectively, if experiments and the develop- less accurately predict the properties of the
ment of theories and analyses are carried out in coated product. Consequently, experiments are
concert. Macosko and Scriven (1984) have called a crucial part of product development activities.
this balanced concept the 'three-legged stool', For many branches of the coating industry, the
and Savage et al. (1992) have named it the following classes of experiments are typical:
'tripartite approach.'
A coated product that can be marketed - a • studying the effects of single solution compo-
photographic film or a floppy disk - for example, nents on the product performance;
is the result of a manufacturing process acting • studying the effects ofcomponent interactions
on and interacting with a coating solution. in single- or multi-layer films;
Typically, the coating solution is composed of • studying unit operations ofthe coating process,
various chemical components and/or emulsions e.g. film formation, film flow on an inclined
and dispersions, e.g. light sensitive emulsions for plane, dynamic wetting, etc;
London. ISBN 0 41206481 2.
210 Experimental methods
• studying product manufacturability in terms manufacturers or software vendors are cited

of capacity (maximum coating speed) and throughout the chapter. The author has either
yield (various coating defects), i.e. determining had personal experience with these methods,
the coating window; instruments, or software packages, or they have
• optimizing product and process performance generally been accepted by industrial and academic
with respect to quality and manufacturability coating engineers. No intention exists to particu-
criteria; larly endorse those manufacturers or vendors
• scaling-up from pilot to production operation; that are cited in the chapter, nor to discriminate
• testing of raw materials. against those that are not listed.
If the discussion in this chapter occasionally
Ifexperiments are essential, so are the methods appears to be biased toward coating processes
and methodologies used to carry them out. The and experimental methods predominantly used
subject of this chapter is to discuss experimental in the manufacturing of photographic products,
methods and methodologies that are relevant for it is because the author had over 15 years of
studying coating flows and their limits with experience in this sector of industry. However,
respect to the uniformity of the coated film. While with a bit of imagination on the part of the
the methods presented will focus on laboratory reader, many of the methods can equally well be
applications, the methodologies, which guarantee applied to other types of coating processes in
an effective use of the methods, can also be other segments of the coating industry.
applied to computer experiments with ease. Further references to experimental methods
Section 7.2 presents not only tools and can be found throughout this book, particularly
instruments (including pilot coating machines in Chapter 3 on wetting, Chapter 4 on surfactants,
and sensors to measure various parameters), but and Chapter 5 on coating rheology.
also methods to gather data effectively (experi-
mental design plans) and evaluate them (quan-
7.2 TOOLS, INSTRUMENTS AND METHODS
tification of film thickness uniformity). Section
7.3 focuses on applications with regard to film The purpose of experiments is to generate
thickness, surface topography, fluid velocity and knowledge in the form of 'know-how' or, prefer-
flow visualization. Section 7.4 is a picture album, ably, 'know-why.' In most industrial companies,
showing visualizations of various flow fields of the research and development environment in
slide coating. The purpose of the photographs is which such experiments are carried out is
to compare characteristic features of the flow controlled by an associated business unit. This
fields as a result of variations in operating and implies the need for effective and efficient
geometrical parameters. Finally, Section 7.5 is experimental work in order to contribute to the
an outlook, offering thoughts on the direction financial success of the business unit. While
of future developments of experimental methods, effectiveness (doing the right thing) is expressed
that may help to advance the understanding of by long-term and short-term strategic and action
coating processes further. plans, efficiency (doing things right) is determined
Obviously, there exists a very large number of by the tools, methods, methodologies and skills
experimental methods that have been or could that are available to carry out experiments.
be applied toward studying coating flows. It is Experimental skills can be learned in school,
not the purpose of this chapter, however, to but above all, they are acquired through hands-
present all these methods nor to compile a on work in the laboratory. Apart from a broad
complete list of references. Instead, only a few knowledge of tools and techniques, experience
techniques suitable to investigate a broad variety combined with theoretical understanding of the
of important coating flow aspects have been area ofinterest are the basis for good experimental
selected. Along the same line, selected equipment work. Equally important are personal qualities of
Tools, instruments and methods 211
the experimentalist such as resourcefulness, for process and product development in the
perseverance, love ofdetail, etc. As far as equipment photographic industry has been published by
and instruments are concerned, they often require Gellrich et al. (1991). According to our own
considerable capital investments. To spend the experience, two types of pilot machines have
money wisely, it is therefore important to know proven themselves extremely valuable for setting
the goals of the experiments and the most up flow visualization experiments, and for
effective way of reaching those goals. investigating the fluid mechanics and physical
principles of various coating flows.
The first one is a simple table-top rig, consisting
7.2.1 PILOT COATING EQUIPMENT
of a (scraped) coating roller and some means of
In most cases, it is not economical to perform delivering the fluid to the coating point. Geo-
experiments on production machines. Instead, a metrical flexibility is the key feature as demon-
set of pilot coating machines is necessary. The strated in Fig. 7.1, where three exemplary
number and types of machines can be deducted configurations are shown.
from the activities (such as those listed in Section The fluid, often a water-glycerin or a polymer
7.1) that need to be performed in a particular solution at room temperature, is recirculated by
environment. Generally, pilot coating machines a suitable means, for example a gear pump, or
should meet the following requirements: a pressurized solution supply vessel. There is no
web involved, so no coating samples can be
• adequate solution preparation facilities and
obtained, nor can the effects of dynamic wetting
delivery systems, see for example Schweizer
as they exist in production coating operations
(1992);
be studied precisely. In spite of these deficiencies,
• more capabilities in terms ofspeed and variation
the setup is a good general learning tool, and it
.ofgeometrical parameters than the production
is perfect for flow visualization experiments.
machine;
Moreover, qualitative features of flow fields such
• multiple coating modes, e.g. extrusion, bead,
as shapes and locations of free surfaces can be
curtain, various roll configurations;
investigated, and conditions for the onset of
• single- and/or multi-layer; various vortices or flow instabilities such as
• narrow coating width; ribbing can be determined quantitatively. Even
• light and/or dark operation (for coating light-
the dependence of the onset of air entrainment
sensitive products);
(dynamic wetting) upon relevant process par-
• setting (gelatinization), curing and drying
ameters can be studied, at least qualitatively.
facilities, if samples are needed.
Last but not least, investment and operating
An overview of pilot coating machines suitable costs are low.
(b) (c)
Figure 7.1 Schematic configurations for table-top pilot coating rigs: (a) slide coating; (b) forward roll coating; (c)
film splitting.
The second type is a narrow-width coating other defects, not only along the lateral edges of
machine with web transport. It can be operated the flow, but also in the supposedly two-dimen-
either continuously or in a loop mode, see sional inner region. In addition, coating modes
schematic layout in Fig. 7.2 as designed in such as gravure, wire rod and spray coating
principle by M. Gueggi and M. Pasquier, senior feature free surfaces with highly three-dimensional
mechanical engineers at Ilford AG. In the structures, at least immediately after the appli-
continuous mode, the coated liquid is neither set cation point and before the spreading, coalescing
nor dried but scraped off the web prior to and leveling processes are completed. The
winding. Dynamic wetting can be studied, and dimensions (wave length) of these structures are
boundaries ofcoating windows can be determined typically small compared with the coating width.
exactly. In the loop mode, an air box is activated In any case, the coating width of a pilot machine
which sequentially supplies cold air for chill- can be minimized in order to minimize investment
ing/setting and warm air for drying. With adequate and operating costs. The maximum coating
mechanical design and process control, the setting width can be estimated from the requirement
and drying steps of a production machine having that 'a few' typical repeats of relevant cross-web
multiple dryer sections can be simulated. If need disturbances must be included; for most appli-
be, a curing/cross-linking step can be added or cations, 10 to 30 cm are sufficient. On a 30 cm
mounted in place of the chilling/setting and/or wide machine operated in the closed loop mode,
drying steps, depending on the particular products additional savings may be possible by coating
that are being investigated. Coating samples can two subsequent 10 cm stripes side by side on the
be obtained, and hence the machine is suitable for same web loop.
studying edge effects, splice coatability, start-ups Coating equipment is worth nothing without
and defects such as cross bars, lines, diffuse the proper coating fluid. In a pilot environment,
streaks, mottle, repellencies, etc. the preferred fluids often have test or model
Apart from edge effects, which are confined to character in that they contain none, or only a
about 1cm on each side, successful coating flows partial number, of chemical additives found in
are essentially two-dimensional. This is true for final products, but they are able to simulate real
many coating modes such as slot, extrusion, slide, coating solutions in terms of their rheological
curtain, roll, blade and dip spin coating. In behavior or other characteristic responses, for
practice, however, the aforementioned coating example with respect to the addition ofsurfactants.
flows are often characterized by unwanted three- One of the simplest model fluids is an aqueous
dimensional features related to instabilities or glycerin solution. It shows Newtonian flow
warm air supply

{[[fI+
00
unwinder winder
Figure 7.2 Schematic diagram of multipurpose pilot coating machine: continuous and closed-loop coating
modes.
behavior up to shear rates of at least 104 s - 1, it different polymer component, for example, latices,
is clean, transparent, odorless, and its viscosity surfactants and thickening agents such as poly-
can easily be adjusted between one and several saccharides or sulfonated polystyroles. Typically,
hundred mPas by changing the glycerin concen- the best type and the optimum concentration of
tration. At high concentrations, viscosity control a polymer additive needed to simulate a real
becomes more difficult, because, owing to the coating solution is found by trial and error. This
hygroscopic nature of glycerin, fluctuations of matching process relies on accurate measurement
the relative humidity in the surrounding air may of fluid properties both for the test and the real
alter the solution concentration and hence its coating solutions. Suitable methods for measuring
viscosity. Glycerin solutions do not gel or cure, static and dynamic surface tensions as well as
which is why they are not suitable for studies rheological properties are described in detail in
where samples of the coated film are needed. Chapters 4 and 5, respectively, as well as in the
However, they can be worked with at room book by Makosco (1994).
temperature, which makes them very attractive
for investigations on pilot coating rigs such as
7.2.2 SENSORS
those shown in Fig. 7.1. Glycerin solutions are
well suited for flow visualization experiments (see The purpose of sensors is to measure the degree
Section 7.3.4), and, at least qualitatively, for of uniformity of the coated film, and to quantify
studying issues offlow stability, e.g., determination relevant parameters of coating flows such as fluid
of the onset of ribbing lines (see Chapters 11 a, velocities, pressures and temperatures. Here, the
11 b, 12a), or the phenomenon of air entrainment. word 'sensor' refers to instruments and methods
If samples of the coated film are needed for suitable for assessing film thickness, surface
further evaluation of the coating uniformity, or topography, fluid velocities and other features
if the real coating solutions exhibit non-Newtonian revealed by flow visualization. Several sensors
flow behavior, then a polymer solution with will be described in detail in Section 7.3.
setting or curing properties is more appropriate Sensors capable of reading relevant parameters
as test fluid. Gelatin solutions, for example, have of coating flows must work in contactless modes
proven useful. Much like glycerin solutions, they so as to not disturb the flow. Depending on the
are clean, transparent, odorless, and their viscosity parameter of interest, they must be able to
can be adjusted over a wide range simply by capture a two-dimensional area of the coating,
changing the gelatin concentration. To prevent or they must be focused on a point. Moreover,
unwanted setting and gelling, the operating they are in motion relative to the flow or the
temperature has to be kept above ambient, at web, either in a parallel or perpendicular direction.
about 40°C. The rheological behavior of gelatin Some features are best detected on-line during
solutions can be characterized with a Carreau-type the coating process, while others are more easily
constitutive equation (Chapter 5; Bird, Armstrong measured off-line from a test sample. On-line
and Hassager 1977). The low shear rate viscosity measurements can be done at the wet end of the
and the time constant that indicates the transition coating machine (immediately after the application
between Newtonian and power law behavior are of the liquid layer), or at the dry end (after the
functions of the gelatin concentration (Blake, drying or curing process). Regardless of the
Clarke and Ruschak 1994). In contrast, the measurement system used, they all must meet
power law index seems to be independent of the very stringent requirements on resolution, par-
concentration. ticularly in the vertical direction, because typical
The value of the power law index, and other wet film thicknesses are on the order of 10 to
rheological characteristics of gelatin or other 1()()() ~m or less, and dry film thicknesses are one
polymeric solutions, such as the level ofelasticity, to two orders of magnitude smaller. In addition,
can be altered by adding small amounts of a film thickness variations must be kept below 1%
in many applications to ensure acceptable product increased knowledge about the subject matter at
uniformity, for example photographic films and hand, and 'investment' refers to time as well as
papers with typical coating widths between 1 and human and material resources spent on such
2 meters. activities.
As will be shown below, most sensors and Work efficiency is determined by many factors.
instruments suitable for coating flow research They include employing a R&D staffwith adequate
use radiant energy, which - ideally - emits from formal education and/or professional experience,
a uniform source, impinges on the coating, is making available properly designed and well
altered by desirable or undesirable features (defects) equipped experimental pilot machines, and offering
in the coating, and is received either in transmission access to powerful computer hard- and software.
or reflection by a detector. In practice, measure- Another factor we particularly would like to
ment systems are found whose wavelengths of point out is called the method of statistical
the radiant energy cover a wide range of the experimental design (SED). This subject was
electromagnetic spectrum, including beta radi- developed several decades ago, and it is extensively
ation, x-rays, micro waves, ultraviolet, visible and described in the literature. Selective examples of
infrared light. Not all types of radiation are textbooks are Draper and Smith (1966), Mont-
equally suited to detect all types of features, gomery (1984), and Deming and Morgan (1986).
because ofdiffraction limited focusing constraints Software packages that have proven themselves
and consequent problems with spatial resolu- useful for PC applications include ECHIP and
tion. An important class of sensors working in Statgraphics for design and evaluation of experi-
the visible wavelength range are cameras comments, and SAS for data analysis. Addresses of
bined with appropriate illumination, see Section these software houses are included in the list of
7.3.4. references.
The state of successful coating flows is not Deming and Morgan (1986) have stated that
only two-dimensional but also steady. In reality, 'statistical treatment is no substitute for good
however, transient situations exist such as start-up data, and good data comes from a well designed
or inception of a defect. In recent years, these set of experiments.' Experimental design is a
transients have caught the interest of coating powerful tool to determine the number and
engineers, because their fundamental understand- placing of experimental conditions to obtain the
ing contributes to yield increases. Consequently, desired information with a minimum of expen-
there is a need for suitable sensors that can either diture of resources.
freeze a rapidly changing situation, or resolve it Even though the method of SED has been
in sufficiently small time intervals. Flash lights proven useful time and time again, we have
and stroboscopes are examples of the former observed that not every engineer or scientist is
category, while the latter one includes traditional readily willing to work with it. There are many
silver halide high-speed photography and modern reasons which explain this reluctance. Sometimes,
high-speed video cameras, see Section 7.3.4. the relatively large number of experiments
requested at the beginning of an investigation
may be a deterrent, because it is hoped that the
7.2.3 EXPERIMENTAL DESIGN
desired information can be obtained with much
METHODS
less effort, i.e. varying just one or two parameters
Any manager of research and development and hoping that the so-determined optimum is
activities, be it in an industrial company or at a not only of local but global nature (home run
university, has the responsibility to assure that syndrome). Sometimes, a problem may be viewed
such activities are carried out efficiently and with too narrow a perspective, thereby missing
generate a maximum 'return on investment.' In important parameters which would have been
this context, 'return' means specific results and necessary to obtain a complete understanding of
the problem. This oversight may then make the as sensitivity analysis of the desired result with
problem look small enough so as to not resort respect to variation of relevant process para-
to a SED but rather hope for a 'home run.' meters.
Reluctance may also come from the fact that a For complex problems, the number of experi-
successful application of experimental design ments needed to obtain the desired information
requires thinking before experimenting, because can quickly increase. For a full-factorial design
questions like 'will the desired information help where all independent parameters are varied
to solve the problem?' and 'can the desired simultaneously over several levels (Montgomery
information be quantified?' must be answered 1984), the number of experiments is equal to the
positively. number of parameter levels raised to the power
On the other hand, the benefits ofexperimental of number of parameters. One way of reducing
design are significant. As mentioned above, the the experimental efforts is using dimensionless
method has the highest return on investment of numbers. According to the n -theorem, or Bucking-
any alternative approach, which is substantiated, ham theorem (Buckingham 1914, Grassmann
for example, by the ease of fitting an empirical 1971), any process or system that can be described
or stochastic model (usually a mathematical by a functional relationship among q parameters,
equation) to the data. This ease is based on the whose units may be given in terms of u fundamental
optimal placing of experimental conditions. The units (i.e. length, mass, time, temperature, electrical
expression 'stochastic' points to a residual term charge, etc.), may also be written as a function
in the model which allows to account for of (q-u) dimensionless groups. Dimensionless
uncertainty. The term 'empirical' describes models representations have the additional advantages
in which the transformation between the dependent of being independent of units, and of indicating
and the independent variables is not known or the ratios of forces that act on a particular
cannot be presumed. The process of model process or system. On the other hand, they are
building itself is empirical, for there are no recipes less transparent for practical applications because
to tell which model is best. Experience has shown the latter are always described by dimensional
that nature is often well approximated, if the parameters.
mathematical equation takes the form of a Another way of reducing the experimental
polynomial or power-law function. Another efforts is applying a sequence of screening and
promising approach is graphical data analysis optimizing. During screening, a coarse experi-
by comparing the experimental curve with a mental design plan is carried out to identify
variety of generic curves that can be found in relevant parameters and discard irrelevant ones.
mathematical handbooks (Abramowitz and Subsequently, an optimum is sought by performing
Stegun 1965, Daniel and Wood 1971). a second, finer experimental design plan with
Proper statistical data treatment and model only the relevant parameters.
building is important because it not only allows Many different experimental design plans
one to correctly identify important parameters have been developed, each of which is charac-
and interactions between parameters, but also terized by particular advantages and dis-
to test the goodness of the model by performing advantages. In contrast to the so called
an analysis of variance (ANOVA). Furthermore, one-parameter method (one parameter is varied
replication of experiments, which is requested in systematically while all other parameters are held
most experimental design plans, provides the constant), methods of the class called factorial
basis for estimating the magnitude of imprecision experiments vary all parameters simultaneously
in the experimental process. Finally, statistical over several levels. Names of well known plans
data treatment and model building is a perfect include simplex design, star design, central
way of condensing the desired information into composite design, full factorial design, fractional
a user-friendly tool for further applications such factorial design, Box-Behnken design, etc. The
reader is referred to the references for detailed 7.2.4 QUANTIFICATION OF COATING

descriptions and optimum choice for a particular UNIFORMITY
application.
Traditionally, statistical experimental design Most experiments are more meaningful, if the
methods refer to experiments performed in the desired information - often some aspect of film
laboratory. This restriction is unnecessary, because thickness uniformity - can be quantified in an
the same methods could very well be applied to objective and reproducible manner. In fact,
experiments carried out on a computer using an quantification of dependent variables is a re-
established mechanistic model. To date, however, quirement for proper evaluation of statistical
the one-parameter method has been preferred in experimental designs. In reality, however, the
most publications where coating flows have been uniformity of coated products is often assessed
examined theoretically by using numerical by visually comparing a sample against a set of
methods to solve the momentum and continuity standards. Typically, there are standard sets for
equations. While this approach allows one to various types ofcoating defects, mottle for example,
visualize and explain physical principles, results, whereby each sample of the standard is assigned
often obtained with considerable computational a numerical value, indicating a different level of
effort, remain of limited practical use, because the defect. Quality assessment ofthis kind is often
they are still presented in a form that does not performed by individuals and can be somewhat
allow interpolation for conditions not specifically subjective.
treated in the study. This is particularly true for The situation described above is rooted in the
complex multidimensional systems, even if highly early years of coating technology because there
developed techniques for tracking desired flow was nothing else available then. Today, in the
states at reduced computational efforts are used, age of highly developed sensor technologies,
for example multiparameter continuation com- coating quality can be measured objectively and
bined with critical point tracking (Yeckel and represented in terms of amplitude and frequency
Scriven 1992). or wavelength, in terms of statistical parameters,
Obviously, we are convinced of the benefits of or in other desirable ways. Usually, the uniformity
statistical experimental design methods. However, of the coated film is captured by an optical
it is not a fool-proof tool that will automatically sensor, and the desired quality information is
solve all existing problems. Instead, as with any extracted by way of image processing and/or
other tool, it must be applied wisely and when suitable data analysis and treatment.
appropriate. In particular, the application of It is often advantageous or necessary to amplify
SED is not a free pass to neglect thinking. Rather, coating defects for better evaluation, either before
it is necessary to state clearly the problem at detection by using, for example, the moire
hand and to develop the objectives of the technique (see Section 7.3.2), or after detection
experiments. Appropriate application of the via digital image processing such as algorithms
methodology refers mainly to that class of for contrast enhancement, etc. If coating processes
experiments where the basic process is well are studied with model fluids, dyes and color
established and the goal is to find optimum filters have proven themselves very useful. Coating
operating conditions for the process. This is in defects can be made more visible, if their color
contrast with another class of experiments where corresponds to the most sensitive wavelength of
the main interest is focused on identifying, verifying the detector.
or visualizing certain physical mechanisms. Here, The uniformity of colored layers in simulta-
as in still other situations, for example those that neously coated multi-layer color photographic
incorporate hysteresis due to multiple flow states materials can be assessed in detail by looking at
(Kistler 1983), the application of SED is more a sample through an appropriate set of color
difficult, if not inappropriate. filters. Unfortunately, experience has shown that
defect patterns of individual color channels, as enhance and evaluate film thickness nonuniform-
they are seen by electronic sensors and expressed ity. While this elaborate and expensive approach
in units of optical density, do not necessarily may be necessary in some applications, it seems,
agree with the superimposed defect patterns of in retrospect, exaggerated for many cases because
all color channels together, as seen through the success can only be achieved with large and
eyes of the quality inspector. This discrepancy long-term deployment of resources.
prevents optimum feedback from Quality Control A simpler but quite successful approach is the
to the Process Development Department, which use of Fourier analysis for treating one- or
in turn may slow down the process ofeliminating two-dimensional data on film thickness uniformity.
coating defects. The discrepancy can be reduced, While this technique is limited to signals that
if the optical density variations per color channel contain some kind of periodicity, or that can be
of a defective coating are translated into coor- converted to periodic signals, it is still very useful
dinates of a colorimetric system, for example the because there is a host of coating defects that
chromaticity diagram as developed by the satisfy this requirement, such as lateral cross
International Commission on Illumination, C.LE. profile, bands of diffuse longitudinal streaks,
(Williamson and Cummins 1983). mottle, ribbing lines, cross bars, etc. Here, periodic
Often, quality attributes of coated materials coating nonuniformities are first captured by a
are judged by subjective perception ofcustomers. suitable sensor, e.g., a CCD array. Next, the
The method called multidimensional scaling discrete (sampled) signal is subjected to a fast
(MDS) is a suitable tool to quantify objectively Fourier transformation (FFT) to represent coating
such quality assessments (Torgerson 1958, Lyne uniformity in terms of amplitude and spatial
and Parush 1983). MDS is a statistically based frequency or wavelength. Today, FFT is readily
metric scaling technique which permits the available as part of various software packages
identification, separation and quantification of for personal computers, for example MathCAD
the factors used by people to judge image quality. or LabWindows. Moreover, an introduction to
Numerical values of these quality factors can be Fourier analysis can be found in the book by
related to image analysis measurements and Lynn (1983).
other known parameters that distinguish the test If a coating defect with a characteristic and
samples. The test procedure requires a controlled dominant frequency is the subject of the inves-
environment with regard to illumination and tigation, it can be isolated by multiplying the
sample preparation. The judges are asked to spectral information with a low-pass, high-pass
compare pairs of test samples, and to indicate or band-pass filter. The cut-off frequencies of the
with a check mark on a continuous scale (from filter are determined either by analyzing the
'exactly alike' to 'completely different') the degree frequency spectrum, by measuring the wave-
of difference seen, as well as to indicate with a length of the defect on the coated sample, or by
check mark on another scale (from 'prefer A' to trial and error. The resulting information may
'prefer B') the preference between samples A and be examined in the frequency domain, or it may
B. The judges are not instructed as to which be transferred back into the original parameter
factors should be used in making their assessment space.
of difference and preference. The power of this technique is demonstrated
There are many roads that lead to the by the following example provided by Rossier
quantification of coating uniformity. Early (1992). A narrow stripe (ca. 3 cm wide, ca. 1.3 m
attempts by the author involved two-dimensional long) ofa full-width sample ofa color photographic
CCD cameras for on-line capture of coating film was uniformly exposed and processed. Then,
quality, mottle for example, and sophisticated the optical density in each of the three color
image analysis hardware in combination with channels cyan, magenta, and yellow was measured
various image analysis algorithms to manipulate, at discrete points spaced apart by about 0.5 cm.
These raw signals represent the widthwise density signals to express the coating uniformity with a
distribution; an example is shown for the cyan single parameter called quality index.
channel in Fig. 7.3. Subjecting this signal to a Consider a generally periodic signal obtained
Fourier transformation, multiplying the resulting from a single scan across the coated web containing
spectrum with a low-pass filter (whereby only bands of diffuse streaks as shown in Fig. 7.5. In
defects with a wavelength of once or twice the Fig. 7.6, a similar signal is depicted schematically;
web width can pass), and performing an inverse it can be characterized by four parameters,
Fourier transformation on the filtered signal, namely the mean value B (DC-component), the
allows one to extract a long-wavelength cross deviation Afrom the mean value (AC-component),
profile which comprises a symmetrical and an the length A. indicating the size of the deviation,
asymmetrical component, see Fig. 7.4. If the and the length C measuring the distance between
low-pass filter is replaced by a band-pass filter, adjacent deviations. Here, W is the width of the
in this case allowing nonuniformities with a final product format.
wavelength between 2 and 10 cm to pass, a defect A model relating these parameters to the
called 'bands of diffuse longitudinal streaks' can coating uniformity can now be constructed by
be visualized, see Fig. 7.5. Here, diffuse refers to considering that the visual perception of an
the edges of the streaks, indicating not a sudden imperfection increases with
but a gradual change in film thickness.
The challenge ofquantifying coating uniformity • increasing 'visibility' of the imperfection, i.e.
lies in reducing the available data, and in the ratio of deviation to mean value, A/B;
representing the degree of coating imperfection • increasing contrast of the imperfection, i.e. the
with as few parameters as possible. In what spatial rate ofchange ofthe signal (slope ofcurve),
follows, a model is developed for deterministic represented in a simple way by the ratio A/A:,
0.75
O.U
0.73
0.72
0.71
-
0.7
~ 0.69
Iii
cQl 0.68
"0
0.67
~
a. 0.66
0
0.65
0.606
0.63
0.62
0.61
0.6
0 20 40 60 80 100 120
Coordinate across the web, [em)
Figure 7.3 Unprocessed signal of optical density distribution on a full-width sample. Horizontal line = mean
density value. From Rossier (1992).
0.04
0.03
0.02
-
.~
(I)
0.Q1
c:
Q)
"0
(ij
.g
8- -0.01
-0.02
---- ------
-0.03
·0.04 +--,---,.--,----,---,.--,----,---.----,----,--.---...,.---1
o 10 20 ~ ~ w w ro w ~ 100 110 120
Coordinate across the web. (em]
cross profile ---- symmetrical part - - ' unsymmetrical part
Figure 7.4 Low-pass filtered signal from Fig. 7.3, showing the densitometric cross profile, and its symmetrical
and asymmetrical components. From Rossier (1992).
0.04
0.03
0.02
-
.~
(I)
0.01
c:
Q)
"0 0
(ij
8" -0.01
-0.02
-0.03
-0.04 +----,--,--,.-----,,--.....---,--.---,------.--.......--.----.--1
o 10 20 ~ ~ w w ro w ~ 100 110 120
Coordinate across the web, [em)
Figure 75 Band-pass filtered signal from Fig. 7.3, showing a defect called 'diffuse longitudinal streaks'. From
Rossier (1992).
terized statistically, for example by computing

the ratio of standard deviation to signal average.
This parameter corresponds to the first term in
C equation (7.1); it is a special case for which the
/\ /~ quality index can only be computed without

distinct frequency information. For the signal in
B
~ Fig. 7.5, the parameter takes the value of
0.008315/0.663 = 0.013.
W
A more informative way of characterizing
stochastic signals is using amplitude and frequency
Spatial Coordinate information extracted from the power spectrum
of the signal, because power spectral information
Figure 7.6 Schematic signal of a nonuniform film does not zero-out for stochastic signals (Shulman
thickness profile. 1970).
Examples ofthis approach, which was developed
• increasing number of imperfections, i.e. the to quantify the severity of mottle, are shown in
ratio of format width to spacing of the Figs 7.7 and 7.8 (Rast 1989). Here, mottle is a
imperfections U1C. sensitometric defect found in photographic
products coated on uneven paper support. Mottle
The quality index, QI, is the product of these
has characteristic wavelengths ranging from 1 to
three factors
10mm; it is thought to be caused by liquid
AAW A2 W leveling after coating due to an egg-carton-like
QI = IiI C = BAC (7.1) surface topology of the paper. Even though
mottle is an area-wide defect, it appears to be
The smaller the quality index, the better is the isotropic, at least in some cases, i.e. its spectral
coating uniformity. The parameters in equation response does not strongly depend on the direction
(7.1) are obtained from appropriate signal of measurement across the web.
processing, for example by computing the mean In Fig. 7.7, the optical density information of
value and the standard deviation of the signal a mottle pattern in the cyan channel of a color
to get A and B, and by performing a Fourier paper as captured by a one-dimensional CCD
analysis (FFT) to get A and C. camera and treated by a FFT is presented in the
A set of eight test samples has been constructed frequency domain in terms of the logarithm of
by superposing stripes of transparent optical the amplitude A, and the square root of the wave
filter sheets having different optical densities and number (T. (T is defined as 1/A, with A being the
widths to form rectangular wave trains with characteristic wavelength of the disturbance in
varying A, B, C and A. The ranking of these millimeters. The lack of a dominant frequency
samples by a group of test persons was in good peak is apparent, indicating the stochastic nature
agreement with the ranking as computed by of the signal. Nevertheless, the mottle pattern
equation (7.1). can be characterized and quantified by two
If the raw signal of the coating uniformity is parameters, namely the extrapolated amplitude
not deterministic but contains a large stochastic A o at wave number (T = 0 (A = (0), and the wave
component, the above described approach for number (T 1j2 corresponding to an amplitude of
quantification does not work, because even though A o/2.
several measurement traces from the same sample A o and (T 1j2 are extracted from the spectrum by
would result in several different (deterministic) approximating the curve with a suitable analytical
signals, their mean value of the deviation from expression. In the example at hand, equation
uniformity is zero. Such signals must be charac- (7.2) is a best fit for mottle application; it has
10.0
9.5
9.0
8.5
~
c
8.0
Q>
"0 7.5
~
c.. 7.0
E
<{
6.5
6.0
5.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Figure 7.7 Frequency spectrum of a coating defect called mottle. A o = 3728.2, (j 1/2 = 0.312. From Rast(1989).
10.0
9.5
9.0
8.5
~
.£ 8.0
Q>
"0 7.5
:::J
~
c.. 7.0
E
<{ 6.5
6.0
5.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Figure 7.8 Frequency spectrum of a coating defect called mottle. A o = 1716.4, (j 1/2 = 1.395. From Rast (1989).
been found by trial and error, and it is shown measures the loss of energy due to selective
as solid lines in Figs 7.7 and 7.8. To evaluate A o absorption by the coated film. For transparent
and <1 1 / 2 , only that portion of the spectrum in films, the energy absorption is described by
terms of wavenumber is used that is relevant Lambert's law, which states that equal paths in
for the defect under consideration; here, the same absorbing medium absorb equal fractions
0.25 mm - 1 < <1 < 1.0 mm - 1, corresponding to of the light that enters them (Longhurst 1973).
1mm<A.<4mm.
(7.3)
A(<1) = A o2-(0'/0'1I2)0-> (7.2)
where 10 is the incident energy, It is the transmitted
As shown in Fig. 7.8 for the yellow channel of energy, H is the coating thickness, and K is the
a different type of color paper, the parameters absorption coefficient. If the absorbing substance
A o and <1 1/ 2 have changed compared with Fig.
is a solution, then its absorption coefficient is
7.7, because process and product related par- directly proportional to the solution concentration.
ameters such as viscosity, density, surface tension, This statement is called Beer's law; it is valid, if
coating speed and layer thickness are different. the absorbing power of a molecule is not affected
Once the product uniformity has been quan- by the proximity of its neighbours (Longhurst
tified, quality criteria or threshold levels must be 1973).
specified to separate acceptable from unacceptable Depending on the frequency or wavelength of
uniformity. This task is dependent on the particular the emitted energy, different measurement methods
application, for example format size, and whether are distinguished. Infrared absorption, microwave
the product is transparent or opaque in the case absorption, and x-ray fluorescence belong to this
of photographic products. This task is also rather
class of measurement technique; they are described
difficult because, beyond the parameters contained in more detail in the following sections.
in equations (7.1) or (7.2), such criteria must
account for the sensitivity of the customer's eyes
to spatial frequencies (Campbell and Robson 7.3.1.1 Infrared absorption
1968) and spectral wavelengths of nonuniformities
This measuring technique is concerned with
(Cornsweet 1970; Kueppers 1973). In practice,
wavelengths A. from the near (0.7 to 3.5 Ilm) and
however, acceptability thresholds are often set
mid (up to 20 Ilm) infrared spectrum. Infrared
subjectively against a particular reference or
energy is produced by any hot or incandescent
'standard' sample. Objective measures can then
object; hence, simple low-watt halogen bulbs are
be defined from measurements on the standard.
sufficient and often used as energy source.
When infrared radiation is passed through a
7.3 APPLICATIONS
film or coating, some of it is selectively absorbed
(Hindle 1984). This loss of energy is due to the
7.3.1 FILM THICKNESS
vibrational exitation of particular molecules
There is an abundance of scientific literature present in the material. Furthermore, the loss of
describing film thickness measurement methods. energy is characterized in the electromagnetic
The material that follows, however, focuses only spectrum by an absorption band whose mean
on a few techniques that have proven themselves wavelength is indicative ofthe absorbing molecule,
useful for fundamental coating flow studies and and whose depth is proportional to the loss of
for routine quality surveillance of production energy and hence to the film thickness. In Fig.
coating processes. 7.9, the absorption spectrum of a polyethylene
An important class ofnon-contacting measure- film is shown. The spectral curve contains a
ment devices utilizes a point source that emits trough (band ofmaximum absorption or minimum
radiation of a particular frequency range in the transmission) with a maximum depression at a
electromagnetic spectrum, and a detector that wavelength of 2.32 Ilm. Examples of absorption
Applications 223
peaks for other hydrocarbon groups include difference of absorption at the two wavelengths,
polyester at 3.4 ~m and polystyrene at 2.1 ~m. must be determined via calibration.
Wet films of aqueous coatings, or equivalently, The infrared method has been highly developed
the water content of a coated film can be gauged by equipment manufacturers, for example Moist-
by using the absorption band of the -0H hydroxyl ure Systems Corporation, or Infrared Engineering
radical at 1.94 ~m. As the absorption is selective Limited. It is possible to measure a great variety
in all cases, the components of the air, including of coatings on metal, paper or plastic substrates.
moisture, do not significantly affect the radiation Moreover, problems associated with interference
at these wavelengths. fringes on very thin and flat films, with light
Once the absorption band of a particular scattering due to pigments and fillers, and with
molecule is known, it can be isolated with an measuring the thickness of a coating whose
optical reference filter (narrow band-pass filter), absorption band is very close to or even
which means that only a narrow portion of the overlapping with the one of the substrate, have
spectrum around the absorbed wavelength is been solved (Hindle 1984; Edgar and Stay 1985).
considered. In Fig. 7.9, the interference filter is However, some critical issues remain that warrant
schematically indicated by a hatched spike at a cautious use of the method; they include
A. = 2.32 ~m. Using a second interference filter coatings on hightly reflective substrates, and
placed outside the absorption band (second influence of the measurement signal by lateral
hatched spike near A. = 2.2 ~m in Fig. 7.9) provides gloss variation in scanned applications. Moreover,
a means of determining what would have been variations in the moisture trapped within the
the incident light level 10 at the absorbing substrate can mask subtle changes due to
wave-length. By applying equation (7.3) to both nonuniformities in the coating.
locations of the interference filters, it can be Characteristic specifications of the infrared
derived that the film thickness is proportional method are listed in Table 7.1 together with
to the logarithm of the ratio of the detected similar data for other measurement methods.
transmitted energy at these two wavelengths The compilation of the table caused some
according to difficulties as there is no consistent way by which
manufacturers specify the performance of their
H= 1 In(111 ) (7.4) equipment. Furthermore, most manufacturers
K 2 -K 1 112 offer several equipment classes, each of which
The proportionality factor, the inverse of the covers a different range of application. Finally,
in most cases, the optimum attainable performance
depends on the particular application and should
be discussed with a qualified specialist. Hence,
the numbers indicated in the table are not
absolute but rather 'ball park' figures. The term
'accuracy' indicated in the table as ± % is a
measure for the vertical resolution of the measure-
ment system relative to its range of application.
7.3.1.2 Microwave absorption

Microwaves, covering wavelengths between 1 cm
2.6
2.1 2.2 2.3 2.' 2.5
and 1 m in the electromagnetic spectrum, have
Wavelength [micron]
long been used to determine the relative moisture
Figure 7.9 Infrared adsorption spectrum of a 50 ~m content of materials. In early applications, that
thick polyethylene film. From Hindle (1984). property could only be obtained by additional
Table 7.1 Performance comparison between film thickness measurement methods
Sampling Sampling Scan

Range Accuracy time area velocity
Method (11m) ±% (s) (mm) (cm/s)
Infrared 2-150 >0.5 >0.2 010 <4.2

Microwave 1000 >1 >1 05-10
X-ray fluorescence depends on several factors 1-10 >5-100 030 <7.5
Capacitance gauge <500 >0.1 >0.001 05 <3
densitometric measurements, because, in the lower ronment, the sample is irradiated with a low-level
frequency range around 1 GHz, the electrical radioactive source (x-rays). An x-ray photon
properties of many materials are influenced by ejects an electron from the inner shell of an atom
ionic conductivity and other effects. However, and an outer shell electron drops to fill the
the need for such extra efforts was eliminated by vacancy. The difference in orbital energy causes
operating microwave guides above about 10 GHz the outer electron to fluoresce and emit an x-ray
(Meyer and Schilz 1980). during the transition (Asoma Instruments 1992).
In later applications, the use of microwaves The energy (wavelength) emitted by the electron
for measuring the thickness of liquid films has depends on the type of transition which has been
also been considered (Roy et al. 1986; Hurley, made, i.e. it is indicative of the type of atom that
Kaufman and Roy 1990). Instruments are typically has been excited. Once the equipment has been
operated either in the attenuation mode or in calibrated, coating thickness is determined by
the phase mode. In the attenuation mode, the measuring the x-ray intensity of excited atoms;
amount of microwave power absorbed by a this procedure is known as the x-ray emission
material is proportional to its water content. On method. The relation between intensity and
the other hand, the phase mode depends on the thickness is typically nonlinear, although it can
measurement of the propagation velocity of be linearized for thin coatings, Fig. 7.10. As the
electromagnetic waves through a material. The
-
waves experience a larger phase delay per unit
./-- -
length of travel in a structure containing material 100
with a high dielectric constant, such as water, 90
than in the same structure with a low dielectric 80

/
value. 70
The response of microwave guide systems as c 60
~
.~ /
a function of the film thickness is slightly nonlinear,
c
~ 50
//
and a calibration procedure is required prior to :g.~ 40
c
I
c
use. Performance specifications are listed in C<
30
Table 7.1. Compared with other methods, 20
~ - I Linear Reaion I
microwave absorption is not suitable to measure 10
I
very thin films. COMPUR Electronic GmbH is /
o
a vendor of microwave moisture meters. o 10 15 20 25 30 35 40
Thickn@$$ (micron]
7.3.1.3 X-ray fluorescence Figure 7.10 Zinc coating on aluminum. Intensity of

zinc coating x-rays relative to the intensity of an
X-ray fluorescence (XRF) is primarily a spec- infinitely thick zinc coating versus the thickness of the
trometric technique where, in a shielded envi- coating. From Asoma (1992).
Applications 225
incident x-rays are absorbed by the coating they law (equations 7.3 and 7.4), and a typical response
can only penetrate a certain distance into the curve is shown in Fig. 7.11 (Asoma Instruments
film. Hence, the application of this method is 1992). If both the coating and the substrate
limited by a maximum film thickness, defined by contain no detectable elements, an excitable
the American Society for Testing and Materials backing specimen may be added, and the
(ASTM A-754) to be 3/11., with 11. being the mass absorption of the x-ray by the substrate and the
absorption coefficient of the coating for the coating can be recorded. However, this approach
detected x-rays. is only reliable if the thickness variations in the
XRF is capable of detecting a great number substrate are minimum.
of atoms ranging from aluminum to uranium. Asoma Instruments is a vendor of XRF
For example, the method has been used to detect equipment. Typical performance characteristics
Ag in photographic films, Fe in coating of are presented in Table 7.1. To achieve good
magnetic recording media, Si and Ti in paper, accuracy, reliable standards for calibration are
etc. However, several x-ray sources are needed necessary, preferably similar to actual coatings.
to cover the entire range, which necessitates a Surface cleanliness and flatness influence the
proper adaptation of the source to a particular attainable accuracy. One drawback to this method
measurement application. The level of radioactivity is the relatively long sample times necessary to
of the source is normally low enough so that no acquire accurate mean values. Consequently, the
special licence to process or operate is required. sample rate is low, which will greatly reduce the
If the coating does not contain any elements maximum scan coverage, if the measuring head
detectable by XRF, its thickness can still be is scanned across a traveling web.
determined by exciting suitable atoms in the
substrate and measuring the absorption and
attenuation of the resulting x-rays by the coating 7.3.1.4 Capacitance gauging
(x-ray absorption attenuation method). Now the The electrical capacitance, C, of a parallel plate
method can be described by the Beer- Lambert capacitor is given by
(7.5)
100
~ 90 \
x~ 80 \ where eo is the permittivity of the free space or
'" "
the dielectric constant of the vacuum (eo = 0.885
e 70 10- 11 A s/V m), er is the dielectric constant of the
~ 60
medium between the plates (for example paper:
.~ 50
tJl
C
ai 40 """ "- ~
er ~ 4.0, Mylar polyester: er ~ 3.1), A is the plate
area and D is the distance between the plates.
---
'0; 30 Capacitance varies with the distance between
.~ 20
"--.. two conductive plates, thus this electrical principle
Qj 10 t---.
c:: can be used for contactless displacement gauging
o between a flat plate sensor and a grounded target
o 5 10 15 20 25 30 plate, or for measuring the thickness of a
Coating weight of Si02 (mg/cm 2)
nonconductive film by moving the film between
the sensor and the grounded metallic backplane.
Figure 7.11 XRF response ofa substrate as a function In the latter case, the capacitance changes
of the thickness of a film containing no elements
according to material thickness, which is pro-
detectable by XRF. Example: Intensity of iron x-rays
as a function of the coating weight of a Si0 2 film. portional to the dielectric constant. After scaling
From Asoma Instruments (1992). for the dielectric constant of the film, the signal
can be converted into a thickness value (McRae ment and accuracy of the obtained results should
1988; Ayers 1993). be sought.
All other parameters being equivalent, the
signal output from a capacitance gauge is
7.3.2 SURFACE TOPOGRAPHY
proportional to its sensor area. This means that
the signal-to-noise ratio, the signal stability, and Surface topography describes the configuration
the vertical resolution improve with larger sensors, of a surface including its relief and the position
at the expense of reduced spatial resolution in of characteristic features. Topography does not
the film plane. A possibility to improve spatial provide any information about what happens
resolution without reducing the sensor size is to below the surface. In contrast, film thickness
use a rectangular 'stripe' sensor instead of the measures the distance between the surface and
usual round sensor. However, as spatial resolution an interface located below the surface. In general,
worsens in the orthoganol direction, this option film thickness cannot be inferred from knowledge
is sensible only in certain applications. A high of the topography. To illustrate this situation,
vertical resolution can only be obtained, if the consider a process where a liquid film is coated
sensor surfaces are placed closely together, typically onto an uneven substrate characterized by a
just a few hundred microns. This fact may render sinusoidal wavyness. Assuming conformal coating,
the application of the technique critical, if the then, immediately after coating, the film thickness
film thickness must be measured on an unsup- will be uniform throughout, but the topography
ported (fluttering) web. Performance specifications of the free surface will show the same nonuniform-
for capacitance gauges are listed in Table 7.1. ity as the substrate. As time proceeds, the liquid
An example of a supplier of such instruments is film will level due to gravity and surface tension
Lion Precision Corporation. forces until a new equilibrium state is reached.
Applications of capacitance probes to measure Now, topographical information will show a
film thickness of coating flows have been shown perfectly flat surface, while the film thickness
to be accurate and convenient, particularly in profile is nonuniform according to the sinusoidal
pilot coating rigs with stainless steel coating rolls wavyness of the substrate.
of relatively large diameters, such as those shown
in Fig. 7.1. Examples of experimental data of
7.3.2.1 Profilometry
such investigations can be found in Spiers,
Subbaraman and Wilkinson (1974), and Thar- Instruments that probe the film surface in a
malingam and Wilkinson (1978). raster-like fashion by way of mechanical or
Capacitance gauges are also used to measure optical sensors are called profilometers. Film
the uniformity of thin gaps and slots as they exist thickness must be inferred by calculating the
in extrusion dies and coating hoppers or between difference between the coated sample and the
pairs ofcylinders in roll coating operations. Here, uncoated reference surface. As this is a rather
the probe is shaped like a button and mounted cumbersome process, profilometers are typically
in a thin shim for support. Capacitec is a supplier not used for film thickness measurements but for
of such instruments. Gaps as narrow as 300 Ilm the assessment of surface topography. Tactile
can be measured with a probe diameter of 4 mm sensors (similar to the pick-up needle of a record
and a claimed vertical resolution of 1 Ilm. Reliable player) make physical contact with the test
measurements depend on a tight fit of the probe surface, hence, they are not suitable for measuring
in the gap; this in turn makes the delicate probe liquid films. Moreover, their measurement rate
susceptible to mechanical damage, if the probe is relatively low compared with optical sensors,
has to be inserted deep into the gap and pulled two of which are worth mentioning and are
across the width of the gap (coating width). In described in more detail below (Blum et al. 1990).
practice, a compromise between ease of measure- The triangulation method utilizes a laser beam
Applications 227
that is focused onto the test surface and viewed A method for determining highly curved surface
at a fixed angle by a position-sensitive detector. profiles of the upper meniscus in slide bead
When the target moves in the direction of the coating has recently been presented by Ikin
incident beam, the light spot moves accordingly (1992). In this 'bead profile tracker', a wire
on the detector, which can be converted into a electrode of 75 Ilm diameter is mounted on a x-
displacement of the test surface relative to the and y-translator driven by stepper motors. The
detector. The method is characterized by high probe is slowly lowered toward the liquid surface,
sampling rates (up to 40 kHz), high vertical and and upon contact, the electrical circuit switches
lateral resolutions (0.5 Ilm and 100 Ilm, respect- the y-translator to up-stroke. The wire must be
ively), a large measurement range (10mm), and lifted far enough to shed the drop which clings
a moderate working distance (3 to 5 cm). Best to the probe due to capillary action. After each
results are achieved on mat or rough surfaces vertical cycle, the probe is displaced horizontally
that are opaque. A drawback leading to measure- by the x-translator to its next measurement
ment errors is attributed to the fact that the point. The surface profile is recorded by mem-
incident light beam may partly penetrate into orizing the y-position on the downstroke at the
the test surface, if the latter consists of an point of initial electrical contact. The method
optically transparent material such as a polymer also allows for correction for the wire radius,
film. In this case, the light seen by the detector depending on whether the wire is leading or
no longer originates from the test surface, but trailing the free liquid surface. Bead profiles
from some unknown location beyond the surface. obtained in this fashion have successfully been
A supplier oflaser displacement meters based on used as initial guesses for finite element calculations
the triangulation principle is KEYENCE Cor- of the bead flow field. Moreover, the technique
poration. can be extended to plotting three-dimensional
In the auto-focus method, a laser beam is effects, for example ribbing lines. A similar setup
focused onto the test surface so that the distance was published by Campanella and Cerro (1984)
between the focusing lens and the surface is held who used a micrometer to measure the film
constant by a dynamic control mechanism. The thickness on a partly submerged rotating cylinder.
reflected beam is imaged via a prism onto a Here, a high impedance meter provided visual
differential photodiode such that both diodes are indication that the micrometer tip had touched
illuminated equally, ifthe distance between surface either the liquid film or the metallic surface of
and focusing lens exactly corresponds to the focal the roll.
length of the lens. If the surface is slightly
displaced, then the two photodiodes are no
7.3.2.2 Moire technique
longer illuminated equally, and the control
mechanism becomes active in order to re-establish Contactless measurement of the area-wide top-
a focused beam. Compared with the triangulation ography of a surface can be achieved with
method, the auto-focus method has lower sampling interference techniques. Classic optical inter-
rates (1 kHz), better vertical and lateral resolutions ferometry requires a coherent laser light source
( <0.01 Ilm and 1 Ilm, respectively), a smaller and a vibration-free environment. The measure-
measurement range « 1mm), and a smaller ment range of application is on the order of
working distance (2 mm). Measurement uncer- 10 Ilm, and the resolution is as high as nanometers.
tainties related to light penetration on transparent These characteristics render the method unsuitable
surfaces are much less of a problem than with for studying liquid film coatings.
the triangulation method, due to the highly In contrast, the moire technique, which is also
focused incident laser beam (Breitmeier 1993). A called mechanical interferometry, does not require
supplier of auto-focus profilometers is UBM coherent light. Its measurement range (depth of
Messtechnik GmbH. topography) extends over five orders ofmagnitude
from about 10 ~m to more than 1 meter with a leads to a periodic deformation of the film
depth resolution of better than 10 ~m, and the surface. Instability is obtained by changing one
characteristic dimension of the object may be of the process parameters, for example increasing
between centimeter and meter. the coating speed, beyond a critical value, while
The development of surface topography by all other parameters are held constant. However,
moire technique started in the 1960s (Nishijima the growth of the amplitude of the deformation
and Oster 1964; Theocaris 1967; Meadows, after the inception is relatively weak, making it
Johnson and Allen 1970) with considerable efforts difficult to determine exactly the onset of the
in medical applications for contouring human instability without the help ofan optical amplifier.
backs (Takasaki 1970, 1979). More recently, In optical interferometry, a fringe pattern of
Kheshgi and Scriven (1983,1991) have described bright and dark areas is produced by superposing
how the method can successfully be applied to two coherent light beams, i.e. beams that have
coating flows. In particular, they have investigated the same wavelength and a constant phase
the leveling of a horizontal film as a function of relation. With moire, similar fringe patterns are
time, the growth of surface waves in a film flowing generated by superposing two mechanical grids
down an inclined plane, deformations in a liquid or optical gratings. The line spacing of the
curtain due to ambient disturbances, and the grating should be large enough to avoid blurring
local thinning of a liquid film due to surface of the image through diffraction effects (i.e.
tension gradients as a result of a surface-active grating pitch less than about 40 linesjmm). If the
contamination. spacing of the two gratings is even, then the
The author of this paper has successfully used spacing of the resulting fringe pattern is also
the moire technique as an optical amplifier to even, Fig. 7.12a. However, if one of the gratings
facilitate the determination ofthe inception point is distorted, so will be the fringe pattern, Fig. 7.12b.
of 'ribbing lines' in slide coating. As described in With respect to coating flows, deformations of
Chapter 11 b, ribbing is a flow instability that liquid surfaces can be used to distort one of the
(a)
(b)
Figure 7.12 Formation of moire patterns: (a) evenly spaced fringes by superposing two uniform grids; (b)
unevenly spaced fringes by superposing an uniform and a disturbed grid.
ALBUM OF COATING FLOWS
(a)
(b)
(c)
Fig. 7.20 Visualization of a liquid layer exiting from a slot onto an inclined plane: effect of
the flow rate on the presence of vortices and the position of the static wetting line. Flow
parameters are listed in Table 7.2.
(a)
(b)
(c)
Fig. 7.21 Visualization of a liquid layer exiting from a slot and merging with another layer on
an inclined plane: effect of the geometry of the downstream corner on the presence of vortices.
Flow parameters are listed in Table 7.2.
(a)
(b)
(c)
Fig. 7.22 Visualization of the bead (liquid bridge) in slide coating: effect of the pressure
difference across the bead on the positions of the static and dynamic wetting lines. Flow
parameters are listed in Table 7.2.
(a)
(b)
(c)
Fig. 7.23 Visualization of the bead (liquid bridge) in slide coating: effect of the lip geometry on
the flow field of the bead. Flow parameters are listed in Table 7.2.
(a)
(b)
(c)
Fig. 7.24 Visualization of the bead (liquid bridge) in slide coating: effect of the application
angle on the flow field of the bead. Flow parameters are listed in Table 7.2.
Applications 229
gratings, which then is superposed with an adjacent equidistant fringes (vertical resolution)
undistorted reference grating to generate the has been calculated by Meadows, Johnson and
fringe pattern that contains the desired information Allen (1970); the result in approximate form is
about the surface topography. given by equation (7.6):
Moire topography can be set up according to
different optical arrangements, which are called ~Z=pH (7.6)
oblique shadow method, reflected image method, D
and refracted image method. Selection criteria
for a particular application include the light Thus, the difference in height between two
source (collimated or diffuse), the characteristics points that are separated by n fringes is obtained
of the surface to be tested (diffusely reflecting, by multiplying the right-hand side of equation
specularly reflecting, or transparent), the relative (7.6) by n. The sensitivity of the method can be
positions of the illuminating and receiving optics tuned by changing either the line spacing of the
(reflection or transmission mode), the sensitivity illuminated grating, p, the distance between the
to movement of the surface to be tested, and the illuminated grid and the observer, H (which is
type of information to be extracted. Details and assumed to be equal to the distance between the
characteristic features ofeach method are described grid and the light source), or the distance, D, by
by Kheshgi and Scriven (1983). which the source and the observer are separ-
Here, the optical lay-out of the oblique shadow ated. With D = 200 mm, H = 50 mm, and
method is presented as an example and sche- p = 0.05 mm, for example, a resolution of 12.5 J.lm
matically depicted in Fig. 7.13 (Meadows, Johnson is obtained.
and Allen 1970). The evaluation and quantitative interpretation
A non-collimated light source is used to of moire fringe patterns that contain information
illuminate the grating, which is casting a shadow about the surface topography is not an easy task.
on the specimen surface. If the specimen surface An example can be found in Kheshgi and Scriven
is distorted, the shadow will be as well. Moire (1983). They measured the leveling (temporal
fringes are obtained by superposing the shadow evolution of the surface topography) of a wavy
with the reference grating. The sensitivity of the liquid film, and they derived a mathematical
method, i.e. the height difference ~Z between reconstruction ofthat film surface from the moire
two points on the surface corresponding to two fringe pattern that was generated by the refracted
image method. Their analysis involves an inte-
Light source o Observer gration which, in practice, may be rather cum-
bersome to carry out. Today, more convenient
methods are commercially available. Breuckmann,
H for example, offers a system in which the fringe
pattern is captured by a CCD camera and fed
into a computer for image analysis. The surface
--
Grating topography is extracted by way of sophisticated
algorithms based on the phase shift method
~~i/!1Jrrm7l7i7l7Jfrr,. . . Specimen (Breuckmann and Thieme 1985), and displayed
surface either as colored contour plot or as three-
dimensional mesh. This added convenience has
a price of several thousand dollars, however.
In summary, the moire technique is a useful
Figure 7.13 Moire technique: schematic diagram of method for generating two-dimensional topo-
the optical configuration for the oblique shadow graphical information ofdeformed liquid surfaces.
method. Its theoretical maximum resolution of about
Illm is adequate for most coating flow applica- inating problems of the coating process, for
tions. Practical resolutions, however, may not be example leaky pipe fittings that lead to air
quite as good due to physical limitations in bubbles in the coating solution;
setting up an optimum optical arrangement as • memorizing events in terms ofsize and position
in the case of detecting the onset of ribbing lines on the web in order to facilitate the elimination
in the nip of a forward roll coating process. of defects in a later process step;
Quantitative information about surface topogra- • identifying 100% defect-free master rolls (the
phy is either difficult to obtain or at a relatively term 'master roll' refers to uncut, coated rolls,
high price for image analysis equipment. The last as they come off the coating machine);
argument points the value of the moire technique • serving as a research tool for studying particular
toward qualitative visualization of deformed coating defects such as ribbing lines and their
liquid surfaces. temporal decay (leveling) after the coating
point as a function of relevant process
parameters (Ikin 1992).
7.3.2.3 Laser scanners
A variety of laser scanner schemes exist, and
Laser scanners are powerful electro-optical systems
they are well described in the literature (Beiser
that measure surface or interface deformations
1974; Pietzsch, Muller and Feige 1988). In essence,
resulting from discrete disturbances. Their main
they comprise emitting optics, receiving optics,
purposes include:
and a computer for data processing and evaluation.
• on-line monitoring of the quality of a coated In a typical commercially available system, the
film (uniformity of surface topography); heart of the emitting optics is a laser whose beam
• triggering an alarm in case ofan event (excessive is reflected off a fast rotating multifaceted mirror
deviation from product uniformity). Owing to wheel (typically eight faces), which causes the
the fast response time, this feature not only beam to scan across the web, Fig. 7.14. Scan
reduces coating waste, but it also provides a rates of up to 3 kHz result in very high linear
means for identifying and subsequently elim- scan velocities of several km/s. The wavelength
(a) (b)
Figure 7.14 Emitting optics of a laser scanner system: (a) simple optical arrangement with rotating multifaceted
mirror wheel; (b) telecentric beam path resulting in perpendicular intersection between beam and web. From
Pietzsch, Muller and Feige (1988).
Applications 231
of the laser light may be chosen between 442 nm acterized by a large (ca. ± 15°) aperture in scan
(blue) and 830 nm (infrared) to optimize product direction and by a small (ca. ± 5°) aperture in
compatibility for a particular application. down-web direction.
In a simple optical arrangement as described In another optical arrangement, the laser beam
in Fig. 7.14a, the incident laser beam would from the web is focused onto the point-like detector
intersect the web at an angle that varies with by way of a lens or a combination of cylindrical
lateral position, hence resulting in unequal lenses and cylindrical mirrors (parabolic mirror
interpretation of coating defects across the web. collector), Fig. 7.15b. This system features a very
To overcome this difficulty, Sick (1972) has small (ca. ± 1°) aperture in both scan and down-web
patented the so called telecentric beam path direction which, in combination with dark- and
which guarantees a perpendicular intersection bright-field apertures, allows the detection ofdefects
between the incident beam and the web at any which produce very small beam deflections such
position across the web, Fig. 7.14b. The size of as longitudinal lines.
the scanning laser spot (spatial resolution) is A bright-field detector receives predominantly
small and may even be varied in sophisticated specularly reflected light and light from strongly
systems. Typically, it has the shape of an ellipse scattering surfaces. Defect-free coating surfaces will
with a cross-sectional area of less than 1 mm 2 • lead to high detector signals while any surface
Despite this small size, the high scan frequencies defect will lower the signal amplitude. Dark-field
allow for 100% scan coverage even for high detectors, on the other hand, have a small plate in
coating speeds of up to 3 m/s. their center that blocks off all specularly reflected
The receiving optics can be arranged in either light and hence gives a low output signal for
reflection or transmission mode, depending on defect-free surfaces. However, weakly scattering
the particular application. Furthermore, they can surfaces or surface defects causing only slight
be classified according to their receiving aperture deflections of the specularly reflected beam will
with respect to the scan and down-web direction, lead to high detector signals. The latter is often the
and according to the receiving angle of the case with sharp longitudinal lines or streaks.
detector, i.e. bright field or dark field (Pietzsch, Laser scanner systems are, above all, good at
Miiller and Feige 1988). These different charac- detecting surface deformations such as scratches,
teristic features of the receiving system can be impressions, creases, wrinkles, fibers, air bubbles,
used to distinguish between different types of inclusions, repellencies, ribbing lines, etc. Detection
defects, for the extent of scatter (distribution of of high-frequency cross lines and longitudinal
the scattered laser light in terms of intensity and lines, particularly if they are located at an interior
angle) depends on the nature and structure of interface of a multilayer film, is more difficult.
the surface deformation on the web. Typically, Detection of a particular defect is facilitated by
the ability of a laser scanner system not only to selecting the optimum aperture and detector type
recognize but also unequivocally differentiate which, in most cases, is arrived at by trial and
between various surface defects depends on the error. Defect classification is achieved by real-time
particular application and must be established data processing using algorithms which examine
by trial and error. 'fingerprints' of defects, for example size, aspect
The so called light-guide rod collector (Sick ratio, orientation of major axes, edge contrast,
1979) is a plexiglass bar that spans across the signal amplitude, rate of repetition, and other
entire web. The laser beam enters the bar, is parameters (Pietzsch, Miiller and Feige 1988;
reflected off a metalized mirror raster and, by Pietzsch and Miiller 1988). Once a defect is
way of total internal reflection, is guided onto a identified, it is displayed on a screen by its specific
photomultiplier which produces an electrical mark, a letter for example. The temporal evolution
current that is proportional to the amount of of this display represents the defect map of the
incident light, Fig. 7.15a. This receiver is char- coated web.
(a) (b)
Figure 7.15 Receiving optics of a laser scanner system: (a) light-guide rod collector; (b) parabolic mirror
collector. From Pietzsch, Miiller and Feige (1988).
Suppliers of laser scanner equipment are, for quent elimination of the defect cause. In some
example, Rheinmetall Sick Optical Inspection instances, this approach requires shutting down
Systems GmbH and Intec Corporation. the coating line to facilitate the search of the
While laser scanners are sophisticated, powerful, defect cause. Such action, in turn, calls for
and very useful instruments for coating process production managers who can withstand the
surveillance, they require not only substantial pressure of short-term profit maximization at the
investment funds (several hundred thousand benefit of sustainable process and product
dollars for a coating width of 1.5 m), but, more development on the basis of prevention and
importantly, a strong commitment from engineers 'quality designed into the product.'
and operating personnel to obtain an optimum The scenario described above is of course an
scanner performance for a particular application. ideal and difficult-to-reach goal. In the meantime,
The task of translating, with any degree of therefore, the use of laser scanners is justified for
certainty, the scanner signals into English 'inspecting quality into the product' by identifying
expressions describing familiar defects like bubbles, and eliminating defects after the coating process
comets, fibers, is not trivial but can be accom- in order to guarantee the delivery of defect-free
plished successfully with sufficient effort and master rolls to customers.
dedication.
The long-term purpose of a laser scanner
7.3.3 FLUID VELOCITY MEASUREMENTS
system should be to put itself out of business.
This operating philosophy is based on the fact Experimental determination of fluid velocities
that a laser scanner is really an element in an has long been a challenge for fluids engineers,
on-line feed back quality control loop, and as and, judging from the small number of publica-
such it detects coating defects after the fact. A tions, the task has been particularly challenging
more desirable operating mode is prevention of in the area of coating flows. Explanations for
defects. Hence, the scanner could be used as an this situation include the small characteristic
instrument toward that goal by using defect dimensions of, and the presence of free liquid
indications to spur the investigation and subse- surfaces in coating flow fields. In addition,
Applications 233
determination of fluid velocities today can be accuracy, linear relationship between measured
accomplished with relatively little effort compared signal and fluid velocity, direction sensitive, free
with experimental approaches, by using com- of calibration, wide velocity range (order to - 3
mercial software packages for computational to 10+ 2 m/s), simultaneous measurement ofthree
fluid mechanics, see also Chapter 9. Finally, the velocity components, relatively expensive, 'not
exact knowledge of the entire velocity field of a so easy to use.'
coating flow may be of limited value for solving In a LOA system whose optical components
practical problems. In spite of the above reser- are arranged according to the dual beam or
vations, velocity information is needed, for fringe mode, two laser beams are focused and
example, to determine velocity gradients, which intersected by a focusing lens to form the so
in turn are indicative of the degree of non- called measuring volume which has the shape of
Newtonian flow behavior of most polymeric an ellipsoid, see schematic diagram in Fig. 7.16.
fluids (see also Chapter 5). Moreover, fluid As a result of the coherent laser light, a fringe
velocities are essential for estimating the magnitude pattern of light and dark areas is produced in
of the transport of surfactant molecules by the measuring volume. A small particle flowing
convection relative to diffusion, in order to through the measuring volume will scatter light
characterize the effects of surfactants added to whose frequency, v (Doppler frequency), is
the coating solution (see also Chapter lId). As proportional to its velocity and indirectly pro-
shown by the examples below, there are other portional to the spacing dX of the light fringes.
instances where the detennination offluid velocities The latter is completely determined by the wave-
is useful. Whether it is advantageous to acquire length ,1.0 of the laser light in the vacuum, the
such velocity information experimentally or by refractive index, n, of the fluid, and the half angle,
way of computer calculations depends on the <1>, of the intersecting laser beams, equation (7.7).
particular application.
One ofthe most elegant experimental measure- dX = ,1.0 (7.7)
ment techniques is called Laser-Doppler Anem- 2n sin (<I»
ometry, LOA. It is an electro-optical system, LOA is an indirect measurement system,
characterized, among others, by the following inferring fluid velocities by measuring particle
features: nonintrusive, instantaneous velocity, velocities under the assumption of no slip.
high spatial and temporal resolution, high According to the above explanations, the particle
Focussing Lens ~ Window

Measuring
-.-.-.-.-.- Volume
I
I
I ~ .•. _...•.-.- .•. _;
I
I
-._---_._._ . . _.j...- _._._._._._._.-
~ - - .-.-.- Cl
I
I
I
j
_....-......................
.......-
j
I _.... -.- .....
Air liqUid
~
f'
Figure 7.16 Laser-Doppler Anemometer: formation of interference fringes in the measuring volume by focusing
and intersecting two laser beams in the test fluid.
velocity V, is given by Small values for Lm are favored by a small

wavelength of the incident light, by thick incident
v= VA O (7.8) beams (use of beam expanders), and by large
2nsin(~) intersection angles. Argon-ion lasers are advan-
tageous compared with He-Ne lasers owing to
The components of a LDA system include a
the emitted light in the blue (476.5 nm) or green
laser source, illuminating optics to produce the
(514.5 nm) range as opposed to the red light at
measuring volume, receiving optics to collect the
632.8 nm. Typically, argon-ion lasers are more
scattered light which contains the velocity
powerful, but also more expensive. Large inter-
information, a light detector and a signal processor
section angles are not easy to realize, because
which is either of a counter or a tracker type. A
they require a lens with a large diameter and a
transparent window in the confinement of the
short focal length. High quality lenses of SLR or
flow system is needed as an entry gate for the
TV cameras with a focal length of 50 mm and
laser beams, ifthe fluid to be measured is a liquid,
smallfnumbers, preferably below 1.0, have proven
see Fig. 7.16. Many optical configurations are
themselves successful for measurements in film
possible with the illuminating and receiving
and slot flows (Oldengarm, van Krieken and van
optics on the same or opposite sides of the test
der Klooster 1975; Schweizer and Widmer 1981).
section. Modern systems use fiber optics com-
The effective size of the measuring volume used
ponents to enhance flexibility and versatility. A
for extracting velocity information can be
comprehensive book describing all aspects of
additionally decreased via electronic means by
LDA was written by Durst, Melling and Whitelaw
s~ppressing low-amplitude portions ofthe Doppler
(1976).
SIgnal. On the other hand, the measuring volume
When applying LDA to coating flows, the
may be enlarged due to lens imperfections (Mues,
dimensions of the measuring volume are of
Hens and Boiy 1989). In summary, the size of
principal concern. Owing to the Gaussian intensity
the measuring volume in a practical application
distribution of laser beams, their diameter, and
is associated with some degree of uncertainty.
hence the boundaries of the measuring volume,
are usually defined by the point where the The small diameters of the incident laser
intensity has a value of e- I or e- 2 of the center beams in the measuring volume that are necessary
to achieve a high spatial resolution of the LDA
intensity. As shown in Fig. 7.16 and derived by
system lead to alignment problems of the optical
Durst, Melling and Whitelaw (1976), the diameter
components. The task of perfectly crossing two
Dm and the length Lm of the effective measuring
narrow beams in space becomes important, and
volume can be calculated by equation (7.9). The
it is crucial to guarantee parallel interference
term 'effective' refers to the measuring volume
fringes. To overcome these difficulties, Oldengarm,
as seen by the photodetector through a small
van Krieken and van der Klooster (1975) propose
aperture (pinhole), placed in front of the detector.
an optical arrangement in which a rotating radial
The effective diameter is smaller than the real
diffraction grating is used as beam splitter and
diameter of the intersection region in order to
frequency shifter, instead of the usual prism and
ensure that the outer region of the fringe pattern,
acusto-optical cell, respectively.
with its poor signal quality, does not contribute
Practical LDA experience of the author includes
to the measurement.
the examination of the fluid mechanics of a
D = 4A of' L = 4A of' mechanically agitated thin film evaporator
m 1tnDe-2COS(~) m 1tnDe-2sin(~) (7.9) (Schweizer and Widmer 1981). In this apparatus,
a liquid film flows down along the inside wall of
Here, f' is the effective focal length of the a vertical cylinder. In addition, the liquid film is
focusing lens in the test fluid, and D -2 is the agitated by several blades that form a rotor
diameter of the incident beam at the foc~sing lens. whose axis is coincident with the axis of th;
Applications 235
cylinder. The tip ofeach blade is slightly submerged Krieken and van der Klooster (1975), and the
in the liquid film, so that a puddle, or so called spatial resolution was claimed to be much less
'bow wave', is formed in front of each blade. The than 40 Jlm. Some of the experimental results
turbulence associated with the flow field in the regarding fluid velocities in the web direction
bow wave is responsible for the enhanced heat were unexpected at first. They could only be
and mass transfer rates typical for this type of explained by postulating a thin entrained air film
evaporator. This flow field in the vicinity of the located just downstream of the apparent dynamic
blade tip is not unlike blade coating, except that wetting line and extending along the web to the
the blade is in motion and the substrate is at rest. position of the real contact line. The dimensions
In another LDA study by the author, velocity of the air film depended on operating conditions
profiles in the metering slot of a coating die were and have been estimated to be of order 100 Jlm
measured as a function ofthe cross-web coordinate, (length) and 10 Jlm (thickness), see also Chapters
in an attempt to test theoretical predictions of 3 and 11 b for cross reference.
outflow profiles for non-Newtonian liquids. The These findings are ofgreat significance, because
die was 43.2 cm wide, and the length of the they shed further light on dynamic wetting, a
metering slot was 4.0 cm; its width was adjustable process for which correct physio-chemical mech-
between 508 and 1016 Jlm. The metering slot was anisms have yet to be established and proven.
formed by the honed metal surface of the die However, the interpretation of some of the results
plate on the one side, and by a 6 mm thick glass by Mues, Hens and Boiy is made difficult by the
plate of optical flatness on the other side. A dual fact that they used an instrument with a spatial
beam backscatter mode was chosen as optical resolution oforder 10 Jlm to study a phenomenon
arrangement, and the resulting length, L m , of the on a molecular scale. Morevoer, one of the
measuring volume according to equation 7.9 was incident laser beams is deflected by the liquid-air
59.7 Jlm, using a He-Ne laser with beam expander interface of the lower meniscus of the coating
and a Canon SLR lens with a focal length of50 mm. bead. As the exact shape and location of this
For a Newtonian fluid (water-glycerin solution), liquid-air interface depend on operating condi-
the parabolic velocity profile in a 660 Jlm metering tions and are not known in general, so is the
slot could be measured perfectly over about 75% location of the measuring volume. Hence, velocity
of the slot width. However, in the vicinity of the information is associated with uncertainty related
metal wall, excessive signal noise due to reflections to spatial uncertainty of the measuring volume.
of the incident laser beams made it impossible It is this very feature which makes the application
to extract velocity information. More interestingly, of LDA questionable for coating flow studies,
rheological and inertial effects on the uniformity particularly if the laser beams enter the liquid
of the lateral velocity profile for Newtonian and through a curved interface of unknown shape
shear thinning fluids (power law index as low as and location. As a consequence, the notion of an
0.51) could be demonstrated. Unfortunately, the entrained air film located downstream of the
validity of all results was jeopardized by a great dynamic wetting line, as is postulated by Mues,
amount of uncertainty in slot width, owing to a Hens and Boiy (1989), remains a conjecture, for
substantial deflection of the glass plate due to the available experimental data do not qualify
fluid pressure in the slot. As can easily be shown as unequivocal proof.
for Newtonian fluids, relative variations in velocity There are alternative techniques to determine
are directly proportional to the negative value fluid velocities experimentally. One is the hydrogen
of relative variations in slot height. bubble flow visualization method described in
Mues, Hens and Boiy (1989) have measured Section 7.3.4. If the voltage applied to the electrode
fluid velocities in the vicinity of the dynamic wire that produces the hydrogen bubbles is
contact line in a slide coating process. The optical pulsed at regular temporal intervals, then com-
arrangement is described by Oldengarm, van bined time streak markers are obtained (Schraub
et al. 1965). Ikin (1992) was able to extract therefore recommend, first, the thorough justifi-
velocity information from hydrogen bubble streak cation of the gathering of velocity information
lines taken in the bead flow field of slide coating and, secondly, the consideration ofcomputational
by using a chopper wheel to modulate the methods as an alternative for generating such
illuminating light beam. information, given the advanced level and the
Another approach is Particle Image Velocirnetry availability ofcommercial codes for computational
(Ply), which allows the simultaneous nonintrusive fluid mechanics; see also Chapter 9 on Computa-
measurement of flow motion at all points in a tional Methods. Still another option is the use
stroboscopically illuminated region (single instant of flow visualization techniques as described
in time). As in laser-Doppler anemometry, fluid below, if qualitative velocity information in the
velocity is inferred by measuring the velocity of form of streamline patterns is sufficient.
light scattering particles which are assumed to
follow the flow path lines. PIV is an optical
7.3.4 IMAGING
technique which involves flow illumination and
photography followed by particle identification Picture taking is a widely used tool in coating
and tracking based on image processing algo- technology, particularly when related to research
rithms. The power of PIV lays in its ability to and development activities. The main attraction
collect large amounts of flow data in a short lays in its communication power by capturing
time. However, the large data quantity leads to qualitative features of coating flow fields and of
the main drawback of the method, namely the coated films. Flow visualization in the visible
long processing times required for image analysis. wavelength range is a popular application of
Overview articles and reports of recent develop- imaging in coating technology, see Section 7.3.5.
ments in this area include those by Adrian (1991) Imaging in the near infrared range (ca. 0.7- 3.0 Ilm)
and Buchhave (1992) who described methods for allows one to 'see in the dark.' This feature is
processing PIV images and discussed practical sometimes used in the photographic industry for
limitations of PIV with respect to measurable process surveillance. Video cameras whose peak
velocities, obtainable accuracy, and spatial sensitivity is in that wavelength range can be
resolution, Grant and Liu (1990) who presented used, for example, to control the curtain coating
a novel method for resolving the directional process by displaying to the operator in the
ambiguity which is common to all particle tracking bright control room if the curtain is still attached
techniques, and Hassan, Blanchat and Seeley to the edge guides. Imaging in the mid and far
(1992) who developed multi-frame tracking and infrared ranges (3-5Ilm and 8-14Ilm) is called
cross-correlation algorithms to determine fluid thermal imaging. It produces temperature maps
velocities. With regard to coating flows, Gaskell of the objects in the picture. Thermal imaging
et al. (1992) have applied a computer particle may be helpful in processes where the fluids are
tracking system to extract velocity information coated at temperatures above ambient. With
from the starved bead of a roll coating process, adequate optics (wide angle lens), a temperature
and Clarke (1995) has used the method to obtain profile across the coated web can be produced
flow functions such as the rate of strain and which, if not flat, may be indicative of lateral
vorticity for the flow field of an impinging liquid film thickness nonuniformities owing to the
curtain. temperature sensitivity of the viscosity of many
In summary, measuring fluid velocities of coating fluids.
coating flows is possible, but not without The choice of hardware in terms of cameras
considerable effort and investment, either in the and lenses is virtually unlimited. In terms oflight
field of electro-optics, or by developing suitable sensitive sensors, traditional silver-halide films
algorithms for electronic image processing. are still the medium with the highest spatial
Depending on a particular problem at hand, we resolution. The lack of immediate display can be
Applications 237
waved by using instant films, but at the expense In contrast, high-speed video systems are
of lower spatial resolution and of poorer picture simpler in terms of hardware, and therefore easier
quality compared with the traditional films. to use, but picture taking rates and resolution
Electronic sensors such as video tubes and one- are lower compared with traditional silver-halide
or two-dimensional CCD-arrays (charge-coupled systems. The EktaPro Intensified Imager Model
device) offer the convenience of instant display, SI (Eastman Kodak Company 1993), for example,
but the spatial resolution does not quite match can record images at rates between 30 and 12000
the one of silver-halide films. For many applica- frames per second. With 192 x 239 pixels the
tions in coating flows, however, the resolution resolution of the sensor is relatively low, but with
ofelectronic sensors and instant films is sufficient. 3200 ISO at 5500 K its light sensitivity is extremely
Owing to the small characteristic dimensions high.
of coating flows, imaging, and in particular flow An application of high-speed photography has
visualization, is often supported by optical aids. been published by Aidun, Veverka and Scriven
Combining a camera with a low-power microscope (1992) who investigated and visualized the details
is necessary when displaying cross-sections of of the dynamic wetting line instability which, so
film flows (see Section 7.3.5). The use of an they argue, is a prerequisite for air entrainment
endoscope has been reported by Ikin (1992), for in coating processes. The authors used a 16 mm
example, who looked at the lower meniscus in silver-halide camera, equipped with a rotary
slide coating, and monitored positions of the prism and continuous film feed, and capable of
static and dynamic wetting lines as a function of recording up to 11000 frames per second (fps).
relevant process parameters. However, with 4 kW of installed illumination the
Still photography as well as conventional maximum picture rate was limited to only 240 fps
movie and video techniques may not be appro- when using a 400 ISO film. Nevertheless, the
priate if unsteady or transient flows are viewed, temporal resolution was judged sufficient relative
for example the start-up of a coating process, the to the maximum coating speed of 0.45 m/s (rotating
passage of a splice, the onset of air entrainment, roller partly immersed in a pool of liquid), and
the impingement of droplets in spray coating, far superior to the resolution of standard video
etc. Instead, high-speed photography is called systems which were also tried unsuccessfully
for in order to resolve the characteristic time because of inadequate frame rates (30 fps).
scale of such flows. While traditional high-speed In another application, Branston et al. (1994)
imaging is a powerful method of observation, visualized the formation of deposits in blade
setting up a proper system requires consideration coating of paper known as 'weeping.' The term
of factors such as velocity of the phenomenon refers to a localized bleeding of coating fluid past
being studied, resolving power of the photographic the blade tip and into wet deposits called 'weeps'
system, magnification, total recording time, type that grow on the otherwise dry tip zone of the
of camera (stationary or moving film, slit or outer surface of the blade. An EktaPro 1000
mirror scanning, shutter design, etc.), synchron- video system (Eastman Kodak Company) was
ization between phenomenon and recording, combined with a long working distance microscope
illumination (continuous, pulsed, or laser light to take pictures at a rate of looofps. This set-up
source), etc. Picture taking rates of 104 to 10 7 allowed the authors to resolve details of the
frames per second are standard, which is more inception, oscillation, growth, partial solidification
than adequate for visualizing phenomena in and detachment of 'weeps' at coating speeds of
coating processes, because their characteristic up to 15.2 m/s.
velocities are relatively low, typically on the In flow visualization, the illumination of tracers
order of meter per second. Relevant aspects of - be they bubbles or flakes - is best achieved
high-speed photography are reviewed in detail with a light sheet that originates from a cylindrical
by, for example, Dubovik (1981). and parallel beam, and that is shaped and
focused by a spherical and a cylindrical lens in and light transmission by optical fibers for easy
sequence. A bulb, e.g. xenon, with an output access to tight spaces. In an application cited by
power of about 150 W is sufficient as light source. Oxford Lasers Inc., the laserstrobe was used to
However, if the light sheet is formed by way of illuminate the flow in the nip between a pair of
slit apertures, most of the available light is rotating rollers.
blocked off, and tracer illuminations may no Taking an image is often only the first step in
longer be adequate. The use of light fibers is a process toward understanding the data that
more convenient in terms of handling than a set has been acquired on the image. Another
of lenses, but due to its strong divergence the important step in this process is data or image
light beam exiting from the fiber bundle is processing. The purpose of image processing is
impossible to focus into a concise sheet, resulting to generate knowledge by extracting quantitative
again in loss of light. Laser beams fanned into information from large data sets or 'fuzzy' pictures.
a sheet by a cylindrical lens have not proven Image processing requires the data to be available
useful, because the Gaussian intensity distribution in digital form, which can be achieved by using
remains in the sheet and leads to a highly uneven digital cameras, e.g. CCD arrays, or by scanning
light intensity in the flow field of interest. An positive or negative images. Classical image
interesting alternative with an even intensity processing techniques make use of a variety of
distribution is offered by Dantec. A laser beam algorithms, many of which have been developed
is reflected off a scanning mirror, thereby creating for applications such as pattern recognition in
a pseudo light sheet. Scanning frequency and aerial images. The algorithms include intensity
divergence angle can be controlled, and the laser distributions, averaging, edge detection, edge
light reaches the mirror through optical fibers. enhancement, low pass and high pass filters,
Continuous lighting may be inappropriate convolution, fast Fourier transformation (FFT),
when viewing unsteady or transient flows. Here, inverse FFT, etc. Nowadays, most algorithms
single or multiple flash illumination (stroboscope) are available on affordable software packages
is an option, possibly in conjunction with that even run on personal computers. Moreover,
high-speed photography in order to either 'freeze' the algorithms are readily adapted to processing
the flow or resolve its characteristic time scale. images of fluid flows (Hesselink 1988; Hesselink,
Typically, continuous high-speed photography Helman and Ning 1992). A general but practical
needs very high levels of illumination (several introduction to image processing may be found
kilowatts), particularly when combined with in the book by Russ (1992).
magnifying optics. The associated heating effects With regard to coating processes, image
must be shielded from the experiments to avoid processing may be used to enhance the visibility
unwanted modification of process parameters, or the form of distinctive patterns of subtle
for example change of viscosity due to increasing defects on the coated web, such as mottle. Digital
fluid temperature. image processing of a different kind is presented
An instrument called a laserstrobe (Oxford in Section 7.4. Flow visualization pictures from
Lasers Inc. 1992) may be able to overcome many the slide coating process have been 'cleaned' from
ofthe problems associated with lighting. Its main undesired reflections that could not be avoided
features include single pulse for still photography during the experiment. The purpose of such a
(pulse duration as short as 30 ns), high-frequency cleaning action is to remove information from
stroboscopic action for high speed photography the picture that strictly lays outside of the fluid
(flash rate of up to 32 kHz), auto synchronization flow cross-section, but that is strong enough to
for compatibility with high speed camera or excessively distract the attention of the viewer.
video systems, laser light available at 511 nm Moreover, equipment contours that are not
(green) and 578 nm (yellow), no emission of visible on the original photo can be redrawn.
infrared light, hence minimum thermal effects, The result is a clear picture with a uniform
Applications 239
background, and with a focus on the unaltered the complicated and curved free surfaces and
cross-section of a fluid flow whose boundaries menisci, but they were not able to resolve details
are distinctly visible. of the flow interior. An overview of these studies
Recent developments in image processing have is given by Schweizer (1988). At about the same
led to real-time imaging. Here, pictures of a time, Scriven and his students at the University
process are captured by way of machine vision of Minnesota began visualizing coating flows by
at rates that are comparable to temporal changes way of computer simulations, see for example
of important process features. The pictures are Saito and Scriven (1981), Kobayashi, Saito and
immediately (real time) evaluated in order to Scriven (1982), Coyle, Macosko and Scriven
provide a feed-back signal for the control loop (1982), and Kistler and Scriven (1983) for an
of the process (Batchelor and Waltz 1993). The overview of the early works. The publication of
above described situation, where cameras are these theoretical results helped to open the way
used to survey the flow of a falling curtain along for industrial companies to demonstrate experi-
its edge guides, is an example of process control mental visualization methods which, as seen in
with real-time imaging, if the picture evaluation retrospect, they had developed years ago. The
results in an alarm as soon as the curtain first quality pictures featuring streamline patterns,
detaches from the edge guide. shapes and locations of free surfaces, static and
dynamic contact angles, and vortices of two
liquid films merging on an inclined plane, and
7.3.5 FLOW VISUALIZATION of the flow field of the 'bead' of the slide coating
process appeared in print in 1988 (Schweizer).
Coating flows are highly two-dimensional flow
Since then, the method has been improved and
fields, characterized by two free surfaces with
applied to many other coating methods by
curved menisci, by one or more interfaces in
numerous authors, for example Sartor (1990)
multi-layer systems, by static and dynamic contact
who studied slot coating, Scriven and Suszynski
lines, and, depending on operating conditions,
(1990) who visualized various premetered and
by vortices or areas of flow separation. The scale
selfmetered coating methods, Chen and Scriven
of the flows is small, typically on the order of
(1992) who compared flow visualization of
100 Jlm. The purpose of experimental flow
multilayer slide coating with results obtained
visualization as discussed below is to
from theoretical modeling, or Ogawa and
• highlight, by way of suitable tracers, charac- Kashiwabara (1992) and Blake, Clarke and
teristic flow features in the plane of interest, Ruschak (1994) who presented pictures ofcurtain
such as free surfaces, interfaces, streamlines, coating, or Clarke (1996) who showed video
stagnation points, separating streamlines, and footage of three different vortices in the flow field
solid surfaces like rollers, substrates, or inclined of an impinging curtain.
planes; The method of choice for visualizing coating
• show how these features change as a function flows is called optical sectioning, because it
of relevant process parameters; introduces tracers in only one plane, namely the
• determine the onset offlow separation or flow plane of interest which contains the desired
instabilities; information. The major advantage of this tech-
• provide realistic boundary and initial condi- nique is its ability to produce concise pictures,
tions for studies in computational fluid revealing flow details down to micron scales, if
mechanics. adequate illumination and magnification are used.
Optical access into the flow is gained through
The first attempts ofexperimentally visualizing a window which is mounted such that it coincides
coating flows were published in the late 1970s with the side confinement of the flowing liquid
and early I980s. They succeeded in displaying film, Fig. 7.17. As the flow fields of interest are
window H2-bubble separator

wall
electrode
Figure 7.17 Flow visualization: optical access into the liquid film and onto the plane of interest through a
side window: (a) three-dimensional edge disturbances due to capillary effects along the side window; (b)
displacement of edge effects by way of a separator wall and feeding extra liquid along the window.
characterized by curved, free surfaces of unknown bubbles generated by electrolysis of aqueous

shape and location, it is practically impossible to solutions, because they are highly visible, if
match the shape of the flow with the shape of properly illuminated, and because they do not
the side window exactly. Consequently, edge contaminate the liquid, which is particularly
effects due to capillary effects will develop in the important in closed-loop systems with recircula-
form of curved free surfaces and possibly curved tion. Thin platinum wires are used as electrodes
interfaces in multilayer films, Fig. 7.17a. Therefore, from which the hydrogen bubbles detach to form
the plane ofobservation must be placed sufficiently the streamlines of the flow. Several quantitative
far away from the window to guarantee truly requirements for a successful application of this
two-dimensional results. A distance of at least technique have been described in the literature
2 cm has proven adequate for a wide range of (Clutter and Smith 1961; Schweizer 1988). They
flow conditions. concern the maximum wire diameter needed to
If the disturbed film surface near the edge guarantee a negligible disturbance effect of the
penetrates below the equilibrium film thickness, electrode, the resulting diameter of the hydrogen
then the upper part of the flow field may not be bubbles necessary to limit the departure of the
visible due to optical distortions (lens effects), bubbles from the true streamlines due to buoyancy
Fig. 7.17a. Ikin (1992) has overcome this problem effects, and the electrical conductivity of the
by mounting a separator wall between the side liquid required to provide adequate bubble
window and the plane of interest and by feeding production rates with inexpensive DC power
extra liquid into this channel, resulting in a supplies.
locally larger film thickness, Fig. 7.17b. The Other interesting tracers are very thin dye
separator wall runs parallel to the side window streams injected through small holes located
and extends to a position just upstream of the along the solid walls of the flow and in the plane
flow field to be visualized. Downstream of the of interest (Schweizer 1988). Dye streams are
separator wall, the two liquid streams merge in suitable for visualizing interfaces in multilayer
such a way that the edge disturbance is still above films, and solid surfaces of coating equipment,
the field of view, and the leveling of the two or, when the liquid detaches from the coating
uneven film thicknesses does not yet affect the equipment, the free surface of the meniscus that
flow field of interest. starts at a static wetting line and ends at the
The most convenient tracers are hydrogen dynamic wetting line. Since the latter streamline
Applications 241
continues to highlight the substrate surface, it H2 -bubble H2 -bubble

also serves as a tool to measure static and electrode electrode
apparent dynamic contact angles.
Depending on the flow field that needs to be
visualized, or a particular detail thereof, there
are numerous ways for mounting wire electrodes
and for placing ports for dye filaments, see Fig.
7.18 for an overview. The main wire extending
through the entire liquid film may be sandwiched
between two plates of a multilayer coating die
or 'hopper', as it is also called (Schweizer 1988),
whereby the sandwich structure provides not
only electrical insulation between the inclined
plane and the wire, but also contains zinc shims
which serve as slave electrodes for cathodic
corrosion protection, Fig. 7.18a. An alternative
for visualizing film flows is to lower the wire
electrode into a small hole in the solid surface
(Ikin 1992). If the surface is metallic, the hole
may be drilled into an insulating material such
as epoxy resin, which fills out a larger cavity in
H 2-bubble
electrode t
the vicinity of the wire electrode, Fig. 7.18b.
Streamlines of internal flows such as slot flow in
extrusion hoppers can be highlighted by pulling
the wire through two holes located in line but (e)
on opposite sides of the slot confining walls
(Sartor 1990), Fig. 7.18d. Visualizing free liquid (f)
surfaces is achieved by using additional wire
electrodes mounted such that their tip just touches
Figure 7.18 Flow visualization: generation ofstream-
the free surface (Schweizer 1988; Blake, Clarke lines by way of (a) and (b) electrolysis along wire; (c)
and Ruschak 1994). Owing to the high current and (e) electrolysis at wire tip; (d) electrolysis along
density at the wire tip, the bubble production wire and injection of dye streams; (I) electrolysis at
rate is also high, resulting in a concise bubble wire tip and injection of dye streams.
stream which, owing to buoyancy forces, tends
to stay in the free surface as it progresses through
the flow field, Fig. 7.18f. lowered into the vortex (Schweizer 1988), Fig.
Vortices, be they attached to solid or free 7.18e. After withdrawal of the wire, the bubbles
surfaces, are divided from the main flow by the remain inside the vortex long enough to capture
so called separating streamline, which starts and the flow patterns on film. For certain operating
ends at stagnation points located on the surface. conditions, however, the flow inside the vortex
Visualizing internal streamlines of vortices is is three-dimensional with a strong helical com-
difficult because, in theory, there is no mass ponent, leading to an immediate drainage of the
penetration, and hence no bubble penetration injected bubbles from the plane of observation,
across the separating streamline, if the flow is and making it impossible to visualize the local
laminar. In practice, it has proven useful to streamline pattern.
discharge hydrogen bubbles from the tip of an Flow features near the exit of slot flows, such
additional wire electrode which is temporarily as the bead of an extrusion coater, can be
visualized by physically combining the hydrogen to 1 zoom ratio), small depth offield, convenient
bubble and dye injection methods, and in focusing mechanism and camera adapters for
particular by injecting dye through the same silver halide or electronic imaging. A viable
holes that hold the electrode wire, Fig. 7.l8d. alternative with similar features but less comfort
While this is convenient from a hardware point is to combine a low power (5 x to 20 x )microscope
of view, our own experience has shown that the objective with a bellows and a camera. Higher
sharpness of the dye filament downstream of the magnification to resolve a particular detail of the
injection port decreases considerably due to the flow field, for example the vicinity of a contact
wire which acts as a disturbance. It is recommended line, can be obtained by mounting an additional
to separate the dye ports and the electrode wire lens in front of the objective lens of the mono
by a large enough distance so as to allow the zoom. At the same time, however, the working
decay of local disturbances (Sartor 1990). distance decreases by the same factor as the
Another possibility to point out contours of magnification increases. In any case, the total
solid surfaces is to cast the electrode wire into magnification achievable with any type of optical
the solid, possibly using an insulating epoxy system depends on the final viewing medium, for
resin, such that its tip is flush with the surface, example a television screen or a photographic
Fig. 7.18c. Depending on the flow orientation print. Typical magnification factors are on the
with respect to gravity, this latter approach may order of 10 to 100.
lead to distorted stream line patterns because of While optical sectioning as described above is
unfavorable ratios between bubble rise velocity a powerful visualization technique and an efficient
and film velocity, owing to the parabolic velocity learning and communication tool, it is not without
profile near the solid wall. limitations. One major difficulty is related to
Adding a small amount of fluorescent dye to visualizing multilayer films whose layers have
the coating solution makes the liquid glow when different physical and optical properties. Dif-
illuminated, which helps to separate it optically ferences in indices of refraction lead to unequal
from the surrounding solid or gas phases. This path lengths of the light, which explains why
in turn is also a means of highlighting contours streamline patterns of multilayer films cannot be
of film flows. focused onto the same image plane in the camera,
A different class of tracers are solid particles owing to the typically very small depth of focus
or flakes dispersed in the bulk of the coating of microscopes. Matching indices of refraction by
solution. While this technique does not allow adding salt to one of the fluids is a possibility.
one to visualize flow cross-sections clearly, it may However, the approach is very tedious and
still give satisfactory results, particularly in renders the viscosity of that fluid a dependent
situations where the above described methods variable. Adjusting the positions of the bubble
are difficult to apply. Examples include Gaskell wires in the different fluid layers to compensate
et al. (1992), who have produced illustrative for differences in optical path lengths is an
pictures of recirculating flows in the nip of a attractive option.
meniscus roll coater, and Schweizer and Scriven Often, the flow visualization method is installed
(1983), who visualized the onset of Gortler-type on an inexpensive table-top apparatus, where the
vortices in curved film flows. liquid is not coated onto a dry substrate but onto
The unaided eye cannot resolve enough detail a scraped roll, see Fig. 7.1. Therefore, any results
in the small-scale flow field of coating flows; influenced by dynamic wetting are burdened with
hence, magnification is essential. A mono zoom uncertainty and must be used with caution.
microscope, e.g. Bausch & Lomb, is the instrument The need for transparent fluids makes the
of choice, because it offers advantages such as method unsuitable for most practical coating
ease of handling, large and variable working solutions. It is therefore important to at least
distance, flexible but sufficient magnification (7 match rheological properties between model
Album of coating flows 243
liquids and real coating solutions by adding substrate (dynamic wetting, circle C in Fig. 7.19).
appropriate polymeric additives. It may also be The emphasis of the picture sequences is on
advantageous to abandon the optical sectioning showing the effects of changing geometrical or
technique and to resort to alternative visualization operating parameters.
methods such as the one described by Ikin (1992) The experimental setup was similar to the one
who used an endoscope and a transparent hollow schematically depicted in Fig. 7.la. The surface
coating roll to observe the positions of the static to be coated was constantly cleaned by a rubber
and dynamic contact lines of the lower meniscus squeegee; it was dry to the touch, but not clean
in slide coating. on a molecular scale. The slide angle, P, was held
To close this chapter, attention is directed to constant at 23°, the radius, R, of the coating roll
well known general references about flow visual- was 53 mm, the slot height, H, was 500 Jlm, and
ization methods, including Merzkirch (1974), and the step height, F, (the off-set between adjacent
several volumes of the Proceedings of the hopper elements) was 1 mm. In all cases, the
International Symposium on Flow Visualization, coating fluid was an aqueous glycerin solution.
for example Veret (1987). In two-layer experiments, both layers contained
the same liquid to avoid differences in refractive
indices. Interfaces were made visible by injecting
7.4 ALBUM OF COATING FLOWS
a thin filament of green (sodium fluorescein,
The capabilities of the optical sectioning method Kodak # 1367499) and/or red (sulforhodamine
are demonstrated in a small 'album of coating B, Kodak # 1367366) dye, or by generating
flows', see Figs 7.20 to 7.24 (between pages 228 hydrogen bubbles off the tip of a platinum wire
and 229). The flow fields are taken from the slide with a diameter ofO.1 mm. Additional geometrical
bead coating process; they show the slot exit flow details are defined in Fig. 7.19. (X is the application
including the position of the static wetting line angle with respect to horizontal, y is the lip angle,
(circle A in Fig. 7.19), the merging of two layers () is the angle of the lip land with respect to the
on the inclined hopper slide (circle B in Fig. 7.19), slide surface, e is the upstream angle of the
and the coating bead, i.e. the transfer of the liquid chamfer at the slot exit, H G is the coating gap,
film from the hopper slide onto the moving L, is the length of the lip land, L 2 is the length
Figure 7.19 Schematic diagram of the slide coating process; forming, merging and coating of liquid films.
ofthe chamfer at the slot exit, R 1 is the upstream uppermost slot of a multilayer slide hopper. The
radius ofthe curved diffusor, R 2 is the downstream flow boundary along the right slot wall appears
radius of the curved diffusor, ~ is the viscosity, as a white line; it has been visualized with a
p is the density, (J is the surface tension, stream of hydrogen bubbles according to Fig.
ti.P = P A - P v is the pressure difference across 7.l8c. The opposite flow boundary in the slot is
the coating bead (vacuum), Ql and Q2 are the marked with a green line (injection of fluorescent
volumetric flow rates per unit width of the lower dye, see Fig. 18d). At the slot exit, the liquid is
and upper layers, respectively and U is the turning to the right around a 90· corner, and it
coating speed. All relevant parameter values for continues to flow as a gravity driven film down
each photograph are listed in Table 7.2. the inclined hopper slide. Here, the white line
In spite of considerable efforts devoted to the from the slot indicates the location of the slide
experimental work in the laboratory, the raw plane, while the green line coincides with the free
photographs did not turn out perfectly. Many film surface. The upstream hopper plate is offset
of them were plagued by excessive reflections by a small distance with respect to the downstream
that distracted the attention of the viewer from plate so as to form a step in the slide surface.
the essential information in the picture. In other Finally, the flow rate, and the corresponding
photographs, details of the equipment contour momentum of the slot flow, is low enough that
(e.g. lip shape, coating roll) were not clearly the liquid is able to turn around the corner
visible. To improve the appearance ofthe various without separating on the downstream side.
flow fields, the photographs have been digitized More importantly, the resulting film thickness on
(Kodak Photo CD System) and manipulated the slide is on the order of the step height, so
with the help of a software package called Adobe that the static wetting line appears to be pinned
Photoshop. Image processing was restricted to to the upstream corner.
removing unwanted reflections outside of the In Fig. 7.20b, the flow rate has been increased
flow cross-section, and to enhancing equipment enough to cause flow separation after the
contours. At no time was the content of the flow downstream corner. Starting at the slot exit, the
field altered. white line follows the path of the separating
In Fig. 7.20a, liquid is flowing upwards in the streamline, thereby outlining the shape and
Table 7.2 Parameter values for flow visualization photographs
Photo <X y b (J R1 R2 L1 L2 HG Jl P (J tJ.P QI Q2 U

# (deg) (deg) (deg) (deg) (mm) (mm) (mm) (mm) (Jlm) (mPa s) (kg/m 3 ) (mN/m) (Pa) (em 2 /s)(em 2 /s) (m/s)
7.20a 0 16.0 1171 68.3 1.24

7.20b 0 16.0 1171 68.3 3.30
7.2Oc 0 28.0 1187 67.4 3.03
7.21a 0 18.0 1175 68.1 3.06 1.19
7.21b 67 5.9 18.0 1175 68.1 3.06 1.19
7.21c 0 2.5 3.5 18.0 1175 68.1 3.06 1.19
7.22a 23 60 0 0.0 508 33.0 1191 67.1 0 1.22 1.56 0.28
7.22b 23 60 0 0.0 508 33.0 1191 67.1 49 1.22 1.56 0.28
7.22c 23 60 0 0.0 508 33.0 1191 67.1 196 1.22 1.56 0.28
7.23a 23 67 23 1.0 254 13.7 1165 68.6 0 0.56 1.96 0.83
7.23b 23 90 0 0.0 254 13.7 1165 68.6 0 0.56 1.96 0.83
7.23c 23 135 -45 1.41 254 13.7 1165 68.6 0 0.56 1.96 0.83
7.24a 46 53 0 0.0 508 24.7 1184 67.6 0 3.49 0.78
7.24b 0 53 0 0.0 508 24.7 1184 67.6 0 3.49 0.78
7.24c -23 53 0 0.0 508 24.7 1184 67.6 0 3.49 0.78
Album of coating flows 245
location of the resulting vortex. Here, the upstream hopper plate, and flow separation is
streamline pattern inside the vortex is not completely suppressed. An eddy, located further
visualized. As a consequence of the increased downstream on the slide, will only appear if the
flow rate, the free film surface above the slot exit flow rate ofthe slot flow is considerably increased.
is bulging out, but the static wetting line is still The flow field of the 'coating bead', i.e. the
pinned to the corner. Experiments have shown liquid bridge between the end of the hopper slide
that excessive bulging leads to a metastable flow and the moving web, is shown in Fig. 7.22a. Two
that will rearrange into a different stable state, layers are being coated, whereby the green, red
if sufficiently disturbed. The result is evident in and white streamlines depict the slide surface,
Fig. 7.2Oc, where, upon tapping on the hopper, the layer interface and the free film surface,
the static wetting line has moved several slot respectively. These streamlines have been visual-
heights upstream along the upstream hopper ized according to Fig. 7.18f. In addition, the
plate, thereby forming a wedge of liquid that is upper layer appears bright, because hydrogen
separated from the exiting slot flow. The flow bubbles have been released according to Fig.
pattern inside the wedge is similar to the classical 7.18b. In contrast, the cross-section of the lower
corner flow described by Moffatt (1964) and layer is dark, as no bubbles have been generated
Taneda (1979), in that the wedge is filled with a intentionally. The purpose ofthe picture sequence
series of vortices, driven by the outer flow, and in Fig. 7.22 is to show the effect of the pressure
rotating in alternate directions. The size and difference (vacuum) across the liquid bridge on
rotational strength ofthe vortices decrease rapidly the location of the lower meniscus and the
in the upstream direction, so that only the dynamic wetting line. In Fig. 7.22a, the pressure
streamlines of the first vortex can be visualized difference across the bead is zero. The lower
(Fig. 7.18e). meniscus (shown as the continuation of the green
In Fig. 7.21a, liquid is exiting from a slot in streamline) is pinned to the tip of the hopper
the same way as described in Fig. 7.20a. In slide (lip), its shape is straight, and the dynamic
addition, a second liquid layer, generated further wetting line is located more than one gap width
upstream on the slide, is merging with the exiting above the corner ofthe lip. The apparent dynamic
liquid to form a two-layer film downstream of contact angle is about 143°.
the slot. The lower white line marks the down- Increasing the vacuum to 49 Pa (Fig. 7.22b)
stream slot boundary and the slide surface. The and 196 Pa (Fig. 7.22c) causes the dynamic
green line follows the upstream slot boundary wetting line to move upstream along the web
and continues as the interface between the two surface. At first, the static wetting line remains
liquid layers. The upper white line is generated pinned to the lip corner. However, as the separating
by hydrogen bubbles according to Fig. 7.18f and angle between the slide surface and the lower
is indicative of the free film surface. As in Fig. meniscus continues to increase, critical conditions
7.20b, the flow rate of the slot flow is too high, will eventually be reached where the so called
resulting in a large vortex downstream of the Gibbs inequality (Gibbs 1906; Kistler 1983) is
exit. Attempts to facilitate the change in flow violated. At that point, the static wetting line
direction and to prevent the onset of separation migrates downstream along the underside of the
by cutting of the sharp 90° corner are shown in lip. As the lip angle in this example is only 60°,
Fig. 7.21 b. Obviously, this measure is not effective, and as the bead flow field is illuminated from
because, with all other parameters held constant, above the hopper slide, the lip itself is casting a
separation cannot be prevented. In fact, the size shadow over the flow field near the static wetting
of the vortex has only marginally decreased. In line, rendering it invisible. Therefore, the green
contrast, Fig. 7.21c shows a slot exit design streamline in Figs 7.22b and 7.22c is only indicative
involving a 'curved diffusor.' Here, the change of the position and shape of the lower meniscus
in flow direction is forced upon the liquid by the near the dynamic wetting line but not near the
static contact line. Additional flow visualization angle is 46° positive, meaning that the application
experiments have revealed that for many operating point is in the upper quadrant of the coating roll
conditions, particularly for small lip angles, the (the application angle is measured against a
liquid near the static wetting line is separated horizontal line through the center of the coating
from the main flow, and a wedge, filled with a roll). In Fig. 7.24b, the application angle is zero,
series of vortices, is formed, much like the one i.e. the tip of the hopper lip is level with the
shown in Fig. 7.2Oc. center of the coating roll. Finally, in Fig. 7.24c,
For the parameter sequence in Fig. 7.22, the the application angle is 23° negative, moving the
apparent dynamic contact angle decreases from application point well into the lower quadrant of
143° to 134° and to 133° as the vacuum is the coating roll.
increased from zero to 49 Pa and to 196 Pa. Similar to Fig. 7.23, the magnitude and location
The effect of the lip shape on the flow field of of the minimum film thickness on the slide and
the bead is shown in Fig. 7.23 for a two-layer the radius of curvature of the upper meniscus
film. The sequence of geometrical changes starts are greatly affected by this geometrical parameter.
with a 'normal' 90° corner (Fig. 7.23a), continues In fact, the impingement angle as determined by
with an upswept or 'ski jump' lip (Fig. 7.23b), the local feature of the lip geometry in Fig. 7.23b
and ends with a downswept or chamfered lip is 113° and equal to the impingement angle in
(Fig. 7.23c). Process conditions such as flow rate, Fig. 7.24a; furthermore, the impingement angles
viscosity and substrate speed have been chosen in Figs 7.23c and 7.24c are also similar (both
such that the characteristic dimension of the bead < 90°). Thus, it appears that desired features of
flow field, e.g., the minimum film thickness on the bead flow field may be obtained in two ways,
the slide, is comparable to the characteristic either by varying the characteristic dimension of
dimension of the lip geometry, e.g., the length of the lip geometry (if it matches the characteristic
the lip land. This is a prerequisite for the lip dimension of the bead flow field), or by varying
geometry to have a significant effect on the flow the application angle. Unfortunately, neither way
field. The effects manifest themselves in different is convenient or cheap, if a particular feature
locations for the minimum film thickness on the should be maintained for several coating appli-
slide and the dynamic wetting line on the substrate, cations with greatly differing process parameters
in different radii of curvature of the upper such as flow rate, viscosity and substrate speed.
meniscus, and in different amounts of 'back flow'
(flow opposite to the web motion near the
7.5 OUTLOOK
dynamic wetting line), as is clearly visualized by
the red streamline marking the interface between Today, the 'tripartite' or 'three-legged stool'
the two layers. According to several patents approach is well established in both industrial
(Jackson 1976; Isayama and Takehara 1979; and academic coating flow research, and experi-
Choinski 1981), upswept lip geometries are mental methods are recognized as valuable tools
favorable to postpone the onset of ribbing lines to advance the fundamental understanding of
to higher coating speeds, see also Chapter 11b. coating processes. Nevertheless, there still is
If, on the other hand, the flow rate is increased room for improving various aspects of these
compared with the conditions in Fig. 7.23, the methods.
liquid volume in the bead, and hence its charac- To this day, the process of dynamic wetting,
teristic dimension, will grow to such an extent for example, is not fully understood. Clearly, the
that the details of the lip geometry will no longer reason is a lack ofclear insight into what happens
have a significant effect on the flow field. at the three-phase contact line. The challenge,
The last picture sequence depicts the effect of therefore, is to develop a visualization technique
the application angle on the bead flow field for that is able to resolve the wetting process on at
a single-layer film. In Fig. 7.24a, the application least a sub-micron scale. Attempts of the author
References 247
in this direction were not successful. The goal experimental work. His commitment and re-
was to investigate dynamic wetting of the slide sourcefulness led to the development of several
coating process by using the maximum magni- new features of the visualization technique. I am
fication of a standard stereo-zoom microscope. also indebted to Rene Sudan, and Rita Hofmann,
Failure was mainly due to insufficient lighting, research physical chemist at ILFORD AG, for
the inability to keep the wetting line in the offering their time to improve the flow visualization
viewing field of the microscope because of slight pictures with digital image processing techniques.
spatial fluctuations of the contact line position,
and the lack of suitable tracers in the liquid/air
interface that approaches the wetting line. REFERENCES
Another aspect offering room for improvement Abramowitz, M. and Stegun, I. A. 1965. Handbook of
does not concern specific experimental methods, Mathematical Functions. New York: Dover Pub-
but their application in general. In a highly lishing.
competitive industrial environment, resources Adobe Systems Incorporated. 1585 Charleston Road,
can be deployed meaningfully, and the return on PO Box 7900, Mountain View, CA 94039-7900, USA.
experimental work can be maximized, if experi- Adrian, R. J. 1991. Particle-imaging techniques for
experimental fluid mechanics. Ann Rev. Fluid
mental techniques are consistently combined
Mechanics. 23.
with statistical experimental design methods as Aidun, C. K., Veverka, P. J. and Scriven, L. E. 1992.
appropriate. Interfacial instability and excessive air entrainment.
Finally, the major role of experimentation is Paper 49a read at 6th International Coating Process
to reveal the fundamental and elemental behavior Science and Technology Symposium, AIChE Spring
of the physical world in general and of coating National Meeting, 29 March-2 April 1992, New
processes in particular, in order to guide the Orleans, LA.
Asoma Instruments, 1992. X-ray Fluorescence-Appli-
development of individual terms in theoretical
cation Summary for Coating Weight Measurement.
models (Churchill 1992). Experimental data, 11675 Jollyville Road, Austin, TX 78759, USA.
however, are recognized as being subject to Ayers, J. 1993. Capacitive Dimensional Measure-
uncertainty and imprecision arising from incom- ment- Theory and Application. Technical Report,
plete definition and control ofthe environment in Lion Precision Corporation, 563 Shoreview Park
which the experiments are conducted (for example, Road, St Paul, MN 55126-7014.
the variation of physical properties with tem- Batchelor, B. G. and Waltz, F. 1993. Interactive Image
Processing for Machine Vision. New York: Springer
perature), and also from measurement errors.
Verlag.
Consequently, care must be taken to increase the Bausch & Lomb, MonoZoom-7 Optical System, Bausch
reliability ofexperimental results, if the important & Lomb Inc., Optical Systems Division, PO Box
role of experimentation is to be carried out 450, Rochester NY 14692-0450.
successfully. Beiser, L. 1974. Laser scanning systems. In Laser
Applications. Volume 2, ed. Monte Ross, pp. 53-160.
New York: Academic Press.
ACKNOWLEDGMENTS Bird, R. B., Armstrong, R. C. and Hassager, O. 1977.
Dynamics of Polymeric Liquids. Volume 1: Fluid
I am grateful to Jacques Rast, manager of Pro- Mechanics. New York: Wiley.
duction Technology, and Pierre-Andre Rossier, Blake, T. D., Clarke, A. and Ruschak, K. 1. 1994.
research engineer, both at ILFORD AG in Hydrodynamic assist of dynamic wetting. AIChE
Fribourg, Switzerland, for their informative input Journal. 40(2): 229-242.
and for providing illustrative examples about the Blum, G., Daams, H. 1., Flegel, H. et al. 1990. Rascher
Zugriff, wirkungsvolle Lenkung; optoelektronische
Quantification of Film Thickness Uniformity. I Messverfahren zur fertigungsintegrierten Qualitats-
am grateful to Rene Sudan, technician at ILFORD sicherung. Technische Rundschau. 50(82): 22-33.
AG, for setting up the experimental flow visual- Branston, R. E., Clark, P. c., Errico, M., Scriven, L.
ization rig and for carrying out much of the E., Sheehan, 1. G., Suszynski, W. 1., Takamura, K.
and Vodniek, J. L. 1994. Weeping in blade coating, 1982. Computer simulation of nip flow in roll
Part I: High-speed video and still macro-photography. coating. In Computer Applications in Applied Polymer
TAPPl Journal. 77(1): 131-138. Science, ed. T. Provder, pp. 251-264. ACS Sympo-
Breitmeier, U. 1993. Schnell abtasten: optische und sium Series 197, Washington: American Chemical
taktile Oberflachenprofilometrie. Kontrolle. (October): Society.
38-42. Daniel, C. and Wood, F. S. 1971. Fitting Equations to
Breuckmann GmbH, Industrielle Bildverarbeitung Data. New York: Wiley.
und Automation, Kunkelgasse 1, D-7758 Meersburg, Dantec Publication No. 9OO2E, Dantec Elektronik,
Germany. Tonsbakken 16-18, DK-2740 Skovlunde, Denmark.
Breuckmann, B. and Thieme, W. 1985. Computer Deming, S. and Morgan, S. 1986. Fundamentals of
analysis of holographic interferograms using phase Experimental Design, Short Course of the American
shift method. Applied Optics. 24: 2145-2149. Chemical Society, Department of Educational
Buchhave, P.1992. Particle image velocimetry - status Materials, 1155 16th Street N.W., Washington, DC
and trends. Experimental Thermal and Fluid Science. 20036.
5: 586-604. Draper, N. R. and Smith, H. 1966. Applied Regression
Buckingham, E. 1914. Physic. Rev. 4: 345-376. Analysis. New York: Wiley.
Campanella, O. H. and Cerro, R. L. 1984. Viscous Dubovik, A. 1981. The Photographic Recording of
flow on the outside of a horizontal rotating cylinder: High-Speed Processes. New York: Wiley.
The roll coating regime with a single fluid. Chem. Durst, F., Melling, A. and Whitelaw, J. H. 1976.
Engng. Sci. 39(10): 1443-1449. Principles and Practice ofLaser-Doppler Anemometry.
Campbell, F. W. and Robson, J. G. 1968. Application London: Academic Press.
of Fourier analysis to the visibility of gratings. J. Eastman Kodak Company. 1993. Motion Analysis
Physiology. 197: 551-566. Systems Division, 11633 Sorrento Valley Road, San
Capacitec, PO Box 819, 87 Fitchburg Road, Ayer, Diego, CA 92121-1097.
MA 01432, USA. ECHIP Inc., 7460 Lancaster Pike, Suite 6, Hockessin,
Chen, K. S. A. and Scriven, L. E. 1992. Multilayer DE 19707.
slide coating: flow visualization and theoretical Edgar, R. F. and Stay, B. J. 1985. Techniques for
modeling. Paper 41i read at 6th International Coating suppressing optical interference errors in infrared
Process Science and Technology Symposium, AIChE film gauging. Infrared Technology and Applications,
Spring National Meeting, 29 March-2 April 1992, SPlE 590: 316-320.
New Orleans, LA. Gaskell, P. H., Malone, B., Savage, M. D. and Thomp-
Choinski, E. J. 1981. Method and apparatus for son, H. M. 1992. Flow visualization study ofmeniscus
coating web. US Patent 4,283,443, August II, 1981, roll coating. Paper 41j read at 6th International
Polaroid Corporation. Coating Process Science and Technology Symposium,
Churchill, S. W. 1992. The role of analysis in the rate AIChE Spring National Meeting, 29 March-2 April
processes. Ind. Eng. Chem. Res. 31(3): 643-658. 1992, New Orleans, LA.
Clarke, A. 1995. The application of particle tracking Gellrich, P., Schliephake, R., Matter, G. and Sturm,
velocimetry and flow visualization to curtain coating. W. 1991. Versuchsbegiessanlagen fur Fotofilme und
Chem. Eng. Sci. 50(15): 2397-2407. Fotopapier. Coating. 3: 76-81.
Clarke, A. 1996. Recirculating flows in curtain coating. Gibbs, 1. W. 1906. The Scientific Papers ofJ. W. Gibbs,
Proceedings ofthe First European Coating Symposium Volume I: Thermodynamics. pp. 314-331. New
on the Mechanics of Thin Film Coatings, eds P. H. York: Longmans, Green and Co.
Gaskell, M. D. Savage and J. L. Summers, Grant, I. and Liu, A. 1990. Directional ambiguity
pp. 32-41. Singapore: World Scientific resolution in particle image velocimetry by pulse
Publishing. tagging. Experiments in Fluids. 10: 71-76.
Clutter, D. W. and Smith, A. M. O. 1961. Flow Grassmann, P. 1971. Physical Principles of Chemical
visualization by electrolysis of water. Aerospace Engineering, Volume 12, International Series of
Engineering 1: 24-76. Monographs in Chemical Engineering. General
COM PUR Electronic GmbH, D-8000 Munich, Ger- editor: P. V. Danckwerts. Oxford: Pergamon Press.
many. Hassan, Y. A., Blanchat, T. K. and Seeley Jr., C. H.
Cornsweet, T. N. 1970. Visual Perception. New York: 1992. PIV flow visualization using particle tracking
Academic Press. techniques. Meas. Sci. Technol. 3: 633-642.
Coyle, D. J., Macosko, C. W. and Scriven, L. E. Hesselink, L. 1988. Digital image processing in flow
References 249
visualization. Annual Review ofFluid Mechanics. 20: Longhurst, R. S. 1973. Geometrical and physical optics.
421-485. 3rd edn. New York: Longman.
Hesselink, L., Helman, J. and Ning, P. 1992. Quantitative Lyne, M. Bruce and Parush, A. 1983. A survey of offset
image processing in fluid mechanics. Experimental and letterpress newsprint print quality. J. Pulp and
Thermal and Fluid Science. 5: 605-616. Paper Science, Trans. of the Technical Section. 9(5):
Hindle, P. H. 1984. Thickness measurement - the 1-7.
infrared method. Paper Technology andIndustry. 25(5). Lynn, P. A. 1983. An Introduction to the Analysis and
Hurley, R. B., Kaufman, I. and Roy, R. P. 1990. Processing of Signals. New York: Macmillan Press.
Noncontacting microstrip monitor for liquid film Makosco, C. W. 1994. Rheology - Principles, Measure-
thickness. Rev. Sci. Instrum. 61(9): 2462-2465. ments and Applications. New York: VCH Publishers.
Ikin, J. B. 1992. Flow visualization studies of a slide Macosko, C. W. and Scriven, L. E. 1984. Coatingflows
bead coating process. Paper 41c read at 6th and coating rheology research: A reviewfor interested
International Coating Process Science and Technology researchers and companies. Department of Chemical
Symposium, AIChE Spring National Meeting, 29 Engineering and Materials Science, University of
March-2 April 1992, New Orleans, LA. Minnesota, Minneapolis, Minnesota, September
Infrared Engineering Limited, Galliford Road, The 20-21, 1984.
Causeway, Maldon, Essex, CM9 7XD, England. MathCAD. Version 2.0. MathSoft Inc., I Kendall
Intec Corporation, 1 Trefoil Drive, Trumbull, CT Square, Cambridge, MA 02139.
06611, USA. McRae, D. J. 1988. Using capacitive sensing for
Isayama, S. and Takehara, N. 1979. Web coating. UK noncontact dimensional gauging. Sensors. October:
Patent Application 2 037 622 A, filing date December 13-20.
11, 1979. Fuji Photo Film Co. Meadows, D. M., Johnson, W. O. and Allen, J. B.
Jackson, B. W. 1976. Apparatus for coating a web. 1970. Generation of surface contours by moire
US Patent 3,993,019, November 23, 1976. Eastman patterns. Applied Optics. 9(4): 942-947.
Kodak Company. Merzkirch, W. 1974. Flow Visualization. New York:
KEYENCE Corporation, 2-13, Aketa-cho, Takatsuki, Academic Press.
Osaka 569, Japan. Meyer, W. and Schilz, W. 1980. A microwave method
Kistler, S. F. 1983. The fluid mechanics of curtain for density independent determination ofthe moisture
coating and related viscous free surface flows with content of solids. J. Physics D: Applied Physics. 13:
contact lines. PhD thesis, University of Minnesota. 1823-1830.
Available from University Microfilms International, Moffatt, H. K. 1964. Viscous and resistive eddies near
Ann Arbor, MI. a sharp corner. J. Fluid Mechanics. 18: 1-18.
Kistler, S. F. and Scriven, L. E. 1983. Coating flows. Moisture Systems Corporation, 117 South Street,
In Computational Analysis ofPolymers, ed. J. Pearson Hopkinton MA 01748, USA.
and S. Richardson, pp. 243-299. New York: Applied Montgomery, Douglas, C. 1984. Design and analysis of
Science Publishers. experiments. New York: Wiley.
Kheshgi, H. S. and Scriven, L. E. 1983. Measurement of Mues, W., Hens, J. and Boiy, L. 1989. Observation of
liquid film profiles by moire topography. Chern. a dynamic wetting process using laser-Doppler
Eng. Sci. 38(4): 525-534. velocimetry. AIChE Journal. 35(9): 1521-1526.
Kheshgi, H. S. and Scriven, L. E. 1991. Dewetting: Nishijima, Y. and Oster, G. 1963. Moire patterns: their
nucleation and growth of dry regions. Chern. Eng. application to refractive index and refractive index
Sci. 46(2): 519-526. gradient measurements. J. Optical Society ofAmerica.
Kobayashi, c., Saito, H. and Scriven, L. E. 1982. Study 54(1): 1-5.
of slide coating by finite element method. Paper 45e Ogawa, S. and Kashiwabara, Y. 1992. Theoretical and
read at AIChE Spring National Meeting, 28 Febru- experimental study of curtain coating flow. Paper
ary-3 March 1982, Orlando, FL. 41 b read at 6th International Coating Process Science
Kueppers, H. 1973. Color: origin, systems, uses. London: and Technology Symposium, AIChE Spring National
Van Nostrand Reinhold. Meeting, 29 March-2 April 1992, New Orleans, LA.
LabWindows 1991. Version 2.1. National Instruments Oldengarm, J., van Krieken, A. H. and van der
Corporation, 6504 Bridge Point Parkway, Austin, Klooster, H. W. 1975. Velocity profile measurements
TX 78730-5039. in a liquid film flow using the laser Doppler
Lion Precision Corporation, 563 Shoreview Park technique. J. Physics E: Scientific Instruments. 8:
Road, St Paul, MN 55126-7014. 203-205.
Oxford Lasers Inc. 1992. Data sheet on laserstrobe. Schweizer, P. M. and Widmer, F. 1981. Die Verweilzeit
33 Nagog Park, PO Box 2620, Acton, MA mittel- und hochviskoser Flussigkeiten im Dunn-
01720-6620. schichtapparat. Verfahrenstechnik. 15(1): 29-33.
Pietzsch, K. and Muller, P. 1988. Automatic surface Scriven, L. E. and Suszynski, W. J. 1990. Visualizing
inspection with laser scanners for quality control in coating flows builds process understanding. Chem.
the production of steel strip. World Steel and Engng. Progress. 9: 24-29.
Metalworking. 9: 240. Shulman, A. R. 1970. Optical data processing. New
Pietzsch, K., Muller, P. and Feige, C. 1988. Luckenlose York: Wiley.
Oberfliichenprufung mit Laserscannern. Metall Sick, E. 1972. Vorrichtung zum photoelektrischen
Oberflqrhe. 42(8): 373-378. Abtasten durchlaufenden Gutes. Patentschrift der
Rast, J. 1989. Spectral analysis of mottle. Unpublished Bundesrepublik Deutschland, DE 1 422 235. Aus-
Technical Report. ILFORD AG, Fribourg, Swit- gabetag October 26, 1972. Erwin Sick GmbH
zerland. Optik-Elektronik.
Rheinmetall Sick Optical Inspection Systems GmbH, Sick, E. 1979. Lichtleitstab. Patentschrift der Bundes-
Machtlfingerstrasse 21, D-8ooo Munich, Germany. republik Deutschland, DE 27 00 027. Ausgabetag
Rossier, P. A. 1992. Quantification of cross-profiles September 20, 1979. Erwin Sick GmbH Optik-
and other coating defects. Unpublished Technical Elektronik.
Report. ILFORD AG, Fribourg, Switzerland. Spiers, R. P., Subbaraman, C. V. and Wilkinson, W. L.
Roy, R. P., Ku, J., Kaufman, I. and Shukla, J. 1986. 1974. Chem. Engng. Sci. 29: 389.
Microwave method for measurement of liquid film Statgraphics STSC Inc., 2115 East Jefferson Street,
thickness in gas-liquid flow. Rev. Sci. Instrum. 57(5): Rockville, MD 20852.
952-956. Takasaki, H. 1970. Moire topography. Applied Optics.
Russ, J. C. 1992. The Image Processing Handbook. 9(6): 1467-1472.
Boca Raton: CRC Press. Takasaki, H. 1979. The development and the present
Saito, H. and Scriven, L. E. 1981. Study of coating status of moire topography. Optica Acta. 26(8):
flows by the finite element method. J. Compo 1009-1019.
Physics. 42: 53. Taneda, S. 1979. Visualization of separating Stokes
Sartor, Luigi. 1990. Slot coating: Fluid mechanics and flows. J. Phys. Soc. Japan. 46(6): 1935-1942.
die design. PhD thesis, University of Minnesota. Tharmalingam, S. and Wilkinson, W. L. 1978. The
SAS. 1987. SAS system for elementary statistical coating of Newtonian liquids onto a rotating roll.
analysis. Version 6. SAS Institute Inc., Box 8000, Chem. Engng. Sci. 33: 1481-87.
SAS Circle, Cary, NC 27512-8000. Theocaris, P. S. 1967. Moire topography of curved
Savage, M. D., Gaskell, P. H., Malone, B. and surfaces. Experimental Mechanics. 7(7): 289-296.
Thompson, H. M. 1992. The relevance oflubrication Torgerson, W. S.1958. Theory and Methods ofScaling.
theory to roll coating. Paper 3ge read at 6th New York: Wiley.
International Coating Process Science and Technology UBM Messtechnik GmbH, Ottostrasse 2, D-7505
Symposium, AIChE Spring National Meeting, 29 Ettlingen, Germany.
March-2 April 1992, New Orleans, LA. Veret, c. 1987. Flow Visualization IV, Proceedings of
Schraub, F. A., Kline, S. J., Henry, J., Runstadler Jr., the Fourth International Symposium on Flow Vis-
P. W. and Littell, A. 1965. Use of hydrogen bubbles ualization, August 26-29, 1986, Paris. Washington:
for quantitative determination of time-dependent Hemisphere Publishing Corporation, and Berlin:
velocity fields in low-speed water flows. J. Basic Springer Verlag.
Engineering, Transactions ofthe ASME. 6: 429-444. Williamson, S. J. and Cummins, H. Z. 1983. Light and
Schweizer, P. M. 1988. Visualization of coating flows. Color in Nature and Art. New York: Wiley.
J. Fluid Mechanics. 193: 285-302. Yeckel, A. and Scriven, L. E. 1992. Muitiparameter
Schweizer, P. M. 1992. Fluid handling and preparation. continuation methods for tracking desired flow
In Modern Coating and Drying Technology, eds states. Paper read at Supercomputing '92 Conference
Edward D. Cohen and Edgar B. Gutoff, pp. 23-61. of the lEEE/ACM, 16-20 November 1992, Min-
New York: VCH Publishers. neapolis, Minnesota.
Schweizer, P. M. and Scriven, L. E. 1983. Evidence of
Gortler-type vortices in curved film flows. Physics
of Fluids. 26(3): 619-623.
ASYMPTOTIC METHODS FOR THE 8
MATHEMATICAL ANALYSIS OF
COATING FLOWS
Marc K. Smith
8.1 INTRODUCTION
the uncertainties in the physics of some portions
As we have seen in Part One of this book, coating of these models bring into question whether
a solid surface with a liquid film is a very complex accurate predictions are indeed possible.
physical process. The behavior of the liquid film In spite of this difficulty, there has been
is influenced in very important and fundamental significant progress in the understanding ofcoating
ways by the physics associated with surface flows because many of the important and
tension and interfacial phenomena, the wetting interesting questions that needed to be answered
ofthe solid by the liquid, fluid rheology, surfactants, allowed for great simplifications in the math-
and the heat and mass transfer and chemical ematical models. Many of these more successful
reactions involved in the drying and curing of coating flows are strongly two-dimensional, and
the film. One of the tasks of a coating engineer so we may confine our attention to a two-
or scientist is to incorporate these physical effects dimensional version of the mathematical model.
into a predictive mathematical model of the The characteristic length scale associated with
coating process. Complete and accurate math- the motion of the liquid in most coating flows
ematical models and accurate numerical compu- is very small, and so the Reynolds number is
tations for simple coating processes are available small. Thus, the flow is laminar and all of the
today. However, this is not the case for the questions and concerns associated with the
mathematical models of the more complicated modeling and calculation of a turbulent flow in
coating processes. These complex models may the liquid do not appear. Finally, all of the
contain many dependent variables and have possible complexities that could appear in a
large sets of nonlinear differential equations and coating flow may not all appear in anyone
boundary conditions. They may be posed on a coating flow, or at least not all in the same area
very complicated geometry. For some effects, of interest. Thus, some of the complex effects
such as wetting and fluid rheology, the fidelity described above can quite often be ignored in a
and accuracy of the mathematical models are calculation designed to answer a question
still debated. These and other difficulties put the concerning one particular region of the flow field.
solution of such complex models out ofthe range For example, the liquid being coated in a particular
ofour mathematical capabilities and may severely process may not display any significant non-
strain our current numerical methods. In addition, Newtonian effects in the region of interest.
252 Asymptotic methods for the mathematical analysis of coating flows
In this chapter, we consider simple two- acceleration terms in the momentum equations.
dimensional coating flows composed of an This low-Reynolds number flow or Stokes' flow
incompressible, Newtonian liquid, and in which is somewhat simpler to handle.
heat and mass transfer effects are negligible. The Another way to produce mathematical results
mathematical model for such a flow is composed is to look in those regions of the model where
ofthe Navier-Stokes equations, the incompress- the flow and the geometry are particularly simple.
ible continuity equation, and the appropriate For instance, the rectilinear flow of a liquid film
stress and kinematic boundary conditions on any down an inclined plane has an exact solution,
solid or free surface in the system. Modern because the parallel flow eliminates the nonlinear
numerical codes of the sort described in Chapter acceleration terms and the difficulties associated
9 have been developed to the point where the with the unknown free surface. Such a simple
accurate solution of these model equations in flow is found in the slide coating, dip coating
many different coating flow geometries is fairly and curtain coating processes for example. It
routine. However, such simulations have their may also be the case that the length scale of the
own limitations. It is often just as difficult for an longitudinal flow variations in a liquid film is
engineer with a numerical code to understand much larger than the thickness of the film. In
the dominant physical interactions simulated by such a case, the flow may be considered as almost
the computations as it is for an engineer looking one-dimensional. This type of simplification is
at the actual physical coating process itself. the basis of lubrication theory, which is used so
Certainly, the computations will provide mega- extensively in coating flows and tribological
bytes of data and detailed pictures of the relevant problems. One of the results obtained from an
flow fields and variables, some of which may be analysis of a thin liquid film using lubrication
difficult or impossible to measure. But the precise theory is a differential equation for the shape of
interplay of the forces that govern the behavior the free surface. In fact, modern numerical
of the flow is sometimes difficult to determine, algorithms have used such results to construct
both numerically and experimentally. It is here the appropriate boundary conditions that match
that a mathematical analysis can contribute most a two-dimensional flow region to a one-dimen-
effectively to the understanding of coating flows. sional flow region (see Chapter 9 for examples
Exact solutions of the full form of the math- of this technique).
ematical model for coating flows are very rare. The simplifications described above are at the
This is due to the complicated geometry of the heart of a collection of mathematical techniques
flows, the nonlinearities of the differential equations known as asymptotic methods. These techniques
and boundary conditions, and the presence of a exploit the smallness or largeness of some of the
free surface or of liquid-liquid interfaces whose parameters in a problem in order to reduce the
locations are unknown. In wetting problems, original equations to a simpler set of equations.
there is the additional complexity of a moving The terms in these simpler equations reflect the
contact line whose location is also unknown (see fundamental balance of forces in the problem as
Chapter 3 for a discussion of this contact-line measured by the small or large parameters. The
problem). To make progress with a mathematical result of the analysis is an approximation to the
analysis of a coating flow in the midst of all this exact solution with an error that is well charac-
complexity, some simplification of the model is terized. The purpose of this chapter is to acquaint
necessary. One way to do this is to exploit the the reader with the basics of asymptotic analysis
presence of a small parameter in the model. For and to show how such techniques can be used
example, in some coating flows the thickness of to understand essential aspects of various coating
the film is less than one millimeter. If this flows.
corresponds to a small Reynolds number for the As a matter of perspective, we note that there
flow, then we can justifiably neglect the nonlinear are many other techniques besides asymptotic
Introduction 253
methods that can be used to develop simplified equations are first integrated over the thickness
theories or to approximate the solutions of of the layer. The resulting integrals are then
differential equations. One of the simplest evaluated using simple approximations or cor-
techniques, as we discussed above, is the removal relations for the dependent variables. The result
of any small terms from the governing equations. is a differential equation that describes the
Two prominent examples of this method, lubri- thickness of the layer as a function of the
cation theory and boundary-layer theory, show remaining independent coordinates. This method
that it can produce very useful mathematical is very useful if the response of the system is fairly
models. The danger of this technique is that the insensitive to the details of the solution across
ad hoc nature of the process could lead to the thickness of the layer. As an example, the mom-
inconsistencies in the removal of the small terms. entum integral equation used in the study of
There is also no way to estimate the errors boundary-layer flows is derived using this
associated with the process. Luckily, many of technique. The derivation is discussed in most
these simplified theories, such as lubrication elementary fluid mechanics textbooks.
theory for example, can be developed in a The methods discussed above are very useful
systematic and consistent manner using the ideas when they work and have lead to significant
of asymptotic analysis, as we will see below. advances in the study of various fluid flows.
Galerkin's method and the various forms of However, in the remainder of this chapter we
the method of weighted residuals are also popular will confine ourselves to the study and use of
ways to find approximate solutions to differential asymptotic methods. This reflects the opinion of
equations. These methods begin by writing the the author that the techniques are some of the
solution of the problem in terms of a finite series more powerful and rigorous of the approximation
of well-known basis functions with unknown methods at our disposal.
coefficients. The coefficients are determined so In the next section, we present a briefdiscussion
that the series satisfies the governing equations of four different asymptotic methods. This will
in some sense, where the 'sense' being used is not be an extensive tutorial on these methods,
what defines the particular method. When a large but rather a look at the kinds of problems that
number of terms are used in the series, the are amenable to such analysis and the kinds of
coefficients are determined numerically. For approximation that result. In the remainder of
example, the finite-element method discussed in this chapter, we examine four different coating
Chapter 9 is based on Galerkin's method. processes and show how asymptotic methods are
Galerkin and weighted-residual methods are used to describe essential flow characteristics in
useful in terms of deriving approximate math- these systems. First, we analyze knife coating
ematical theories when only a small number of using standard lubrication theory, which exploits
coefficients are used. For example, in some the quasi-one-dimensional flow of the liquid
transient problems with one spatial dimension under the knife. Next, we use forward roll coating
the method of weighted residuals produces a to show how the standard lubrication theory fails
small set ofcoupled ordinary differential equations because of the difficulty in determining the
for the time-dependent coefficients that works correct boundary conditions for the flow near
well in terms of describing the response of the the free surface of the film that appears in the
original system. The work of Joo and Davis nip of the rolls. We then describe how this failure
(1992), which is discussed in Section 8.6.1 below, is resolved by using a slowly varying domain
contains an example of this approach. perturbation method together with singular
The integral method is another easy and perturbation techniques to recover lubrication
effective approximation technique that is often theory at leading order and to derive the
used in problems involving boundary layers or appropriate boundary conditions for this flow.
thin liquid films. In this method, the governing The next problem we consider is that of a liquid
film rising from or falling into a static liquid bath. if similar simplifications would be successful in
This model is relevant to the dip coating process other systems. Another problem with such an
and also to certain areas of roll coating. Here, informal method is that there is no way to
domain perturbation techniques lead to nonlinear systematically refine the mathematical results or
equations that can predict the shape of the free to determine the error associated with the analysis.
surface of the film. Finally, we consider the flow We can remove these difficulties by formalizing
of a liquid film on an inclined plane, which occurs the simplification process described above. This
in some regions of slide arid curtain coating. The can be done using several well-defined techniques
vital question of the stability of the free surface known collectively as asymptotic methods or
and of the interfaces in a multilayered film is perturbation methods. The basis ofeach technique
addressed using regular perturbation and domain is to start with the dimensionless governing
perturbation techniques. These four case studies equations, and then to exploit the relative sizes
highlight the current usage of these asymptotic ofthe dimensionless parameters in a certain way.
methods and the advantages they can bring to In the following four subsections, four asymptotic
the understanding of coating flows. methods are described. Each method is applied
to a simple heat transfer problem in order to
show how it is used, to demonstrate its fundamental
8.2 ASYMPTOTIC METHODS characteristics, and to show how it isolates the
When faced with a complex physical system, fundamental physics involved in the model. Heat
scientists and engineers often resort to simplifi- transfer problems were chosen for these examples
cations of the mathematical model in order to because they are relatively simple and they lend
produce an approximation that could be useful. themselves to clear interpretations in terms of
Successful simplifications are based on a good the physics they represent. The details of the
physical understanding of the system and a mathematics used in each perturbation method
knowledge of the relative sizes of the forces, is not described in great depth. For that, please
fluxes, velocities, etc. in the problem. These refer to the many good books on this subject,
simplifications can be done on an informal basis some of which are the standard references of
by noting the dominant terms in the governing Nayfeh (1973) and Van Dyke (1975), and the
equations and neglecting the smaller ones. The more recent book by Hinch (1991).
resulting simplified theory is then supplemented
with appropriate boundary conditions and solved. 8.2.1 THE REGULAR PERTURBAnON
The model is a success if it agrees with the METHOD
relevant physical experiments.
The reduction of a complex mathematical Consider the differential equation
model to one that displays just the dominant
physical effects of interest to the engineer is the
<jJxx + <jJx + 6<jJ = 0, <jJ(0) = 0, <jJ(I) =1 (8.1)
very simplest kind of asymptotic analysis. The where the letter x subscript denotes a derivative
problem with this approach is that some of the with respect to x and 161 « 1 is a small parameter.
simplifications and/or boundary conditions are This equation can be interpreted in terms of a
entirely ad hoc in nature. Even if they were one-dimensional heat transfer problem with a
constructed from sound physical reasoning, they dimensionless temperature given by <jJ(x). The
may lack a clear mathematical justification. One domain of the problem is a layer of liquid of unit
result of this is that any agreement between the thickness and the independent coordinate x
results of such a model and an experiment may measures the dimensionless distance through the
be entirely fortuitous. The model may not lead layer. The first term represents conduction through
to any fundamental insights into the physics of the layer, the second a uniform downward
the process and it may not be useful in determining convection, and the third a small amount of heat
Asymptotic methods 255
generation proportional to the local temperature
and of magnitude e. Ignoring the fact that we
can solve this equation exactly, we use the regular ,
I
;"
perurbation method to find an approximation 0.8 _.- ~,~", i

to the solution. If>
£ =O.5y',
First, we expand the function 41 in a power /,' I
0.6 _. ,
/,'
series in e as follows /,
/,'
41 = 410 + e41l + ... , /,'
,
/,' i
0.4 ! /" .._. ,_.. -- ._.
This is substituted into the differential equation , /0 Exact solution
and then terms of the same power of e are
collected and set equal to zero (see Nayfeh (1973)
pp. 2-4 for more details). The result is a sequence
0.2
-~-'--"--'r'--'-'--r-'--'''- .L.... --.
of differential equations that determines each of i
the 41i-terms in the expansion. If the mathematical 0
0.4 0.6 0_8
solution of the original differential equation is 0 0.2
difficult or impossible, then one would hope that x

each problem in this sequence is easily solved.
For the differential equation (8.1) above, the
approximation to the solution is Figure 8.1 A comparison of the results of a regular
perturbation analysis of equation (8.1) for two values
41 = 1 - e-
1-e-
x
1
+ e{(ll-e- 1
+ e- )(1 - e-
1 1-e
X
1
)
of e. The solid line is the exact solution and the broken
lines are the perturbation approximation (8.2).
-X +_;1-X)} + ...
1
(1 (8.2)
this equation as a heat transfer problem, we
might assume that conduction through the layer
is very small, and that the primary balance is
This approximation is compared with the
between the uniform downward convection and
exact solution for e = 0.5 and e = 0.8 in Fig. 8.1.
the internal heat generation. Following this
Note that even though these values of e are not
assumption, we apply the regular perturbation
small, the approximate solution compares quite
method as described above and obtain the
well to the exact solution. The difference in the
following approximation,
exact solution for these two values of e is not
noticeable on the graph. The error in the
approximation is uniformly small over the entire
41 = e1-X{l + e(l- x)} + ...
domain of interest and it approaches zero This approximation, known as the outer
uniformly as e becomes smaller. This kind of expansion, satisfies the boundary condition at
behavior in the error is a characteristic of a x = 1, but it does not satisfy the boundary
regular perturbation problem. condition at x = O. The reason is that when we
set e = 0 in the original differential equation, it
is reduced from a second-order to a first-order
8.2.2 THE SINGULAR PERTURBATION
differential equation. Thus, the outer expansion
METHOD
can only legitimately satisfy one of the original
Now consider equation (8.1) above, but with the boundary conditions. This reduction of order is
small parameter in front of the first term in the a signal that we are faced with a singular
equation, perturbation problem.
The regular perturbation method failed because
e41xx + 41x + 41 = 0, 41(0) = 0, 41(1) = 1 (8.3)
we implicitly assumed that the term e41xx
where lei « 1. In terms of our interpretation of representing heat diffusion is uniformly small in
the whole interval as we let e ---> 0, and this is not

true. There is a small region near the point x = 0,
E =0.01
known as the inner or boundary-layer region,
where this second-derivative term becomes very
large and the regular perturbation approximation
breaks down. In this boundary-layer region, heat
diffusion is balanced by the downward convection
of heat, and heat generation has only a small
effect. The region outside the boundary-layer
region is called the outer region and the dominant
balance here is between heat generation and ._. - . _._._._-_._. _ ; _ _;--_ _._._._ 1' .
convection. Heat diffusion is small in the outer ,j
region. ·····~~_·_·_·_·_·_·_·_·~·_·_·_··_·_· __·_·t'-_·_.._·_·_·_. _.
The perturbation analysis is modified to correct j
j
j
this problem by making what is known as a o~~~'-'-~~'--'-~~'--'-~-'---''---'-~.........,
;
stretching formulation of the x-variable in order o 0.2 0.4 0.6 0.8

to look locally within the boundary layer near x
x = O. The transformation is simply x = x/em,
Figure 8.2 A comparison of the results of a singular
and the power of e is determined in this case by perturbation analysis ofequation (8.3) for three different
balancing the second-derivative term with the values of e. The solid lines are the exact solution, the
first-derivative term in the differential equation short-dash lines are the perturbation approximation
to obtain m = 1. The solution of the resulting (8.4), and the long-dash lines are the two-term outer
second-order differential equation is then expansion.
approximated using the regular perturbation
method to obtain the inner expansion. The two is just slightly deformed away from some regular
constants that arise are determined from the shape that is easy to deal with. For example,
boundary condition at x = 0 and by matching consider the following two-dimensional heat
to the outer expansion for large values of x. The conduction problem for the dimensionless tem-
inner and outer expansions are then combined perature 1J(x,y) as a function of the Cartesian
to form what is known as a composite expansion, coordinates x and y,
which is uniformly close to the exact solution
over the entire domain. See Nayfeh (1973) V 2 1J = 0, 1J(x,O) = 0, 1J(x, h(x)) = 1 (8.5)
pp. 110-22 for more details of this procedure. and let 1J and h be periodic in the x-direction.
For this problem, the composite expansion is This boundary-value problem describes heat
1J = e 1 - {1 + e(l
X
- x)} conduction in a layer of some material with a
planar surface held at a fixed temperature and
- e1 - x /£{1 + x + e} + .. , (8.4) an upper surface located at y = h(x) held at a
This approximation is compared to the exact higher temperature. The numerical solution of
solution for three different values of e in Fig. 8.2. this problem is quite routine these days, but
Here, we clearly see the development of the mathematical solutions can be found only for
boundary layer as e gets small. some specific forms of the upper surface. The
simplest case is the planar upper surface h = 1,
and the solution is just one-dimensional heat
8.2.3 THE DOMAIN PERTURBATION
conduction with a linear temperature profile 1J = y.
METHOD - SMALL DEFORMATION
Now, consider an upper surface that is a small
One form of a domain perturbation problem perturbation of a plane as described by
arises when the boundary of a region of interest h(x) = 1 + eh 1 (x), where lei « 1 and Ih11 ~ 1. In
Asymptotic methods 257
general, to approximate the solutions for this Note that the deformation of the isotherms is
class of small-deformation, domain perturbation localized near the upper surface because of the
problems we first expand the boundary condition somewhat large value of a.
on the perturbed surface in a Taylor series about The first two terms in the expansion of the
the unperturbed surface. In the present example, boundary condition on the upper surface represent
we expand the upper boundary condition about the linearization ofthe upper boundary condition
the plane y = 1 to obtain about h = 1. Thus, the two terms shown in the
approximation (8.6) represent the solution to the
¢(x, h) = ¢(x, 1) + ¢y(x, 1)eh 1
original problem linearized about h = 1. This
+ ¢yy(x, l)e 2 hi/2 + ... = 1 kind oflinearization is used to study the stability
where letter subscripts denote partial differenti- of solutions and to obtain boundary conditions
ation. Next, we use the regular perturbation for some initial-value and boundary-value prob-
method, expand ¢ in a power series in e, substitute lems. Specific examples of the use of this kind of
it into equation (8.5) and the above boundary linearization are discussed in Sections 8.5 and
condition, group terms of like powers of e, and 8.6 below.
solve the resulting sequence of boundary-value
problems. For the specific case of hi = cos(a,x), 8.2.4 THE DOMAIN PERTURBATION METHOD
where the wave number a is an arbitrary constant, - SLOWLY VARYING DEFORMATION
the resultant asymptotic approximation is
Another kind of domain perturbation problem
¢= y - e cos(ax) sinh(a,y)/sinh(a) + ... (8.6) occurs when the position of the boundary varies
The first or leading-order term of this approxi- very slowly in some direction. Consider the heat
mation represents pure one-dimensional heat conduction problem (8.5) above with the upper
conduction. The second term is the first-order surface given by y = h(x), where x = ex and
correction for the slightly deformed upper surface. lei « 1. We impose no further restrictions on the
We show the isotherms corresponding to this magnitude of h except that it be physically
approximation in Fig. 8.3 for a = 2n and e = 0.1. realizable, i.e., h > 0. The dependence of h on the
variable x together with the smallness of the
parameter e means that the upper-surface position
changes very slowly in the x-direction. Given this
y gradual variation, we might expect that locally,
the solution looks like pure one-dimensional
heat conduction through a layer of constant
thickness h. This result is obtained in a more
formal way when we rescale the x-variable in the
differential equation in order to introduce the
variable x, and then use the regular perturbation
method. We let %x = e%x and obtain the
boundary-value problem
.....
ol.------."tO•.S ---+----O::-l.."..S- - - - + 1 X
e2 ¢xx + ¢yy = 0, ¢(x,O) = 0, ¢(x,h) = 1 (8.7)
with ¢ and h periodic in the x-direction. Here,
Figure 8.3 The isotherms computed from the per- letter subscripts denote partial differentiation.
turbation approximation (8.6) for a small-amplitude The solution of this problem is approximated
domain perturbation analysis of heat conduction using a regular perturbation expansion of ¢ in
through a slab of material given by equation (8.5). The
terms of powers of e2 • At leading order, we obtain
amplitude of the surface displacement is s = 0.1. The
darker line is the upper surface of the slab. ¢Oyy = 0, ¢o(x,O) = 0, ¢o(x, h) =1
The solution is <Po = y/h, which is the pure respect to the linear temperature field under
one-dimensional heat conduction solution we crests of the upper surface. This effect is due to
expected. the lateral conduction of heat from each side of
The problem at the next order is the crest on the upper surface toward the center
of the crest. This increases the temperature under
<Pl yy = -<Pox;:' <Pl(.X,O) = 0, <pl(x,h) =0 the crests and lowers the isotherms.
Physically, this problem is heat conduction and The slowly varying domain perturbation
variable heat generation inside a layer of uniform method is very important in the study of film
thickness h, in which the boundaries are held at flows and coating flows because the length scale
the same temperature. The heat generation effect, for variations in the film thickness is often much
represented by the term <Poxx' is actually an larger than the mean thickness of the film. In
asymptotic model for the lateral conduction of fact, this is the main assumption used in the
heat in the x-direction due to the slow variation theory of lubrication. We study this theory and
in the thickness of the layer. how it is used in two important coating flows in
The two-term approximation to the solution the next two sections.
of the original problem (8.7) is
<P = y/h + 1: 2 {(2h; - hh )/h 3} (h 2 y - y3)/6 + ...
xx
8.3 KNIFE COATING AND LUBRICATION
(8.8) THEORY
The isotherms for the specific case of We begin our examination of asymptotic methods
h = 1 + L\ cos(l:x), with L\ = 1/2 and I: = n/3 are for the analysis of coating flows by introducing
shown in Fig. 8.4. The large value of I: was chosen a method that is not, in the strict sense, asymptotic.
to accentuate the first-order correction of the This method is standard lubrication theory, and
isotherms in this figure. Note, however, that even it is a set of simplifying assumptions that has
for this value of I: the mean layer thickness to been used successfully to describe many thin-film
surface wavelength ratio is 1/6, which implies coating problems. In this section, we apply the
slowly varying upper surface position. The dashed theory to knife coating and briefly discuss its
isotherms correspond to the linear leading-order success and some key results. In the following
temperature field. The solid isotherms contain section, we apply the same theory to roll coating
the first-order effects of the surface deformation. and show how it fails to describe the response
We see that these isotherms are lowered with of the system adequately. Once we see this failure,
+---=.iL;=:===+----=-t---=!======c-~x
-6 -4 -2
Figure 8.4 The isotherms for a slowly varying domain perturbation analysis of heat conduction through a
slab of material given by equation (8.7). The solid lines are the isotherms computed from the perturbation
approximation (8.8). The dashed lines are the leading-order isotherms representing pure one-dimensional heat
conduction. The amplitude of the surface displacement is ~ = 1/2 and the wave number is e = re/3. This
wavenumber gives a mean thickness-to-wave1ength ratio of 1/6. The darker line is the upper surface of the slab.
Knife coating and lubrication theory 259
we use formal asymptotic methods to derive However, away from the ends of the passage the
lubrication theory, to demonstrate why it fails, flow under the knife will be nearly parallel if the
and to show how it can be repaired to produce angle between the knife tip and the lower substrate
accurate results. There are of course many other is small. These nearly parallel solid surfaces form
coating flows in which lubrication theory can be a flow passage that has a slowly varying thickness
applied, such as flexible-blade coating, membrane H in the flow direction. In this situation, we can
coating, squeeze-roll coating and slot-fed knife assume that variations in the flow in the
coating. See Chapters lla and 12 for further X-direction are small compared with variations
discussions of these other systems. in the Y-direction. We also expect that the
For simplicity, we consider the simple knife- velocity in the X-direction is much larger than
coating geometry shown in Fig. 8.5. The end of the velocity in the Y-direction. Finally, we neglect
the knife and the coating substrate form a thin the body force oriented in the Y-direction as we
converging passage through which the coating anticipate that the hydrostatic pressure it produces
liquid flows. The flow is driven by the lower will be much smaller than the viscous lubrication
substrate as it moves to the right relative to the pressure in the thin liquid film under the knife.
knife at the speed U*, and it forms a thin liquid The governing equations for the flow under
film on the moving substrate downstream from the knife but away from the ends reduce to the
the knife as shown in the figure. We assume that following set of equations when the small terms
the liquid is Newtonian with viscosity Il and is described above are neglected:
incompressible and that the flow is two-dimen-
0= -Px + IlU yy (8.9a)
sional. We use a Cartesian coordinate system
fixed to the knife with the origin at the lower 0= -Py (8.9b)
substrate, the X-axis along the plane of the lower
Ux+Vy=O (8.9c)
substrate, and the Y-axis normal to this substrate.
The surface of the knife tip is inclined at a very U = U*, V = 0 on Y = 0 (8.9d)
small angle with respect to the plane of the
U = 0, V = 0 on Y = H(X) (8.ge)
moving substrate.
The governing equations for the flow under where letter subscripts denote partial differenti-
the knife are the Navier-Stokes equations and ation. The set of assumptions used to obtain
the continuity equation together with the appro- these equations is known as lubrication theory.
priate boundary conditions on the solid surfaces, The first two equations (8.9a, b) are the momentum
at the inlet to the flow passage and in the film equations for a laminar, parallel flow in a channel
downstream from the knife. The flows near the with no gravity. These equations are viewed to
inlet and at the outlet to the passage are fully hold locally at any point in the flow passage
two-dimensional and difficult to determine. under the knife. The continuity equation (8.9c)
connects the small variations of the longitudinal
velocity due to the slowly changing thickness of
Knife
the flow passage with the small normal velocity.
The normal momentum equation (8.9b) says
that the pressure is constant across the height of
the passage. This allows one to integrate the
longitudinal momentum equation (8.9a) to find
the parabolic longitudinal velocity
u· U = 2~ P (y 2 - HY) + U*(1 - Y/H) (8.10)

X
Figure 8.5 A schematic of the knife coating process. The continuity equation is now integrated
across the height of the passage. The second gravitational body force in the flow under the
integral in this equation is zero because of the knife for this very reason, we also neglect it at
no-slip boundary conditions on the two solid the two ends. We can now integrate the differential
surfaces. When the first term is integrated and equation (8.13) to obtain the pressure under the
the no-slip boundary condition is applied we knife
obtain
P(H) = 6J.LU*b (H l - H)(H - H 2 ) (814)
Qx = 0, Q = f~ U d Y (8.11)
(Hi - H;) H2 •
where H land H 2 are the thicknesses of the inlet

Here, Q is the net flow rate in the passage. The and the outlet to the flow passage under the knife
integrated continuity equation (8.11) says that respectively and b is the thickness of the knife
the flow rate is constant along the length of the in the flow direction.
passage. This fact can also be deduced from a This pressure can be integrated to give the net
simple control-volume, conservation of mass upward force on the knife due to the flow beneath it
analysis for this flow. 2
Integrating the longitudinal velocity (8.10) F = 6J.LU*J {In(H l ) _ 2(H l - H 2 )}
across the thickness of the passage gives (H l - H 2f H2 (H l + H 2 )
H3 1 (8.15)
Q = - 12J./x + 2. U*H (8.12) Substituting the pressure (8.14) into the flow rate
given by equation (8.12) yields a simple expres-
Substituting this into the global mass conservation sion for the flow rate beneath the knife
equation (8.11) yields
U*H H
(H 3 P xh = 6J.LU*H x (8.13) Q = ----=-....::.
l 2
(8.16)
Hl + H2
This equation is called the Reynolds equation.
And finally, the thickness H co of the liquid film
It was first devised by Reynolds (1886) for the
far downstream from the knife is determined
more general case of longitudinal motion of the
from the mass-balance relation Q = U* H It is
lower boundary, longitudinal and normal motion
X)'
of the upper boundary, and both longitudinal HH l 2

and transverse variations in the thickness of the H = ---=---=- (8.17)
co H l +H 2
liquid film. This equation is also discussed by
Lamb ([1932] 1945, Sec. 330a). The standard lubrication theory just described
The Reynolds equation (8.13) is a second-order is in fact a simple form of a slowly varying
differential equation for the pressure variation domain perturbation analysis, which we demon-
under the knife. Integrating this requires a strate in the next section. The theory is very
specification for the shape of the knife tip. In our successful in predicting the flow rate under the
example of Fig. 8.5, the knife tip is a straight knife when applied to the knife coating of a
line. Furthermore, we need two boundary viscous liquid with no significant viscoelastic
conditions on the pressure. If the inlet and the effects. The success of the method lies in the fact
outlet to the flow passage are completely immersed that the flow and the pressure under the knife
in liquid, then the appropriate boundary condition are well-characterized by the quasi-one-dimen-
is zero pressure (atmospheric) at each end. There sional flow away from the ends of the flow
is a small hydrostatic pressure at these ends to passage. Near the inlet and the outlet to the
consider, but any observed pressure is negligible passage the flow is two-dimensional and the
compared with the large lubrication pressure lubrication approximation fails, but the failure
that is seen under the knife. As we neglected the is local. A full two-dimensional calculation of the
Forward roll coating 261
flow under the knife would produce a pressure a free surface bounding two liquid films of equal
field that is only slightly different from that thickness attached to the two rolls. Far to the
computed with lubrication theory and the right of the film-splitting point, each liquid film
differences would be localized near the inlet and is in rigid-body rotation with its respective
the outlet regions of the flow passage. These roll.
differences would not significantly influence the We will write down the full set of governing
flow rate under the knife (provided the gap equations and boundary conditions for this flow
thickness and the knife-tip angle remain small) in the next subsection, but for now we note that
and the difference in the computed normal force the curved geometry and the free surface make
on the knife would also be small. The simple this a very difficult problem. The first steps
boundary conditions for the pressure at the ends toward a mathematical understanding of this
of the flow passage used in this lubrication theory process were made using the simplifications of
do an adequate job in modeling the effect of the lubrication theory. In many industrial roll-coating
two-dimensional flow in the end regions on the processes the roll radius is very large compared
quasi-one-dimensional flow under the knife but with the gap between the rolls. The gap close to
away from the ends. In the next section, we the nip of the rolls (the point of minimum gap
describe an analysis in which similar simplified width) actually looks like a channel with a slowly
boundary conditions are not so successful, and varying thickness and so in this region lubrication
we discuss the reasons for this failure. theory is valid. The difficulty with this approach
is the proper treatment of the free surface that
appears at the film-splitting point. It is not clear
8.4 FORWARD ROLL COATING
that lubrication theory is valid in describing the
flow near this free surface, but at a minimum the
8.4.1 AN AD HOC LUBRICATION ANALYSIS
location of the film-splitting point must be close
A schematic of the symmetric, forward roIl- to the nip to be inside the region where lubrication
coating process is shown in Fig. 8.6. The two theory is valid. In this subsection, we describe
rolls have equal diameters and rotate in opposite the lubrication model for the flow in the nip of
directions with the same surface speed U*. The the rolls and discuss two sets of ad hoc boundary
rotation of the rolls drags a liquid through the conditions that have been used to model the flow
gap between the rolls from the left to the right. near the film-splitting point. We contrast the
On the right side, the liquid splits apart to form results of each model and then discuss the
reasons for their failure.
The Cartesian coordinate system we use has
its X -axis directed along the symmetry plane of
Upper roll the two rolls, the Y-axis is normal to this plane,
and the origin is at the nip of the two rolls as
shown in Fig. 8.6. The half-gap thickness is H(X).
The governing equations for the flow are simplified
Free su rf"ce using lubrication theory to produce the same
lubrication equations found in the previous
problem of knife coating, i.e., equations (8.9a-c)
together with the no-slip boundary condition on
the rolls. The solution of these equations yields
Lower roll the horizontal velocity
Figure 8.6 A schematic of the forward roll coating

(8.18)
process.
and the following differential equation for the To the right ofthe nip, we can only reasonably
pressure expect lubrication theory to hold up to the
U* Q)
p x = 3p. ( H2 - H 3
film-splitting point, and maybe not even that far,
because of the two-dimensional nature of the
flow in this region. The location of the film-
splitting point 0 = Om and the flow rate through
where Q = f{; U d Y is one half of the flow rate the gap q are both unknown. They are determined
through the nip. by posing two boundary conditions for the
We can integrate this pressure equation once pressure at the film-splitting point. A variety of
we have an expression describing the surface of different pairs of boundary conditions have been
the roll in the nip region. To within the proposed, usually based on ad hoc, physical
approximation used in the lubrication theory, arguments about the nature of the flow field near
the required expression is the parabola this point. These boundary conditions have been
discussed by Savage (1977), Dowson and Taylor
(1979), Ruschak (1985), and Coyle, Macosko and
Scriven (1986). As it turns out, none of these ad
where R is the radius of the roll and H 0 is the hoc pairs of boundary conditions have lead to
minimum gap half-width. The spatial transform- very accurate results from the lubrication analysis
ation tan 2 0 = X 2 /2H oR applied to the gap half- of roll coating. To highlight this difficulty, we
width (8.19) yields H = H 0 sec 2 O. After substituting describe two cases: the Swift-Steiber conditions
this into the differential equation for the pressure and the Hopkins flow-separation conditions.
and writing the result in dimensionless form, we The Swift-Steiber boundary conditions are
obtain
Here, the dimensionless volume flux q scales with where X m is the horizontal position of the
U*H o' the pressure scale is p.U*/Hof., and film-splitting point. These conditions were orig-
f. = ...)(Ho/R) is a small parameter measuring the inally proposed for a roll-coating process com-
variation in the thickness of the gap. posed of a single roll on a stationary flat plate
The pressure boundary condition to the left (see Fig. 8.6 and replace the horizontal symmetry
of the nip (the inlet side) is the same as in knife plane with a rigid flat plate). The liquid film
coating. We assume that the space next to the separates from the flat plate to the right of the
rolls is fully immersed in liquid and ignore the nip forming an air cavity between the liquid film
small hydrostatic pressure that is present compared and the rigid plate. The pressure and the pressure
with the large lubrication pressure between the gradient inside the air cavity are zero. If we
rolls. Thus, the pressure goes to zero as x --+ - 00, assume that these conditions hold inside the
orin terms ofO,p( -n/2) = o. With this boundary liquid film at the film-splitting point, we obtain
condition, we integrate equation (8.20) to obtain the boundary conditions (8.22). For the single-
the pressure in the gap roll/flat-plate problem, a zero pressure gradient
p(8) -{1( n) 1
= 3...)2 2 8 + 2" + 4sin28
in the liquid implies that the horizontal velocity
profile is linear at the film-splitting point, which
seems reasonable. However, in forward roll coating
[8"3( 2n) 1. I.]} a zero pressure gradient at the film-splitting point

-q 0+ +4sm20+32sm48 implies that the horizontal velocity is uniform,
as shown by equation (8.18). This is because the
(8.21) mid-plane symmetry condition is one of no shear
stress rather than a rigid plane with no slip. A -e- Fully Immer ed
uniform horizontal velocity is not expected to -e- Swift-Steiber
be a very good approximation to the flow field - - Flow Separation
--x--Ca=oo
at the film-splitting point and so one should not
.+ Ca = 0.01
expect very good results with the Swift-Steiber
0.6
conditions.
The second of the Swift-Steiber conditions
leads to an equation for the flow rate q = sec 2 8m , 0.4
Pressure
and the first applied to equation (8.21) together
with this value of q produces a nonlinear equation 0.2
for the location of the film-splitting point. This

is easily solved numerically to obtain 0
8m = 0.443554 and q = 1.22575.

The Hopkins flow-separation conditions are ·0.2
based on the idea that the film-splitting point is
a true flow-separation point. Therefore, the ·0.4
horizontal velocity, the tangential shear stress
°
and the pressure are all zero at this point, which
is located on the mid-plane Y = and horizontally
at the position X = X m' These conditions are
-0.6
-0.5 -025 0
9fT[
0.25 0.5
au
U(Xm,O) = 0, aY(X m, 0) = O,P(X m) = ° Figure 8.7 The liquid pressure in the nip of the
forward roll coating process for different models. The
(8.23a-c) solid lines are based on three different ad hoc models
for the downstream pressure boundary conditions: 0,
In the roll-coating problem, symmetry ensures a Swift-Steiber; 0, Hopkins flow separation; 0, fully
zero shear stress on the mid-plane. The zero- immersed. The dashed lines are based on the results
velocity condition together with the horizontal of an asymptotic solution of the flow in this process.
velocity (8.18) imply that the pressure gradient Numerical values for the flow rate were obtained from
at the film-splitting point is the results of Ruschak (1982): +, Ca = 0.01; x, Ca = 00.
(8.23d) model misses the negative pressures to the right

of the nip. As a point of reference, the pressure
This pressure-gradient condition leads to the distribution when the right side of the nip is fully
flow rate q = sec 2 (8 m )/3. The zero-pressure immersed in the liquid is also plotted in Fig. 8.7.
condition applied to equation (8.21) together For this case, q = 4/3.
with this value of q produces a nonlinear equation One deficiency of both of these ad hoc models
for the location of the film-splitting point. The is that the surface tension of the free surface is
numerical solution of this equation yields not included. The experiments of Taylor (1963)
8m = 1.04017 and q = 1.30152. showed that the surface tension is an important
The pressure distributions between the rolls parameter that influences the flow rate through
as a function of 8/n for these two ad hoc models the nip of the rolls. The necessary modification
are shown in Fig. 8.7. Comparing these with the is to replace the zero-pressure boundary condition
distributions obtained with a more accurate with a capillary pressure boundary condition at
asymptotic method described in the next subsec- the film-splitting point P(X m) = -a/Rm(Xm).
tion shows that the separation model is at least Here, a is the surface tension of the free surface
qualitatively correct, while the Swift-Steiber and Rm(X m) is the mean curvature of the free
surface at the film-splitting point (defined as boundary conditions used with the lubrication
positive when the free surface is oriented as in flow fail to compensate for this defect.
Fig. 8.6). The difficulty with this new condition As the location of the film-splitting point used
is that the radius of curvature of the free surface in the lubrication analysis is intimately dependent
is unknown and so an approximation is needed on the two-dimensional flow in the film-splitting
for the free-surface shape. One approximation region, we should have been suspicious of any
that has been used is to assume a local shape for simple attempt to construct an effective boundary
the free surface at the film-splitting point, such condition. This kind of intimate dependence
as a parabola or a circular arc, connect it to the between two-dimensional flow effects and a
geometry of the thin films on the two rolls, and lubrication flow variable is a signal that an ad
then use the resultant free-surface shape to hoc model may not be adequate.
compute the radius ofcurvature at the film-splitting In contrast, lubrication theory does work well
point. This approximation is extremely ad hoc in describing knife coating. The reason for this
and we should expect the accuracy of the modified is that we only needed to specify the zero-pressure
models to be fairly crude. In fact, Coyle, Macosko boundary condition at the inlet and outlet points
and Scriven (1986) showed that the separation to the flow passage under the knife. This condition
model with the capillary pressure boundary does an adequate job in communicating the effect
condition incorrectly predicts the trends for the of the two-dimensional flows in these regions to
flow rate through the gap and the location of the lubrication flow under the knife. (In the next
the film-splitting point as a function of the subsection, we show that this condition also has
capillary number Ca = f.lU*/a. a clear mathematical justification). In addition,
The reason for the failure of these ad hoc the locations of the inlet and outlet points are
models to describe the roll-coating process fixed, and so there is no intimate dependence
accurately is that the two-dimensional flow at between the two-dimensional flows near these
the film-splitting point is a controlling influence points and the locations used in the lubrication
on the behavior of the system as a whole. The analysis. Thus, we do not receive the signal of
ad hoc models attempt to reduce the entire effect potential difficulty that we received in the case
of this two-dimensional flow into a pair of of roll coating.
effective pressure boundary conditions using just Another pitfall in using these ad hoc boundary
simple physical assumptions and qualitative conditions is seen in the work of Savage (1982).
reasoning about the flow. The zero-pressure He analyzed the film-splitting problem in asym-
boundary condition is adequate as shown in the metric roll coating, in which two rolls of different
next subsection, but the pressure-gradient and diameters are used. Savage assumed that the
the capillary-pressure boundary conditions that stagnation streamline separating the flow going
have been used do not contain enough information onto the upper and lower rolls also separated
about the intricate behavior of the two-dimensional two recirculation cells that were both attached
flow to model its effects adequately on either the to the free surface. However, numerical calculations
lubrication flow between the rolls or on the by Coyle, Macosko and Scriven (1986) for
location of the film-splitting point. For instance, asymmetric roll coating showed that this picture
when the capillary number is small enough a ofthe flow field is completely wrong. The dividing
pair of small recirculation cells exists near the streamline for the flow lies below both of the
free surface in the film-splitting region and the recirculation cells, and only one of these cells is
size of these cells depends strongly on the capillary attached to the free surface. Even though the
number. While lubrication theory can model an asymmetry in this numerical model was provided
infinitely long recirculation cell, it cannot ad- by the different surface speeds of two equally
equately represent the effect of these small sized rolls, one would expect the same kind of
two-dimensional recirculation cells. The ad hoc behavior for equal speeds and differently sized
rolls as in the problem of Savage. The lesson equations and continuity. Gravity is ignored.
here is that ad hoc models of this kind may Standard boundary conditions on the solid surface
inadvertently miss many of the important features of the roll and on the free surface of the liquid
ofthe flow field because the original assumptions are used. As the flow is symmetric, we consider
used to construct the models don't allow such only the flow in the upper half of the system.
features to develop. If these features turn out to The parameter e = J(HoIR), where H o is the
be major determining factors in the flow, then half-gap thickness at the nip and R is the radius
the ad hoc model will never faithfully reproduce of the rolls, is a measure of the slope of the
the observed behavior of the system. upper-roll surface near the nip. It is typically
So, if the ad hoc models used in the lubrication small in roll coating, say e < 0.1. Such small
analysis of roll coating are not accurate enough, values of e imply that the gap thickness between
how does one proceed with the analysis? One the rolls near the nip changes very slowly as one
way is to do a complete numerical simulation of moves through the nip. Thus, we may analyze
the process. Following this path, we would incur this problem using the slowly varying domain
all of the computational expenses involved with perturbation method.
the large number of elements or nodes needed The appropriate scales for this analysis are the
to produce accurate results. In addition, we standard lubrication scales. Vertical distances
would lose the simple representation of the flow are scaled with the minimum half-gap thickness
given by lubrication theory that is accurate away at the nip H 0' and horizontal distances are scaled
from the film-splitting point. An alternative with the much larger length scale Hole. This
approach is to use a slowly varying domain horizontal length scale comes from forcing the
perturbation analysis coupled with a singular horizontal variation of the roll surface to appear
perturbation analysis. The slowly varying domain at leading order in the equation for the surface
perturbation analysis provides a good approxi- of the roll. Horizontal velocities are scaled with
mation to the quasi-one-dimensional flow near the surface speed of the roll U* and vertical
the nip (we actually recover lubrication theory velocities are scaled with eU*. This velocity scale
to leading order). The singular perturbation is derived from balancing the two terms in the
analysis isolates the flow in the two-dimensional continuity equation and it is a reflection of the
region near the film-splitting point from the small vertical velocities that are expected in the
lubrication flow near the nip. Then as we described nip. Finally, the pressure is scaled on a large
in Section 8.2.2, these two flow fields are connected viscous pressure scale J1U*leH o in order to
through a well-defined matching process. An balance the pressure gradient and the viscous
asymptotic analysis of this kind for symmetric, terms in the horizontal momentum equation
forward roll coating was first done by Ruschak near the nip.
(1982), and the extension to asymmetric roll The resulting dimensionless equations ofmotion
coating was presented by Coyle, Macosko and are
Scriven (1986). The former analysis is described
next. For more information on roll coating, see eRe {uu x + vU y } = - Px + e2 uxx + Uyy (8.24a)
Chapter 12a. 3
e Re{uvx + vVy } = -P y +e 4
vxx +e
2
vyy (8.24b)
Ux + vy = 0 (8.24c)
8.4.2 A SINGULAR PERTURBAnON ANALYSIS
Here, Re = p U* HoiJ1 and letter subscripts denote
The geometry of symmetric forward roll coating partial differentiation.
is shown in Fig. 8.6. The liquid is isothermal, The equation for the upper-roll surface y = h(x)
incompressible and Newtonian with constant is
density p and viscosity J1. The flow is steady and
the governing equations are the Navier-Stokes
The no-slip boundary condition on the upper- the condition that the pressure in the liquid falls
roll surface is to atmospheric
p--+O as x--+ _e- 1 (8.24p)
where t r is the unit tangent vector to the surface This is appropriate because we have ignored
of the roll. gravity and have assumed that the region to the
Symmetry conditions are imposed on the left of the nip is fully immersed in liquid.
mid-plane y = 0, As the parameter e is very small, consider the
limit of the horizontal momentum equation
Uy = 0, v = o. (8.24f,g) (8.24a) as e --+ O. The disappearance of the viscous
term Uxx in this limit signals that we have a
The free surface of the liquid film is described singular perturbation problem as discussed in
by the function y = 17(X). The tip of this free Section 8.2.2. We should expect the existence of
surface, called the film-splitting point, is located an inner region, or internal layer, of small
at the point x = X m • Thus, we require horizontal extent in which this viscous term is
(8.24h,i) important. Given our physical understanding of
roll coating, it is reasonable to suppose that the
The boundary conditions imposed on the free inner region contains the film-splitting point.
surface are the normal- and tangential-stress Thus, the asymptotic analysis of the flow between
balances and the kinematic condition, the rolls starts with an application of the regular
perturbation method to equations (8.24) to find
-p + 2e 2 {v y - 17x(U y + e2vJ + e217;Ux}N-2 the equations for the leading-order approximation
= -eCa- 1 K to the flow in the outer nip region. Then a
stretching transformation at the film-splitting
(u y + e2vx)(1 - e217;) + 2e217x(Vy - uJ = 0 point is applied to equations (8.24) to determine
v = u17x (8.24j-l) the governing equations for the inner film-
splitting region. A second application of the
where N = (1 + e217;)1/2, the curvature ofthe free regular perturbation method on these stretched
surface is K = -e217xxN-3, and the capillary equations is used to determine the equations for
number is Ca = jlU*/a. the leading-order approximation to this inner
Far to the right of the film-splitting point, the flow. A second outer region is the rigid-rotation
liquid film is in rigid-body rotation with the roll of the liquid far to the right of the film-splitting
point. The flow field here is known exactly.
(u,ev) = (1 + e2 r)t r for 0 ~ r ~ h oo (8.24m) Matching these three flows leads to the appropriate
hoo = e- 2 {[(1 + e2 - e217)2 + e2x 2 r/ 2 - 1} boundary conditions that allow us to compute
fully the flows to leading order in both the inner
(8.24n) film-splitting region and the outer nip region. In
particular, matching the pressure in the outer
Here, r is a radial coordinate measured from the nip region with that in the film-splitting region
surface of the roll and h oo is the final thickness yields the asymptotic boundary conditions for
of the film on the roll. In this rigid-rotation the pressure that replaces those in the ad hoc
region, the free surface of the liquid is a circular models used in previous lubrication analyses of
arc concentric with the roll. Its slope is roll coating. From these flow solutions, a final
(8.240) composite approximation that is uniformly valid
everywhere can be constructed if desired. To
Finally, far to the left of the nip as x approaches begin the regular perturbation analysis of the
the radius of the rolls, we only need to specify outer nip region, we expand the dependent
variables in terms of the small parameter e as Indeed, it is precisely the next higher correction
follows that allows us to estimate the error associated
with the approximation at any given order and
u = Uo + WI + ... v = Vo + WI + ... and
to develop the parameter bounds that establish
P = Po + ePI + ... the domain of validity of the approximation. The
boundary value problems produced in this way
We substitute these expressions into the governing would resemble the leading-order problem, but
equations (8.24) and obtain the usual sequence they would have additional forcing terms formed
of boundary value problems. from the solutions to the lower-order problems.
The leading-order equation set is For example, the horizontal momentum equation
o = - Pox + U Oyy (8.25a) at the next order is
o = - PO y (8.25b)
uox + VOy = 0 (8.25c) The terms on the right-hand-side of this equation
are forcing functions that represent the convective
h(x) = 1 + x 2 !2 (8.25d) acceleration of the liquid at the leading-order
Uo = 1, Vo = hx ' on y = h (8.25e,f) velocity. Inertia was neglected at leading order
and so these terms force another purely viscous
U Oy = 0, Vo = 0, on y = 0 (8.25g,h) flow that corrects the leading-order flow for this
Po ---+ 0 as x ---+ - CI) (8.25i) deficiency. As one might expect, carrying this
procedure to higher orders can quickly become
Note that equations (8.25a-c) are identical to very cumbersome. Fortunately, symbolic man-
the scaled form of equations (8.9a-c), which were ipulation software can be used to remove some
the result of standard lubrication theory applied of the drudgery and the potential for error in
to knife coating. The boundary conditions posed this procedure.
on the upper and lower surfaces are different, Rather than continue with the analysis of the
but these result from the different coating models outer nip region, we now consider the inner
being examined, and not from the asymptotic region that exists near the film-splitting point
method used to obtain them. This equivalence x = x m • The appropriate stretching transformation
demonstrates that standard lubrication theory is for the x-coordinate in this region is x = (x - xm)!e.
obtained as the leading-order problem from a Physically, the new variable x is a horizontal
slowly varying domain perturbation analysis. coordinate scaled with the length scale H 0' which
One estimate ofthe range of validity of the theory is the same scaling used for the vertical coordinate.
is given by the asymptotic condition that the In addition, we define new inner dependent
next term in the series is small, i.e., lei « 1. variables as follows
The solution of equations (8.25) was presented
above during our discussion of standard lubri- u(x,y) = u(x,y), v(x,y) = w(x,y), and
cation theory applied to the flow through the p(x,y) = e -1 p(x,y) (8.26)
nip of the rolls. The pressure field is given by
equation (8.21). The flow rate q in this equation The new vertical velocity has effectively been
is as yet unknown. scaled on the surface velocity of the roll U*, the
We could continue with this method and same scale used for the horizontal velocity. The
proceed to the next-order problem and to higher new pressure variable is equivalent to using the
orders if desired. This would bring in small viscous pressure scale J.l U*! H 0 instead of the
corrections to the flow and pressure fields due large lubrication pressure scale used in the outer
to the effects of inertia, other viscous effects, and nip region. Both of these scales are the more
a more accurate representation ofthe roll surface. reasonable ones to use when we consider that
the flow near the film-splitting point is two- The slope of the free surface of the liquid in this
dimensional and bounded by a free surface. region is
After performing this transformation and re-
scaling, the governing equations become the
ijx = e(x m + d)(1 + e2 - e2ij)-1 (8.270)
following, We use the regular perturbation method to
approximate the solution of this inner problem.
Re {uu x + VUy} = - Px + uxx + Uyy (8.27a)
The dependent variables are expanded as follows
Re{uv x + vV y} = -P y + vxx + Vyy (8.27b)
(8.27c) P = Po + ePI + and
The equation for the upper-roll surface, now ij = ijo + eij 1 + ..
defined as y = h(X) is
The resulting leader-order problem is
h = 1 + e- 2 - e- 2Jl - e2(x m + d)2 (8.27d)
The no-slip boundary condition on the upper- (8.28a)
roll surface is
- -
(u, -
v) = (1, h x)/(1 + hi"'2x )1/2 = -t
r (8.27e)
(8.28b)
where ~ is the tangent vector to the upper-roll
surface. (8.28c)
The symmetry conditions on the plane y = 0 are The upper-roll surface is the horizontal plane
(8.27f,g) (8.28d)
The free surface of the liquid film is described The no-slip boundary condition on the upper-
by the function y = fi(x). The film-splitting point roll surface is
is located at x = 0 and we require that
(8.28e)
(8.27h,i)
Symmetry conditions on the mid-plane are
The boundary conditions imposed on the free
surface are UOy = 0, Vo = 0 on y = 0 (8.28f,g)
- P + 2{v y - ijx(u y + vx) + ij; ux} N- 2 = - Ca- 1 K The conditions at the film-splitting point are
(u y + vx)(1 - ij;) + 2ijx(v y - ux) = 0 (8.28h,i)
v = uijx (8.27j-l) The free-surface boundary conditions are
where N = (1 + ij;)1/2 and the curvature of the - Po + 2 {V Oy - fiox(u oy + vox) + fi~x uox }
free surface is K = -ijuN-3. -- 2 = - Ca - I -K
X N0 o
Far to the right of the film-splitting point, the
rigid-body rotation of the liquid film with the (U Oy + vOx )(1 - fi~x) + 2fiox(voy - UOx ) = 0
roll is given by (8.28j-l)
(u,v) = (1 + e2r)~ for 0 ~ r ~ h oo (8.27m) where No = (1 + fi~x)I/2 and the curvature of the
free surface is Ko = -fioxxN~3.
hoo = e- 2 {[(1+ e2 _ e2ij)2
In a standard singular perturbation problem,
+ e2 (x m + d)2r/ 2 - I} (8.27n) the leading-order equations for the inner problem
are solved exactly in terms of several unknown flow of the outer nip region. Simplifying the
constants. These constants are then determined inner-flow equations for a parallel flow as x-+ - <X)
by matching the inner expansion to the outer yields
expansion. However, the leading-order equations
(8.28) are the full Navier-Stokes equations posed 0= -POx + UOyy , 0 = -Poy
on a domain with an unknown free surface and uo(hm ) = 1, UOy(O) = 0
cannot be solved exactly. To solve these leading-
order equations numerically, we need the appro-
priate boundary conditions for the inlet and
The solution is Uo = ~ Pox(y2 - h;,) + 1, where
outlet to the flow domain. These conditions are POx is a constant. This constant can be determined
found by doing the asymptotic matching now. in terms of the half-flow rate through the nip,
The matching conditions to leading-order are
very simple. For a given inner flow variable u(x)
f:
qo = m uody. We find that POx = 3{hm - qo}/h;,·
Thus, the matching condition for the flow to the
and an outer-flow variable u(x), we must have left of the film-splitting point is
u(x-+ (0) = u(x-+ x m +) for matching to the right
and u(x-+ -<X» = u(x-+x m _) for matching to
the left. The reader should be warned that these
conditions are appropriate only for a leading- (8.28p)
order match. If higher-order corrections are
desired, then a generalized matching procedure Also note that the flow rate leaving the film-
is needed as described by Van Dyke (1975), splitting region on the right must be the same
Nayfeh (1973) or Hinch (1991). as the flow rate entering the region on the left.
The inner flow at the film-splitting point has This implies the dimensionless relation h ro = qo'
simplified somewhat because the upper-roll surface We now have a fully-posed, leading-order flow
is now a flat, horizontal plane as shown by problem in the inner region. Because of how the
equation (8.28d). This occurs because the flow boundary of the upper roll has simplified, the
passage in physical variables has a very small flow looks like that of a liquid past a large
slope (as we said above, this small slope is the two-dimensional bubble in a parallel channel of
basis for analyzing this flow using the slowly- width 2fi m , as shown in Fig. 8.8. Here we have
varying domain perturbation method). If we just used a moving reference frame fixed to the bubble
look at a small region near the film-splitting (the free surface of the liquid film) and the upper
point, the small slope is imperceptible and the and lower solid surfaces (the rolls) are moving
upper-roll surface appears to be flat and horizontal. past the bubble at the dimensional surface speed
Far to the right of the film-splitting point the
liquid film is in rigid-body translation with the
upper roll and the free surface of the liquid is
"-
parallel to the upper-roll surface. Thus, the
appropriate matching condition for the inner ______
u"
- __ -.1.
- ~C
problem is the rigid-body translation of a uniform
film on the moving horizontal flat surface, Llquid
0, ~ for
(u o , v~) :
'10 - hm
~o ~ ~ hm }
h ro , '1ox - 0
as x-+ <X) --...
r- /I
(8.28m-o)
To the left of the film-splitting point, the inner Figure 8.8 The simplified inner problem near the
flow must match to the locally parallel, lubrication film-splitting point in forward roll coating.
U*. For simplicity, we rescale the y-variable used pressure distribution in the outer nip region
in equations (8.28) so that the channel has a given by equation (8.21) produce the following
dimensionless width of two. In doing so, the equation for the location of the film-splitting
matching condition for the flow to the left of the point 8m.
film-splitting point becomes
~(8m +~) + ~Sin28m - iis
8m[~(8m +~) +~sin 28m+32 sin 48mJ

2
1
sec
o ~ y ~ 1, as x-+ - 00, (8.29)
X
= 0 (8.30)
where qs = qolhm is one-half of the flow rate
through this rescaled channel. Given qs = qs( Ca, Re) from the numerical solution
The leading-order inner-flow problem (8.28) is ofthe inner region, equation (8.30) is solved using
now solved numerically using one of a variety a simple root-finding routine. Once 8mis known,
ofdifferent techniques, some of which are described the half-flow rate through the nip is easily
in Chapter 9. The results of such a computation calculated from the equation q = qs sec 2 8m.
include the velocity field, the pressure field, and Some specific results from solving equation
the shape of the free surface (see Coyle, Macosko (8.30) were obtained using the data for qs presented
and Scriven 1986). However, the only result by Ruschak (1982) for Re = O. For Ca = 00,
needed to complete the asymptotic analysis of iis = 0.41 and we find 8m= 0.972 and q = 1.29.
the flow in the outer nip region is the flow rate. For Ca = 0.01, iis = 0.052 and we find 8m= 1.372
For a channel scaled to a unit half width the and q = 1.33. These results show that the flow
half-flow rate is qs = qs( Ca, Re) and so the half-flow rate is fairly insensitive to the capillary number
rate through the inner region is qo = qs hm. over the range of 0.01 to infinity and that the
To complete the asymptotic analysis of this film-splitting point moves away from the nip as
flow the pressure and the flow rate in the outer the capillary number decreases (or as surface
nip region and the inner film-splitting region are tension increases). The pressure distributions in
matched. The pressure in the inner region is the nip for these two capillary numbers are
Pinner(X) = epo + O(e 2 ) and the pressure in the shown in Fig. 8.7. Further results and discus-
outer region is Pouter(x) = Po + O(e), where the sions of roll-coating flows are found in
inner and outer pressures have been scaled the Chapter 12a.
same. The different orders of the leading-order A final note on the ad hoc boundary conditions
terms in these expressions are a result of the used with the standard lubrication analysis
different scalings originally used in each region. discussed in Section 8.4.1 is in order. Both the
The appropriate matching condition to the left Swift-Steiber model and the Hopkins flow-
for the pressure is Pinner(X -+ - (0) = Pouter separation model use the same zero-pressure
(x -+ X m_). The result to leading order is boundary condition. We now see that this
Po(x = x m) = 0 or Po(l~m) = O. The relation be- condition has a firm mathematical justification
tween the variable 8 and the half-gap thickness as it was derived using the asymptotic matching
near the nip as defined below equation (8.19) process described just above. The pressure-gradient
shows that hm = sec 2 8m. So, matching the flow boundary conditions used in these two models
rates through the outer nip and the inner film- lead to an expression for the flow rate through
splitting region is simply q = qo = qs hm = the nip in the form q = qs sec 2 8m, where qs = 1
qs sec 2 8m · in the Swift-Steiber model and qs = 1/3 in the
The two matching conditions Po(8 m) = 0 and Hopkins flow-separation model. Equation (8.30)
q = iis sec 2 8m together with the leading-order shows that the parameter iis' which physically is
Rising or falling film flow - a domain perturbation analysis 271
the flow rate in the scaled inner film-splitting 8.5 RISING OR FALLING FILM FLOW - A
region, is the only parameter controlling the DOMAIN PERTURBAnON ANALYSIS
location of the film-splitting point. As qs is
A liquid film rising from or falling into a liquid
determined completely by the solution to the
bath is a common occurrence in many film
leading-order, inner problem, we see that the
coating processes. In dip coating, the rising film
dynamics of the flow in the film-splitting region
formed on a solid substrate that is dipped into
is a major controlling factor for the flow in the
a liquid bath and then withdrawn is the desired
roll-coating process as a whole. Comparing the
liquid coating. Falling films occur in roll coating
values of qs from the ad hoc models with the
where the roll enters the liquid bath. In this
results of Ruschak (1982) shows that the ad hoc
section, we analyze rising and falling films
boundary conditions do not approximate the
according to the geometries shown in Fig. 8.9.
behavior of the flow in this inner region very
For the sake of simplicity, we confine ourselves
well at all. Even including the effect of surface
to a vertical surface or wall moving upward or
tension into the ad hoc models does not improve
downward with respect to a large container of
the situation as shown by Coyle, Macosko and
liquid with the speed U*. Two length scales are
Scriven (1986). It is not surprising then that the
immediately apparent: the thickness of the liquid
ad hoc lubrication models are ineffective.
We have seen in this section a very successful
application of the singular perturbation method.
The strength of this method is that it can isolate
HO
and describe a flow that occurs over a very small
length scale compared with the length scale of
the main flow. This characteristic is usually, but
not always, signaled by a small parameter in
front ofa highest-order derivative in the governing
differential equations for the flow. Boundary-layer
or interior-layer flow problems of this kind are y
difficult for numerical analysis precisely because
X Liquid
of this disparity in length scales. The singular
perturbation method we have described has also (8)
been used successfully in many other problems.
It has been used to describe the viscous bending
of the free surface that occurs very near to a Ho
moving contact line (see Chapter 3 for a discussion
of this dynamic wetting problem). It has resolved
the controversy regarding the inclusion of gravity v-l l g
in the analysis of the dip coating process (see the

discussion at the end of Section 8.5). And of
course, the singular perturbation method has
been extremely useful in fluid mechanics problems y
such as the high-Reynolds-number boundary-layer X Liquid
flow over a solid surface (this is important in the
(b)
short-dwell coater described in Chapter 12d), the
low-Reynolds number flow past a sphere, Ekman-
layer flow on a rotating plane, etc. Van Dyke Figure 8.9 A schematic of the rising and fallin~ fil~
problems: (a) a monotonic film rising from a lIqUid
(1975) discusses a number of these problems in bath; (b) a film falling into a liquid bath with surface
greater detail. waves.
film far above the bath H 0 and the width of the Ux + vy = 0 (8.31c)
container. We assume that the container is large
u(x,O) = -N, v(x,O) = 0 (8.31d,e)
enough so that the free surface becomes flat far
from the moving surface, but before the container -p
2
+ 26 {V y - hAu y + vx~ + h;uJN- =
2 -K)
wall. Therefore, the width of the container is not
(u y + vx)(1 - hx ) + 2hAv y - uJ = 0
a relevant length scale for this analysis. If there
was no film on the vertical surface and it was hI + qx = 0
motionless, the liquid bath itself would be static on y = h(x,t) (8.3lf-h)
and the free-surface shape would be governed by
the equation ofcapillary statics. The force balance Here, N = (1 + h;)1/2 and the curvature of the
between surface tension and gravity represented free surface is K = - hxx N - 3. The three parameters
by this equation gives rise to yet another length in these equations are the Reynolds number
scale called the capillary length La = (a/pg)1/2. Re = p2gH~/J1.2, the upward wall speed
We are particularly interested in rising or falling N = J1.V*/pgH~, and the film thickness
liquid films in which H 0 «La' For such thin 6 = HO/L a .
films, we expect that the flow in the film will We wish to solve this problem for the case
have only a local effect on the shape of the free 6 « 1 and 0(1) values of the Reynolds number
surface near where the film enters or leaves the and the wall speed. When 6 = 0, we see that all
liquid bath. The free-surface shape near the bath of the highest-order derivatives in equations
should be close to that computed by capillary (8.31) disappear. As we saw in Section 8.2.2, this
statics. For the film above the bath, it has been is a signal of a singular perturbation problem.
observed that the free-surface shape changes very We should expect different length scales in both
slowly along the direction of the moving wall. the x- and y-directions for the flow in the liquid
The analysis of a thin-film flow of this kind is film near the point where it either enters or leaves
done using the slowly varying domain perturbation the bath, and that both of these scales will be
technique. This is described next. smaller than the capillary length La'
The governing equations for the rising or falling To find an approximation for the flow in the
film flows shown in Fig. 8.9 are normalized with outer bath region, we ignore the boundary
the following scales. All lengths are scaled on the conditions near the moving wall as this is where
capillary length La as the primary region of inter- we expect the rapid variations in velocity to
est is governed by this length scale. The velocity occur. The equations describing the leading-
scale is V s = pgH~/J1.. This represents a balance order behavior in the outer bath region away
between the vertical component of gravity and from the moving wall are obtained by letting
the viscous shear in a film of thickness H o and it 6 = 0 in equations (8.31). The result is
is a measure of the velocity entering the region

0= -Px + 1,0 = -Py, p(x,h) = K,
of interest. Time is scaled on a convective scale
based on this velocity and the capillary length, which reduce to one differential equation for the
t s = La/V s' Pressure is scaled on the capillary curvature of the free surface, the equation of
pressure Ps = a/La' Finally, the volume-flux scale capillary statics
is Qs = VsLa' With these scalings, the governing
(8.32)
equations and boundary conditions become:
To all higher orders the result is zero velocity
6Re{u l + uU x + vU y} = -Px + 1 + 62 (U xx + uy,)
and pressure (remember that we do not impose
(8.31a) the wall boundary conditions on the outer
problem). Thus, the solution of equation (8.32)
6Re{ VI + uV x + vV y} = - Py + 62 (V xx + vy,)
is the complete and exact solution of this outer
(8.31b) problem. This equation is most easily integrated
in arc length coordinates along the free surface. one given our scaling Ii = hie, we must rescale
From this formulation, one integration gives the the x-coordinate as x = xle 1 / 3 • This is equivalent
relation Cm = K 2 + 2 sin 4>, where 4> is the angle to the length scale 1= (aR olpg)l/3 used by
between the tangent to the free surface and the Wilson and Jones (1983).
x-axis and Cm is a constant that parameterizes The scale for the velocity in the longitudinal
the solution. This parameterization was presented direction remains the same as before. With this
by Ruschak (1978). For a horizontal free surface scale and the new seatings for the independent
far from the moving wall, Cm = 2. variables, the horizontal velocity component
The inner region near the moving wall has must be rescaled as v = vle 2 / 3 in order to balance
two domains of interest. There is one boundary the two terms in the continuity equation. The
layer on the wall beneath the free surface of the effect of the capillary pressure gradient is felt in
bath that generates a weak recirculation flow in the longitudinal momentum equation through
the main part of the bath. It allows the liquid to the pressure-gradient term. To ensure that this
satisfy the no-slip conditions on the moving wall term balances the viscous term and gravity, we
and the recirculating flow conforms to the static- must rescale pressure as p = ple 1 / 3 • The volume
bath, free-surface shape governed by equation flux in the film should scale according to the film
(8.32). We are not concerned with this flow. The thickness, and so ii = qle. Finally, the time scale
second inner region is the rising or falling film is adjusted to represent the convective time based
on the moving wall above the free surface of the on the new x-length scale, t = tle 1 / 3 •
bath. We will focus on this second region and The complete stretching transformation yields
develop an equation for the free-surface profile the following form of the governing equations
of the film that will match to the free-surface and boundary conditions
shape in the nearly static bath.
To examine the film-flow region next to the e2 / 3 Re {u; + uU,;;
+ vUy }
moving wall, we stretch the y-coordinate with - + 1 + e4/3 u,;;,;; + U yy
-- - p,;; (8.33a)
the transformation ji = Yle and Ii = hie, where
the tilded quantities are order one. This is e2 Re{v; + uV,;; + vV y }
equivalent to changing the seatings for these = -p-y_ + e8 / 3 v__
xx
+ e413 v__
yy
(8.33b)
variables from the capillary length L" to the
thickness of the film far above the bath R o. The U,;; + vy = 0 (8.33c)
stretching transformation must also include a
rescaling of some of the other variables in this
u(x,O) = - N, V(X,O) = ° (8.33d,e)
flow. The appropriate seatings are determined -p + 2e4 / 3 {v y -li,;;(uy + 8 4 / 3 v,;;} + 84/3~ U,;;}N- 2
by considering the physical forces of interest in
the film. The free surface in both rising and falling = -K
films is almost parallel to the vertical wall far (u y + e4 / 3v,;;)(1 - e4 / 31i;) + 284 / 31i,;;(Vy - u,;;)
above the bath, but then it curves away from the =0
wall to form the free surface of the bath. The
curvature of the free surface near the bath gives Ii; + ii,;; = 0
rise to a capillary pressure gradient in the film
on ji = Ii(x,t) (8.33f-h)
such that the lowest pressure occurs near the Here, N = (1 + e4 / 3 ~)1/2 and the curvature of
point where the free surface curves away from the free surface is K = -li,;;,;;N- 3 .
the wall. This pressure gradient accelerates the In equations (8.33), the ratio of the x-length
flow in this region and could possibly thin the scale to the y-Iength scale is e- 2/3, which is very
film. We wish to retain the effect ofthis capillary large as e -+ O. This means that for any dependent
pressure gradient at leading order in the flow variable, such as the free-surface position, the
equations. To make K x = (-h xx N- 3 )x of order length scale for variations in the x-direction is
much larger than the length scale in the y-direction. The solution of this leading-order problem is
Thus, the scales used above are the appropriate
ones to use in setting up this problem for a slowly Uo = F(hY - y2/2) - N (8.36a)
varying domain perturbation analysis. The end Vo = -FA hy 2/2 - y3/6) - F hxy 2/2 (8.36b)
result of such an analysis on this problem is a
differential equation for the free-surface position Po = K, qo = Fh3/3 - Nh (8.36c,d)
h. We derive this equation by expanding the F = 1- K: x (8.36e)
dependent variables in an appropriate power
series in e as follows, From the next order in the expansion, we obtain
Re {uo t + uou ox + VOU Oy } = - PIX + U1yy '

(8.34a,b) 0= - Ply' U1x + V1y = 0 (8.37a-c)
P= Po + e2 / 3pl + "', ii = qo + e2 / 3ql + ... U 1 (O) =0, V 1 (O) =0 (8.37d,e)
(8.34c,d) Pl (h) = 0, U 1y (h) =0 (8.37f,g)

Here, the normal-stress boundary condition (8.37f)
We now substitute the expansions (8.34) into
represents zero pressure on the free surface of
the equations of motion (8.33) and expand the
the film. All of the small terms that would have
result. The leading-order equations from this
normally appeared here when the curvature is
expansion are
expanded are missing because they were included
o= - Pox + 1 + UOyy' 0 = - POy' uOx + VOy = 0 to leading order. The result of this boundary
condition and the horizontal momentum equation
(8.35a-c) (8.37b) is that the pressure in the layer at this
order is zero. The longitudinal velocity in the
Uo(O) =- N, vo(O) =0 (8.35d,e)
film is a parallel shear flow with no-slip at the
Po(h) = K, uoy(h) = o. (8.35f,g) rigid wall and no tangential stress on the free
surface. The flow is driven solely by the inertia
This is standard lubrication theory for the flow of the leading-order flow field. The horizontal
in the thin film (see equations (8.9) and (8.25) for velocity in the film is just that required to satisfy
comparison). The pressure is constant across the continuity.
thickness of the film and is determined by the The solution of the flow field at this order is
local curvature ofthe free surface. The longitudinal easily obtained using symbolic manipulation
velocity is determined as a locally parallel shear software, or with somewhat more difficulty, by
flow driven by the capillary pressure gradient hand. Of direct interest to the remainder of this
and gravity. There is no slip on the moving wall analysis is the longitudinal flow rate
and no tangential stress on the free surface. Note 2 -5 37 i"'7 5 -4-
ql = Re { -15 F i h - 840 FF x n - 24 Fh hi
that we have kept the entire curvature term in
- 40F h hx + 15Fxh N + 24Fh
3 2 -6 - 2 -5 5 -4 -
the normal-stress boundary condition (8.35f) hxN}.
even though it contains some small terms that
(8.38)
could be formally neglected at this order. In
doing so, we follow Wilson (1982), Wilson and With equations (8.36d) and (8.38), we now have
Jones (1983), Kheshgi (1989), and Kheshgi, Kistler an expression for the flow rate in the film to two
and Scriven (1992). There is no difficulty in orders in e, i.e., ii = qo + e2/3 q l' Inserting this
keeping this full curvature expression, and it into the kinematic condition for the free surface
leads to a uniform approximation for the free- hi = - iix produces an evolution equation for the
surface profile as we will see below. free-surface position. This equation can be used
to study the transient motion of the film profile always looks like that of a static bath. However,
as well as its stability. our main goal in this section is to obtain a
In the remainder of this section, we confine uniformly valid representation of the free-surface
ourselves to the study of steady rising or falling shape from the horizontal surface of the bath to
film flows. For this case, the evolution equation the uniform film on the vertical wall far above
reduces to qx = 0 and so the flow rate in the film the bath. The matching of the flow fields in the
is constant, a result in keeping with a simple film and in the bath is not necessary for this
mass balance argument in the film. The constant purpose, and so we do not attempt it.
is evaluated by noting that far above the free The matching of the film profile given by
surface of the bath the film is parallel to the equation (8.39a) with the shape of the static open
vertical moving wall. This means that fi = I, bath occurs as the film thickness increases.
K = 0, F = 1, and q = 1/3 - N. The expression Consider the inertialess film profile equation
q = qo + e2/3 ql = 1/3 - N leads to thefollowing generated by ignoring the inertial terms in equation
film profile equation (8.39a), i.e., set Re = O. In the limit of fi- 00,
this inertialess film profile equation reduces to
1
- Kx + 1 - fi3 + 3N ( fi31 1)
- fi2 Kx = K x = 1, which is exactly the equation for
the free surface of a static bath. Thus, the
matching of the thin-film inner region to the
+ e2/3Re{ - ~ FF _fi4 - ~ F 2 fi3fi_x static-bath outer region occurs automatically
280 x 40
because we retained the small terms in the
--} = 5 expression for the curvature that would have
+ SFxn
21""2
N + gFhhxN 0 ordinarily been neglected to leading order in a
standard asymptotic analysis of this film flow.
(8.39a) The solution of the inertialess form of equation
where again (8.39a) yields a composite approximation to the
free-surface position that is uniformly valid over
F = 1- Kx and K = -Fi xx (1 + e4 / 3ii;)-3/2 the entire domain of interest. This conclusion
does not hold, however, when the inertial terms
(8.39b,c)
are included. To properly incorporate inertial
Equation (8.39a) is a fourth-order, nonlinear effects into the film profile equation, equation
ordinary differential equation describing the profile (8.39a) must be modified. The problem introduced
of the film above the bath. It is solved subject by the inertial terms is best illustrated by
to the boundary conditions for a uniform film considering the boundary conditions needed to
far above the surface ofthe bath, and the solution solve the film profile equation (8.39a).
is matched as fi gets large to the shape of the Two of the four necessary boundary conditions
static open bath given by equation (8.32). Once are posed far above the surface of the bath where
the film profile equation is solved, we can use the film is uniform. These are fi = 1 and fix = 0
the result to compute the pressure and velocity for x- - 00. As the film profile equation was
fields within the film. To match these quantities constructed to enable an automatic matching to
to those within the bath, we would need to the free surface of the static bath, the remaining
compute the flow in the bath driven by the two boundary conditions are posed far from the
moving wall and by the film (if it is falling). This moving wall where the film merges with the flat,
calculation is simplified to some degree as the horizontal free surface of the static, open bath.
analysis of equations (8.31) for the flow in the These boundary conditions are fix - 00 and
outer bath region indicated that these flows do fixx = 0 as fi - 00. The numerical solution of the
not influence the shape of the free surface of the film profile equation (8.39a) with these boundary
bath to leading order, i.e., the free-surface shape conditions uses the shooting method. We start
far above the bath where the film is flat and the largest real part is the dominant root. The
integrate this equation toward the bath as an exponential with this root will grow the fastest
initial value problem. During the integration, we (or decay the slowest) as we move down the film
simultaneously compute the parameter toward the bath. Thus, it governs the behavior
cm = K 2 + 2 sin</>. Recall that Cm was the sole of the small-amplitude perturbation of the film
parameter characterizing the free-surface shape far above the surface of the bath. We write the
in a static bath as given by equation (8.32). As film profile using this dominant exponential in
h gets large and the film profile matches to a the form h = 1 + AReal {exp(mdx)}, where A is
static-bath, free-surface shape, Cm should approach a small amplitude parameter and md is the
a constant. Varying the initial conditions, and dominant root and we ignore the other three
for a rising film the film thickness parameter e, roots. With this film profile and its first three
will yield the appropriate set of conditions that derivatives, we now have the four required initial
correspond to the open bath configuration cm = 2. conditions for the film in terms of only one
There are several difficulties with this numerical unknown constant A. The film profile equation
integration. The first one is in identifying the (8.39a) is then integrated from some fixed initial
four initial conditions for the film far above the point x_co far above the surface of the bath. The
bath. One attempt is to use the conditions for a constant A, and for a rising film the film thickness
vertical uniform film, i.e., h = 1, hx = 0, hxx = 0 parameter e, is varied until the solution matches
and hxxx = 0 for x- -00. However, a uniform to the open bath configuration.
film is also an exact solution of equation (8.39a) The roots ofthe characteristic equation (8.40b)
and using these initial conditions in the numerical are easily approximated if we treat 2 / 3 e Re
as a
integration will produce the solution h = l. To small parameter. The first three roots of this
overcome this difficulty, we use a small-amplitude equation are close to the three cube roots of
domain perturbation in the region far above the 3(N - 1). The character of these roots change
bath. We define a small perturbation to the free with the wall speed. For films with N < 1, there
surface h = 1 + h1 , where Ih 11« 1, substitute this is one negative real root and a complex conjugate
into the film profile equation (8.39a), and then pair with positive real parts. For films with
linearize the equation for small h1 • The resulting N > 1, there is one positive real root and a
ordinary differential equation is complex conjugate pair with negative real parts.
h lXXx + 3(1 - N) The fourth root of the characteristic equation is
a singular root because the small parameter
h1 + e2/3 Re(- }870 + ~ N )h lXxxx Re
e2 / 3 multiplies the m4 term. We use the singular
perturbation method to obtain the following
+ e2 / 3 Re(-~ _
+ ~N)h 1x = 0
40 8
(8.40a)
leading-order approximation of this root
The solutions of this equation have the exponential

form h 1 = exp(mx), where m is a characteristic m::::; -
37 - -2 N
( 280 5
)-1 (e2/3 Re)- 1 (8.41 )
exponent. Substituting this form for h 1 into
equation (8.40a) produces a fourth-order char-
acteristic equation for m,
For small values of 2 3
/ e Re
and
m+ 3(1 -
3
N) + e Re(-E.
2 3
/ + ~N)m4
280 5
N > 37/112 = 0.33036, the singular root is very
large and negative. One of the three cube roots
of 3(N - I) has the larger real part and so it
+ e2 / 3 Re(- ~ + ~8 N)m =
40
0 (8.40b) would be the dominant root. Thus, the singular
root has little effect on the linear behavior of the
Of the four roots of this equation, the one with film far above the bath.
For small values of £;2/ 3 Re and N < 37/112, Hansen (1987) considered a falling film on a
the singular root from the characteristic equation moving wall formulated as a two-dimensional
is very large and positive. It is now the dominant Stokes flow in the liquid. His results indicated
root, and it describes a rapid exponential increase that the film is well-described by lubrication
in the film thickness as we approach the bath. theory. A falling film with inertia on a stationary
However, the fact that this root is proportional wall was investigated by Ruschak (1978). He
to £; - 213 implies that the x-variable should be computed the film flow based on lubrication-type
rescaled by this same amount. The original equations, and his results indicated that inertia
stretching transformation of the governing behaved as a small perturbation to the flow. We
equations (8.31) for the inner film region used conclude from these two studies that the physical
the transformation x = X/£;1 / 3. An additional breakdown of the film profile equation described
£; - 2/3 scaling means that we should have used above probably does not occur.
the transformation x = x/£;. This is the same The only remaining explanation for the break-
scaling used for the y-variable, and physically it down of the film profile equation is now
represents the film thickness far above the surface mathematical, and this is found through an
of the bath. After this new stretching transform- examination of how the singular root was
ation for the inner film region is completed (this generated. Recall that the film profile equation
also includes new scales for the pressure and the (8.39a) is essentially an equation for a constant
y-component of velocity), we find that the flow rate in the film,
leading-order approximation to the governing
equations are the full, two-dimensional
1/3 - N = q = qo + £;2 / 3q1 + ...
Navier-Stokes equations and the complete The flow rate qis approximated with an asymptotic
free-surface boundary conditions. There is no expansion in terms of the free-surface shape
simplification! function ii. For this expansion to be asymptotically
The implication of this result is that the film valid, the second term should be much smaller
profile equation (8.39a) breaks down on either than the first. We found the first three roots of
a physical or a mathematical level when the characteristic equation (8.40b) as regular
N < 37/112 owing to the appearance of the perturbation expansions that are close to the
positive singular root. A physical breakdown cube roots of 3(N - 1). When the linearized
means that the equation no longer describes the free-surface shape based on these roots is
physics of the flow in the proper way. This often substituted back into the film profile equation,
happens in an asymptotic analysis. From the the equation qo = 1/3 - N is satisfied to leading
scaling argument described above, we may suspect order. The remaining small terms from this
that the film connects with the bath through a equation balance with the small terms generated
two-dimensional flow region that scales with the from the second term of the expansion £;2/ 3 q 1 .
film thickness H 0 in both directions. This kind Thus, the expansion for the flow rate q based on
of flow violates the original scalings used in the these roots is a valid asymptotic expansion.
asymptotic development of the film profile When the linearized free-surface shape based on
equation. A slowly varying, lubrication-like film the singular root is substituted back into the film
flow above the bath could possibly be matched profile equation, the leading-order balance is
to this two-dimensional flow, or it may not even between the first term qo and the second term
exist. Within this two-dimensional flow region, 3
£;2/ q l' Such a balance destroys the underlying
inertia should also be important as the singular asymptotic structure ofthe expansion. This balance
root is derived from the inertial terms in the film also destroys the automatic matching of the film
profile equation. To verify this conjecture, a two- profile to the static bath shape that was seen to
dimensional calculation of a falling film with leading order. On a more physical level, balancing
inertia is needed. the two terms qo and £;2/ 3 q 1 means that we are
attempting to balance inertia with the capillary profile equation in which the effect of this
pressure gradient and the viscous stresses to spurious solution is removed. Using the slowly
leading order. The film profile equation breaks varying, domain perturbation technique, he
down because this balance can only be made in derived a film profile equation that is equivalent
the context of a two-dimensional flow, as shown to equation (8.39a) when his different scalings
by the scaling argument above. This balance does are taken into account. He then developed an
not work for the lubrication-type flow that is alternative form of this equation using the idea
examined in a slowly varying domain perturbation that the basic force balance in the film is between
analysis. the capillary pressure gradient and the viscous
The singular root from the characteristic shear stresses, and that inertia should always act
equation (8.40b) is now recognized as a spurious as a small perturbation to the flow. As a first
root. It has no physical implications for this film approximation then, he ignored inertia and
flow because it is a mathematical artifact produced enforced the leading-order flow rate condition
by violating the asymptotic structure of the film qo = 1/3 - N to produce the inertialess film
profile equation. We should just ignore it. Once profile equation
this is done, the three remaining roots from the
characteristic equation (8.40b), which are close
to the cube roots of 3(N - 1), control the entire - Kx + 1 = F = ii13 - 3N ( ii3
1 - iiI2) (8.42)
linear response of the film. For N> 1, the
positive real root is dominant and we should see Substituting the expression for F from the right-
a monotonic increase in the thickness of the film hand side ofequation (8.42) into the inertial terms
as we move towards the bath. For N < 1, the ofequation (8.39a) yields the alternative equation
complex conjugate pair of roots with positive
real parts are the dominant roots and we should
expect to see an oscillation of the free surface
with an increasing amplitude as we move toward
the bath. This kind of wave-like structure has
been observed in falling films in the experiments
of Cook and Clark (1973). Such observations
lend further support to this mathematical expla-
nation for the breakdown of the film profile
equation.
Ignoring the spurious root in the initial This film profile equation is an asymptotically
conditions for the integration of the film profile consistent, third-order differential equation de-
equation is easy enough. However, this root must scribing the shape of both rising and falling films.
also be ignored in the differential equation itself. It is asymptotically consistent because the
If we integrated the film profile equation (8.39a) substitution (8.42) for F converts the fourth-
with the appropriate initial conditions, we would derivative term in the film profile equation
eventually see the spurious solution contaminate (8.39a) into a first-derivative term. This is clearly
the entire numerical solution when the spurious seen if we linearize the alternative film profile
root is positive, i.e., for N < 37/112. As we move equation (8.43) about ii = 1 to obtain
toward the bath, we would see a rapid increase
in the film thickness, and we would probably not h wx + 3(1 - N)h l
be able to match to any static bath configuration
for large values of ii. This behavior, of course, is + B2/3 Re -6(1
- -
5 7
-17 N
21
+ N 2) h-
Ix
entirely unphysical.
Kheshgi (1989) developed an alternative film = 0 (8.44a)
The corresponding characteristic equation is the open bath. Their approach, as described by
Levich (1962), was to apply standard lubrication
m 3+ 3(1 - N) + e2/3 Re-6(1- - -17 N + N 2) m theory to the slow, steady-state motion in a rising
5 7 21 film. In terms of the slowly varying domain
=0 (8.44b) perturbation analysis described above, their results
are the leading-order approximation for the flow
Here the fourth-derivative term in the linearized in the film given by equations (8.36). The only
perturbation equation and the fourth-order term difference is that they used the asymptotically
in the characteristic polynomial do not appear correct leading-order form of the surface curvature
as they do in the original equations (8.40). K ~ - hxx ' while in the present analysis, the small
One can also show that in the limit of h-+ co, terms in the curvature were retained in order to
the full film profile equation (8.43) reduces to the produce a uniformly valid differential equation.
equation for a static bath Kx = "x
= 1. Thus, the Equation (8.36d) for the flow rate in the film with
solution of this film profile equation matches this modification of the curvature and written in
automatically to the shape of a static bath. This dimensional form is
solution represents a composite approximation
Q = - U*H + (pg + uH xxx )H 3 /3Jl
for the film profile that includes the effects of
capillary pressure, viscosity and inertia. This result is equivalent to the film profile
Kheshgi (1989) linearized both the original and equation given by Levich (1962, equation (133.12)).
the alternate film profile equations about a (Recall that a positive U* is a velocity in the
uniform film thickness, found the corresponding - X -direction, which is the negative ofthe velocity
characteristic equations and then computed the defined by Levich).
roots for a rising film with no gravity. The To solve this third-order differential equation,
dominant root governing the behavior of the film four boundary conditions are needed because
was compared with the exact results of Higgins the final film thickness is also unknown. Three
(1982) who computed the root by linearizing the of these conditions correspond to a flat uniform
full Navier-Stokes equations. The alternative film far above the bath. The fourth condition is
equation (8.43) produced more accurate results obtained by explicitly matching the shape of the
for the rising-film problem over a wider range thin film on the vertical, moving wall to the shape
of Reynolds numbers than did the original film of the static free surface in the open bath. This
profile equation (8.39a). In this comparison, the must be done because of the simple curvature
spurious root did not affect the results because term employed in the analysis. Levich (1962) did
it is never the dominant root for a rising film. this matching by first computing the static free-
Based on this success, Kheshgi, Kistler and surface shape of an open bath in which the free
Scriven (1992) used the alternative film profile surface was tangent to the vertical wall at its
equation (8.43) to compute the film profiles for point of contact (the liquid film on the wall is
both rising and falling film flows. They showed so thin that its thickness was ignored). Then, he
how the film profile is affected by changes in the ignored gravity in the film profile equation and
bath configuration, i.e., changes in the parameter matched the curvature of the thin film for large
em that parameterizes the free-surface shape of a values of thickness to the curvature of the static
static bath. Their results for the open bath free surface at its point of contact with the
configuration followed the expected behavior moving wall. With these four boundary conditions,
discussed above. he was able to compute the thickness of the film
Landau and Levich (1942) did the classical on the moving wall far above the bath.
derivation of a film profile equation for the The matching condition employed by Landau
rising-film problem in order to obtain the final and Levich (1942) works well when gravity is
thickness ofthe film on the moving wall far above ignored in the thin film. In this case, it is an
example of the simple leading-order asymptotic be found in the review by Ruschak (1985) and
matching that was used in the roll-coating problem in Chapter 13.
of Section 8.4. However, the matching condition Finally, we mention that the slowly varying
does not work at all when gravity is included in domain perturbation method is not the only way
the thin film because the curvature of the film to derive useful film profile equations. A number
as the thickness gets large does not approach a of other techniques are available. One method
constant. This difficulty was resolved using a is just to neglect small tenns from the governing
fonnal singular perturbation analysis by Ruschak equations to obtain an approximate set of
(1976) and more completely by Wilson (1982) differential equations. A second method is to
who obtained a two-term approximation for the integrate the longitudinal momentum equation
final thickness of a rising film in terms of the across the film and to approximate the longitudinal
small capillary number, Ca = j1.U*/(J. velocity with a simple expression or perhaps a
Other workers have considered falling films in correlation of some kind. The work of Ruschak
various special limits or approximations. Wilson (1978) described above uses both of these
and Jones (1983) considered a falling film with techniques. These methods are discussed in slightly
no inertia and used a singular perturbation more detail in the introduction to this chapter.
analysis to compute a series of waves that Higgins et al. (1977) and Higgins and Scriven
appears on the free surface of the film as it (1979) have also discussed these two methods
approaches the bath. Ruschak (1978) examined and compared them with asymptotic methods
a falling film with inertia. In his analysis, he within the specific context of their use in deriving
removed small tenns from the governing equations film profile equations.
by assuming a lubrication-like flow in the film
above the bath. He then approximated the
longitudinal velocity in the film with a second-
8.6 THE STABILITY OF INCLINED LIQUID
order polynomial, developed the corresponding
FILM FLOWS
horizontal velocity, and integrated the longitudinal
momentum equation to obtain a nonlinear Consider a unifonn liquid film flowing steadily
equation for the free-surface position. He solved on a solid substrate under the action of gravity.
this equation numerically and found waves on This situation occurs in the dip coating process
the free surface ofthe film above the bath. Hansen far above the free surface of the liquid bath, in
(1987) considered a falling film on a wall that some portions of the flow domain in slide
was either stationary or moving downward into coating, and in other coating flow processes. In
the bath. He assumed a pure Stokes flow in the these situations, the steadiness of the film flow
film and used a boundary integral method to is an important factor to consider. An instability
compute the shape of the free surface. He also of the flow would lead to wave-like motions in
found waves in the free surface of the film above the free surface of the film that could in turn
the bath. In addition, for a stationary wall and produce unacceptable nonuniformities in the
low values of the surface tension, he saw a thickness of the final coating. Thus, it is important
depression in the free surface of the bath near to consider the stability of the associated coating
the wall such that the free surface there was flows during the design of any quality coating
actually lower than the free surface of the bath process. The first, and probably the easiest,
far from the wall. The depression deepened for analysis that should be done is a linear stability
a moving wall as the speed of the wall into the analysis.
bath increased until a kink appeared on the free A linear stability analysis of any flow is based
surface. At this point, the calculation was stopped on a technique somewhat akin to the regular
because of a loss in accuracy. Further discussion perturbation method. The small parameter in
of various dip-coating analyses and results can this case is the amplitude of a disturbance to the
The stability of inclined liquid film flows 281
flow under consideration, such as the falling film
in the above coating processes. Usually this flow
field, called the basic-state flow, is known as an
exact solution of the governing equations. The
mathematical details of the stability analysis are
simpler if this is the case, but a numerically
determined basic-state flow can also be used (see
Chapter 9 for examples of this kind of analysis).
Once the basic state is known, the method
develops a set of disturbance equations that are Figure 8.10 A schematic of a single liquid layer
linearized and solved. The solution of these flowing down a rigid inclined plane.
equations characterizes the temporal and spatial
behavior of a disturbance to the basic state and
provides information on when the amplitude of the inclined plane with the X-axis embedded in
a disturbance is expected to grow or decay in the plane and aligned in the flow direction and
either time or space. with the Y-axis normal to the plane. All lengths
The stability analysis has one additional are scaled using the undisturbed thickness of the
complication in a free-surface flow that is not film H 0' the velocity scale is based on a balance
seen in flows with just rigid boundaries. The between the longitudinal body force and the
free-surface position in these flows is a dependent viscous shear stress Us = pgH~ sin(p)/Jl, time is
variable and must be determined as part of the scaled on the convective scale t s = Ho/U s ' and
solution. However, as the free-surface position pressure is based on the viscous scale Ps = JlUs/H o.
ofthe basic-state flow is known, we can use a small- The governing equations and boundary conditions
amplitude, domain perturbation analysis to become
determine the perturbations to the system,
including those to the free surface.
In the next two subsections, we apply these (8.45a)
asymptotic techniques to single- and multi-
layered, falling film flows. The techniques are Re {v, + uVx + vV y } = - Py - cot(P) + Vxx + Vyy
discussed first, and then the results are discussed (8.45b)
in terms of the underlying physical processes
reflected in the analysis. For a more complete (8.45c)
discussion ofthe techniques oflinear and nonlinear u(x,O) = 0, v(x,O) = 0 (8.45d,e)
stability theory, see the standard texts by Drazin
and Reid (1981), Joseph (1976), and Chandrasekhar - p + 2{vy- hAu y + ~x) + h;ux } N- = _-ca-
2 1K)
([1961] 1981). Further examples of stability
(u y + vJ(l - hx ) + 2h x (vy - ux ) - 0
problems in various coating flows are discussed
in Chapters 9, lIb and lId. v = h, + uh x
on y = h(x,t) (8.45f-h)
Here N = (1 + h;)1/2 and the curvature of the
8.6.1 A SINGLE LIQUID LAYER FLOWING
free surface is K = -h xx N- 3 • The two para-
DOWN AN INCLINED PLANE
meters are the Reynolds number Re = p2gH~
Consider a single liquid layer flowing down a sin(p)/Jl2 and the capillary number Ca = pgH~
smooth rigid plane inclined at the angle p with sin (P)/(J.
respect to the horizontal as shown in Fig. 8.10. The basic-state flow is an exact solution of
We use a Cartesian coordinate system fixed to these equations in the form of a uniform, parallel
film flow. This solution, marked by an overbar, governing equations that are products of two or
is more disturbance quantities and obtain a linearized
set of equations. In effect, we expand the
u=y-y2/2, v=O, p=cot(P)(I-y), n=1 disturbance quantities in a regular perturbation
(8.46a-d) expansion in terms ofa small parameter measuring
the amplitude, and then only consider the
The stability ofthis basic-state flow is examined leading-order problem. The difficulty in applying
by assuming that there exists some disturbance boundary conditions on the unknown free-surface
to the system, and then writing all of the flow position is resolved by using the small-amplitude
variables in terms of the basic state and this domain perturbation method and linearizing the
disturbance. For example, the longitudinal velocity result. This method allows us to write the
is written as u = U + u', where primes denote the free-surface boundary conditions in terms of
disturbance quantities. These disturbance quan- linear conditions evaluated at the basic-state
tities are substituted into the original governing free-surface position, which is known. The resulting
equations, which are then simplified. The result set of linear partial differential equations and
is a set of nonlinear disturbance equations boundary conditions is
Re{u; + uu~ + u'u~ + v'uy + v'u~}
= -p~ + u~x + U~y (8.47a) (8.48 a)
R e {Vr' + UV
-'} " ,
x = - Py + Vxx + Vyy (8.48 b)
(8.47b) u~ + v~ = 0 (8.48c)
u~ + v~ = 0 (8.47c) u'(x,O) = 0, v'(x,O) = 0 (8.48d,e)
u'(x,O) = 0, v'(x,O) = 0 (8.47d,e)
-p - p' + 2{v~ - h~(uy + u~ + v~) y=1

+ (h~)2U~}N-2 = -Ca-1K'
(u y + u~ + v~)[1 - (h~)2] (8.48f-h)
+ 2h~(v~ - u~) = 0 To solve these linear partial differential dis-
v' = h; + uh~ + u' h~ turbance equations, we note that the nonconstant
coefficients in the equations (which are all
on y = h(x,t)
(8.47f-h) basic-state quantities) depend only on the y-
coordinate. Therefore, the general form of the
Here, N = [1 + (h~fr/2 and the curvature of solution contains exponentials in the x- and
the free surface is K' = - h~xN - 3. t-variables. (This is motivated by recalling that
Given the appropriate initial conditions, these the general form of the solution of a linear
equations describe the temporal spatial develop- ordinary equation with constant coefficients in
ment of the disturbance in the layer. However, the independent variable x is exp(mx), where m
they are just as difficult to solve as the original is a constant.) On a physical level, this means
set of governing equations and so it appears that that an arbitrary disturbance in the film can be
we have gained nothing at this point. We make written as the sum of an infinite number of small
further progress by assuming that the amplitude amplitude traveling waves ofarbitrary wavelength
of the disturbance is small compared with the and frequency. The amplitude of these waves
basic state. Thus, we let u', v' « U, p' «P, and may grow or decay in time or space. This kind
n
h' « = 1. We now neglect all terms in the of structure and the linearity of the equations
allows us to break up the disturbance into its which in normal-mode form is 'II = Real
individual components and to solve the disturb- {4>(y)exp[ia(x - ct)]}. From the standard defi-
ance equations for only a single traveling wave. nition of the stream function we write the velocity
This kind of analysis is called the method of components as u = D4> and v = - ia4>. The
normal modes (see Drazin and Reid 1981; Joseph pressure is eliminated in the two momentum
1976). For example, we write the longitudinal equations (8.50a,b) by differentiating (8.50a) and
velocity in the form then combining the two equations, the pressure
in the normal-mode free-surface boundary con-
u' = Real {u(y) exp[ia(x - ct)]} (8.49)
dition (8.50f) is eliminated by using the momentum
Here, a is the real-valued wave number of the equation (8.50a), and the free-surface disturbance,
disturbance and c = Cr + iCi is a complex number. 11, is written in terms of the function, 4>, by using
The phase speed of the disturbance wave is the the kinematic boundary condition (8.50h). The
real part cr and aC i is the exponential growth final result is
rate of the disturbance. The complex function
u(y) is the depth-dependent amplitude of the (D 2 - ( 2)24> - iaRe{(u - c)(D 2 - ( 2)4> - Uyy 4>}
disturbance wave. The normal-mode form just =0 (8.51a)
described is designed to examine the temporal
stability of a traveling disturbance wave because 4>(0) = D4>(O) = 0 (8.51 b,c)
the complex number C in front of the time
variable allows the wave to grow or decay in D 24>(I) + {_ uyy (l) + a2}4>(I) = 0
(u(l) - c)
time. We can also write the normal-mode form
to allow bounded oscillations in time and growth (8.51d)
or decay in the x-direction by letting ac be real
D 3 4>(I) - {iaRe[u(l) - c] + 3a 2}D4>(I)
and a be complex. These spatial instabilities are
described in more detail in Drazin and Reid (1981). 2 1
+ {ia[Cot(p) + a Ca- ]}4>(I) = 0
Substituting the temporal form of the solution (u(l) - c)
for the dependent variables (8.49) into the linear
disturbance equations (8.48) yields the normal- (8.51e)
mode form of the disturbance equations Equation (8.51 a) is known as the Orr-Sommerfeld
Re{ia(u - c)u + uyv} = -iap - a2u + D 2u equation.
In either form, the disturbance equations (8.50)
(8.50a) or (8.51) are a fourth-order system of linear
2 2
iaRe(u - c)v = -Dp - a v + D v (8.50b) ordinary differential equations with non-constant
coefficients that constitute an eigenvalue problem
iau + Dv = 0 (8.50c)
for the complex eigenvalue c. The eigenvalues
u(O) = 0, v(O) = 0 (8.50d,e) depend on the four parameters a, Re, {J and Ca.
2 Numerical codes today can easily integrate these
-p + 2Dv = pyll - a ca- 1l)
1
equations and compute the eigenvalues. One
Du + iav = - uyyll on y= 1 major difficulty occurs when a Re is large. In this
v = ia(u - c)1I case, the differential equations become stiff and
increased grid resolution or special techniques
(8.50f-h)
such as orthonormalization are needed to deal
Here, we use the operator D" = d"jdy". The letter with the situation. Spectral methods provide an
subscripts on the basic-state quantities also accurate and effective way to solve these disturb-
represent derivatives with respect to y. ance equations. They are described by Canuto
This system of disturbance equations is often et al. (1988, Section 6.4) in the framework of the
written in terms of a disturbance stream function, Orr-Sommerfeld problem. These methods can
provide the complete eigenvalue spectrum of a
truncated system, which is very useful in deter- Re
mining the most dangerous eigenvalue as defined
below. The paper by Scott and Watts (1977)
describes a special-purpose code called SUPORT
that uses the shooting method and orthonormal-
ization to solve stiff two-point boundary value
problems. (The current version of this public
domain code, BVSUP, is a member of the
SLATEe library maintained on the WorId Wide Stable
Web with the URL http://gams.nist.gov/.) An
eigenvalue for the above disturbance equations
is computed by combining the BVSUP code (or 0:
its equivalent) for the integration of the equations
with a root-finding routine supplied by the user. Figure 8.11 A schematic of a neutral curve for a
The result is an accurate calculation of an two-dimensional fluid flow.
eigenvalue given a good enough initial guess.
However, the fact that only one eigenvalue is
computed at a time is a serious disadvantage. disturbance with the wave number IXe and with
One must be careful to determine if the most the phase speed Cre = cr(IXe,Ree).
dangerous eigenvalue has indeed been found. The stability calculation for an inclined falling
The stability characteristics of the system are film shows that the film is susceptible to two
defined by the behavior of the most dangerous different types or modes of instability.The first
eigenvalue. This eigenvalue is the one with the is a shear-flow instability similar to that seen in
largest imaginary part. It corresponds to the a parallel flow inside a two-dimensional channel.
disturbance mode that grows the fastest or There is only a small deflection of the free surface
decays the slowest. For fixed values of the of the film for this mode. A detailed numerical
parameters fJ and Ca, the computations described analysis of this instability was presented by
above are used to find this eigenvalue as a Floryan, Davis and Kelly (1987) for the angles
function of the wave number and the Reynolds ofinclination 0.5' < fJ < 4°. The critical conditions
number in the form Cr = cr(IX,Re) for the real part are affected by changes in the surface tension
and Ci = ci(IX,Re) for the imaginary part. The and the angle of inclination of the film, but
neutral curve for the system is determined by roughly the critical Reynolds number is on the
finding the locus of points for which ci(IX,Re) = O. order of 7400 or larger, the wavelength of the
A schematic of one such curve is shown in Fig. critical disturbance is from two to six times the
8.11. This curve divides the (IX,Re)-plane into the mean film thickness, and the phase speed of the
shaded region above the curve where Ci is positive critical disturbance is from 16 to 26% of the
(unstable modes) and the region below the curve free-surface speed of the film in the base state.
where Ci is negative (stable modes). The minimum These rough ranges also seem to hold for fJ > 4°.
of this neutral curve (IXe,ReJ is called the critical The mechanism for the instability lies in the
point or onset point of the instability. For inertial Reynolds-stress term Re uyv' found in the
Re < Ree, all possible disturbances decay in time longitudinal momentum equation (8A8a). This
and so the system is stable. For Re> Ree, term describes how energy is fed into the
disturbances within a finite range of wavenumbers disturbance through the motion of high-momen-
will grow in amplitude and so the system is tum liquid from the upper part of the film
unstable. The Reynolds number Ree characterizes downward into the low-momentum region ofthe
the onset of instability in the system to a film near the wall.
The second mode of instability seen in an expand the free-surface variable h. Instead we
inclined film flow takes the form of a free-surface normalize the solution by choosing h = 1.
deflection or wave. A defining feature of this The expansions (8.52) are substituted into the
instability is that the wavelength of the critical disturbance equations (8.50) and a sequence of
disturbance is much larger than the film thickness. flow problems are produced. The leading-order
In fact, Yih (1963) showed that the wavelength flow problem is
ofthe critical disturbance is infinite in an infinitely
long liquid layer. In a finite coating device, D2uO = 0, uo(O) = 0, Du o(1) =1 (8.53a-c)
disturbances oflarge, but not infinite, wavelength Dpo = 0, Po(1) = cot(fJ) (8.53d,e)
compared with the thickness of the liquid layer
have been observed. For this reason, the mode The first-order flow problem is
is called a long-wave instability. Numerical DV t + iu o = 0, Vt (0) = 0, Co = u(l) + iV l (1)
computations by Floryan, Davis and Kelly (1987)
and De Bruin (1974) have shown that the critical (8.54a-c)
Reynolds number for the long-wave instability
D2u l = ipo + iRe(u - co)u o
is less than the critical Reynolds number for the
shear-flow instability when fJ > OS. This means + Re UYV l , u l (0) = 0, DU t (1) = 0
that for these angles of inclination the long-wave
(8.54d-f)
mode is unstable while the shear-flow mode is
stable. Thus, long waves are the more dangerous The necessary equations from the second-order
mode for fJ > OS. As angles of inclination greater flow problem are
than half a minute occur in typical coating
processes, the long-wave instability is of particular DV 2 + iU t = 0, v2 (0) = 0, cl = iv 2 (1)
importance to coating engineers. Numerical (8.55a-c)
integration of the linear disturbance equations
has been used to find the neutral curves and the After solving these three flow problems for the
critical conditions that completely characterize velocity field, equations (8.54c) and (8.55c) are
this instability, as discussed by Floryan, Davis used to determine the eigenvalue. The result is
and Kelly (1987) and De Bruin (1974).
Rather than proceeding numerically, the long- C = 1 + ai { - ~cot(fJ) + 125 Re} + ... (8.56)
wavelength character of the observed disturbances
allows us to look for simple approximate forms The critical Reynolds number is found by setting
of the critical conditions using the regular the imaginary part of this eigenvalue to zero,
perturbation method. This method was first used
on the inclined-film-flow problem by Yih (1963). Rec = 5 cot(fJ)/2 (8.57)
We first expand the dependent variables in terms To verify that the Reynolds number given by
of the small wave number a. These expansions are equation (8.57) is indeed the critical value, we
should continue with the perturbation procedure
a= Uo + aUt + a 2 u 2 + ... (8.52a)
and compute the two-term expansion for the
2
(j = aV t + a v 2 + ... (8.52b) Reynolds number. It has the form Rec = 5 cot
(fJ)/2 + a 2R 2, where R 2 = R 2(cot(fJ),Ca- t ). Be-
P= Po + aPl + a 2p2 + ... (8.52c) nney (1966) found R 2 using a different asymptotic
C = Co + aCt + a c 2 +...
2
(8.52d) procedure. His result shows that R 2 is positive
for fJ > 4.1' and Ca- t = 0 (zero surface tension).
As equations (8.50) represent an eigenvalue This verifies that for this parameter range, the
problem for c and the solution is known only to minimum of the neutral curve for the long-wave
within a multiplicative constant, we do not instability is given by equation (8.57) and it
occurs at ex = O. For p < 4.1', R 2 is negative and ~peed

there is a local maximum at ex = O. Most probably,
there is also a local minimum very close to this Interface
point, although this has never been examined in deflection
such fine detail. The numerical calculations of
the disturbance equations (8.5 1) by Floryan,
Davis and Kelly (1987) suggest that the critical
Reynolds number is given by (8.57) for P> 0.5'.
However, these calculations did not consider
wavenumbers small enough to observe the
behavior suggested by Benney's result (1. M.
Floryan, personal communication, June 1994). Figure 8.12 A schematic of the motion in a disturbed
The critical Reynolds number (8.57) is not very liquid film. The upper straight gray line is the undisturbed
large except when Pis very small. In fact, Ret = 0 free-surface position. The upper solid line is the
disturbed free-surface position. The leading-order
for a film flow down a vertical wall. Thus, a longitudinal flow rate in the film is proportional to
vertical film flow is always unstable to a long- the deflection of the free surface. The upper wavy gray
wavelength disturbance. We also note that the line is proportional to the longitudinal flow rate in
critical Reynolds number does not depend on the film. It is phase shifted to the left due to inertial
the capillary number to this order of the effects in the flow. The gray box in the layer is a control
approximation because the long wavelength of volume showing the interfacial deflection at leading
the disturbance implies an extremely small order.
curvature of the free surface. Surface tension
appears in the next term of the expansion for the the normal-mode form of the disturbance tan-
critical Reynolds number. The phase speed gential-stress boundary condition (8.50g), exactly
determined from the eigenvalue (8.56) shows that cancels the shear stress in the basic-state flow at
the disturbance wave travels at about twice the the disturbed location of the free surface. Taken
free-surface speed of the film in the base state. together the basic state and the leading-order
Aside from these closed-form results for the flow perturbation form an almost parallel shear
critical Reynolds number and the eigenvalue c, flow in the thin film in which the free surface is
the use of the regular perturbation method in subjected to a small amplitude, long-wavelength
the inclined film-flow problem allows us to disturbance. When the free surface is deflected
interpret the motion of the unstable mode in the upward by this disturbance, the increased longi-
film in terms of readily identifiable forces and tudinal body force on the liquid in the film
accelerations. A detailed discussion of this kind increases the longitudinal velocity in direct
was presented by Smith (1990). We present a proportion to the magnitude of the deflection.
more concise version next. Thus, the velocity disturbance U o is exactly in
Consider a thin liquid film flowing down a phase with the free-surface deformation.
rigid inclined plane as shown in Fig. 8.12. The A small elemental control volume in the film
basic-state flow in the film is a parallel shear flow located to the right of a disturbance crest is
driven by the longitudinal component of the shown in gray in Fig. 8.12. The flow rate entering
gravitational body force. Now consider a long- the left side of this control volume is larger than
wavelength disturbance to the free surface of this the flow rate exiting from the right side, because
film. The disturbance equations (8.53a-c) indicate the flow rate increases with the deflection of the
that the leading-order longitudinal flow pertur- free surface. Conservation of mass within the
bation behaves like a parallel shear flow, but one control volume implies that the free surface of
driven by a perturbation shear stress on the free the film is deflected upwards. The same argument
surface. This shear stress, given by -u yy (l)J1from applied to a control volume to the left of the
disturbance crest shows that the free surface there we actually see a small reduction in the longitudinal
is deflected downwards. The net result of these velocity compared with the leading-order value
motions is that the disturbance moves to the on the right side of the crest. As the left side of
right as a wave with a phase speed faster than the crest passes the fixed point, the free-surface
the liquid velocity in the film. Thus, the leading- deflection decreases and the flow decelerates.
order longitudinal velocity perturbation in the Inertia resists this motion and we see a small
film controls the phase speed of the long- increase in the longitudinal velocity on the right
wavelength disturbance. side of the crest. The net effect is a phase lag
The leading-order pressure perturbation gov- between the free-surface deflection and the
erned by the disturbance equations (8.53d,e) is longitudinal velocity, as shown in Fig. 8.12.
simply the extra hydrostatic pressure in the layer The effect of this phase lag on the deflection
due to the normal component of gravity. When of the free surface is seen with another control-
the free surface is deflected upwards, the extra volume analysis. The control volume includes all
liquid above the mean position of the free surface of the liquid underneath a disturbance crest and
increases the hydrostatic pressure underneath the between the two node points. The inertial phase
disturbance crest. This pressure perturbation is lag in the film flow slows the longitudinal velocity
exactly in phase with the free-surface deformation. of the liquid at the front node point and speeds
The instability ofa long-wavelength disturbance up the velocity at the rear node point creating
to the film is a direct result of the longitudinal a net voume flow into the control volume. The
flow perturbation given by equations (8.54d-f). result of this volume flow is an increase in the
These disturbance equations describe this flow free-surface deflection, which then causes an
as a parallel shear flow driven by the leading-order increase in the phase lag of the longitudinal
pressure field and the inertial effects of the velocity, which then further increases the net
leading-order shear flow. The po-term represents volume flow underneath a disturbance crest, and
the pressure gradient established by the variation so on. Here, the unsteadiness of the flow and the
along the layer of the leading-order pressure inertia of the liquid have combined to form a
disturbance. This pressure disturbance is greatest destabilizing feedback mechanism.
under a free-surface crest and lowest under a The advection ofinertia exerts a similar influence
trough. The resulting pressure gradient pushes on the flow in the film. The term i Re it U o in the
liquid away from the crest and towards the disturbance equation (8.54d) represents the effect
troughs. The result is a leveling of the free surface; oflongitudinal advection. Consider the flow field
a stabilizing effect measured by the -cot(P)/3 at a particular instant in time. On the right side
term in the growth rate (8.56). ofa disturbance crest, the liquid is being advected
The three terms proportional to the Reynolds longitudinally from a region of higher
number in the longitudinal disturbance equation velocity at the crest to a region of lower
(8.54d) represent the effects of inertia associated velocity at the trough of the disturbance. The
with the leading-order flow field. The effect of inertia of the liquid resists this deceleration
the unsteady term -iRecouo is seen by considering resulting in a small increase in the longitudinal
the motion in the film under a crest of the velocity of the liquid to the right of the crest.
free-surface disturbance wave. As the crest The opposite effect is seen on the left side of the
approaches a fixed point in the film from the left, crest causing a small decrease in the longitudinal
we observe an increase in the film thickness and velocity. The net effect is to advance the phase
to leading order an increase in the longitudinal between the free-surface deflection and the
velocity because these two quantities are exactly longitudinal velocity in the film. Applying the
in phase. However, the inertia of the liquid previous control-volume analysis to this case, we
prevents it from accelerating fast enough to see that this phase advance has a stabilizing effect
follow the free-surface deflection exactly and so on the film.
The term Reuyv t in the disturbance equation not be very accurate ifit occurs at a wave number
(8.54d) represents the effect of advection by the that is not small enough. If this second deficiency
flow normal to the film. On the right side of a occurs, the results violate the original asymptotic
disturbance crest, the normal velocity of the assumption that the wavelength ofthe disturbance
liquid in the film is towards the free surface. This is large and so the theory may not be applicable
motion advects liquid with a small longitudinal, to the problem at all.
basic-state velocity from near the solid surface The first two of the deficiencies noted above
towards the faster-moving liquid near the free are removed by incorporating nonlinear effects
surface. The inertia of the liquid resists this into the stability theory for an inclined film flow.
acceleration causing a small decrease in the One of the first of such investigations was Lin
longitudinal velocity of the liquid to the right of (1969) who did a weakly nonlinear asymptotic
the crest. Similarly, the liquid to the left of the analysis on the full Navier-Stokes equations.
crest increases its velocity. As we saw for the The method he used is based on the fact that for
unsteady term, this is equivalent to a phase lag a Reynolds number just above the critical value
between the free-surface deflection and the there are two very different time scales in the
longitudinal velocity. The effect is destabilizing problem. The first time scale is based on the
to the flow in film. characteristic speed of the basic-state film flow.
The contribution of each of these inertial The second time scale is much smaller and is
effects to the imaginary part of the eigenvalue based on the growth rate of the unstable
(8.56) is 25Re/120 for the unsteady effect, disturbance. Lin used these two time scales in a
- 11Rell 20 for longitudinal advection, and regular perturbation analysis and computed
2Rel120 for normal advection. The unsteady several terms in the expansions to obtain a
inertial effect is the dominant one and it is nonlinear equation for the amplitude of a
destabilizing according to the mechanism de- disturbance wave. Gjevik (1970) also did a weakly
scribed above. The flow driven by these inertial nonlinear analysis, but his results were based on
effects competes with the stabilizing flow produced the fully nonlinear evolution equation for the
by the longitudinal variation of the hydrostatic shape of the free surface that was first obtained
pressure in the film, which is measured by by Benney (1966). This equation was derived
cot(fJ)l3. If the Reynolds number is large enough, using the slowly-varying domain perturbation
the inertial effects win, the film is unstable and method in the same way as our development of
the free-surface disturbance wave grows in the film profile equation in the first part of
amplitude. Section 8.5. From this equation, Gjevik obtained
The linear instability of a single Newtonian a nonlinear amplitude equation for a small
liquid layer flowing down an inclined plane is disturbance near the critical point. The more
now well understood. One deficiency of this recent work of Chang (1989) considered a
theory is that it is linear and so it only describes higher-order form of Benney's nonlinear evolution
the initial development of a disturbance. The equation for the free surface. He analyzed this
stability theory for long-wavelength disturbances equation near the critical point using modern
has two additional deficiencies. At the critical ideas of bifurcation theory. All of this nonlinear
point, the most dangerous mode has an infinite work showed that a linearly unstable long-
wavelength and such a disturbance cannot exist wavelength disturbance on the liquid film can
in any finite system. Secondly, when the Reynolds grow and equilibrate to a finite-amplitude wave
number is above the critical point the maximum of permanent form. This behavior has also been
growth rate occurs for nonzero wave numbers. seen in the fully nonlinear, numerical calculations
We must compute additional terms in the of Joo, Davis and Bankoff(t991). Three-dimen-
expansion for the growth rate in order to estimate sional numerical calculations for the flow in the
this maximum, and even then the estimate may film have been done by Joo and Davis (1992).
This work showed that the two-dimensional of the solid wall do not affect wave propagation
finite-amplitude wave of permanent form is itself in the uniform-flow region of the film. However,
unstable to a three-dimensional disturbance of their experiments showed that the surface waves
long enough wavelength. The film evolves to an on the liquid film had the same frequency as the
irregular three-dimensional pattern that does not wall oscillations, and that they evolved according
equilibrate to a pattern of permanent form, at to a linear, spatial stability theory. The conclusion
least to within the extent of this calculation. they made was that the wall oscillations forced
Additional work on liquid film flows has the development of the surface waves in the
examined how additional physical and geometrical entrance region to the flow where the liquid is
effects influence the instability. A heated film that introduced onto the inclined wall. Once formed,
includes the effects of thermocapillarity and these waves propagate through the uniform-film
evaporation has been studied both linearly and region without any further interaction with the
nonlinearly in the long-wave limit by Joo, Davis oscillating wall.
and Bankoff(1991). Evaporation may lead to the In the coating of polymers and adhesives, the
rupture of the film and to the breaking of surface non-Newtonian nature of the liquids is an
waves. Thermocapillary forces may promote important consideration. Shaqfeh, Larson and
significant local thinning of the film, enhance Fredrickson (1989) considerd the linear stability
wave breaking and accelerate the rupture process of a film flow composed of an Oldroyd-B liquid
brought about by evaporation. The effect of flowing down an inclined plane. They formulated
inertia for larger Reynolds numbers in the film the stability problem in terms ofa spatial instability
flow has been studied by Prokopiou, Cheng and and solved the linear disturbance equations
Chang (1991) using an integral technique to numerically. This work showed that the viscoelastic
develop the evolution equation and to study its nature of the liquid does reduce the critical
nonlinear behavior. This work has shown that Reynolds number, as seen in earlier studies.
inertia causes the disturbances in the liquid film However, for Reynolds numbers greater than
to rapidly evolve into finite-amplitude solitary about 7, viscoelastic effects actually reduce the
waves. A thin-film evolution equation for the maximum growth rate in the system. This implies
flow on the outside of a circular cylinder with that the viscoelastic destabilization of the film
strong surface tension effects has been derived may not be observable.
by Frenkel (1993). A numerical study of this The pure viscoelastic mode of instability seen
equation for an axi-symmetric liquid film coating in a film with no inertial effects (zero Reynolds
a vertical fiber has been done by Kerchman and number) has a growth rate an order of magnitude
Frenkel (1994) and Kalliadasis and Chang (1994). smaller than that of the inertial instability. This
This work showed that if the film thickness is suggests that a pure viscoelastic instability may
large enough the film will evolve to form a drop never be observed in an actual coating device
on the fiber, although the mechanism through because there may not be enough time for a
which this is achieved is different in each of these disturbance to grow to a discernible size. The
papers. This result should be of direct interest review article by Larson (1992) contains a further
to coating engineers involved in the coating of discussion of the work on the instability of
wires and fibers. viscoelastic liquids in single-layer and multi-layered
The effect of equipment vibration on the film flows.
growth of surface waves propagating on a liquid Lastly, the study of surfactants and how they
film has been studied by Weinstein, Baumlin and alter the stability characteristics of a liquid film
Servant (1993). The geometry they considered flow is of direct interest to coating engineers.
was a single liquid layer flowing down an inclined Surfactants affect the inertial instability in the
solid wall from an upstream entrance port. They layer directly and they introduce new instability
showed mathematically that in-plane oscillations mechanisms of their own. The effects ofsurfactants
on the flows found in coating processes are A.~){(U~) + v~»)[1 - (11~»)2] + 211~)(V~) - u~»)}
discussed in detail in Chapter lId.
= A.~+l){(U~+l) + v~+l»)[1 - (11~»)2]
8.6.2 A MULTI-LAYERED LIQUID FILM + 211~)(V~+1) - U~+l))} (8.58d)
FLOWING DOWN AN INCLINED PLANE
Here, for each interface we have N(i) =
The production of photographic film involves {I + (11~)f}1/2,thecurvatureK(i) = -11~~(N(i))-3,
the simultaneous application of many different and . the capillary number Cali) = J1. s
liquid layers onto a substrate. One process that Us/all), for which J1. s and Us are a reference
can accomplish this task is slide coating (see viscosity and velocity respectively and a(i) is the
Chapter 11 b). The flows occurring in this process surface tension of the i-th interface. The dynamic
are very complex, but in several regions there is viscosity ratio for each layer is A.~) = J1.(i)/J1.s'
a rectilinear flow of a multi-layered liquid film where J1.(i) is the dynamic viscosity of the i-th
on a rigid, inclined plane surface. As the uniformity layer. Finally, we pose a kinematic condition for
of the final coating depends to some degree on the motion of this interface
the stability of the flow in these regions, the (5.58e)
question of the stability of multi-layered film
flows must be addressed. Given this set of governing equations and
A linear stability analysis of N liquid layers boundary conditions, one proceeds just as in the
flowing down a rigid, inclined plane is the same case of a single-layer film flow. A basic state of
in principle as that of a single liquid layer. The steady, rectilinear motion is found, it is disturbed
mathematical model is formulated by writing the by an infinitesimal, normal-mode perturbation,
Navier-Stokes and continuity equations for each and a set of linear disturbance equations is
individual layer. The boundary conditions for derived. These equations are solved either
the rigid plane and for the free surface of the film numerically for arbitrary wave numbers, or
are the same as in the single-layer case. At the approximately in the limit of small wave numbers
interface between any two adjacent layers, we using the regular perturbation method. There
pose a total of five continuity conditions. In the are no mathematical problems associated with
following, we present only these interfacial this analysis, but there are practical difficulties.
continuity conditions as they are the major For N liquid layers, the normal-mode stability
difference between the single-layer and the equations form a system of 4N ordinary differential
multi-layered film flow models. equations. This increases the complexity of the
Let the interface between layer i and layer numerical solution and the algebra associated
i + 1 be given by y = ll(i)(X,t). For each of the with the perturbation technique.
N-l interfaces, we impose the condition of We expect the same types of behavior in a
continuity of velocity across the interface multi-layered film flow as we saw in the single-
U(i) = U(i+1), vIi) = V(i+1) (8.58a,b) layer film flow. There is still a shear mode of
instability that occurs for large values of the
where the superscript (i) refers to the individual Reynolds number. For lower Reynolds numbers,
liquid layers. We also impose the normal- there is a separate mode of instability associated
and tangential-stress balances with each of the N-l interfaces and one with the
- p(i) + 2A.~) {v~) - l1~)(u~) + v~») free surface. The motion in each liquid layer and
+ (11~»)2u~)}(N(i»)-2 the mechanism of the instability for the long-
wave interfacial and free-surface modes can be
= _p(i+1) + 2A.~+1){V~+1) -l1~)(U~+l) described as we did for the single-layer film flow.
+ v~+ 1») + (11~»)2U~+ l)}(N(i»)- 2 _ (Cd i »)-l K(i) However, this description is more complex for
the multi-layered film because the motion of the
(8.58c) interfaces and the free surface are all coupled
together. This increases the difficulty ofdescribing viscous liquid is next to the wall. When the
the force balances at work in the unstable motion Reynolds number is zero so that inertia is not
of the film. important and there is no surface tension the
A number of authors have examined the linear interfacial mode is always unstable. The unstable
stability of a multi-layered film flow down an disturbances have a long-wave character when
inclined plane. The effect of density stratification the upper-layer thickness is more than about
in two inclined liquid layers with different twice the thickness of the lower layer. When the
thicknesses was studied by Kao (1965a, 1965b) upper layer is thinner than this, the unstable
using long-wave asymptotics. He showed that disturbances have a short-wave character. In-
the free-surface mode is more stable when the creasing the surface tension of the free surface
upper liquid is lighter than the lower liquid. This and of the interface stabilizes these short waves.
effect is more pronounced as the upper layer gets If the surface tension of the interface is large
thicker. When the upper liquid is heavier than enough the short-wave instability is completely
the lower liquid, the free-surface mode is more eliminated, as expected. Adding inertia to the
unstable, but the effect does not depend all that system stabilizes the long-wavelength disturbances
much on the relative thicknesses of the two so that they are completely stable for large
layers. The interfacial mode that appears in the enough Reynolds numbers. Short waves are
two-layer system is always stable when the upper destabilized by inertia but stabilized by surface
liquid is the heavier liquid. When the lower liquid tension. For this case when the less viscous liquid
is the heavier liquid but the two densities are not is next to the wall, it is not possible to have both
too different, the interfacial mode is always the interfacial mode and the free-surface mode
unstable. In this case, the free-surface mode could stable at the same time.
be stable if the Reynolds number is smaller than When the more viscous liquid is next to the
the critical value, and so the film flow would be wall and inertia is negligible, the interfacial mode
dominated by the instability of the interfacial is always stable to long waves and neutrally
mode. If the lower liquid is much heavier than stable to very short waves. Any amount of inertia
the upper liquid, the interfacial mode is stable in the system destabilizes all of the disturbances.
for Reynolds numbers less than some critical Thus, long waves are stable for Reynolds numbers
value. If the upper layer is thin enough, the less than a critical value, and short waves are
critical Reynolds number for the free-surface always unstable. Adding enough surface tension
mode is less than the one for the interfacial mode to the interface completely stabilizes the short
and the instability of the free-surface mode waves so that all disturbances are stable for
dominates the system. Reynolds numbers less than a critical value. The
The effect of density and viscosity stratification critical wave number in this case is nonzero and
and of different layer thicknesses in a two-layer may not be small.
inclined film flow was studied by Kao (1968) in A three-layer inclined film flow was considered
the long-wave limit. A two-layer flow in which by Akhtaruzzaman, Wang and Lin (1978), Wang,
the liquids had different thicknesses and viscosities Seaborg and Lin (1978), and Weinstein and Kurz
but the same density was examined more recently (1991). These authors gave results for the effects
by Chen (1993), who did numerical computations ofdensity and viscosity stratification and varying
for all wavenumbers. His results showed that the layer thicknesses. In contrast to the single-layer
stratification does not alter the character of the and the two-layer problems, the instability of a
free-surface mode. It is still a long-wave instability three-layer film flow can arise both at leading
with a critical Reynolds number above which order and at first order in the expansions for the
the mode is unstable. However, the character of growth rate.
the interfacial mode reacts strongly to the viscosity The complex rheology of modern coating
stratification. Consider the case in which the less liquids has prompted investigators to examine
non-Newtonian liquids in multilayered film flows. influence of a small amount of inertia in the liquid
Weinstein (1990) considered a two-layer inclined and the small curvature of the roll on the loading
film flow with shear-thinning liquids using the of the knife. The singular perturbation method
Carreau constitutive relation for the viscosities. has been used successfully in computing the
He showed that the free-surface mode ofinstability viscous bending of the free surface that occurs
behaves as if it sees some kind of an average very near to a moving contact line (see Chapter
viscosity in the film except at very small disturbance 3 for a discussion ofthis dynamic wetting problem).
frequencies. This is expected given the mechanism Domain perturbation and asymptotic techniques
for this mode of instability as discussed above have been used to study the spreading of a
in the context of a single-layer film flow. In droplet on a solid surface by Hocking (1983) and
contrast, the interfacial mode depends strongly to study the effects of insoluble surfactants in dip
on the local viscosities near the interface. It is coating by Park (1991). And of course, linear
possible to either stabilize or destabilize this stability theory has been applied to many coating
interfacial mode through alterations of the flows of interest (see Chapters 9, lt b, ltd, and
rheology of the two shear thinning liquids in the 12a for some more examples).
film. See the review by Larson (1992) for a more Lubrication theory is probably one of the
complete discussion of the instabilities seen in oldest methods of approximation used in fluid
multilayered inclined film flows of viscoelastic mechanics. It is often employed to study the flow
liquids. in thin liquid films in which the thickness of the
film varies slowly in the flow direction. In these
situations, the flow in the film is quasi-one-
8.7 CONCLUSION
dimensional. The original formulation of lubri-
Asymptotic methods are powerful tools that can cation theory by Reynolds (1886) was nothing
be used to obtain approximate solutions to more than a simplification of the momentum
fluid-flow problems of direct interest to coating equations based on the quasi-one-dimensional
engineers. These methods should be considered nature of the flow. We have seen in this chapter
whenever a small parameter exists in the problem, that lubrication theory can be formally derived
or when there are two or more length or time as the leading-order problem in a slowly varying
scales for the flow, one of which is much smaller domain perturbation analysis of the flow in a
than the others. An asymptotic method is successful thin liquid film.
when it generates a simplified problem that is Standard lubrication theory is very successful
solvable, either mathematically or numerically. in the knife coating of a viscous liquid with no
The resulting approximation to the exact solution significant viscoelastic effects. The most important
is useful because it displays the primary balance characteristics of the system, the flow and the
among the dominant forces in the problem. pressure in the gap beneath the knife and the
Asymptotic approximations can also be used to normal lifting force on the knife, are all computed
augment numerical computations by simulating easily and accurately. The obvious deficiency of
boundary conditions at infinity, and by repre- the theory is that the two-dimensional flows in
senting the flow field in critical regions that the small inlet and outlet regions to the gap under
would have otherwise forced an expensive increase the knife are not well characterized. However,
in the resolution of the numerical method. asymptotic techniques show that the effect of
Besides the examples discussed in this chapter, these flows on the lubrication flow through the
asymptotic methods have been used in many gap takes the form oftwo zero-pressure boundary
other coating flow problems. Hsu et al. (1985) conditions imposed at the ends of the flow
studied a coating process in which a knife was passage. These boundary conditions successfully
placed upon a large rotating roll. They used replace a more detailed knowledge of the two-
regular perturbation techniques to examine the dimensional flows in the inlet and outlet regions
Conclusion 293
of the gap. This implies that the details of these of these is the simple zero-pressure condition.
flows are relatively unimportant to the overall The other specifies the flow rate through the nip
behavior of the system. region and it depends on an accurate calculation
The application of standard lubrication theory of the flow in the inner region about the film-
to roll coating is not so successful. The flow in splitting point. The ad hoc boundary conditions
the region of the nip of two large rolls is quasi- used above do not approximate this flow-rate
one-dimensional and it is accurately described condition very well and so the results of the ad
by the Reynolds equation. However, it is not hoc models are unacceptable.
clear how to formulate the boundary conditions Rising and falling liquid films appear in dip
for the lubrication flow at the film-splitting point. coating and in other coating processes where a
Not only is the flow in this region two dimensional, wetted roll or surface moves into or out of a
but the location of this region is also unknown. nearly static liquid bath. The free surface of the
Various ad hoc boundary conditions based on a liquid film far enough above the surface of the
number of different physical arguments have bath is parallel to the underlying moving surface,
been proposed and used to create lubrication and it is of interest to determine how the
models for roll coating. None of these models thickness of the film varies as it joins to the
accurately predict the location of the film-splitting surface of the bath. For thin liquid films, these
point or the flow rate past the rolls as a function variations in the film thickness typically occur
of the capillary number. In hindsight, one should over a length scale that is much larger than the
have expected such difficulties because the average film thickness. Thus, the film profile can
two-dimensional flow near the film-splitting point be computed using a slowly varying domain
is a critical influence on the location of this point. perturbation technique. The result of such an
It would be surprising if these ad hoc boundary analysis is a third-order, asymptotically-consistent,
conditions actually did contain enough informa- nonlinear differential equation describing the
tion about this flow to allow accurate predictions. profile of the film. Because some of the normally
The asymptotic resolution to this dilemma is neglected small terms in the curvature of the free
the realization that roll coating must be solved surface were kept during the development of this
as a singular perturbation problem. The two- equation, its solution is a uniformly valid
dimensional flow at the film-splitting point is the approximation of the film thickness from far
inner or interior-layer problem, and it is found above the bath all the way to and including the
using numerical methods. The quasi-one-dimen- free surface of the static bath itself. The numerical
sional flow in the nip region of the rolls is an solution of this equation does a good job in
outer flow, and it is described using lubrication describing the film profile for both rising and
theory. The rigid-body rotation of the film after falling liquid films. This calculation shows that
the film-splitting point is a second outer flow, when the upward velocity of the underlying
but with an exact solution. The flows in these surface is larger than some critical value, the film
three regions are matched together to form a profile will join to the static bath in a smooth
uniformly valid approximation to the flow over monotonic way. When the upward velocity is
the entire domain. One interesting feature of the less than this critical value, the profile develops
inner problem is that the matching to the outer a series of waves as it joins to the bath. This
flows must be done first in order to supply the wave-like structure occurs for all falling films (a
appropriate boundary conditions for the numerical falling film has a net flow rate into the bath),
calculation of the two-dimensional flow in the film- and even for some rising films.
splitting region. One set of results from this The last problem considered in this chapter
analysis is a pair ofeffective boundary conditions was the linear stability analysis of one or more
for the outer lubrication flow in the nip region liquid layers flowing down a rigid inclined plane.
that are applied at the film-splitting point. One Linear stability theory is essentially the leading-
order problem from a regular perturbation analysis to the film. In a multilayered film flow, there is
based on the small amplitude of a disturbance a separate long-wave mode of instability for the
to a known basic state. The normal difficulty free surface and for each of the interfaces in the
associated with the unknown position of the film. The property stratification occurring among
interfaces of the liquid layers is removed in this the different liquid layers influences the free-surface
stability analysis by using the small-amplitude mode to some degree, but the essential character
domain perturbation technique. The interfacial ofthis mode is still the same as in the single-layer
boundary conditions are then applied at the film flow. However, this same stratification is the
known basic-state positions and the small ultimate cause ofthe interfacial modes of instability.
deviation of the interfaces from these positions Each interfacial mode is influenced very strongly
become additional dependent variables in the by the fluid property discontinuities occurring
problem. at that interface. Some specific examples of this
The linear stability of a single liquid layer complex phenomena for a two-layer film flow
flowing down an inclined plane was discussed in with density and viscosity stratification were
detail. The appropriate linear disturbance equa- discussed.
tions have been solved numerically for arbitrary The coating flows discussed in this chapter
wave numbers. The results show that there is a were chosen to display how asymptotic techniques
shear-mode of instability occurring for large can be used to provide accurate approximations
values of the Reynolds number, and a long-wave and useful information about a flow without
mode that appears for relatively small values of resorting to full numerical calculations. These
the Reynolds number except when the inclined techniques can also provide boundary conditions
plane is almost horizontal. Thus, in most coating for a numerical computation on those surfaces
flow applications, the long-wave mode would where the flow must join to another flow that is
occur before the shear mode. The long-wave well-characterized mathematically. The most
mode was then examined using a regular per- critical part of these asymptotic analyses is the
turbation expansion of the disturbance equations scaling of the governing equations. It is here that
for small wavenumbers. There are two advantages the dominant balances and the required small
of such an analysis over a numerical calculation parameters are introduced into the problem.
of the disturbance equations. First, it yields Once this is done, the actual application of the
algebraic expressions for the stability character- asymptotic method is more or less routine. The
istics of the layer that are easily computed. art of scaling is where the physical insight of the
Secondly, the mechanism of the instability can coating engineer into the nature of the coating
be clearly described by examining the force and flow under investigation enters the analysis. This
mass balances represented by the governing is where the engineer will have the greatest
equations at each order in the expansion. Current influence on both the success and the utility of
research on single-layer inclined film flows involves the analysis.
the study of the nonlinear response of the film
and the influence of additional effects of interest
to coating engineers and others, such as heating,
ACKNOWLEDGMENTS
surfactants and non-Newtonian rheology.
The linear stability analysis of a multilayered I would like to thank the editors Stephan F.
film flow is not fundamentally different from that Kistler and Peter M. Schweizer, and also Haroon
of a single liquid layer. The same linear analysis S. Kheshgi for their careful editorial work on
and small-amplitude domain perturbation tech- this chapter. The chapter is better because of it.
niques are used. The problem gets more compli- I would also like to thank my colleagues Stephen
cated of course, both in terms of its analysis and H. Davis, Prateen V. Desai and Cyrus K. Aidun
in terms of its response, as more layers are added for their useful comments on a draft of this chapter.
References 295
REFERENCES Hinch, E. J. 1991. Perturbation Methods. Cambridge:
Cambridge University Press.
Akhtaruzzaman, A. F. M., Wang, C. K. and Lin, S. Hocking, L. M. 1983. The spreading of a thin drop
P. 1978. Wave motion in multilayered liquid films. by gravity and capillarity. Q. J. Mech. Appl. Math.
J. Appl. Mech. 45: 25-31. 36: 55-69.
Benney, D. J. 1966. Long waves on liquid films. J. Hsu, T. c., Malone, M., Laurence, R. L. and Middleman,
Math. Phys. 45: 150-155. S. 1985. Separating forces in blade coating of viscous
Canuto, c., Hussaini, M. Y, Quarteroni, A. and Zang, and viscoelastic liquids. J. Non-Newtonian Fluid
T. A. 1988. Spectral Methods in Fluid Dynamics. Mech. 18: 273-294.
New York: Springer-Verlag. Joo, S. W. and Davis, S. H. 1992. Instabilities of
Chandrasekhar, S. [1961] 1981. Hydrodynamic and three-dimensional viscous falling films. J. Fluid
Hydromagnetic Stability. Reprint. New York: Dover Mech. 242: 529-547.
Publications. Joo, S. w., Davis, S. H. and Bankoff, S. G. 1991.
Chang, H.-C. 1989. Onset of nonlinear waves on falling Long-wave instabilities of heated falling films: two-
films. Phys. Fluids A. 1: 1314-1327. dimensional theory of uniform layers. J. Fluid Mech.
Chen, K. 1993. Wave formation in the gravity-driven 230: 117-146.
low-Reynolds number flow of two liquid films down Joseph, D. D. 1976. Stability of Fluid Motions. Berlin:
an inclined plane. Phys. Fluids A. 5: 3038-3048. Springer Verlag.
Cook, R. A. and Clark, R. H. 1973. An analysis of the Kalliadasis, S. and Chang, H. -Co 1994. Drop formation
stagnant band on falling liquid films. Ind. Eng. during coating of vertical fibers. J. Fluid Mech. 261:
Chem. Fundam. 12: 106-114. 135-168.
Coyle, D. J., Macosko, C. W. and Scriven, L. E. 1986. Kao, T. W. 1965a. Stability of two-layer viscous
Film-splitting flows in forward roll coating. J. Fluid stratified flow down an inclined plane. Phys. Fluids.
Mech. 171: 183-207. 8: 812-820.
De Bruin, G. J. 1974. Stability of a layer of liquid Kao, T. W. 1965b. Role of the interface in the stability
flowing down an inclined plane. J. Eng. Math. 8: of stratified flow down an inclined plane. Phys.
259-270. Fluids. 8: 2190-2194.
Dowson, D. and Taylor, C. M. 1979. Cavitation in Kao, T. W. 1968. Role of viscosity stratification in the
bearings. Ann. Rev. Fluid Mech. 11: 35-66. stability of two-layer flow down an incline. J. Fluid
Drazin, P. G. and Reid, W. H. 1981. Hydrodynamic M echo 33: 561-572.
Stability. Cambridge: Cambridge University Press. Kerchman, V. I. and Frenkel, A. L. 1994. Interactions
Floryan, J. M., Davis, S. H. and Kelly, R. E. 1987. of coherent structures in a film flow: Simulations
Instabilities of a liquid film flowing down a slightly of a highly nonlinear evolution equation. Theoret.
inclined plane. Phys. Fluids. 30: 983-989. Comput. Fluid Dynamics. 6: 235-254.
Frenkel, A. L. 1993. On evolution equations for thin Kheshgi, H. S. 1989. Profile equations for film flows
films flowing down solid surfaces. Phys. Fluids A. at moderate Reynolds numbers. AIChE J. 35:
5: 2342-2347. 1719-1727.
Gjevik, B. 1970. Occurrence offinite-amplitude surface Kheshgi, H. S., Kistler, S. F. and Scriven, L. E. 1992.
waves on falling liquid films. Phys. Fluids. 13: Rising and falling film flows: viewed from a first-
1918-1925. order approximation. Chem. Eng. Sci. 47: 683-694.
Hansen, E. B. 1987. Stokes flow down a wall into an Lamb, H. [1932] 1945. Hydrodynamics. Reprint. New
infinite pool. J. Fluid Mech. 178: 243-256. York: Dover Publications.
Higgins, B. G. 1982. Downstream development of Landau, L. D. and Levich, V. G. 1942. Dragging of a
two-dimensional viscocapillary film flow. Ind. Eng. liquid by a moving plate. Acta Physicochimica
Chem. Fundam. 21: 168-173. URSS. 17: 42-54.
Higgins, B. G. and Scriven, L. E. 1979. Interfacial Larson, R. G. 1992. Instabilities in viscoelastic flows.
shape and evolution equations for liquid films and Rheol. Acta. 31: 213-263.
other viscocapillary flows. Ind. Eng. Chem. Fundam. Levich, V. G. 1962. Physicochemical Hydrodynamics.
18: 208-215. Englewood Cliffs, NJ: Prentice-Hall.
Higgins, B. G., Silliman, W. 1., Brown, R. A. and Lin, S. P. 1969. Finite-amplitude stability of a parallel
Scriven, L. E. 1977. Theory of meniscus shape in flow with a free surface. J. Fluid Mech. 36: 113-126.
film flows. A synthesis. Ind. Eng. Chem. Fundam. 16: Nayfeh, A. H. 1973. Perturbation Methods. New York:
393-400. John Wiley.
Park, c.-W. 1991. Effects of insoluble surfactants on G. H. 1989. The stability ofgravity driven viscoelastic
dip coating. J. Colloid Interface Sci. 146: 382-394. film-flow at low to moderate Reynolds number. J.
Prokopiou, T., Cheng, M. and Chang, H.-c. 1991. Non-Newtonian Fluid Mech. 31: 87-113.
Long waves on inclined films at high Reynolds Smith, M. K. 1990. The mechanism for the long-wave
number. J. Fluid Mech. 222: 665-691. instability in thin liquid films. J. Fluid Mech. 217:
Reynolds, O. 1886. On the theory of lubrication and 469-485.
its application to Mr. Beauchamp Tower's experi- Taylor, G. J. 1963. Cavitation of a viscous fluid in
ments, including an experimental determination of narrow passages. J. Fluid Mech. 16: 595-619.
the viscosity of olive oil. Phil. Trans. Roy. Soc. Van Dyke, M. 1975. Perturbation Methods in Fluid
London. 177: 157-234. Mechanics. Stanford, CA: The Parabolic Press.
Ruschak, K. J. 1976. Limiting flow in a pre-metered Wang, C. K., Seaborg, J. J. and Lin, S. P. 1978.
coating device. Chern. Eng. Sci. 31: 1057-1060. Instability of multi-layered liquid films. Phys. Fluids.
Ruschak, K. 1. 1978. Flow of a falling film into a pool. 21: 1669-1673.
AIChE J. 24: 705-709. Weinstein, S. J. 1990. Wave propagation in the flow
Ruschak, K. 1. 1982. Boundary conditions at a liquid/air of shear-thinning fluids down an incline. AIChE J.
interface in lubrication flows. J. Fluid Mech. 119: 36: 1873-1889.
107-120. Weinstein, S. J., Baumlin, J.-M. and Servant, J. 1993.
Ruschak, K. J. 1985. Coating flows. Ann. Rev. Fluid The propagation of surface waves in flow down an
Mech. 17: 65-89. oscillating inclined plane. AIChE J. 39: 1113-·1123.
Savage, M. D. 1977. Cavitation in lubrication. Part 1. Weinstein, S. J. and Kurz, M. R. 1991. Long-
On boundary conditions and cavity-fluid interfaces. wavelength instabilities in three-layer flow down an
J. Fluid Mech. 80: 743-755. incline. Phys. Fluids A. 3: 2680-2687.
Savage, M. D. 1982. Mathematical models for coating Wilson, S. D. R. 1982. The drag-out problem in film
processes. J. Fluid Mech. 117: 443-455. coating theory. J. Eng. Math. 16: 209-221.
Scott, M. R. and Watts, H. A. 1977. SUPORT - A Wilson, S. D. R. and Jones, A. F. 1983. The entry of
computer code for two-point boundary-value prob- a falling film into a pool and the air-entrainment
lems via orthonormalization. SIAM J. Nurn. Analy. problem. J. Fluid Mech. 128: 219-230.
14: 40-70. Yih, c.-S. 1963. Stability of liquid flow down an
Shaqfeh, E. S. G., Larson, R. G. and Fredrickson, inclined plane. Phys. Fluids. 6: 321-334.
ADVANCES IN COMPUTATIONAL 9
METHODS FOR FREE-SURFACE FLOWS
Kostas N. Christodoulou, Stephan F. Kistler and
P. Randall Schunk
9.1 INTRODUCTION
catastrophic hydrodynamic instabilities. It is
As in many areas of science and technology, often aggravated by edge effects, and might be
computational mathematics performed on high- triggered by large transients of the sort encountered
speed, large-memory computers has become an during start-up or splice passage. Catastrophic
essential ingredient of modern coating process coating failure may also stem from the inability
research and development. Traditionally, most of the coating liquid to displace sufficient air
advances in the manufacture of imaging films, previously in contact with the dry substrate.
magnetic storage and many other precision In the quest to identify the mechanisms by
coatings have been based on extensive experi- which defects form or coating processes fail,
mentation. A predominantly empirical approach scientifically designed experiments are indispens-
was sufficient to achieve a remarkable level of able, for they are the ultimate test for any
technological perfection. Nowadays, however, hypothesis or theory (cf Chapter 7). Experiments
rapidly intensifying competitive pressures demand alone, however, often fail to elucidate fully cause-
ever increasing coating speeds, thinner and more and-effect relationships. They can also be very
uniform coatings (often multilayered), reduced costly. Conventional methods of mathematical
defect levels and diminished waste. Further analysis can furnish valuable insights into essential
refinements in coating technology become in- mechanisms, but are usually restricted to idealized
creasingly intricate and put a premium on flow configurations and limited parameter ranges
analyzing the physical mechanisms that determine (cf Chapter 8). Practically relevant analyses of
success or failure of a coating process. Imperfec- realistic coating processes often require advanced
tions in a coated layer may arise from a wide numerical methods that solve the full set of
variety ofmechanisms, including inhomogeneities Navier-Stokes equations and associated bound-
or foreign matter in the coating solutions, ary conditions for complicated free-surface flows.
disturbances induced by equipment vibration or State-of-the-art computer-aided analysis can
uncontrolled air flow, various wetting phenomena, explain many hydrodynamic mechanisms that
and also flow instabilities that grow in the control the outcome ofcoating processes; suggest
coating device but partially decay further down- design of process and product for improved
stream. Complete coating failure, on the other coatability; visualize details of the flow field that
hand, almost exclusively arises from are inaccessible to the experimentalist due to
London. ISBN 0 412 06481 2.
298 Advances in computational methods for free-surface flows
geometric constraints or non-transparent liquids; the challenges listed above face serious math-
evaluate many more configurations than could ematical difficulties. Several are inherent in solving
realistically be tested in experiments, thereby the Navier-Stokes system, even on fixed domains:
saving cost and time; allow systematic optimiz- the incompressibility constraint couples the
ation, thereby further narrowing the choice of velocity components among themselves and with
prototypes that need to be built; and permit pressure, thereby precluding the use of purely
exploration of parameter ranges that may be explicit time-marching procedures; the viscous
inaccessible with existing pilot equipment. term has a broad eigenvalue spectrum that
The objective of the present chapter is to give makes the equation system stiff, and hence adds
a comprehensive overview of computational to the need for implicit treatment; the advection
methods suitable for the analysis of viscous term makes the equations nonlinear; additional
free-surface flow of the sort found in industrial nonlinearities - and even changes in type of the
coating processes. The chapter focuses on four governing equations - arise when the coating
key challenges in computational flow mechanics liquid is non-Newtonian. Further difficulties arise
which, by and large, have been met with algorithms from the free surfaces of liquid against air,
available today. The first is to identify and interfaces between immiscible liquids, as well as
calculate possible film-forming flows that produce deformable boundaries such as tensioned webs,
uniform coatings. The second is to assess the flexible blades, or rubber rolls. The location of
stability of the predicted flows to small disturb- all these features is unknown a priori, making
ances in order to establish parameter windows the equation system severely nonlinear, even
ofsuccessful coater operation. The third challenge when inertia is negligible, the liquid is Newtonian,
is to compute the sensitivity of stable flow states and the deformable materials are Hookean solids.
to small upsets in order to identify parameter In addition, their shape is rarely a small
ranges within the coating window in which perturbation from a configuration that fits a
coating performance will be optimal. And the standard coordinate system. In fact, coating flow
fourth is to predict how fast or whether at all geometries can undergo dramatic variations as
the desired flow states recover from large upsets flow parameters vary or time evolves, including
of the sort induced during start-up or splice formation of cusp-like corners in the interface
passage. A further challenge is to predict whether and even changes in topology. Additional diffi-
the calculated flows displace sufficient amounts culties, both physical and mathematical, stem
ofthe air previously in contact with the substrate. from static contact lines where the free surfaces
This challenge remains unanswered to date because separate from solid coating devices, and dynamic
ofincomplete understanding ofthe wetting physics. wetting lines where the coating liquid appears
Another challenge is to make sure that successful to advance over a previously dry substrate.
operation of a steady, uniform coating flow is Contact angle hysteresis and sharp edges can
not fouled by the inability of the edge regions to inhibit the migration of static contact lines, and
survive satisfactorily at the desired conditions; thereby allow multiple steady flow states. Con-
this also has received insufficient attention to ventional continuum theory breaks down alto-
date. Last but not least, a final challenge is to gether at dynamic contact lines. Ad hoc boundary
extend computational coating-flow models to conditions are frequently used to substitute for
include mechanisms other than those accounted a refined physical model of the wetting dynamics,
for by the standard Navier-Stokes system, namely but compromise the predictive capabilities of
viscoelastic stresses from non-Newtonian behavior, even the most sophisticated coating-flow models
surface-tension gradients due to surfactants and/or advanced to date. Further complications arise
mass transfer, and elastic deformations of solid from the extremely wide ranges in deformation
members confining the flow. rates encountered in coating flows, especially
Computational methods that seek to address close to the apparent singularities associated
Introduction 299
with contact lines. They tax the capabilities of years ago, were locked away from analysis -
the best rheological instruments available to especially their stability to small transitory
date, and are sure to put to test the most disturbances, their operability in the space of
sophisticated constitutive models and algorithms parameters, and their sensitivity to ongoing
for viscoelastic flows. Polymeric and surface-active disturbances. Many of the advances have relied
additives that are commonly employed as coating on finite-element techniques (Strang and Fix
aids present further challenges in computational 1973). The governing equations are discretized
analyses, for they can give rise to steep, boundary- on boundary-confining meshes that adapt to
layer-like concentration gradients. In addition, changes in flow geometry independently of the
their equilibrium and transport properties are underlying liquid flow. Such a mesh movement
often unknown. strategy accurately tracks irregular flow domains
To date, no algorithm has been advanced that in both steady-state free-boundary problems and
successfully copes, in a robust manner, with all transient moving-boundary problems. Together
the difficulties inherent in coating-flow analysis. with finite-element discretization, boundary-
Thus many problems posed by practicing coating conforming meshes also simplify considerably
engineers are still far from routine solution. the imposition offree-surface boundary conditions.
Nonetheless, significant advances in computational The strategy is sometimes called arbitrary
mathematics and computer hardware have made Lagrangian-Eulerian (ALE) because the mesh
accessible to rigorous theory many practically is neither tied to a fixed Eulerian frame of
relevant coating flows which, ten or even five reference, as in one special class of numerical
PERFECTLY UNIFORM
PRODUCT
Steady 2-d Flow
EVOLUTION OF 2-D UPSETS BEAD BREAK-UP, FLOODING

ONSET OF RIBBING
START-UP, SPLICE PASSAGE CROSS-WEB BARRING
Stability w.r.1. 2-d Stability w.r.1. 3-d
Unsteady 2-d Flow Disturbances
Disturbances
OPTIMIZATION
Process Design Parameters
Product Design Parameters - COATING WINDOWS
Structure of State Space
Tracking of Critical Points
-
OPTIMIZATION SENSITIVITY TO
ONGOING DISTURBANCES
Operating Parameters 3-d Frequency Response
(Controls)
Quality window
Figure 9.1 Overview of finite-element analysis of coating flows.

methods for free-surface flow analysis (cf Section most precision coating configurations, the desired
9.4.1), nor does the mesh movement follow the coating flows are uniform across the width of
fluid motion or the Lagrangian frame of reference the coater except for unavoidable edge regions.
associated with it, as in another class of methods For that reason, Sections 9.2.1 through 9.2.4
(cf Section 9.4.2). focus primarily on two-dimensional flows. Most
ALE-type finite-element methods are the main of the computational techniques presented,
emphasis of the present chapter. Section 9.2 however, can be generalized to three-dimensional
describes all the basic steps that make up a flows - at least in principle - and selected
modem finite-element algorithm for coating flows comments are interjected on how to do so.
(for an overview, see Fig. 9.1). An earlier review The second challenge - a full stability analysis
by Kistler and Scriven (1983) summarized the of steady, two-dimensional flows - is also within
state-of-the-art as it stood more than ten years the reach of available, albeit more sophisticated
ago, focusing on robust 'spine' parameterization finite-element methods. Section 9.2.5 describes
of free surfaces and interfaces. That review also methods that predict the growth or decay of
discussed the efficient coupling of flow and two-dimensional disturbances as a flow parameter
free-boundary computations via Newton's method. varies. The methods can detect the onset of
The present review provides an update, testifying instabilities that produce cross-web patterns on
to the considerable progress that has been made coated layers, such as coating bead break-up,
since. Section 9.2.1 briefly recounts the governing 'barring' or 'chatter', and bead flooding. Section
equations and boundary conditions, and then 9.2.6 reviews methods that monitor the fate of
describes their finite-element discretization by three-dimensional disturbances as a coating
numerically stable combinations of the approxi- condition is varied and cross-web wavelength is
mation spaces. Section 9.2.2 reviews generalized scanned. These methods can predict the onset of
algebraic and elliptic mesh generation schemes instabilities such as ribbing, bead break-up into
that endow state-of-the-art finite-element algo- rivulets and diagonal chatter or 'chevrons'. In
rithms with the flexibility to accommodate both types of analysis, the main problem boils
complicated free-surface flow geometries and down to solving a large, sparse, generalized and
automatically adapt to significant changes in singular eigenproblem for at least the leading
flow geometry. Section 9.2.3 revisits the advantages eigenvalues and corresponding models that
of linearizing the resulting set of discretized dominate flow stability. Section 9.2.7 explains
equations by Newton-like methods. It also methods that accurately compute a critical point
comments on the effectiveness of recently devel- detected by linear stability analysis (i.e., the
oped iterative matrix solvers for the resulting critical parameter and critical state for two-
system of sparse linear equations. Section 9.2.4 dimensional disturbances and, in addition, the
discusses computational strategies for initial critical disturbance wavenumber for three-di-
start-up when no prior solution is available, and mensional disturbances), and systematically trace
continuation in one or several parameters for its path (i.e., critical boundary) as a second para-
advancing from the very first start-up solution meter varies in order to map out the operability
to realistic solutions that are of practical interest. diagram of the process. The same section also
Systematic continuation may also lead to the presents a novel extension of Abbott's (1978)
discovery of limit and bifurcation points which method for accurately computing the critical
can provide preliminary insights into the stability conditions at which the flow loses translational
of the computed solution branches. Altogether, invariance to the ribbing instability. Operability
Sections 9.2.1 through 9.2.4 describe the tools analysis is useful to determine optimal process
necessary to meet the first challenge in computa- designs that give maximum operating latitude.
tional coating flow mechanics - the calculation The third challenge - investigating the sensitivity
of steady flows that yield uniform coatings. In of two-dimensional coating flows to omnipresent
Finite-element methods 301
disturbances within the window of stable coater undergo extremely large distortions - including
operation - can be addressed with computer-aided changes of topology as may be associated with
methods of frequency response analysis. Section the break-up of a liquid layer or jet into rivulets
9.2.8 describes such methods for ongoing two- or droplets - is the forte offully explicit free-surface
dimensional disturbances; they are rather straight- computations on fixed Eulerian grids (Section
forward extensions of steady-state finite-element 9.4.1), and also Lagrangian rezoning methods,
methods. Section 9.2.8 also introduces a recent especially what are called free-Lagrangian tech-
extension that handles three-dimensional dis- niques (Section 9.4.2). Such algorithms may be
turbances of a single cross-web wavenumber. of particular value when only limited a priori
Sensitivity analysis is a powerful tool to identify knowledge of the free-surface flow configuration
the subregion of optimal coating operating is available and intelligent design of a boundary-
conditions within the coating window in which conforming mesh is impossible. It is tempting,
steady-states are the least sensitive to disturbances furthermore, to invoke various transient algo-
and therefore produce the highest-quality coating. rithms to perform ad hoc stability analyses of
The fourth challenge - tracking the conse- coating flows. Such analyses must be performed
quences oflarge-amplitude transients - necessitates with care, however, because the computed
fully nonlinear time-dependent analysis. Section predictions depend strongly on the initial state
9.2.9 recounts methods of fully implicit time- from which integration in time is started.
integration. It also briefly comments on alternative Moreover, transient analyses fail altogether to
formulations with semi-implicit integration. Apart elucidate the structure ofnonlinear solution spaces.
from finding the most efficient method for solving Apart from methods that offer a true alternative,
the stiff differential/algebraic system, the key other methods for computational free-surface
issues are choosing self-consistent initial conditions flow analysis more closely resemble boundary-
and designing appropriate error estimates for confining finite-element methods for base-flow
step size control. Transient analysis has the analysis. Among these are finite-difference methods
potential to follow the evolution of growing combined with mapping to a rectangular domain
disturbances to a final defect, and is needed to (Section 9.4.3a) and spectral-element methods
simulate start-up, 'skip-out' and splice passage. (Section 9.4.3b). While the former offer no
To date, however, none of these problems has compelling advantage, the latter share many of
been tackled satisfactorily, largely because ALE- the hallmarks of finite-element methods yet may
type finite-element approaches have great difficulty provide improved accuracy. Boundary-element
in tracing very large deformations. methods would also be very attractive for
Section 9.3 takes up special topics in the coating-flow analysis, especially of three-dimen-
physics of coating flows that are not always sional situations, if it were not for their inability
included in standard algorithms, namely the to cope effectively with flow-internal nonlinearities
elusive physics ofdynamic wetting (Section 9.3.1), (Section 9.4.4).
the computational challenges of non-Newtonian
behavior (Section 9.3.2), the care needed to
9.2 FINITE-ELEMENT METHODS
account for the impact of surface-active and
volatile additives accurately (Section 9.3.3), and
9.2.1 COATING FLOWS AND THEIR
the complexities inherent in elastohydrodynamic
DISCRETIZATION
computations (Section 9.3.4).
Section 9.4 reviews alternative schemes for
9.2.1.1 Governing equations and boundary
free-surface flow computations, most developed
conditions
in contexts quite different from coating and many
suitable only for transient calculations. Indeed, Chapter 2 provides a comprehensive introduction
fully transient analysis of free-surface flows that to the hydrodynamics that govern viscous
free-surface flows of the sort encountered in Navier's slip boundary condition near the apparent
coating operations. Figure 9.2 recounts the dynamic contact line (which can be replaced by
governing equations in the dimensionless form a discontinuous velocity field, see Section 9.3.1);
most frequently used in finite-element simulations (9.8) specifies the angle ofcontact of the liquid/air
of coating flows, exemplified here for the case of interface with the moving substrate surface; (9.9)
two-dimensional slide-coating flow of two New- is Gibbs' complementarity condition at the static
tonian liquid layers A and B of viscosities /lA and wetting line where the liquid/air interface meets
/lB' densities PA and PB' and surface tension (J with the stationary die wall and 8 == cos - 1 (D' Ds );
(for an in-depth discussion of slide coating, refer and (9.10) is the condition of stress continuity at
to Chapter lIb): (9.1) is the momentum equation, the interface (approximated here as a mathematical
(9.2) the Newtonian constitutive relation and surface) between miscible layers, A and B.
(9.3) the continuity equation; (9.4) and (9.5) are In the equations listed in Fig. 9.2, u is the
the kinematic condition and traction balance at velocity measured in units of U, a characteristic
free surfaces; (9.6) is the condition of no slip at velocity, such as web speed; length is measured
solid surfaces (away from contact lines); (9.7) is in units of L, a characteristic length such as gap
FLOW POMAIN
Re (~~ + u· V u) = V. T + St f (9-1)
OUTFLOW BOUNDRY
T = -pI + Vu + (VujT (9-2) n· T = 0 (9-13)
V.u=O (9-3)
t = ts (9-14)
INFLOW BOUNDARY
u = - 1 (2y_ y2) i (9-11) SOLID BOUNDARY - NO SLIP
2
u = Us (9-6)
h=1 (9-12)
SLIP REGION - NAVIER SLIP

FREE SURFACE ts' (u - us) = ~slip tsns:T (9-7)
n.(u - x) =0 or Singular Elements
n·T=2H. n + npa (Discontinuous Velocity)
Ca
INTERFACE DYNAMIC CONTACT LINE
n·(u-x)=o n· n s= cose D (9-8)

n· (u - x) =0 (9-4)
n.(T A - ~TB)=O (9-10)
u = Us (9-6)
(Discontinuous velocity)
STATIC CONTACT LINE
(es-e) hg = 0,
(9-9)
hs <: 0, e-9s <:0
Figure 9.2 Governing equations and boundary conditions for free-surface flow in two-layer slide coating.
clearance or, as in Fig. 9.2, the thickness H o of velocity component and the normal traction are
the film on the inclined slide; time in units of imposed. Prescribing the normal velocity com-
LIU; i and j are the unit vectors in the x and y ponent is also feasible, but can result in spurious
directions; f is the unit vector in the direction of wiggles unless the normal component at the
gravity; T the stress tensor and p the pressure, inflow is not fixed at the same time.
both measured in units of J1.UIL; I the unit dyadic; To shrink the length of the computational
Re == pQIJ1. the Reynolds number; Ca == J1.Ula domain further where a grid needs to be deployed
the capillary number; St == gL 2 piJ1.U the Stokes and the full Navier-Stokes system is solved,
number, where g is the force of gravity per unit improved inflow and outflow conditions can be
mass and Q is the volumetric flow rate per unit constructed from approximate descriptions of
width in the cross-web direction; Us the dimen- the capillary hydrodynamics, such as the lubri-
sionless velocity at solid boundaries; 0 the local cation approximation or the Young-Laplace
unit normal to the boundary; H the mean equation (cf Chapters 2 and 8), or more rigorous
curvature of the interface; Os and t s the unit perturbation expansions (Chapter 8). Such ap-
normal and tangent to the solid surface; Os the proximations quite often reduce the governing
static contact angle; hs the dimensionless wetted equations (9.1)-(9.6) into a single nonlinear
distance from the corner to the contact line; 0D ordinary differential equation for the film thickness
the dynamic contact angle; and PSIiP the dimen- profile (e.g., Christodoulou and Scriven (1989)).
sionless slip coefficient. PSlip, 0D and Os can be If the approximations apply throughout the flow
regarded as empirical parameters that have to domain, this equation can be solved numerically
be inferred from experiments (see Section 9.3.1, for the film thickness (e.g., Ruschak (1978)). More
and also Chapter 3). often, however, this equation is linearized and
Because computational resources are finite, then integrated once to provide boundary
there is a need to truncate the computational conditions of the third kind (Robin) (e.g., Bixler
domain upstream and downstream of the region (1982); Christodoulou and Scriven (1989)). A
of main interest, the so-called flow forming zone. third alternative is to match the nonlinear equation
Artificial boundary conditions must then be for the film profile in upstream or downstream
imposed at the resulting inflow and outflow flow regimes to the full Navier-Stokes system
boundaries. To limit the influence of these for the flow forming zone, and solve the two
conditions, the boundaries are most often placed equation sets simultaneously (e.g., Kistler and
sufficiently far upstream and downstream of flow Scriven (1983); Kistler (1984). Malamataris and
forming zones, where the flow relaxes to almost Papanastasiou (1991) advocate imposing the
fully developed film or channel flow. In Fig. 9.2, traction o· T and free surface slope at outflow
for example, equation (9.11) specifies the two boundaries as calculated from the discretized yet
components of velocity and (9.12) the free surface still unknown flow field and free surface profile.
location at the inflow boundary. At the outflow This so called 'free boundary condition' amounts
boundary, a natural condition of the form (9.13) to imposing no boundary condition at all. It
results in zero traction if the flow is indeed fully yields well-posed problems for hyperbolic prob-
developed where the condition is imposed (the lems in which disturbances and even discontinuities
effect of gravity is assumed negligible here); and propagate downstream without any upstream
(9.14) specifies the slope of the free surface. Most influence. It also looks promising for elliptic
commonly, at inflow boundaries, the free surface problems in which the entire equilibrium solu-
location and either all velocity components or tion depends on the boundary conditions on
the tangential velocity component and the normal all the boundaries. For the latter, however,
traction are specified. At outflow boundaries, the ramifications and possible limitations of the
preferably the slope of the free surface and either 'free boundary condition' need to be studied
the traction in both directions or the tangential further.
9.2.1.2 Finite-element discretization Primitive variable formulation (u- P)

In the u-P formulation, the system of equations
In flow regions of viscous free surface flows where (9.1) and (9.3) with boundary conditions (9.4)-
the flow domain fits no standard coordinate (9.14) (see Fig. 9.2) is discretized as is, using a so
system, the finite-element method is the technique called mixed interpolation. Velocity, pressure,
ofchoice for finding steady and transient solutions and position are expanded in terms of simple
of the system (9.1)-(9.14). The main idea is to polynomial basis functions <pj(~, ,,), tll(~,,,) and
approximate the unknown flow field (i.e., velocity, <pl(~,,,) which are constructed on a standard
pressure, flow geometry, etc.) in terms of a element, for instance a square with local ~ and
collection of particularly simple, polynomial basis " coordinates in the ranges - 1 ~ ~ ~ 1 and
functions. These are nonzero on only a few
-1 ~" ~ 1:
subdomains of the flow domain, in contrast to
the traditional functions of mathematical physics,
(9.15a)
which span the entire domain. The partial
differential or integral equations are converted
into ordinary or algebraic equations for the P= L Pk (t) til (~, ,,), (9.15b)
coefficients of the basis functions by forming their k
weighted residuals. In the Galerkin method the
weighting or test functions are the same as the x= L Xl(t)<pl(~,,,) (9.15c)
I
basis functions. Although this choice is optimal
for linear second-order elliptic problems, other Equation (9.15) is written here for two-dimensional
choices such as Petrov-Galerkin methods may flows; for three-dimensional flows, the expansions
be better for hyperbolic and convection dominated are of the same form but the basis functions
problems (e.g., Brooks and Hughes (1982)). <pj(~,,,,o. ",k(~,,,,O and <p1(~,,,,O depend on a
As in the solution of the Navier-Stokes third coordinate (. In both cases, the basis
equations on fixed domains, finite-element analysis functions are defined only locally (i.e., within
offree-surface flows can start from several different each element) and need only have limited
formulations of the equations depending on the continuity between elements. The coefficients OJ'
treatment of the incompressibility constraint (ei Pk and Xl are the unknowns. With the exception
Gresho (1991)): the primitive-variable or velocity- of those variables for which the basis functions
pressure (u- P) formulation, which has been used are discontinuous at inter-element boundaries,
almost exclusively for free-surface flows; the the coefficients represent nodal values of the flow
penalty function formulation, which is very field and the flow geometry. They are time
attractive because it allows elimination of the dependent in transient analyses and constant in
pressure unknowns and thereby significantly steady-state analyses.
reduces the dimension ofthe discretized equation In finite-element solutions of the Navier-
set; the pressure Poisson equation formulation Stokes equations for free-surface flow problems,
(PPE), which is gaining popularity because it as for fixed boundary problems, not every
allows semi-implicit time integration; and the combination of basis functions for velocity and
streamfunction-vorticity ('I'-w) formulation, pressure gives stable results. The key requirement
which is very popular in finite-difference algorithms is that the interpolations satisfy the so called
of the sort described in Section 9.4.3, but rarely LBB condition, a compatibility condition named
used with the finite-element method, especially after Ladyzhenskaya (1963), Babuska (1971) and
for free surface problems. The first three formu- Brezzi (1974). It ensures that the discrete pressure
lations are described in detail in what follows. field is not polluted by nonphysical oscillations
For the 'I'-w formulation, the interested reader - so called spurious pressure modes - and the
is referred to Gresho (1991). discretized system of equations is stable, i.e., its
solution converges with mesh refinement. Com- the divergence theorem to the momentum
monly used combinations are nine-node biquad- equation:
L
ratics for ljJi and four-node bilinears or linear
discontinuous functions for ljJk (see Huyakorn et
al. (1978». Finding stable combinations for R~== {T·V4>i+[Re(Ii+(u-x)
viscoelastic flows, with or without free surfaces,
is still an area ofactive research (see Section 9.3.2).
Equal-order interpolation is most commonly
·Vu) - Stf]4>i}dA - f
aA
n·T4> i ds
used for nodal position and velocity, as in (9.17)
equation (9.15c). The mapping of each element
of the boundary-conforming mesh in the physical
domain onto the standard domain (~, I}) on a
fixed rectangular grid - provided by equation
R~ == L ljJkV·udA = 0 (9.18)
(9.15c) - is then called iso-parametric. For interior

nodes away from free surfaces, sub-parametric R~==f m
n·(u-x)4> ds=O (9.19)
mapping with lower-order basis functions is also oAr
possible. In fact, it is advantageous in conjunction
with elliptic mesh generation (Christodoulou and Here M, C and K denote momentum, continuity
Scriven (1992); see also Section 9.2.2). and kinematic residuals; A denotes the computa-
In transient problems, the mesh usually deforms tional domain, oA its entire boundary, and OAf
at each time step to follow the motion of free all free-surface boundaries. Introducing the
boundaries in an arbitrary Lagrangian- Eulerian expansions (9.15) into the integrals (9.17)-(9.19)
manner that preserves the topology of the initial turns the latter into a system of algebraic and
mesh configuration yet is independent of the ordinary differential equations - or just algebraic
underlying liquid flow (see also Section 9.4.2). equations in the case of steady state calculations
Essential to the formulation of the discretized - for the unknown coefficients in the basis
transient problem is proper accounting for the function expansions. The integrals (9.17)-(9.19)
relative motion between the deforming grid and are sometimes referred to as the weak form
the flow. It is readily taken care of via the iso- because the original set of governing equations
or sub-parametric mapping (9.15c) (see Lynch and boundary conditions is satisfied only in an
(1982». Time derivatives a/at at a fixed Eulerian average, integral sense rather than being exactly
location are transformed to time derivatives at valid at every point in the domain.
fixed (iso)-parametric coordinates, denoted by an
over-dot: Imposition of boundary conditions: Essential
ou/ot = Ii - x· Vu (9.16) boundary conditions (BC), usually at solid or
inflow boundaries, are imposed by replacing the
Here, x is the mesh velocity. The term x· Vu is, corresponding weighted residual equations (9.17)
of course, the same as the extra term that arises or (9.19) with the desired velocity or free-surface
in so called arbitrary Lagrangian-Eulerian specifications. Natural boundary conditions,
methods - which recently have received much usually at free surfaces and outflow boundaries,
'hype' in the context of free-surface flow compu- are imposed through the boundary integral
tations (see Section 9.4.3). SMn·T 4> i ds in (9.17). For example, the stress
The governing equations (9.1), (9.3) and (9.4) boundary condition at the free surface is imposed
are transformed into a discretized system of by substituting the right-hand side of (9.5) into
ordinary differential equations by multiplying the traction boundary integral of (9.17) and
with the appropriate basis functions, integrating applying the surface divergence theorem (Ruschak
over the computational domain, and applying 1980). For two-dimensional (2D) problems,
equation (9.17) becomes: possible, as mentioned in Section 9.2.3). The
R~ = L [(Reu' Vu - Stf)</i
kinematic condition is a hyperbolic equation
and, hence, a Galerkin weighted residual of the
f
form (9.19) can be unstable. However, in a system
of partial differential equations, it is the type of
1
+ V</Ji'T]dA + -C d</Ji ds
t -d the entire system and not the type of the individual
a OAr s
equations that determines the numerical treatment
1. . required. Indeed, difficulties with the Galerkin
- Ca (</J't 1 - </J'to) (9.20)
formulation of the kinematic condition mostly
arise at high capillary numbers, and are especially
The end-point terms in (9.20) are 'shell forces'
likely close to static and dynamic contact lines
due to surface tension. Their directions to and
where spurious wiggles can totally destroy the
t 1 have to be specified at inflow and outflow
solution. Kistler and Palmquist (1990) advocated
boundaries, and also at contact lines. Specifying
an alternative formulation to (9.19) that insists
a contact angle in the weak form (9.20) rather
explicitly on the net mass flux between neighboring
than through an essential condition works well
nodes being zero. Where applicable, this approach
at static separation lines that are free to migrate
completely eliminates wiggles. Soulaimani et a1.
along a solid, but is problematic at dynamic
(1991) made use of a Petrov-Galerkin method
wetting lines (see Christodoulou and Scriven
to eliminate spurious free-surface wiggles. Likewise,
(1989». Wambersie and Crochet (1992) tried to fight the
In three-dimensional (3D) problems, the bound-
wiggles with a streamline-upwinding/Pet-
ary integral of the traction on free surfaces oAr
rov-Galerkin formulation (Brooks and Hughes
. f .
in equation (9.17) is a surface integral:
1982) for the kinematic Be.
f
System (9.17)-(9.19) is closed by suitable mesh
1
n·T</J'dA = -C (VII'en)n</J'dA generation equations to be introduced in Section
oAr a oAr
9.2.2 below (e.g., equation (9.27». These equations
+ r
lOA r
npa </JidA (9.21) determine the location of interior mesh nodes as
well as the distribution of nodes along the
boundary. The position of the boundary in the
Application of the surface divergence theorem
direction of its normal is, of course, determined
(Weatherburn 1927, p. 123) to the first term on
by a distinguished equation that captures the
the right-hand side of (9.21) lowers the order of
physics of the problem, such as equation (9.19),
spatial derivatives (Ruschak 1983), so that
or a parametric equation for the (solid) boundary
f oAr
</Ji(V'n)ndA = r
loA r
VII </JidA - 1- </Jimds
Jc
shape.
Integrals in (9.17)-(9.19) and the mesh gener-
ation equations are usually evaluated by Gaussian
(9.22) quadrature. The final result is a system of
Here s is the arclength along C, the curve differential and algebraic equations of index two
bounding oAr, and m is the unit tangent to the (Petzold (1982); see also Section 9.2.9),
free surface oAr along C that is normal to C. As
R(t, y,y) = 0 (9.23)
in the two-dimensional case, equation (9.22) is
substituted into (9.21), and (9.21) in turn into (9.17). where y == [u i , Vi' Pj' X" y,]T is the vector of time-
The system (9.17)-(9.19) - the formulation of dependent coefficients, and y == dy/dt the vector
choice when full Newton iteration is used (see of their time derivatives. Section 9.2.3 addresses
Section 9.2.3 below) - singles out the kinematic methods for finding steady state solutions of the
BC as the 'distinguished equation' of change for system (9.23); Section 9.2.9 discusses time stepping
the free-boundary location (other choices are methods.
Penalty function formulation for pressure are discontinuous at inter-element

boundaries so that (9.24), weighted with the
In the penalty function formulation the equation
pressure basis functions on each element, gives
p=A.V·u (9.24) rise to algebraic equations that can be solved as
an independent set. In this way, the discrete
replaces the continuity equation (e.g., Bercovier continuity and momentum equations are solved
and Engelman (1980), Kheshgi and Scriven (1985». sequentially rather than simultaneously (see, for
Here A. is a large, empirically chosen number and example, Kheshgi and Scriven (1985».
p is an approximate pressure which approaches
the exact pressure in the limit A.---+ 00. Exact Pressure Poisson equation (PPE) formulation
incompressibility V· u = 0 is only enforced in
that limit. The penalty parameter must be chosen In the PPE formulation, an elliptic equation is
carefully: too small a value of A. results in derived for pressure by taking the divergence of
inaccurate results; too high a value causes matrix the momentum equation (9.1) and using the time
conditioning problems. Kheshgi and Scriven derivative of the continuity equation (9.3) to
(1985) suggested varying A. locally with element eliminate the divergence of the acceleration:
size to reduce rounding error, and were the first
(9.25)
to tackle successfuly free-surface flows with the
penalty method. The most successful variant of Equation (9.25) can replace the continuity equation
the penalty method, though, is the Uzawa to form, together with (9.1), a system that is
algorithm (see Fortin and Glowinski (1983» equivalent to (9.1) and (9.3). The new system has
which by-passes the difficulties of specifying a index one (Petzold (1982), see also Section 9.2.9)
variable penalty parameter altogether. As an and, hence, is easier to solve than (9.1) and (9.3).
application of the augmented Lagrangian method, For the pressure Poisson equation (9.25) to be
the algorithm includes iterative correction steps free of spurious solutions, however, the incom-
that ensure the continuity equation is satisfied pressibility condition must be satisfied on the
to computer precision regardless of the value of domain and its boundary. The main advantage
the penalty parameter A.. Fortin and Fortin (1985) of the PPE formulation, however, is that it allows
extended Uzawa's algorithm from Stokes' flow the separation of the velocity calculation from
to the full Navier-Stokes system. the pressure calculation and therefore makes
When equation (9.24) is used to eliminate possible the use of semi-implicit methods (Gresho
pressure from the momentum equation before 1991). The typical semi-implicit method uses an
discretization, reduced (i.e. lower order) integration explicit treatment of the advection terms and an
must be applied to the A.-dependent terms in the implicit one for the viscous term. The objective
Jacobian matrix. Otherwise, the trivial solution is to reduce the intense coupling between the two
(i.e. u = 0) may be computed as A. becomes large. terms and thereby permit the sequential solution
Reduced integration effectively lowers the order of smaller and linear systems for the individual
of pressure interpolation but does not satisfy the velocity components, usually by iterative methods.
LBB condition. It is therefore subject to parasitic Pressure, of course, is an inherently implicit
'pressure modes.' In an alternative and preferable variable and is obtained by solving the pressure
formulation, called the consistent or discrete Poisson equation (9.25). A long-standing contro-
penalty method, the Galerkin weighted residuals versy about the correct boundary condition for
of the momentum equation (9.1) and penalty- the pressure in (9.25) was resolved by Gresho
continuity equation (9.24) are formed as usual, and Sani (1987). That condition is the Neuman
respecting the requirements on velocity and condition that results from forming the component
pressure interpolation - and therefore satisfying of the momentum equation normal to the
the LBB condition. The basis functions of choice boundary.
In practice, the so called consistent pressure meshes for free surface flows can be divided into
Poisson equation (CPPE) is preferred over direct two steps. The first step is to parameterize the
discretization of equation (9.25) (Gresho 1991; location of the nodes that reside on the free or
Gresho and Sani 1987). This slightly different moving boundaries; and the second is to calculate
form is obtained from the discretized momentum values of the nodal coordinates in the interior
equation and the time derivative ofthe discretized of the computational flow domain from the
continuity equation. It ensures that the discrete values at the boundaries. The goal is to keep in
velocity field is incompressible at the boundaries check the discretization errors in the solution of
and the normal accelerations are zero at the the hosted physical problem. Finite-element size
walls, and allows the use ofdiscontinuous pressure is of course of main concern - especially in
basis functions that satisfy the LBB condition regions where increased resolution is demanded
(cf discrete penalty method, above). Care must by the physics of the flow - but subdomain angles
also be taken when using PPE formulations to grossly deviating from orthogonality and rapid
ensure they satisfy the solvability conditions of local changes of subdomain size can also cause
the original u- P formulation. Otherwise, spurious significant discretization errors. There are two
solutions can be obtained (Gresho 1991). For a ways to generate meshes ofacceptable orthogonal-
recent application of the CPPE formulation to ity and smoothness: algebraic mesh generation,
three-dimensional free-surface flow problems at which interpolates the location of interior nodes
high capillary numbers, see Wambersie and from the boundary nodes with the aid ofalgebraic
Crochet (1992). blending functions (Smith 1982; Thompson, Warsi
and Mastin 1985); and differential (elliptic) mesh
generation, which computes the mesh points by
9.2.2 FREE BOUNDARY PARAMETERIZAnON solving a system of partial differential equations
AND MESH GENERAnON for the mapping from the rectangular computa-
Most numerical methods for computational fluid tional domain to the deformed physical domain.
dynamics discretize the governing equations on
a mesh (grid) of nodes (points). Finite-element
9.2.2.1 Algebraic mesh generation
methods organize the nodes in a large number
of small subdomains, or elements. For an accurate The simplest form of algebraic mesh generation
solution of free-surface problems like coating for free-surface flows is based on one-dimensional
flows, so-called boundary-conforming meshes parameterization of the physical coordinates of
that deform in response to changes in free- the boundary nodes. The earliest finite-element
boundary shape are particularly well suited. schemes for two-dimensional flows relied on a
Element boundaries coincide with free surfaces one-dimensional representation that simply saves
and/or interfaces, and hence become computa- the interface position as a discrete set of heights
tional entities that expressly and accurately track above the axis in a standard coordinate system,
the position and shape of the free boundaries. In most often a Cartesian one, Le., Yi = h(x i ) (Nickell,
addition, curvilinear coordinate systems defined Tanner and Caswell 1974; Zienkiewicz 1977;
on the boundary-conforming elements greatly Frederiksen and Watt 1981). To handle highly
simplify the application of boundary conditions curved menisci, Saito and Scriven (1981) and
and the solution of the hosted set of partial later Ungar and Brown (1985) combined Cartesian
equations. By mapping the curvilinear coordinates and polar representations. A more flexible
on a fixed rectangular grid, all computation can one-dimensional free-boundary representation
in effect be performed on rectangular coordinates came with Ruschak's (1980) generator-line method
regardless of the shape or movement of the in which the free surface is parameterized by its
physical boundaries. distance along straight lines from fixed boundary
The task of generating boundary-conforming points (supports). The position of the boundary
supports and the orientation of the lines was and associated unidirectional algebraic mesh
arbitrary but had to be prescribed a priori. In generation are computationally efficient, and
the similar, but more flexible method of spines have been widely used in solving coating flow
(Kistler and Scriven 1983, 1984), the boundary problems. The methodology, however, is easily
supports (base points) and line directions are not applied only for strip-like domains such as those
necessarily fixed but can be made to follow encountered in extrusion, thin-film flows, and
predicted, salient features of a free surface flow. simple bead-coating flows like curtain or slide
The method of spines, originally developed for coating. Non-strip-like domains require tedious
two-dimensional flows, has been generalized to patching of several boundary representations,
handle three-dimensional coating flows (e.g., and may necessitate multidirectional algebraic
Katagiri (1992)) and extrusion flows (e.g., interpolation. Even for strip-like domains, simple
Karagiannis, Hrymak and Vlachopoulos (1988); algebraic generator lines require some a priori
Legat and Marchal (1992): Wambersie and knowledge of the solution and a large number
Crochet (1992)). of input data. For large deformations, moreover,
In conjunction with one-dimensional free- they are prone to yield excessive mesh distortions,
surface parameterization, algebraic mesh gener- intersect within the flow domain, or cause
ation typically relies on interpolation of the multivalued free-surface representation (e.g.,
domain-internal nodal positions along a single Christodoulou and Scriven (1989). To avoid such
direction, namely that of coordinate or generator problems, skillful user interaction is often needed
lines which form two opposing element boundaries. as time evolves or parameters vary during a
Most often, simple algebraic proportions (e.g., simulation.
Nickell, Tanner and Caswell (1974); Ruschak Kistler and Palmquist (1990) advanced a
(1980); Kistler and Scriven (1983); Karagiannis, generalized algebraic mesh generation scheme
Hrymak and Vlachopoulos (1988)) or explicit that liberates free-boundary parameterization
algebraic mapping functions (e.g., Ettouney and from the limitations inherent to generator lines
Brown (1983); Ungar, Ramprasad and Brown and, hence, can tackle two-dimensional free-surface
(1988)) are used to place the nodes. In the flows of great complexity and track large
so-called streamlined finite-element method, two deformations with much less user interaction
element boundaries still consist of generator than earlier algebraic schemes. The scheme permits
lines, but the other two are made to form the flow region of interest to be divided into
streamlines (Papanastasiou, Macosko and Scriven quadrilateral subregions or macro-elements of
1985). This method is most useful for viscoelastic arbitrary connectivity. The edges of each macro-
flows, especially those governed by integral element can be made into 'free boundaries' -
constitutive equations (Papanastasiou, Scriven some physical boundaries such as free surfaces,
and Macosko 1987), but cannot be applied to interfaces, deformable elastic members, etc., and
flows with recirculation. The method is also others merely auxiliary internal boundaries. The
well-suited for three-dimensional extrusion prob- position of the nodes along these edges is
lems (e.g., Yokoi and Scriven (1989); Ellwood, parameterized by their two Cartesian coordinates.
Papanastasiou and Wilkes (1992)). Legat and One is determined from a 'distinguished' boundary
Marchal (1992) devised an alternative scheme for condition as in conventional, one-dimensional
three-dimensional extrusion problems that con- free-boundary parameterization schemes. The
strains only free-surface nodes to move along other is calculated from an additional, user-selected
prescribed directions; the interior remeshing relies or adaptive constraint that controls the spacing
on a generalized but still linear algebraic rule to between boundary nodes. Key to the flexibility
relate the interior node displacement via the and adaptability of the method are the extra
Eucledian spacing to that of the boundary. conditions that control the position ofthe vertices
One-dimensional free-surface parameterization of the macro-elements. These conditions can be
chosen so as to make the mesh automatically significant changes in flow geometry occur, and
follow essential features of the flow. They can the ultimate failure of a mesh with a given
even be designed so as to allow algebraic mesh topology to follow extreme distortions in flow
generation to replicate many essential character- geometry.
istics of elliptic mesh generation. Figure 9.3 In the version presented by Kistler and
illustrates, for the case of heel formation in Palmquist (1990), the generalized algebraic
curtain coating (see also Chapter 11c), the ability procedure was restricted to at most two free
of the generalized algebraic scheme to track boundaries along opposing edges of each macro-
dramatic changes in flow geometry. Figure 9.4 element, and relied on unidirectional interpolation
demonstrates the same capability for the formation to distribute internal nodes between those two
of a rolling inflow bank in forward kiss coating boundaries. However, the procedure can readily
(see also Chapters 12a and 12c), which involves be augmented to accommodate four free bound-
not only free surfaces but also deformable aries per macro-element and to employ multi-
substrates. Figure 9.3, however, also reveals some directional algebraic interpolation. A widely used
of the shortcomings of algebraic mesh generation form of the latter is the so called transfinite
- and most other deforming mesh schemes - interpolation (Gordon and Hall 1973). It applies
namely the need to deploy a mesh with a a unidirectional interpolation in each direction
predetermined structure based on some a priori and then adds a correction term for the corners.
knowledge of the flow geometry, the necessity When the coefficients in algebraic blending
to change some mesh control parameters as functions are made to depend not only on the
(a) (c)
Figure 9.3 Finite-element meshes and associated flow fields for impingement flow in curtain coating, generated
by the algebraic technique of Kistler and Palmquist (1990): (a) UIV = 1.0, where U is the coating speed and
V is the vertical impingement speed; (b) UIV = 0.25; (c) UIV = 0.15 (Re == pQIIl = 2.5, where p is the liquid
density, Q the volumetric flow rate per unit coating width, and Il the liquid viscosity; Ca == IlVI(J = 10, where
(J is the surface tension; note that the three sets of figures are not drawn to the same scale).
(a)
I~ " I I
(b)
UWRAP = 0.1 0
UWRAP 1.0 0
UWRAP = 1.65°
Figure 9.4 Finite-element meshes and associated flow fields for the inlet zone of forward kiss coating, generated
by the algebraic technique of Kistler and Palmquist (1990): (a) overview of flow configuration; (b) details of
the inlet zone as a function of the angle of wrap of the tensioned web around the coating roller (for purposes
of clarity, the mesh plots on the right-hand side are not drawn to scale). Re = PQ/Jl = pUHo/Jl = 3.0,
Ca= JlU/u = 10, Ho/R = 0.006, EcP/[12(1 - y2)JlUH o2] = 1.11, TjJlU = 20000, where is H o the coming film
thickness, R is the roller radius, E is Young's modulus, b is the substrate thickness, y is Poisson's ratio, and
T is the tension in the substrate per unit width.
physical coordinates of the boundaries but also 9.2.2.2 Elliptic mesh generation
their derivatives, algebraic mesh generation can
even ensure orthogonality of the elements at the Differential elliptic mesh generation locates the
boundaries (e.g., Smith (1982». To date, however, mesh points by solving a system of partial
no applications ofsuch methods to free-boundary differential equations for the mapping from the
problems have been reported. computational to the physical domain. Compared
with algebraic techniques, it is usually more variable. The most widely accepted of such error
expensive but it provides inherent smoothness of indicators quantify the smoothness, orthogonality
the mesh, less danger of singularity and automatic and concentration of the mesh. Brackbill and
local control of mesh properties. Saltzman (1982a, 1982b) proposed the func-
A variety of elliptic mesh generation methods tional
II
have been reviewed by Thompson and co-workers
(1982, 1985) who focused on two-dimensional
flows - like the discussion that follows here. The [(Ve)l + (V'1)l] dx dy
simplest method locates two-dimensional meshes smoothness
e
II
by solving a pair of Laplace equations for the
e
and '1 coordinates, V2 = 0 and V2 '1 = 0 (Winslow
+ Ao [Ve'V'1]2J 3 dxdy
1967). This system satisfies an extremum principle
in the physical space and, hence, precludes orthogonality
II
overlapping coordinate lines. Other methods
e
solve a Poisson system V 2 = p(e, '1) and V 2 '1
+ Aw wJdxdy (9.26)
= Q(e, '1), where P and Q are specified spacing
control functions (e.g., Thompson, Thomas and node spacing
Mastin (1974); Thomas and Middlecoff (1980);

Thomas (1982)). Applications to water-wave
Here J = x~y~ - x~y~ is the Jacobian of the
mapping from the physical to the computational
problems have been reported by Shanks and domain. The mesh results from solving a pair of
Thompson (1977), Haussling (1982), Aston and second-order, quasi-linear partial differential
Thomas (1982), and to mold-filling flows by equations that are the Euler equations of the
Subbiah, Trafford and Giic;:eri (1989). A disad- problem of minimization of (9.26). The three
vantage of the Poisson-generated meshes is lack functionals in (9.26), however, are dimensionally
of control of orthogonality in the interior, inhomogeneous and, as a consequence, the Euler
particularly when the physical region is curved equations can change type and become ill-posed.
or patched. Orthogonal meshes, generated by This was pointed out by Kreis, Thames and
stretching of conformal ones (e.g., Potter and Hassan (1986), who addressed the problem by
Tuttle (1973); Eiseman (1982); Ryskin and Leal rescaling the weights Ao and Aw • Christodoulou
(1983)), overcome this difficulty but are not and Scriven (1992) chose an orthogonality
suitable for highly distorted domains. Behrens functional that scales with smoothness and has
et al. (1987) generated meshes for mold filling the same functional form as in the discretization
flows by using the system V:.~x = 0 and error (Thompson, Warsi and Mastin 1985). The
V; y = O. This and related systems do not satisfy
,,~ resulting mesh generating system is
any extremum principles in the physical space
and, therefore, pose the danger of coordinate V· {(J[(x: + y:)/(x; + y;)] + Bs ) Ve}
lines overlapping in general configurations
B1 0 2 2
(see Thompson, Warsi and Mastin (1985), + y~)f(m 0
p.205). oe
- - -In[(x~
J
=
Mesh generation can be viewed as an opti- (9.27)

mization (variational) problem for the mapping
that minimizes the discretization error in the V· [(J[(x; + y;)/(x: + y:)] + Bs )V'1]
numerical solution of the physical problem. The

discretization error is generally formidable to _ Bl ~ In [(Xl + yl)g('1)] = 0
} 0'1 ~ ~
evaluate, however. The usual approach is to
employ simple error indicators tailored after the Here f(e) and g('1) are manually set stretching
error in the first spatial derivative of a dependent functions that control the concentration of nodal
points and 8 1 and 8 2 are their weights; 8s weighs and replacing the 11-equation with a Poisson
the smoothness of the mesh. The most important equation. The resulting method, called the
advantage ofthe system (9.27) over Brackbill and mixed-mapping method (MMM), is similar to
Saltzman's equation (9.26) is that the resulting the two-step procedure of Visbal and Knight
meshes are nearly orthogonal, even for patched (1982). Because the generating system is aniso-
domains of the sort often needed for coating flow tropic, its application to more complex geometries
simulations. with free surfaces that conform to both coordinate
With sub-parametric mapping in the interior directions is difficult. A more general modification
of the mesh, i.e. straight-sided elements, elliptic was proposed by de Santos (1991), who employed
mesh generation is only about 10% more expensive a concentration functional based on equi-
than spine parameterization, as was first dem- distribution ofa weight function in each coordinate
onstrated by Christodoulou and Scriven (1992). direction. According to the experience of one of
As an added bonus, the nonlinearity of the us (KNC), de Santos' method shares a disadvan-
Jacobian of the mapping from the physical to tage with other methods based on Laplace's
the computational domain is less severe, and with equations in that it cannot achieve the desired
it the nonlinearity ofthe weighted residuals (9.17) distribution close to highly curved bound-
and (9.18), resulting in improved accuracy of the aries (Thompson, Warsi and Mastin 1985,
numerical integration. Sub-parametric mapping p. 192).
can equally well be applied to algebraic mesh Elliptic mesh generation can be refined to
generation for improved accuracy, but there is become adaptive, i.e. automatically to place
no cost advantage because the number of nodal nodal points where accurate resolving of the flow
position unknowns remains the same. Sub- is most important. With (9.27), adaptive meshes
parametric mapping is an example of a more can be generated by making the control functions
general and attractive two-level strategy: a coarse fW and g(11) depend on critical features, such as
elliptically generated mesh of macro-elements directional derivatives of the flow field or curvature
with well-controlled orthogonality; and a finer of free surfaces. An alternative is to add to the
mesh within each macro-element, efficiently mesh generating system a concentration functional
generated with algebraic techniques (see also (e.g., Brackbill and Saltzman (1982a, 1982b);
Morice (1983». Demkowicz and Oden (1986); de Santos (1991».
Figure 9.5a illustrates sequences of steady Saeger, Davis and Scriven (1990) proposed a
states ofslide-coating flow, with meshes generated generalization of (9.27) to three dimensions and
by Christodoulou and Scriven's (1992) method. applied it to generate finite-element meshes of
Figure 9.5b illustrates how a single mesh topology three-dimensional elemental structures for self-
can successfully capture either liquid bridges that assembling surfactant systems. Contrary to the
pin at the sharp die edge or ones that wet the two-dimensional case, orthogonality in three
underside of the die lip. By virtue of the degrees dimensions is a strong constraint (no infinity of
of freedom of every node in the mesh, elliptic solutions). For that reason, the system of Saeger,
mesh generation is particularly well-suited for Davis and Scriven (1990) had to make compro-
design and shape optimization problems. mises between smoothness, orthogonality and
Tsiveriotis and Brown (1992, 1993) adopted adaptivity. Adaptivity functions based on estimates
Christodoulou and Scriven's (1992) method to of the solution curvature proved to be more
highly deformed interfaces in solidification robust than functions relying on magnitudes of
problems, but failed to generate suitable grids the raw (i.e. unweighted) residuals. Attempts by
because of a sign error in the control parameters one of us (KNC) to apply similar adaptive-mesh
8 1 and 8 2 which, for small 8s ' resulted in an strategies to coating flows were encouraging only
ill-posed system. They still managed to get for problems without singularities. More research
solutions by discarding the concentration terms is needed to identify effective indicators ofsolution
U w = 98.5cm/s Uw = 62.4 cm/s
(a)
z
E
E ..
WETIING
I ..
NONWETIING
If
0
i=
iii 0
0
0..
W
z ·2
::J
u
i=
cl: I
I-
(/) -4
-600 ·400 ·200
..
til
't:l
I
I
W
..J
l'
I
120
~
z
cl:
I-
U
cl:
I- 100
Z
0
U
u 80
i= -600
cl: ·400 -200
l-
(/)
(b) BACK PRESSURE Pa
Figure 9.5 Finite-element meshes and associated flow fields for slide-coating, generated by the elliptic technique
of Christodoulou and Scriven (1992). (a) Sequence of steady states for different web speeds (pressure difference
liP = -161 Pa, coating gap H GAP = 0.347 mm, liquid viscosity Jl = 8.717 mPa s, liquid density p = 1130kg/m 3 ,
surface tension (J = 70.1 mN/m, imposed dynamic contact angle eD = 160·, slide inclination 30·); the last state
is past a turning point - see Fig. 9.7. (b) Lip wetting versus non-wetting states (same parameters as in (a),
except that Uw = 1.67 m/s and the pressure difference liP varies).
error in patched meshes, especially when static Trial method
or dynamic contact lines are part of the flow.
Unfortunately, elliptic mesh generation in its In two-dimensional viscous free-surface flows
current form places restrictions on grid structure. like coating flows, the traction condition (9.5)
In fact, both algebraic and elliptic mesh motion and the kinematic boundary condition (9.4)
strategies require structured meshes on some furnish three scalar boundary conditions, one
level, with unstructured aspects arising only in more than would be required to solve the
the domain decomposition, e.g., into macro- Navier-Stokes equations on a fixed domain. The
elements, as in the Kistler and Palmquist (1990) third boundary condition - sometimes referred
scheme. As for the elliptic mesh generation to as the distinguished condition - is necessary
approach, once the conceptual hurdle is made to determine the unknown position of the free
to placing mesh points onto the reference domain surface. That is, if one of the free boundary
without restricting the grid structure to lie upon conditions is neglected, the governing system can
constant coordinate surfaces, unstructured grids be solved on a fixed domain. This is the principle
can probably be accommodated. Sackinger et al. of the trial method, common in the earliest
(1996) have recently proposed a natural follow- finite-element schemes for viscous free-surface
on that removes all restrictions on grid structure, flows (e.g., Nickell, Tanner and Caswell (1974)).
and hence can take advantage of automatic mesh The trial method assigns, at each iteration step,
generators. Their approach is to employ the a fixed shape to the free surface and computes
differential equations governing the displacements the flow field, ignoring the distinguished boundary
of an elastic solid deforming in a Lagrangian condition. Next, that boundary condition, along
fashion. The distinguished conditions (e.g., the with the current estimate of the flow field, is used
kinematic boundary condition, equation (9.19) to update the free surface location, and the cycle
are apportioned to boundary loads on the is repeated to convergence. The key advantages
pseudo-solid and are satisfied when the solid of this method - a special case of fixed point
reaches static equilibrium. Sackinger et al. have iteration (also known as Picard iteration, successive
demonstrated the pseudo-solid mesh motion substitution, or functional iteration: see Isaacson
technique for several coating problems and have and Keller (1966) and Ortega and Rheinboldt
found that unstructured grids allow for efficient (1970) - are its simplicity and, compared with
refinement around static and dynamic contact the methods discussed below, its large radius of
lines and greatly reduced user-interaction time. convergence. Its main drawback is the slow,
Essentially, for the particular case of a reversible linear rate of convergence inherent in fixed-point
elastic deformation, the pseudo-solid approach iteration. Worse, the method may fail to converge
is equivalent to using elliptic mesh generation altogether depending on the choice ofthe boundary
equations to perform a mapping with the condition condition used to update the free surface and the
of minimizing a strain energy functional. value of the capillary number. Silliman and
Scriven (1980) showed that at low capillary
numbers the normal-stress scheme converges
whereas the kinematic scheme does not, and
conversely at high capillary numbers. At inter-
9.2.3 STEADY-STATE SOLUTIONS
mediate capillary numbers, i.e. Ca = 0(1), which
are quite often of most practical interest, there
9.2.3.1 Iteration procedures
is a range where both schemes converge but do
The discretized system of equations (9.17)-(9.19) so slowly and sometimes only with the aid of
combined with, for instance, the mesh generation artful under-relaxation. Wambersie and Crochet
equations (9.27) is severely nonlinear. Finding (1992) recently applied the trial method to three-
steady-state solutions therefore requires iteration. dimensional extrusion flows, extending the
decoupling algorithm of Gresho and Chan (1990) The main disadvantage offull Newton iteration
and relying on a pseudo-transient approach to is that the Jacobian matrix as well as its
reach the desired steady state. They only considered decomposition into an upper and lower triangular
cases with Ca ~ 0(1) and chose the kinematic matrix needed for Gauss elimination (so called
equation as the distinguished condition. This LV-decomposition; see Section 9.2.3.2 below)
choice is consistent with Silliman's findings and, must be recomputed at every iteration. The
as expected, their method converges. computational cost can be reduced considerably,
however, by quasi-Newton methods (Engelman
Newton's method and Sani 1983). For simple algebraic mesh
generation schemes, another disadvantage of
The tedious iteration between the flow-field Newton's method and its modified versions is
calculation and free-surface update can be avoided that the variations of the residuals with respect
by simultaneously solving all the equations, to all free surface degrees of freedom involve
including equations for the position of the nodes integrals that span a large portion of the domain.
on the free surface and, if applicable, additional This non-local structure drastically increases the
nodes in the interior. Newton's method (e.g., band-width of the Jacobian matrix and, thereby,
Isaacson and Keller (1966); Ortega and Rheinboldt precludes the use of standard finite-element codes,
(1970)) linearizes the complete equation set (9.23) in which an equation associated with a nodal
around the current estimate of the solution, yk, point receives contributions only from neighboring
to yield elements. A further difficulty in applying Newton's
aR method arises from the complicated manner by
ay (yk)~y = - R(yk) (9.28) which the free-surface unknowns enter the
discretized equations (see Kistler and Scriven
and adds the solution ~y = [yk+ 1 - yk] of the (1984)). In fact, computing and coding the
resulting linear system to the current estimate yk variations with respect to free boundary unknowns
in order to provide an updated estimate of the analytically can be a tedious programming
flow field, including the position of the free challenge. Even small errors in a single entry to
surface(s). The iteration is repeated until conver- the Jacobian matrix can severely hamper the
gence. The structure of the Jacobian matrix overall rate of convergence. For the same reason,
aRjay, with many of its entries related to free on the other hand, the rate of convergence
surface unknowns, has been described in detail becomes a useful test of the correctness of the
by Kistler and Scriven (1984). Their description code that computes the residual and Jacobian
is readily generalized to elliptic mesh generation entries.
schemes (Christodoulou and Scriven 1992). There are three alternatives to analytically
In contrast to the trial method, Newton's deriving and manually programming the Jacobian
method has superb convergence properties. In matrix. The first is to compute its entries by
fact, it converges quadratically and, hence, requires numerical differentiation of the finite-element
only four or five iterations in most cases, the residuals. Caution is in order, however, in
number being independent ofthe flow parameters estimating a suitable step size so as to manage
(Ruschak 1980; Saito and Scriven 1981). Besides, properly the trade-off between approximation
the Jacobian aRjay of the converged solution error and round-off error. The second alternative
contains invaluable information for parameter is to calculate the analytical Jacobian by using
continuation (Section 9.2.4), stability analysis to a symbolic manipulator such as MACSYMA
transient disturbances (Section 9.2.5), direct (Pavelle and Wang 1985). The resulting computer
tracking of critical points (Section 9.2.7), and code can be made efficient by optimizing compilers
sensitivity analysis to ongoing disturbances that recognize and remember repeatedly occurring
(Section 9.2.8). sub-expressions. The third alternative is to leave
the differentiation to a preprocessor or compiler (1) Gauss elimination with or without pivoting
that uses the newly arising techniques ofautomatic to generate the LV factorization of A, where L
differentiation (e.g., Griewank and Corliss (1992». is a lower triangular and U an upper triangular
matrix; (2) forward reduction of the vector b, i.e.
solution of the system Lz = b, which can be
Total linearization method (TLM)
performed at the same time as the LV factorization;
In contrast to Newton's method, which linearizes and (3) back-substitution, i.e. solution ofUx = z.
the discretized equation set, the TLM linearizes As A is sparse, a reduced storage mode such
the continuous weak form before discretization as the skyline method (see Hasbani and Engelman
- as first done by Dupret (1982) - and then (1979» can be used. The main drawback of this
simplifies the linearized form by means of the method is that no pivoting can be performed.
free-surface boundary conditions (see Kruyt et This limits accuracy and stability, especially close
al. (1988) for details). As a result, the influence to singular points of the nonlinear system.
of the unknown position of the free surface(s) is A class of more robust and widely used
confined to boundary integrals, and hence the Gauss-elimination methods for the solution of
TLM is readily incorporated into standard sparse linear systems resulting from finite-element
finite-element codes for fixed domains. To do so, discretization is that of frontal solvers, first
the distance of the next approximation of a free applied to unsymmetric matrices by Hood (1976,
surface from the current one is made an additional 1977). Frontal solvers rely on the fact that Gauss
unknown function, and is expanded into one- elimination of a variable can commence immedi-
dimensional finite-element basis functions. The ately after its last appearance during the process
distance is usually measured along the normal ofassembling the discretized equations, i.e. before
to the current estimate of a free surface, although assembly of all the residuals and Jacobian entries
other definitions can be used, for instance spines. is complete. LV factorization can thus be confined
In principle, the TLM is much easier to implement to a small subset of equations associated with
than Newton's method while retaining a super- nodes on a 'front' that propagates through the
linear rate ofconvergence. However, only limited mesh as Gauss elimination proceeds. For enhanced
evidence for its effectiveness is available to date stability, the frontal method can be combined
(Kruyt et al. 1988; Cuvelier and Schulkes 1990). with full, partial or threshold pivoting among a
In particular, its advantages in conjunction with set of fully assembled equations that are carried
elliptic mesh generators are not clear; and its along with the front. Reid (1981) reviewed such
applicability to stability analysis has not been algorithms, including element reordering methods
explored. to minimize front width and sub-structuring
techniques leading to multi-frontal algorithms.
More recent work has focused on enhancing
9.2.3.2 Solution of linear equations
performance on specific computer architectures,
All the solution algorithms introduced above e.g. Duff (1984), Duff, Erisman and Reid (1986)
produce one or more matrix equations of the and Benner (private communications 1984, 1985).
form Ax = b that have to be solved at each Habashi, Nguyen and Bhat (1991) suggested that
iteration. The matrix A is usually large (5 to 50 frontal-like methods optimized for vector-parallel
thousand or more), sparse, banded and unsym- processing on CRAY supercomputers are superior
metric. to other sparse-matrix solution algorithms.
Quite often, there is a need to augment the
system of partial differential equations with one
Direct matrix solvers
or more global (or dense) constraint equations.
In most conventional algorithms, a direct method A prominent example comes from pseudo-
is used to find the solution in three distinct steps: arc-length continuation (see Section 9.2.4 below).
Global equations destroy the bandedness of the properties of the coefficient matrix (e.g., symmetry,
coefficient matrix, making standard sparse tech- positive definiteness, sparseness).
niques inapplicable. Thomas and Brown (1987) For symmetric coefficient matrices, successful
proposed a modification of the frontal solver to iterative techniques are well established. Most
cope with added global constraints in finite- popular is the conjugate gradient (CG) method,
element equation systems. In our experience, a actually a semi-iterative method. It is based on
much simpler yet equally efficient alternative is minimization of a certain error norm and is
to assign the global constraints to an extraneous closely related to the Lanczos method for tri-
node that belongs to every element in the mesh. diagonalizing a matrix. An attractive feature of
The global constraint equations become complete the CG method is that it requires no estimation
only after the assembly of the last element in the of parameters. The conjugate residual method
mesh and are therefore carried in the front (CR) is quite similar, differing mainly in the error
throughout the elimination. functional used.
The number of operations required for the For unsymmetric systems, the most successful
most efficient direct solution of a banded linear class ofiterative methods is that of preconditioned
system increases as the square of the bandwidth polynomial iterative methods, with two important
and as a linear function of the number of subclasses: Chebychev-like methods and conju-
equations. The resulting computational cost is gate-gradient-like methods. Chebychev methods
tolerable for most two-dimensional flows, and (Manteuffel 1977) rely on minimax polynomials
direct solvers have been found adequate for such and require a priori or adaptive estimates of the
applications. However, the cost usually becomes eigenvalue spectrum of the coefficient matrix.
unacceptably high for realistic 3D problems with Moreover, their iteration parameters are inde-
their inherently much larger bandwidths. In- pendent of the initial guess and so there is no
creasing user demands for fully 3D simulations benefit from a good initial guess. Chebychev
has stimulated a resurgent search for methods methods are not suitable when the spectrum of
whose computational expense grows at a slower the coefficient matrix includes the origin. Con-
rate than direct solution. Currently emerging jugate-gradient-like methods generalize the CG
iterative techniques may be able to deliver the method to unsymmetric matrices. The most
needed break-through. popular class uses projection methods onto
special Krylov sub-spaces. The most noteworthy
among these, in turn, are conjugate gradient
squared techniques (Sonneveld 1989) and the
Iterative matrix solvers
stabilized bi-conjugate gradient methods (Van
The key idea behind iterative solvers is to der Vorst 1992), which are similar to CG but use
compute a sequence of approximate solutions by different residual norms; and generalized minimum
means of simple and hence computationally residual techniques (Saad and Schulz 1986),
efficient matrix operations, starting from an which create an orthogonal sub-space ofspecified
initial guess and ultimately converging to the size with Gram-Schmidt orthogonalization. The
exact solution (for a fairly recent review, see last are often considered to be the most robust,
Engelman and Hasbani (1990)). The main ad- but Howard, Connolley and Rollet (1990) found
vantage is that the computational expense is that, for flow problems, bi-conjugate gradient
independent ofthe matrix bandwidth. Moreover, and conjugate-gradient squared methods work
many implementations require only the product better.
ofthe coefficient matrix and a vector, rather than Preconditioning, i.e. pre-multiplying the matrix
the coefficient matrix itself. The main drawback equation by a conditioning matrix, is often
of iterative methods, on the other hand, is the essential to achieving tolerable convergence rates
strong dependence of their performance on with iterative matrix solvers. The intent is to
reduce the condition number (or the spread of is the basis ofa noteworthy finite-element solution
the spectrum) of the resulting linear system. The technique called the segregated solver approach
preconditioning matrix should be sparse, easy to (e.g., Haroutunian, Engelman and Hasbani (1993)),
compute, and easy to invert. Commonly used which first treats the convective terms explicitly
strategies include incomplete factorization to make the equations linear, then uses the
methods based on classical matrix splitting such consistent PPE formulation to decouple pressure
as Jacobi (diagonal scaling), Gauss-Seidel or from velocity, and finally solves the pressure
successive overrelaxation (SOR), and incomplete equation and each individual component of the
LV decompositions (see Jackson and Robinson momentum equations separately and successively
(1985)). Multigrid methods (see Brandt (1977) by iterative techniques.
and Sonneveld, Weissling and de Zeeuw (1985)
may also be viewed as a special form of 9.2.4 INITIALIZATION AND PARAMETER
preconditioning. CONTINUATION PROCEDURES - SOLUTION
With the exception of a few isolated cases (e.g., FAMILIES
Einset and Jensen (1992); Carey, Wang and
Joubert (1989); Howard, Connolley and Rollet 9.2.4.1 Standard continuation
(1990)), even the most robust iterative methods
Newton's method and its variants are currently
perform poorly at the core of Newton or Picard
preferred for computing coating flows. Their
procedures for solving fully coupled Navier-
superior rate of convergence, however, can only
Stokes problems. In particular, the matrix systems
be realized in a small neighborhood ofan isolated
resulting from the mixed velocity-pressure (u-P)
solution, called the basin of attraction. The
formulation are indefinite and poorly conditioned
boundary of the basin of attraction has a very
(Engelman and Hasbani 1990), often causing
complex structure and is often offractal dimension
convergence that is erratic and distressingly slow.
(Peitgen and Pruefer 1984). Outside the basin,
Achieving robust convergence properties for
usually characterized by the radius ofits inscribing
general Navier-Stokes problems remains one of
(hyper)-circle called the radius of convergence,
the outstanding challenges in numerical analysis.
the iteration diverges or oscillates indefinitely.
At present, the key to successful application
For that reason, Newton-like methods are
of iterative methods to incompressible Navier-
preferably used in combination with a continuation
Stokes problems comes from highly specialized
technique which produces an initial estimate
matrix preconditioners that depend on detailed
from one or more known solutions at nearby
knowledge of the matrix structure (e.g., Einset
parameter sets.
and Jensen (1992)). Resorting to transient analysis
The most popular continuation technique is
is another way to overcome poor performance of
first-order or Euler-Newton continuation (Keller
iterative solvers (see for example Mittal and
1977). Given a solution Yo == y(P0) of the governing
Tezduyar (1992); Strigberger et al. (1993)). The
equations at a parameter value Po, an approxi-
idea is to take advantage of iterative solvers
mation of the solution at a nearby parameter
thriving on a good initial guess - which transient
value Po + (jp is predicted along the tangent
analysis with small enough a time increment
vector to the solution branch at Yo from
delivers at each time step. Also, the smaller the
time step the more diagonally dominant the oy
matrix system, because the time derivatives occur
Yo(Po + (jP) = y(P o) + op(jP (9.29)
on or near the diagonals. For this strategy to be
The derivative oy/oP is calculated from the
effective, of course, one has to avoid solving the
solution of the linear system
continuity equation which lacks time derivatives,
for instance by solving the pressure Poisson oR oy oR
equation (PPE) instead (see Section 9.2.1). This oy (y(P 0)' Po) oP = - oP (9.30)
Here oR/oy (y(Po), Po) is the Jacobian matrix at

Yo' whose LU decomposition is already available IIxll
from the preceding Newton iteration. Therefore,
the additional cost of continuation amounts only
to one re-solution (i.e. solution of two triangular
systems) with the different right-hand side
-oR/oP.
Another popular continuation technique,
BIFURCATION
applicable when converged solutions at more POINT
than one point along a solution branch are
available, uses an interpolation polynomial p
through these points to produce a good estimate
for the solution at a new point (e.g., Jackson and
Winters (1984)). Usually only two or three previous Figure 9.6 Schematic of turning points and bifurcation
points are used because higher-order interpolation points, which are singular points of a two-dimensional
polynomials tend to become oscillatory and free surface flow; and Hopf bifurcation points which
are regular, or non-singular (lIxll denotes a norm of
eventually unstable. Sophisticated methods for the solution, such as arc-length or another variable
selecting the parameter step-size along the solution that changes monotonically in the vicinity of the
branch have been proposed by Heijer and critical point; P is the bifurcation parameter; the circles
Rheinboldt (1981) and Rheinboldt and Burkardt past the Hopf point, or rather their distance from the
(1983), for example. Often, however, simple steady-state branch, designate the amplitude of the
heuristic strategies suffice, for instance doubling bifurcating, time-dependent solution).
or halving the step size depending on the
convergence history (e.g., Jackson (1987)). Similar interpolation allows stepping around turning
techniques can be used to adjust the time-step points (e.g., Jackson and Winters (1984)), but the
size in integration ofthe transient equations, with singularity of the Jacobian continues to hamper
additional controls for the time truncation error convergence. A more robust procedure is to
(see Section 9.2.9). introduce a new parameter which increases
monotonically, such as the arc-length s along the
solution family, defined by
9.2.4.2 Continuation around singular points
Newton's method is guaranteed to converge (y - YoHy - Yo) + (P - P O )2 = (s - so)2
provided that the initial guess for Yis sufficiently (9.31)
close to the solution and the Jacobian matrix is
nonsingular. Convergence becomes difficult or More commonly, a 'pseudo-are-length', defined
impossible, however, close to points where the by a linearizerd version of (9.31), is used (see
Jacobian matrix becomes singular. Prime examples Keller (1977) for details). Equation (9.31) or its
are turning points where a solution branch turns linearized version is added to the original equation
back on itself, or bifurcation points where one set and the augmented system is solved for (y, P),
solution branch meets with another; see Fig. 9.6. with s serving as the continuation parameter.
So called Hopf points that mark the onset of The Jacobian of the augmented system is
spontaneous oscillations usually cause no con- nonsingular at simple turning points (Keller
vergence problems because the Jacobian is 1977), but remains singular at simple bifurcation
nonsingular there. points. Arc-length continuation can still be used
First-order continuation can still be used to to step over the latter, but special techniques
step over bifurcation points (Keller 1977) but may be necessary to ensure convergence and to
fails at turning points. Quadratic or higher order switch branches reliably (Keller 1977).
Abbott (1978) suggested choosing as pseudo- and no-shear stress conditions applied at free
arc-length a component of the solution vector, surfaces. The resulting solution is often a good
preferably the one that locally changes the fastest enough initial estimate for the free-surface flow,
along the solution branch. This amounts to especially when under-relaxation is used for the
switching the parameter for a component of the first few Newton steps. Another alternative is to
solution vector. Such parameter switching can start from an ad hoe initial guess but to precede
be invaluable not only for stepping over turning Newton iteration by a few steps of successive
points, but also for getting a first solution when substitution which, for a proper choice of the
the parameter range for which a two-dimensional distinguished boundary condition at free surfaces,
steady state exists is unknown (e.g., vacuum usually has a larger radius of convergence.
range in slot coating). It is also useful when static
contact lines of coating flows approach sharp 9.2.5 LINEAR STABILITY ANALYSIS WITH
corners, where Gibbs' condition (9.9) applies RESPECT TO 2D DISTURBANCES
(pinning the contact line to a sharp edge by
introducing a geometrical or sometimes compo- Steady, 2D states of coating flows are often
sitional discontinuity is a common practice in unstable to 2D and 3D disturbances over wide
industry for achieving a uniform line across the ranges of their characteristic parameters. Some
width of a coating die). Because Gibbs' condition of the disturbed flow fields are steady, others are
is a 'unilateral constraint', it has discontinuous periodic, others may be chaotic in time, and most
derivatives with respect to the contact line location result in unacceptable nonuniformities. For
and hence causes convergence difficulties for the practical applications, predicting the limits ofthe
Newton iteration. Abbott's continuation method operating parameters past which the flow is
eliminates the difficulties by fixing the contact unstable is therefore an essential ingredient of a
line at a given location and computing the comprehensive coating-flow analysis.
required vacuum, for instance. Even though a fully transient analysis of the
complete system of nonlinear equations may be
needed to investigate stability to all types of
9.2.4.3 Initialization disturbances (see Section 9.2.9), an asymptotic
analysis with respect to infinitesimally small,
Continuation methods are useful when a solution
two-dimensional disturbances is sufficient to
at a nearby point is known. When solving a new
predict the critical parameters at the onset of
problem, however, such an initial solution is
many coating flow instabilities. Like conventional
usually not available. To get a very first solution,
procedures of linear stability analysis in hy-
one can sometimes simplify the problem into one
drodynamics (e.g., Drazin and Reid (1981); and
whose solution is known or easy to find, and
also Chapter 8), finite-element stability analysis
then apply a continuation technique to proceed
decomposes the disturbances into their normal
to the sought first solution of the original (i.e.
modes. Two-dimensional disturbances ofvelocity,
not simplified) problem. During the course of
pressure, and free surface position can be
such 'strategic start-up continuation', flow par-
represented by the same set of basis functions as
ameters, boundary conditions and even the
the steady base flow, qi(~, 1/), t/l(~, 1/), qi(~, 1/) (ef
geometry of the solid boundaries can be varied.
equation (9.15)), and exponentials in time:
Christodoulou and Scriven (1989), for example,
managed to obtain a converged solution of u(~, 1/ t) = L [u? + eu; exp(At)] qi(~, 1/)
slide-coating flow starting from a known solution
of curtain-coating flow (Kistler and Scriven p(~,1/,t) = L [p~ + ep~exp(At)]t/Jk(~,1/) (9.32)
1983). An alternative is to solve for the creeping k
flow (zero Reynolds number) with the free
boundaries temporarily fixed and the kinematic
x(~, 1/, t) = L [x? + ex; exp (At)] 4>' (~, 1/)
I
Here u?, p~, x?, are nodal coefficients that is singular because the continuity residuals and
describe the undisturbed steady state; u;, p~, x;, essential boundary conditions do not involve
are nodal parameters that describe the disturbance time derivatives of the nodal unknowns. The
mode; e is the amplitude of the disturbance; and singularity ofM gives rise to 'infinite' eigenvalues
A = Ar + iA j is the complex growth rate whose that correspond to unbounded speeds of propa-
real part determines whether a disturbance decays, gation of sound waves in an incompressible
in which case Ar is negative, or grows, in which medium. There are as many of them as equations
case Ar is positive. The imaginary part Aj is the with no time derivatives, and all should be
temporal frequency of the disturbance. eliminated before the most dangerous ones can
To derive the discretized stability problem, the be computed.
representations of the disturbed fields (9.32) are Christodoulou and Scriven (1988) advanced a
inserted into the weighted residuals (9.17)-(9.19), method that successfully addresses all three
and also the weighted residuals of the mesh challenges and efficiently computes leading
generation equations, e.g., equation (9.27), when eigenvalues and their eigenvectors. The method
the elliptic technique is used. The resulting is based on Arnoldi's algorithm with
expressions are then linearized with respect to e. Schur-Wielandt deflation, developed by Saad
These steps are complicated because the nodal (1989). In contrast to the simple power method
velocities u, pressures p and coordinates x all (e.g., Wilkinson (1965)) and the subspace iteration
depend on e. The end result is a large, sparse, method (Stewart 1978) - which rely on the same
unsymmetric generalized eigenproblem of the form iterative principle as Arnoldi's method but discard
all previous estimates of the eigenvectors except
Jell = AMell (9.33)
the very last one - Arnoldi's algorithm uses the
The elements of the Jacobian matrix J == results from all previous iterations to form the
aRjay (Yo, Y= 0), already defined at (9.28), best approximation of the desired eigenvectors.
measure the sensitivity of the residuals of the The result is an enormous speed up of the rate
Navier-Stokes equations and boundary condi- of convergence of the modes corresponding to
tions to changes in the nodal unknowns (Le., the eigenvalues in the outer part of the spectrum
velocities, pressures, positions). The elements of (Saad 1980).
the mass matrix M == -aRjay(yo,Y = 0) measure To eliminate the infinite eigenvalues mentioned
the sensitivity of the same residuals to changes above, Christodoulou and Scriven (1988) used a
in the time derivatives of the nodal unknowns few steps of 'rational acceleration', a scheme that
(liquid accelerations, velocities of nodal points). filters out the corresponding modes from the
Solutions of (9.33) are eigenvectors ell, which initial vectors (cf Goldhirsch, Orszag and Maulik
represent possible modes of instability, and (1987)). The idea is to perform a few steps of
eigenvalues A, which are their corresponding inverse iteration on the initial vector, much like
complex growth rates. initializing the time integration of differential-
Finding solutions of (9.33) is challenging for algebraic systems by a few steps of backward-
several reasons. First, the eigenproblem is differences (e.g., Kheshgi and Scriven (1984); see
unsymmetric. Therefore, eigenvalues and eigen- also Section 9.2.9). Inverse iteration requires the
vectors may be complex. Second, the size of the LV-decomposition of the Jacobian matrix or of
full eigenproblem rules out computing all eigen- a linear combination of the Jacobian and mass
values. Only the most dangerous (or leading) matrices. To extract the most dangerous eigen-
eigenvalues, Le. the ones of algebraically largest values, Christodoulou and Scriven (1988) used
real parts, should be sought. Most existing approximately exponential preconditioning by
methods for partially solving such problems find rational transformation. This transforms the
eigenvalues ofsmallest modulus and not the most eigenproblem to one whose eigenvalues are the
dangerous ones. And third, the mass matrix M exponentials of the original eigenproblem, and
thereby maps leading eigenvalues to ones of interfacial tension is zero and the power law
largest modulus, which are then easily calculated indices of the two layers are equal. In such cases,
by the Arnoldi algorithm. leading eigenmodes exist at high wavenumbers
When the imaginary part of the most dangerous that cannot be resolved by the Arnoldi method.
mode is large compared with the modulus of the
second leading modes, convergence ofthe method
outlined above is slow. Christodoulou and Scriven
9.2.6 LINEAR STABILITY ANALYSIS WITH
(1988) found that once a crude approximation
RESPECT TO 3D DISTURBANCES - ONSET
to the leading mode has been found in such cases
OF RIBBING
of slow convergence, the complex shift approach
of Parlett and Saad (1985) can resolve the leading Even ifa steady flow is immune to 2D disturbances,
mode efficiently. Used by itself, the complex shift it can be unstable to 3D disturbances that break
approach carries the danger of missing the the translational symmetry and make the flow
leading mode, especially in cases of frequent nonuniform in the third, transverse direction.
mode overtaking. The same is true for Natarajan's The most prominent three-dimensional coating-
(1992) multiple shift-and-invert, restarted Arnoldi flow instability is ribbing, a set of evenly spaced,
algorithm which employs automatic shift selection. stationary free-surface waves in the cross-
Christodoulou and Scriven (1988) applied their web direction that results in a corduroy-type
algorithm to analyze the stability of slide coating pattern in the down-web direction. It often arises
flows to disturbances that leave the flow uniform in thin liquid layers applied by brushing, spreading,
in the cross-web direction. Instabilities usually or rolling (for a recent review, see Coyle (1992)
arise at turning points, where solutions of the and also Chapter 12a). Ribbing is also common
steady-state equation set cease to exist, or at in various pre-metered bead coating devices (for
Hopf bifurcation points, where the flow begins details, see Chapters 11a and 11 b), which can
to oscillate spontaneously. Figure 9.7, taken from also suffer from many other three-dimensional
Christodoulou (1990), illustrates the most dan- instabilities, including rivulets (break-up of the
gerous part of the eigenvalue spectrum for a slide film flow into individual strands or 'streamers'),
coating flow at selected states along a parameter diagonal chatter (or 'chevrons') and even air
family with varying web speed. More generally, entrainment (for the last, see also Chapter 3).
the main findings of Christodoulou and Scriven To predict fully 3D flow states with finite
(1988) are: at low web speeds and/or high flow amplitudes in free-surface elevation, 3D flow
rates a turning point marks the onset of bead models are needed. At present, such models still
flooding; at insufficient pressure differences tax the most advanced algorithms available. To
('vacuum') a Hopf point marks onset of sponta- predict the onset of instabilities that break the
neous bead oscillations; and at large gap clearances translational symmetry of 2D flow, however,
a Hopf point is followed very closely by a turning linear stability analysis by the Galerkin/finite-
point. Static liquid bridges, in contrast, break at element method is again sufficient. The method-
a single turning point (Higgins 1980). ology follows standard procedures of stability
In another illustrative example, Anturkar, analysis, relying on Fourier expansions in the
Papanastasiou and Wilkes (1991) applied Chris- third dimension to describe 3D disturbances (e.g.,
todoulou and Scriven's (1988) algorithm to Davey (1962); DiPrima and Rogers (1969); Meyer
calculate the stability of 2D coextrusion flows of (1969)). Bixler (1982) and Ruschak (1983) were
viscoelastic liquids through long converging dies the first to show that, with such expansions,
and compared critical values of the flow-rate analysis of stability to 3D disturbances can be
ratio between two layers with those predicted by performed on the same 2D mesh as analysis of
a simplified Orr-Sommerfeld-type analysis. They the base flow that is uniform in the transverse
found good agreement in all cases except when (cross-web) direction.
E 1.4
E
(J)
(J)
w 1.0
Z
~
!:? 0.6
:I:
I-
:E
..J STABLE
u:: 0.2
..J
«
Z
u:: ·0.2
0 40 80 120 160 200
WEB SPEED em/s
(a)
3 3 3
0
0
J J
0
0 8 0
-1 ~ 0
0
0
0 0
-3
167.3 em/s 98.5 em/s 69.9 em/s
3 3 3
J 0
0 0
)
0
o 0 0
0
o 0
-1
-1 0 -1 0
0
62.4 em/s 71.3 em/s
(b)
Figure 9.7 Evolution of final film thickness (a) and associated flow field together with leading part of eigenvalue
spectrum (b) for slide coating along a parameter family of varying web speed at a constant flow rate of
1.93 cm 2/ S• Past the turning point at 62.4 em/s, the family turns back toward larger values of web speed (pressure
difference tlP = 161 Pa, coating gap H OAP = 0.347mm, liquid viscosity J.l = 8.717mPas, liquid density
p = 1130kg/m 3 , surface tension (J = 70.1 mN/m, static contact angle 80 = 60·, imposed dynamic contact angle
8D = 160·, slide inclination 30·; from Christodoulou (1990)).
The key idea is to represent 3D disturbances generalized, unsymmetric eigenproblem for the
by 2D finite-element basis functions multiplied finite-element coefficients of the normal modes,
by Fourier basis functions of wavenumber N in <1>, and the corresponding eigenvalues, A:
the cross-web direction, i.e.
J 3D <I> = AM 3D <I> (9.35)
u(~,,,,(,t) = L: [u? + I:U;' D(NOexp(At)] qi(~,,,) with J 3D ==oR/oY(Yo,y=0;N) and M 3D ==
i
-oR/oy(yo, y = 0; N). The main difference from
p(~,,,, (t) = L: [pZ + I:p~cos(N Oexp(At)] ljl(~,,,) 2D stability is that here the Jacobian and mass
k
matrices are functions of the disturbance
x(~,,,,(,t) = L: [x? + I:X;cos(NOexp(At)] </J'(~,,,) wavenumber N, and so are consequently the
I
calculated eigenvectors and eigenvalues. The
y(~,,,,(,t) = L [y? + I:Y; cos (NOexp(At)] </J'(~,,,) most dangerous disturbances must therefore be
I found by calculating growth rates over a wide
z (~, ", " t) = ( (9.34) range of wavenumbers. As wavenumber is raised,
the growth rate ofthe most dangerous mode may
where D == [( ii + jj) cos(N 0 + kk sin(N (n All pass through a maximum but, eventually, it will
other quantities are defined as in (9.32). decrease because capillarity dampens the growth
Ruschak (1983) adopted a computational of short waves. The search for the maximum
domain of infinitesimal thickness in the cross-web growth rate must be repeated at different values
(z) direction. For weighting functions, he employed of a flow parameter of particular interest. The
the 2D ones used in solving for the base flow. onset of a 3D instability is signalled by the
To reduce the order of derivatives in the capillary maximum in the growth rate versus wavenumber
term of the boundary traction integral at the free reaching zero.
surfaces, Ruschak applied the surface divergence As in the two-dimensional case, solving (9.35)
theorem (Weatherburn (1927), p. 123; recently for the eigenvalues A and corresponding three-
rediscovered by Ho and Patera (1991)). Bixler dimensional modes <I> is a major computational
(1982), in contrast, chose a computational domain challenge. Ruschak (1983), and later Coyle,
consisting of the 2D flow domain extended over Macosko and Scriven (1990), took an elegant
one wavelength in the cross-web direction. He short-cut by omitting the time derivative terms
constructed weighting functions of the form in the momentum equation. This procedure
qi(~, rO D for the momentum equations (9.1), statically condenses the large eigenproblem to
ljlR,,,) cos(N () for the continuity equation (9.3), one of much smaller dimension, equal to the
and </J'(~,,,) cos(N 0 for the kinematic condition number of kinematic residual equations. The
(9.4). At free surfaces, Bixler applied the tensor latter can be solved with standard subroutines
form of the surface divergence theorem (9.22) to available in various software libraries. However,
the capillary term of the boundary traction in the resulting approximation, the computed
integral, which again reduced the order of growth rates are correct only for non-oscillatory
derivatives. instabilities at vanishingly small Reynolds numb-
In either Bixler's or Ruschak's approach, the ers. Bixler (1982), in contrast, solved the full
disturbed fields (9.34) are then inserted into the eigenproblem (9.35) without simplifications by
weighted residuals (9.17)-(9.19) and the resulting using inverse subspace iteration (Stewart 1978).
expressions are linearized with respect to I: (in Christodoulou (1990) calculated the most dan-
more modern algorithms with elliptic mesh gerous modes and their growth rates by the
generation, the same procedure is applied to the deflated Arnoldi algorithm ofSection 9.2.5 without
weighted residuals of equation (9.27)). Chris- making any approximations either.
todoulou (1990) showed that both Bixler's and To date, most published applications of 3D
Ruschak's approaches result in the same large, finite-element stability analysis have been aimed
at predicting the onset of ribbing. Quite some For forward roll coating, for instance, it misses
time ago, approximate lubrication-flow analyses several complex modes that become more unstable
of simple configurations such as wedge spreaders than the ribbing mode.
(Pearson 1960), forward roll coating (Pitts and Christodoulou (1990) applied his method also
Greiller 1961) and journal bearings (Savage to slide coating. He predicted that ribbing sets
1977a, 1977b) had made plain that the adverse in when the sub-ambient pressure ('vacuum')
pressure gradient under the meniscus is the root applied under the liquid bridge is reduced, coating
cause of the instability (see also Chapter 12a). speed is decreased, or flow rate is diminished.
Quantitative predictions of the onset of ribbing, Increasing gap clearance stabilizes the flow with
however, were compromised by uncertainties in respect to ribbing when the lower mensicus pins
the lubrication approximation, the most serious at the die edge, but destabilizes the flow when
being the arbitrariness of the boundary conditions the liquid wets the lip underside and the lower
imposed near the meniscus, where the flow is meniscus takes a position past the edge. In
generally far from rectilinear. Bixler (1982), who contrast to lubrication-type flows under spreaders
applied full finite-element stability theory to a and between rollers, the effect of surface tension
liquid film emerging from a diverging coating is small and non-monotonic in the parameter
knife, was the first to compute the onset of ranges examined, and the physical mechanisms
ribbing without any ad hoc assumptions. Unfor- responsible for ribbing in slide coating are still
tunately, because of an elusive error in the not fully understood (see also Chapter 11 b).
derivation of the stability matrices of the 3D Using the same method, Do and Christodoulou
flow, Bixler - and Coyle (1984) after him for the (1992) computed the stability of curtain coating
case of ribbing in forward roll coating - overes- flow (see also Chapter ltc, which discusses
timated the critical capillary number by about curtain coating in detail). They found that 2D
20%. Ruschak (1983) carefully tested his finite- liquid curtains are linearly stable to both 2D and
element stability method against standard results 3D disturbances, with the 2D ones being the
on the stability ofa stagnant liquid layer and Hele- most dangerous. As Weber number (i.e. the ratio
Shaw flow. In spite ofthe success of his algorithm, of inertial forces to capillary forces) was reduced
he later showed that the much simpler lubrication- below We < 2- where, according to earlier pub-
flow theory can accurately predict the onset of lications, a falling liquid curtain is unstable (e.g.,
ribbing in slowly diverging flow sections if it is Lin (1981)) - the most dangerous growth rate
complemented with accurate boundary conditions approached zero but never became positive.
obtained from steady, two-dimensional finite- Further reductions in flow rate caused the lower
element analysis of the 'inner' meniscus region part of the curtain to be pulled by the web while
(Ruschak 1985). Coyle, Macosko and Scriven the growth rate started decreasing. In a recent
(1990) successfully compared their finite-element Orr-Sommerfeld-type analysis, Lin, Lian and
predictions of the onset of ribbing between Creighton (1990) also found that a liquid curtain
counter-rotating rolls with experimental data as is stable to 2D varicose and sinuous disturbances
well as Ruschak's (1985) theory. Christodoulou ofeither the temporally or spatially growing type,
(1990) validated his algorithm against Ruschak's thereby disproving Lin's earlier result (1981) that
(1983) results for a stagnant layer and Hele-Shaw was incorrectly based on Gaster's (1962) trans-
flow at zero Reynolds number. He then proceeded formation. Recently, Finnicum, Weinstein and
to show that, towards higher Reynolds numbers, Ruschak (1993) reported observations of stable
the discrepancy between his exact analysis and liquid curtains in a wide range of flow conditions,
Ruschak's (1 983) quasi-static condensation method including We < 2. They attributed the observed
increases. At high Reynolds numbers Re = 0(10), stability at low We to nonlinear effects that may
the approach used by Ruschak (1983) and Coyle, stabilize disturbances after they grow to a finite
Macosko and Scriven (1990) may fail altogether. amplitude. According to the analyses by Lin,
Lian and Creighton (1990) and Do and Chris- conditions (Melhem and Rheinboldt 1982): (1)
todoulou (1992), however, it is the occasionally the tangent vector to the solution branch is
observed instability for We < 2 (low flow rates) perpendicular to the parameter axis; (2) the
rather than stability that arises from nonlinear Jacobian has a zero eigenvalue; (3) the Jacobian
effects (i.e. large amplitude disturbances such as has a nonzero right null-vector; (4) the Jacobian
those caused by bubbles). has a nonzero left null-vector; or (5) the deter-
minant of the Jacobian vanishes. For Hopf
bifurcation points, the extra condition is the
9.2.7 OPERABILITY ANALYSIS TO 20 AND
existence of a purely imaginary eigenvalue. The
3D DISTURBANCES-COATING WINDOW
additional equation set for turning point tracking
As a consequence of the nonlinearity of the is of approximately equal dimension as the
governing system, at a given set of parameters, original problem, and for Hopf-point tracking
a coating flow may have no steady state, one the set is twice as large. Efficient critical point
steady state, or multiple steady states. As tracking algorithms avoid solving linear systems
parameters are varied, multiplicity and stability twice or three times as large as the steady-state
of states change at certain parameter values - set - as they would arise from direct application
usually called critical points. In designing and of Newton's method - by taking advantage of
operating coating systems, it is important to the block structure of the augmented system.
know how many steady states exist at a given One Newton iteration for finding turning points
set of parameters, whether these states are stable reduces then to solving two linear systems of the
to disturbances, and how they are connected in same dimension as the steady-state equation set,
the space of parameters. This, in turn, puts a and only twice that dimension for Hopfbifurcation
premium at devising computational schemes that points. Altogther the cost of computing a critical
automatically locate critical points in order to point is only a few times that of computing a
construct operability diagrams efficiently, which steady state.
are invaluable in technological applications. Continuation methods can be applied to the
With continuation and linear stability analysis, augmented system in order to track paths of
critical points can be located approximately. critical states, and to step over bifurcations,
Constructing operability windows that are of points ofisola birth and cusps. The last are simple
practical interest, though, requires locating critical turning points of the augmented system itself and
points in one parameter more accurately and are especially important for the dynamic response
tracking their paths as other parameters are of the system. Examples of applications to flow
varied. An efficient means to do so is to augment problems with fixed boundaries have been given
the steady-state equations with an additional by Cliffe (1983), Cliffe and Spence (1984), Winters,
equation set that uniquely characterizes the Cliffe and Jackson (1987) and Jackson (1987).
critical point, and to solve the augmented set Ungar and Brown (1982) used Abbott's (1978)
with a Newton-like method that simultaneously method to compute turning points and bifurcations
finds the critical parameter value and the critical of a rotating cylinder, a capillary statics problem.
flow state. Such critical point tracking algorithms Christodoulou (1990) tried Abbott's as well as
have been reviewed by Melhem and Rheinboldt Poenisch and Schwetlick's (1981) algorithms to
(1982), and also by Schwetlick (1984) who focused follow turning points of 2D slide coating flow,
on constrained optimization problems. and uncovered a wealth of structure in the
parameter space. He also proposed an extension
of Abbott's methods for computing Hopf points
9.2.7.1 2D disturbances
that is easier to implement in existing finite-element
To locate turning points, the steady-state equations codes than a method proposed by Jepson
can be augmented with one of the following (1981).
9.2.7.2 3D disturbances state <l>x is the greatest among all the entries of
$. Solution of the system (9.36) gives the critical
~hen a coating flow loses stability to 3D
state y, critical parameter value P, nullvector $
disturbances before it becomes unstable to 2D
and critical disturbance wavenumber N. Chris~
ones - ~s at the onset of the ribbing instability
todoulou, Scofield and Scriven (1990) used a
(see SectIOn 9.2.6) - computing the corresponding
rather inefficient extension of Abbott's method
symmetry-breaking bifurcation points and tracing
that required eight LV decompositions and
out their paths in parameter space is more
several back-solutions per Newton iteration. An
important than computing and tracking turning
algorithm similar to Poenisch and Schwetlick's
or Hopf bifurcation points of the 2D flow states.
(1981) would require only two LV decompositions
Cliffe (1983) was the first to consider extended
per iteration but is more difficult to imple-
systems for the direct computation of bifurcation
ment.
points at which stability of translationally invariant
C~ristod~ulou, Scofield and Scriven (1990)
flow states is lost. Werner and Spence (1984)
applIed theIr method to predict the maximum
proposed more efficient systems that capitalize
allowable pressure difference across a slide coating
on the symmetry of the underlying steady flow.
bead without ribbing. They found that this upper
Applications have focused on the axisymmetric
vacuum limit increases with web-to-slide angle
Taylor flow between concentric rotating cylinders
in the range between O· and 90· and has a
and the onset of Benard convection in a liquid
maximum at 90· or above, depending on the final
layer heated from below (e.g., Cliffe (1983); Cliffe
coated film thickness. These trends corroborate
and Spence (1984); and Winters, Cliffe and Jackson
experi~ental measurements by Burket, Conaghan
(1987)). In these cases, however, the bifurcating
and Hushburg (1984). Christodoulou, Scofield
flows are still 2D (or axisymmetric) and the
and Scriven (1990) also examined the effect of
methods used are not directly applicable to the
w~ter redi~tribution between the layers oftwo-Iayer
ribbing instability. Christodoulou, Scofield and
slIde coatmg, keeping the total amount of water
Scriven (1990) proposed a system for tracing
and thus the dryer load constant, and found that
paths of 'pitchfork' bifurcations that break the
viscosity stratification raises the upper vacuum
translational invariance of a 2D flow. Their
limit. Francis et al. (1990) applied the same
augmented system consists of the equations for
method to slot coating flow and found that the
the 2D steady flow, the eigenvalue problem of
upper vacuum limit increases with die-to-web
the linear stability analysis to 3D disturbances
angle at the downstream meniscus.
(with a 2D mesh and a single Fourier mode in
To date, quantitative comparisons between
the cross-web direction, see Section 9.2.6), and
theoretica~ predictions for the onset of ribbing
the condition that the maximum growth rate at
and exp~nments have been published only for
the most dangerous wavenumber vanishes:
roll coatmg (cf Chapter 12a). For slide coating
R(y,P) =0 unpublished work of one of us (KNC) suggest~
that computed values of the upper vacuum limit
(J 3D - A.M 3D) $ = 0 agree with experiments to within 20-30%, even
<1>, == e; .$ = 1 (9.36) when dynamic and static contact angles are
taken to be constant. The lower vacuum limit
dA./dN = 0 in contrast, is highly sensitive to the contac~
A.=O angles imposed as boundary conditions and
reliable predictions cannot be made unless
Here, the eigenvector is normalized with accurately measured angles are available - or
~, == ~O, .... , 1, ]T, the unit vector whose rth entry much needed theories are advanced for predicting
IS umty and all others are zero; the distinguished contact angles from first principles (see also
index r is chosen so that at the last computed Section 9.3.1).
9.2.8 SENSITIVITY ANALYSIS TO 2D AND 3D disturbance;F == oR/oP(yo,y = 0) is the sensitivity
DISTURBANCES - QUALITY WINDOW vector of the weighted residuals with respect to
E.ven if a coating flow is stable so that transitory the parameter P that is forced to oscillate' and
dIsturbances damp out with time, it can be z is the complex nodal response vector. Sensi~ivity
measurably sensitive to ongoing disturbances at each frequency w is then computed as the
coming from mechanical vibrations, pump pul- magnitude of the response compared with the
sations, speed variations, inhomogeneities in the magnitude ofthe disturbance. To avoid complex
web surface, etc. Such sensitivity is of practical arithmetic, for which frontal solvers are not
significance because it translates directly into commonly available, the real and imaginary
coating quality and, hence, ought to be predicted parts of (9.37) can be separated - at the expense
as part of a complete analysis. Tracking the effect of doubling the dimension of the linear system.
oflarge-amplitude disturbances requires transient Reordering the equations in pairs of real and
analysis of the full nonlinear equation set (Section imaginary parts minimizes the bandwidth, only
9.2.9). Such analysis can be quite costly, especially doubling it from steady-state analysis.
when sensitivity to a wide range of disturbances Van Abbenyen, Christodoulou and Scriven
and frequencies needs to be studied. For small- (1989) demonstrated the utility of their method
amplitude disturbances, transient analysis of the for slide coating. For parameter disturbances up
linearized equations set is adequate and more to 20% of their steady-state values, frequency-
efficient than the nonlinear transient analysis response predictions, expressed as an amplitude
(Kaneko and Scriven 1988), but shares some of ratio, compared well with solutions of the fully
its major shortcomings. The most atttractive and nonlinear transient system, as obtained earlier
efficient method is frequency-response analysis, by Katagiri and Scriven (1986). The cost of the
quite similar to the analysis of small-signal linear frequency response analysis, however, was
response of nonlinear electrical systems. about two orders of magnitude lower than that
of the transient analysis. Both approaches detected
the same two resonant frequencies, one associated
9.2.8.1 2D disturbances with the upper free surface and the other with
In the finite-element based frequency-response the lower one. Frequency-response analysis also
analysis of Van Abbenyen, Christodoulou and uncovered an important interfacial mode in
Scriven .(1989) (see also Christodoulou (1990)), two-layer flow and showed that forcing different
small dIsturbances of a particular parameter, parameters excites distinct eigenmodes of the
usually one entering a boundary condition, are base flow. Finally, the analysis made clear that
decomposed into their Fourier spectrum, and so the susceptibility to forced disturbances strongly
are the resulting perturbations of the coating depends on whether the lower mensicus remains
flow. The responses of velocities, pressures and pinned at the corner or not (see also Chapter 11 b).
nodal positions are discretized by the same set Van Abbenyen et al. (1990) and Bussmann,
of basis functions as is used to represent the Hoffmann and Beck (1992) attempted to verify
steady 20 flow. The Navier-Stokes system computed frequency responses with experiments,
governing the perturbed transient system reduces also for slide coating. Agreement with experiments
to a linear system ofcomplex algebraic equations was acceptable only if the vacuum imposed in
of the form: the model was adjusted to ensure the computed
and measured dynamic contact lengths matched.
(J + iwM)z = -F (9.37) The disagreement between theoretical and ex-
perimental frequency response at identical vacuum
Here J and M are the Jacobian and mass levels was attributed by these authors to uncer-
matrices of the linear stability analysis, already tainties in modeling dynamic wetting, unknown
defined at (9.33); w is the frequency of the external liquid rheology at the high deformation rates
close to contact lines, and surface-tension gradient systematic coating-flow analysis. It is useful not
effects ignored in the model. only to predict coating windows, but also to
assess the sensitivity to disturbances of every
steady state within a coating window and
9.2.8.2 3D disturbances
therefrom delineate a quality window where
As transitory disturbances can be nonuniform in resulting film-thickness uniformity is acceptable.
the cross-web direction, so can ongoing and As an example, Fig. 9.8 (taken from Hackler
steady disturbances. Examples include spot-like et al. (1992» shows a computed coating window
inhomogeneities on a web surface, cross-web for a single-layer slide coating, complemented
variations of the pressure difference ('vacuum') with contours of sensitivity to two different
imposed over a coating bead, and damaged disturbances in imposed pressure difference
surfaces and edges (scratches and nicks) ofcoating ('vacuum'): one is a steady, 3D disturbance (Fig.
dies - which all can result in unacceptable spot- 9.8a); the other is 2D but oscillatory (Fig. 9.8b).
or streak-like defects. The method of frequency These disturbances correspond to two peaks
response analysis can be extended to predict identified beforehand in the 3D response surface
cross-web flow and film-thickness variations in of the amplitude ratio as a function of frequency
addition to ones in the machine direction. To and wavenumber. Figure 9.8 illustrates that,
describe the perturbations, Hackler et al. (1992) contrary to a belief widely held among practi-
combined Fourier expansions in both time and tioners, parameter regions of least sensitivity to
cross-web coordinate with finite-element basis disturbances - and therefore best product quality
functions in the two space coordinates of the - do not necessarily coincide with regions of
underlying uniform flow. As in linear stability maximum coating latitude, defined as the difference
analysis to 3D disturbances (Section 9.2.6), only between upper and lower 'vacuum' limits (the
a single (representative) Fourier mode was used lower limit, usually arising from air entrainment
in the transverse direction. The Navier-Stokes or bead breakup, is taken here to be zero because
system was thereby transformed into a set of experiments suggests it is typically well below
linear algebraic equations similar to (9.34) but the upper limit and accurate theoretical predictions
now depending on two parameters, the disturbance are still difficult for reasons indicated above).
frequency wand the cross-web wavenumber N: Figure 9.8 also makes evident that the region of
least sensitivity to one disturbance can be quite
(9.38) different from that to another. Hackler et al.
Here J 3D and M 3D are the Jacobian and mass (1992) concluded that, to maximize the benefit
matrices already defined at (9.32); F 3D : ; from computational sensitivity analysis, the
8R 3D /8P (Yo,)' = 0) is the sensitivity factor of the prevalent ambient disturbances ought to be
weighted residuals of the 3D equation set with identified and measured in experiments first.
respect to the parameter P that is forced into
oscillation; and z is the discretized response vector
9.2.9 TRANSIENT ANALYSIS
which, here, depends on both wand N. Hackler
et al. (1992) computed the maximum amplitude Time-dependent analysis of the full, nonlinear
ratio of the response to the forced disturbance by equation set (9.23) is indispensable to study the
direct search in the (w,N) space. Their results for large-amplitude transients associated with start-
slide coating show that 3D disturbances can be up, splice passage and termination of coating.
amplified even when 2D ones are strongly damped, Time-dependent analysis is also useful to inves-
especially close to the onset of ribbing. tigate the fate of liquid films after coating (see
The combination of linear stability analysis, Chapter 6), such as the leveling of nonuniformities
direct tracking of critical points, and frequency resulting from moderate (i.e. non-catastrophic)
response analysis is a comprehensive strategy for instabilities in the coating bead, the local break-up
180 180 (b)
160
140
'""'
'""'
e
('II
e
('II
E E 120
::l ::l
::l ::l
U
U ('II
>
('II
> 100
'0
100
"0 ('II
('II Qj
Qj
!Xl !Xl
80 80
60
200 250 300 200 250 300

Web Speed (em/sec) Web Speed (em/sec)
Figure 9.8 Coating window of single-layer slide coating, showing 'upper vacuum limit' as well as contours
of sensitivity to perturbations in the imposed pressure difference ('vacuum'): (a) steady perturbation (w = 0)
and wavenumber N = 30/cm; (b) two-dimensional perturbation with w = 45 Hz and N = 0 (from Hackler et
al. (1992».
and de-wetting ofcoated films due to contaminants boundary conditions make the mass matrix
or defects and, ofcourse, drying and solidification. aRjay singular, and the linkage between velocity
A complete time-dependent analysis can also be and time derivatives renders 'mass-lumping'
a 'poor man's' way of performing stability and questionable (Engleman and Sani 1983).
sensitivity analysis. In fact, such analysis may be Differential/algebraic equation systems (DAES)
the only way to predict and track the complete like (9.23) can be quite different from ordinary
flow breakdown that results from catastrophic differential equation systems (ODES) and nu-
instabilities. Transient analysis, combined with merical methods for solving ODES do not
iterative solution of the matrix problem at each always work for DAES (Petzold 1982). DAES
time step, can furthermore be a pragmatic, yet are classified according to the degree ofsingularity
successful approach to tackling large, three- of the mass matrix aRjay. The minimum number
dimensional free-surface flow problems (Ro and of times that all or part of the constraints of a
Patera 1990; Wambersie and Crochet 1992). DAES must be differentiated with respect to time
Algorithms for integrating the nonlinear in order to obtain an ODES is called the index
differential and algebraic equation system (9.23) of the DAES. The index has important implications
in time must cope with the fact that the time for the numerical solution of the system, which
derivatives yenter the system implicitly, although encounters serious difficulties when the index is
usually linearly. In the u- P formulation, the time two or higher with regard to: (I) finding initial
derivatives cannot be solved for explicitly - and conditions that satisfy not only the algebraic
implicit schemes are therefore mandated - because equation subset, but also the hidden algebraic
the algebraic constraints of continuity (9.3), mesh constraints, i.e. constraints that would come up
generation such as (9.27), and various essential during the reduction to a differential system; (2)
developing stable integration algorithms; and derivatives for the next predictor step are evaluated.
(3) developing estimates of the time-truncation Depending on the predictor and corrector used,
error. special procedures may be needed for the first
The Navier-Stokes system discretized with few time steps. The Adams- Bashforth/trapezoid-
the u- P formulation has index two (Petzold rule integration, for instance, is best initiated with
1982). In free surface problems, the nodal position three or four backward-difference steps to generate
unknowns appear explicitly in the mesh generation the time 'history' needed for the second-order
equations, e.g. (9.27), as do their time derivatives scheme and to ensure the initial trajectory is
in the kinematic condition (9.4). As only one devoid of spurious transients that can result from
time-differentiation yields time derivatives of all inappropriate initial data for the index-two system
nodal position unknowns, the index (two) of the (e.g., Gresho, Lee and Sani (1980); Kheshgi and
Navier-Stokes system is not affected by the Scriven (1984)). Another key to efficient time
presence of free surfaces. integration is automated time-step size control
Christodoulou and Scriven (1992) successfully that sets each step so as to keep the local time
employed the differential-algebraic solver DASSL truncation error within a user-specified tolerance
by Petzold (1982), thereby demonstrating that (Gresho, Lee and Sani 1980; Engelman and Sani
sufficiently sophisticated package solvers can 1983; Kheshgi and Scriven 1984). With the
provide an easy yet effective means to perform predictor/corrector scheme and automated time-
the integration in time of (9.23). Most finite- step control, Newton's method often requires
element codes for viscous free-surface flow analysis, only one iteration per time step for convergence.
however, incorporate a finite-difference discret- In addition, the Jacobian may not have to be
ization in time as integral part of the algorithm reformed every time step, analogously to quasi-
(Frederiksen and Watts (1981 ), early on pioneered Newton methods in steady-state analysis (cf
a finite-element discretization of both time and Section 9.2.3). As an added boon from judicious
space which, more recently, is gaining popularity time-step limitation, the matrix systems become
again: see, for example, Mittal and Tezduyar more and more diagonally dominant, even in
(1992)). State-of-the-art algorithms use pre- fully implicit schemes, and thereby maximize the
dictor-corrector schemes, many with a second- effectiveness of iterative solvers (e.g., Liou and
order Adams- Bashforth rule as predictor and a Tezduyar (1991); see also Section 9.2.3).
second-order unconditionally stable trapezoid Finite-element algorithms with fully implicit
rule as corrector (Gresho, Lee and Sani 1980; time integration have been applied to a wide
Engelman and Sani 1983; Kheshgi and Scriven variety of free-surface flow problems, including
1984). Simpler schemes employ first-order ex- large-amplitude waves traveling down inclined
trapolation and implicit Euler correction (Keun- plates (Kheshgi and Scriven 1984); leveling and
ings 1986). A typical algorithm recurrently executes de-wetting phenomena (Kheshgi and Scriven
five essential steps. First, the predictor extrapolates (1988); see also Chapter 6 in this volume); the
the flow field, free boundaries and internal node response of coating beads to step changes in
locations, the time derivatives being known from coating conditions (Kheshgi and Scriven 1984;
the preceding time steps. Second, the finite-element Katagiri and Scriven 1986); frequency response
method reduces the partial differential equations of coating beads to externally imposed forced
to a set of ordinary differential equations. Third, disturbances (Katagiri and Scriven (1986); see
the corrector discretizes the set in time and also Section 9.2.8, and Chapters 11b and 11 c);
reduces it to a nonlinear system of algebraic periodic oscillatory flow states that ensue past a
equations. Fourth, Newton's method solves the Hopfbifurcation point as predicted independently
system simultaneously for the displacement of by linear stability analysis (Katagiri and Scriven
the free boundaries and internal mesh nodes 1986; Christodoulou and Scriven 1988); and the
together with the flow field. And fifth, the time catastrophic transient response of a coating bead
to small parameter changes near a limit point 9.2.10 INVERSE PROBLEMS - PROCESS AND
(Christodoulou and Scriven 1992). In spite of PRODUCT DESIGN AND OPTIMIZATION
these successful applications, the capabilities
and limitations of implicit finite-element A coating system consists of coating hardware
algorithms have not been fully explored. Arche- (process) and a coated film on a substrate (product).
types of transient coating phenomena such as Traditionally, process and product design are
start-up or splice passage, for instance, remain pursued in two independent stages. The goal of
formidable computational challenges that have process design is typically to determine the
not been satisfactorily tackled. Frederiksen and hardware parameters for best coatability of a
Watts (1981) investigated the inception of dip range of predetermined products. The goal of
coating from a pool with their space-time product design is to determine the physical
finite-element formulation, but their results are parameters (rheology, surface properties, etc.) of
based on coarse meshes and must be considered a new product for best product performance and
preliminary. coatability in a given process, the latter quite
To enhance the efficiency of time-dependent often even playing a secondary role. Computa-
free-surface flow analysis, especially for 3D cases, tional analysis of coating flows is usually fitted
semi-implicit methods can offer some distinct in as a tool for understanding the complex
advantages (Ho and Patera 1990; Wambersie interplay of the forces that control the process
and Crochet 1992). Most of these methods are and predicting process effects on product quality.
based on the pressure Poisson equation (PPE) Such a subdivision of the global process and
formulation introduced in Section 9.2.1 (see product optimization problem is prone to result
equation (9.25)). For free-surface flow problems, in suboptimal solutions.
the PPE formulation is usually combined with The real challenge is to design the whole flow
Picard iteration (e.g., Wambersie and Crochet system to satisfy certain quality, productivity or
(1992)). De-coupling free-surface and flow-field cost constraints (see also Chapter 15 in this
calculations avoids any extra complications from volume). This goal can only be achieved by
the presence of free surfaces. solving 'inverse problems' - which are more
As another alternative, Krylov subspace pro- difficult than direct problems because they are
jection methods ought to be extended to transient often ill-posed and thus require regularization
viscous free surface flows, for they hold significant techniques.
promise for improved efficiency. Until recently Design and analysis of coating processes gives
these methods were used only for linear stability rise to three kinds of inverse problems. The first,
analysis (see Section 9.2.5), but now they are and simplest, is to predict a control or design
gaining popularity for the calculation of steady parameter so that the corresponding flow satisfies
and also transient states of fixed boundary flows a given condition. An example is the problem of
(see Edwards et al. (1994)). The key idea is to finding the maximum pressure difference
project the discretized transient system on Krylov ('vacuum') applied across a coating bead that
sub-spaces that are constructed by the Arnoldi leaves the static contact line pinned at a die tip.
process from evaluations of the Navier-Stokes Pinning the contact line there is often advan-
residual and Jacobian without inversion of the tageous as it precludes thickness variations
viscous operator. Incompressibility is handled resulting from contact lines that are not straight
via the PPE formulation in the vein ofsemi-implicit across the width of the die, and may also
schemes. Nonlinearities result in an integral minimize the susceptibility of the coating bead
equation which is solved by functional iteration. to external disturbances (see Section 9.2.8). A
The final result is a small system that is efficiently single control or design parameter can be found
solved through exact exponentiation, or so called by augmenting the governing equation set with
'exponential propagation', of the coefficient matrix. a global (or local) constraint which provides the
additional equation necessary to calculate the 'fractal' solutions with infinitely detailed structure.
unknown parameter value. This procedure is Special regularization techniques may be required
exactly Abbott's (1978) method of continuation to impart desired smoothness to the solution of
(see Section 9.2.4). convex optimization problems (e.g. Chung (1985)).
The second inverse problem is that ofdie shape
optimization to achieve desirable characteristics
9.3 SPECIAL TOPICS
of the global flow field. Examples of shapes that
are readily amenable to this sort of optimization
9.3.1 CONTACT LINES
include the cross-sections of distribution cavities
to achieve a uniform flow profile across the width An essential element ofindustrial coating processes
of the coater (see Chapters 10 and 15), or the is continuous film formation at a static contact
shapes of extrusion slots that result in extrudates line where the free surface of the coated layer
of a given shape or thickness profile (e.g., Yokoi separates from a solid coating device; static
and Scriven (1989); Wambersie and Crochet wetting often also influences or controls the
(1992)). In these cases, the governing systems position of the upstream mensicus of a coating
need to be augmented by several equations that bead (see, for example, Chapters lla and llb in
result from the discretization of the differential this volume). When a contact line is free to
equation that expresses the desirable flow con- migrate along a solid surface, its position is set
dition. by the static contact angle Os' For the purpose
The third, and most difficult, inverse problem of macroscopic flow computations, Os can be
is to identify a single flow or geometrical parameter considered a material property (see Chapters 2
that results in optimal coatability, as quantified and 3). At a sharp corner, the contact line has a
for example by widest operability latitude for propensity to remain pinned. The critical condition
changes in 'vacuum' or lowest sensitivity to for either advance or retreat is Gibbs' inequality
disturbances. To accomplish this, a nonlinear condition (see Chapter 3) which can be written
eigenvalue problem and additional constraints as a set of complementary conditions of the form
have to be solved together with the original (9.9) (Christodoulou and Scriven 1992), as listed
governing system, as described in Section 9.2.7. in Fig. 9.2. Most published, computational analyses
A yet more difficult case arises when not a single of coating flows force a static contact line to be
parameter but the shape of an entire piece of either pinned or free to migrate a priori, and
hardware needs to be optimized, as for example require cumbersome remeshing procedures to
the shape of a coating lip in a slide coater that accommodate transitions from one wetting
ensures widest latitude in 'vacuum'. The sought configuration to another (see, for instance, Kistler
after shape must then be parameterized as an and Scriven (1994)). Christodoulou and Scriven
unknown free boundary much in the same way (1992), in contrast, devised an elegant scheme
as a :ree surface. Again, the extra equations that imposes the conditions (9.9) as a unilateral
required for determining its location arise from solver constraint and deploys several collapsed
the optimality condition. Application of such finite elements at the sharp corner automatically
optimization to coating flows appears to not to accommodate solutions in which the meniscus
have been reported to date. Well-established advances past the corner (see Fig. 9.5b). Evidence
techniques for optimization of solid structures published to date indicates that Gibbs' inequality
(e.g., Chung (1985); Haftka and Grandhi (1986); applies in the presence of flow as it does under
Kikuchi et al. (1986)) could possibly be adopted, static conditions (Kistler and Scriven 1994), but
but may have to be advanced considerably to further comparisons between computed predic-
handle the complicated solution spaces ofcoating tions and experiments are needed to establish
flows. Some optimization problems may be unequivocally under what conditions successful
nonconvex, producing non-smooth and often contact line pinning is independent of the
Special topics 335
surrounding flow. In addition, care must be boundary condition at the putative wetting line
taken to account properly for the well-known - even though apparent dynamic contact angles
contact angle hysteresis that afflicts static wetting observed in experiments vary with coating
on nonideal surfaces encountered in most coating conditions and, in complete theories, ought to
devices (see Chapter 3). Hysteresis in static be predicted variables. Worse yet, even if the
wetting line position may also arise from purely proper value of the local contact angle were
hydrodynamic effects, i.e. multiplicity of steady- known, the singularity at the wetting line is so
state solutions (e.g., Kistler and Scriven (1994)). severe that the boundary condition for the angle
The latter can be predicted with suitable com- cannot be satisfied in rigorous mathematical
putational procedures, outlined in Section 9.2.4 treatments.
and 9.2.7. To cope with the difficulties in modeling
A special case arises at static separation lines dynamic wetting, five distinct procedures have
of interfaces between two miscible fluids in been adopted in finite-element codes for coating
multilayer coating. Because miscible fluids inter- flows. The first restricts the analysis to conventional
diffuse, an inter-layer forms that has no interfacial continuum theory at macroscopic length scales.
tension and, hence, a contact-angle boundary It removes the multivalued velocity at the wetting
condition is inappropriate to locate the separation line by ad hoc boundary conditions that allow
line. Scanlan (1990) enforced continuity in pressure perfect slip at the line (i.e., the liquid velocity at
across the interface where it intersects the solid the wetting line is the same as that of the line)
as an additional condition to complete the and partial slip nearby, and prescribes a contact
equation set. Attempts to impose continuity of angle that is supposed to represent an apparent
total stress have been less successful because dynamic contact angle of the sort observed in
deformation rates are very small near the experiments, possibly taken from an empirical
separation line (Dheur and Crochet 1987; Scanlan correlation. This approach is by far the most
1990). A contact angle condition and an extrapo- common and permits computed predictions of
lation of the interface to the solid surface have realistic coating configurations (e.g., Christodoulou
also been tried (Dheur and Crochet 1987), but and Scriven (1989)). However, the approach
both are ad hoc procedures that lack a physical defeats a major purpose of performing computer-
basis. Away from the immediate proximity of the aided analysis, for it presupposes one of the most
stagnation line, nonetheless, computed solutions important aspects of the coating flow dynamics.
appear to be insensitive to the particular boundary The imposed angle can significantly affect the
condition chosen (Dheur and Crochet 1987). predicted flow state as well as its sensitivity and
Another central mechanism in coating flows stability. Moreover, imposing an apparent dy-
that requires particular attention in computer- namic contact angle can be more subtle than is
aided analysis is dynamic wetting, the process commonly appreciated. When the mesh is refined
by which the coating liquid displaces most of the to yield converged macroscopic solutions that
air previously in contact with the substrate (for are consistent with the local slip-flow problem
comprehensive reviews, see Chapter 3, and also posed, the angle of meniscus inclination at 'visible'
Kistler (1993)). Assuming that the liquid/gas distances from the solid may be different from
interface meets the solid surface at a wetting line that imposed at the putative wetting line (see
where the advancing liquid completely displaces Tilton (1988) for a case in point). The assumption
the receding air, conventional hydrodynamic of perfect slip at the wetting line leads to another
continuum theory predicts a multivalued velocity disconcerting feature in computed coating-flow
that gives rise to a well-known, unphysical force solutions, namely abrupt liquid deceleration
singularity. In addition, when surface tension immediately upstream of the wetting line. It is
effects are significant, hydrodynamic theory most likely unphysical and particularly conse-
requires that a contact angle be specified as a quential in simulations that seek to account for
the transport and influence of surface-active and Zvan 1991); for an extended summary, see
materials (Schunk (1989); see also Section 9.3.3 also Kistler (1993). Computed values of apparent
and Chapter ltd). contact angles (Zhou and Sheng 1990; Kistler
The second approach overcomes the need to and Zvan 1991) corroborate results from asymp-
specify perfect slip at the putative wetting line totic theory (Voinov 1976; Cox 1986). Results
by exploiting the weak form of the Galerkin from both types of analysis, in turn, agree with
weighted residual equations which - in an ad hoc selected experimental data - at least up to
manner - can accommodate multi-valued veloc- Ca ~ 0(0.1) (Kistler 1993). Computed velocity
ities at dynamic wetting lines without catastrophic fields for slide coating at Ca ~ 0.16 (Kistler and
consequences for the flow nearby. The first Zvan 1991) furthermore agree with laser-Doppler-
attempt allowed multiple unknowns at the same velocimetry data for the acceleration of liquid
node (Kistler 1984). More refined schemes collapse toward the dynamic wetting line (Miies, Hens
several nodes onto the same point (Schunk 1989; and Boiy 1989), even though the finite-element
Christodoulou and Scriven 1990), a standard model insists on perfect slip at the line. Above
procedure in fracture mechanics to capture the Ca ~ 0(0.1), however, where many industrial
stress singularity at crack tips (Bathe 1982). To coating processes operate, systematic deviations
accurately reproduce the multivalued velocity set in between computed and measured apparent
field and associated 1/r singularity in rate of dynamic contact angles. Thus, single-phase
strain at dynamic wetting lines, at least two hydrodynamic theories with ad hoc assumptions
elements ought to be collapsed, one at each side at the wetting line cannot be relied upon to
ofthe wetting line, and the pressure basis function predict the impact of the macroscopic flow
must be modified to explicitly account for the configuration (see also Chapter 3).
1/r pressure singularity (Christodoulou and Scriven The fourth approach attempts to predict the
1990). Computed predictions suggest that the impact of the macroscopic flow on apparent
velocity along the free surface retains a nonzero dynamic wetting behavior by restricting the
value of 80-90% of the web speed down to the analysis to high capillary numbers. In the limit
wetting line, in good agreement with the analytical Ca » 1, the interface cannot transmit the influence
'outer' solution of Moffat (1964) and also experi- to the local wetting physics upstream, and apparent
mental measurements (Miies, Hens and Boiy dynamic contact angles 00 become dependent
1989). The dynamic contact angle, however, remains variables that can be computed without specifying
an empirical input parameter. Moreover, the use a boundary condition for a local contact angle
ofcollapsed elements to simulate a 'rolling motion' (Kistler 1984). Predicted flow conditions for
at a dynamic wetting line remains an ad hoc curtain coating at which 00 comes close to 180·,
procedure, for it does not yield solutions that for example, agree qualitatively with those
converge with systematic mesh refinement. measured at the onset of catastrophic air
The third approach adheres to the use of entrainment (Kistler 1984); for further details, see
standard, ad hoc boundary conditions at a putative also Chapter ltc. Nevertheless, the practical
contact line, i.e. prescribed contact angle and relevance of this limiting regime needs to be
local slip flow, but imposes a sub-microscopic explored further.
static angle rather than a macroscopic apparent The fifth approach seeks to resolve the
one, uses a dimensionless slip length well below submicroscopic physics of dynamic wetting. It
0.001, and systematically refines the mesh to incorporates a refined model for the local air
permit a local analysis that attributes dynamic displacement mechanism right into a macroscopic
variations in an apparent contact angle to viscous coating-flow computation, and thereby attempts
bending of the menisucs at length scales invisible to provide realistic boundary conditions for the
by the naked eye or a low-power microscope macroscopic flow. In high-speed coating, en-
(Lowndes 1980; Zhou and Sheng 1990; Kistler trainment and subsequent collapse of a thin layer
Special topics 337
of air is a plausible mechanism (Miyamoto 1992); tutive models and incorporating them into realistic
for more details, see also Chapters 3 and Hc. In computer simulations remains one of the most
steady-state flow calculations, it is easiest to prominent unresolved issues in computational
ignore the likely collapse of the air film and coating-flow mechanics.
restrict the simulation to steady, two-dimensional Almost all non-Newtonian coating formulations
two-phase flow under the influence of Van der experience a reduction in viscosity with increasing
Waals and possibly electrostatic forces between shear rate, and most numerical studies of
a liquid and a smooth solid as well as compres- non-Newtonian coating flows have been limited
sibility and slip flow in a thin air layer. For slide to generalized Newtonian models for viscosity
coating, a preliminary analysis of this sort as a function of the second invariant of the rate
produced free-surface shapes which, at macro- of deformation tensor (cf Chapters 2 and 5). The
scopic length scales, appear to resemble those in most useful example is perhaps the well-known
realistic flows (Christodoulou and Scriven 1984). Carreau equation (Bird, Armstrong and Hassager
A complete analysis, however, would have to 1987) which adequately describes shear-dominated
address stability and maybe three-dimensional processing flows of suspensions and polymer
surface topography. These are formidable chal- solutions (Bird 1976). For liquids that exhibit
lenges which, to date, have precluded quantitative solid-like behavior at very low deformation rates,
predictions of apparent contact angles or critical a Bingham-type yield model may have to be used
air-entrainment velocities that agree with experi- instead (Bird, Armstrong and Hassager 1987), or
mental data. may be combined with the Carreau model at
higher deformation rates (Nelson, Kistler and
Olmsted 1990). The classical power-law model,
9.3.2 NON-NEWTONIAN LIQUID BEHAVIOR
which is a degenerate case of the Carreau equation,
Although much of the published literature on may sometimes be sufficient to capture the
coating flows focuses on simple, Newtonian essence of shear-thinning viscosity behavior. It
liquids, most liquids coated in industrial practice can cause difficulties, though, in flows with
come in the form of particle suspensions, polymer stagnant zones because the viscosity diverges
solutions, polymer melts or other nonideal where the shear rate vanishes. Contrary to a
mixtures that exhibit significant non-Newtonian widely held belief (Coyle, Macosko and Scriven
behavior. Moreover, even liquids that appear to 1987; Pranckh and Scriven 1990), however, even
be Newtonian when tested in conventional the more sophisticated Carreau model and its
rheological equipment may exhibit nonlinear relatives can run into serious if not catastrophic
material responses when subjected to the very computational difficulties when naIvely solved
intense and highly localized shear, extension and with Newton's method (Tanner and Milthorpe
mixed deformations that are characteristic of 1983): in regions of low deformation rates, not
most coating flows. In slide and other bead viscosity but its higher-order derivatives with
coating devices, for example, shear rates easily respect to deformation rate diverge - in violation
reach 105 s - 1, and extension rates 104 s - 1 (see, of one of the necessary conditions for Newton's
for example, Schunk and Scriven (1990)). Even method to converge (cf Dahlquist and Bjork
in the comparatively slow dip coating process, (1974)). A much more slowly converging successive
surface tension gradients can induce shear rates substitution of a pseudo-Newtonian viscosity or
in excess of 1000 s -1 (cf Chapter 13). other specialized iteration procedures are often
Non-Newtonian liquid behavior can drastically the only recource (Engelman 1982; Gartling and
alter the success or failure of a coating process, Phan-Thien 1984; Karagiannis et al. 1989),
as is well established through experimental especially for small power-law indices or vastly
observation (see Chapters 5 and 12a for specific different high- and low-shear viscosities. For
examples). However, designing suitable consti- liquids with a distinct yield point and associated
discontinuous material behavior, added com- the formation of thin stress boundary layers; (2)
putational complexities arise from the need to non-integrable stress singularities from either
track yield surfaces, another form of free bound- physical features in the flow domain or the
aries whose position is unknown a priori (Gartling mathematical nature of the constitutive equation;
and Phan-Thien 1984; Beris et al. 1985). In (3) incompatability of stress and velocity inter-
conventional finite-element algorithms, these polants (failure to satisfy the LBB condition);
complexities can be avoided in an ad hoc manner and (4) the hyperbolic character of the governing
by replacing rigid plastic regions with a liquid equations and boundary conditions that are not
of finite, but very high viscosity below a preset compatible with the characteristics of the gov-
threshold in deformation rate (Bercovier and erning equations. These errors can lead to spurious
Engelman 1980), or by the use of a modified bifurcations and turning points in the solution
Bingham-type constitutive relation that approxi- space, and also an artificial change of type of the
mates the transition from yielded to unyielded discretized system that leads to a loss of the
regions in a continuous manner (Papanastasiou evolutionary character and Hadamard (short
1987). wave) instabilities (Crochet 1989; Joseph 1990;
Many coating formulations, especially solutions Keunings 1991). The difficulties become increas-
including high-molecular-weight additives, exhibit ingly apparent - and challenging - as the natural
a viscosity that not only falls as the rate of shear relaxation time of the liquid rises in comparison
increases but also rises as the rate of extension to the characteristic time of the flow (the ratio
increases. Viscosity rise in steady-state extension between the two time scales is commonly referred
can be represented by an analogous Carreau to as the Deborah number; cf Chapter 5). For
equation (Fuller et al. 1987). Schunk and Scriven integral constitutive equations, the difficulties are
(1990) incorporated into a simulation of slide compounded by the need to trace the deformation
coating a constitutive equation that is a linear history throughout the flow domain, which
combination of a Carreau equation for shear becomes especially involved for flows with
thinning with the same type of equation for recirculations.
extension thickening. Their work suggests that Several recently developed algorithms are able
such an inelastic constitutive equation may be to cope with at least some of these difficulties.
able to account for certain extensional, elastic- One group ofalgorithms for differential viscoelastic
like effects in dilute polymer solutions and other models relies on a 4 x 4 sub-element interpolation
mildly viscoelastic coating formulations (cf Hyun ofstress and non-eonsistent stream-line upwinding
and Ballman (1978)), but quantitative verification of the convective term of the upper-convective
of this conjecture is still outstanding. stress derivative (Marchal and Crochet 1987;
For many coating liquids with measurable Crochet and Legat 1992). Even though effective
elasticity, realistic flow simulations will probably and successfully extended to free-surface flow
require complete viscoelastic constitutive models, problems (e.g., Dheur and Crochet (1989)), this
which come in either differential or integral form scheme alters the constitutive equation by
(see, for example, Bird, Armstrong and Hassager introducing artifical diffusivity and accuracy is
(1987)). In spite of significant progress over the limited to first order. In an alternative approach,
last decade, reviewed not too long ago by Crochet King et al. (1988) were the first to directly address
(1989) and Keunings (1989, 1991), incorporating the issue of change of type by using an 'explicitly
viscoelastic models into computational fluid ellipitic' form ofthe momentum equations (EEME)
mechanics algorithms necessitates highly-spe- designed to maintain its elliptic character inde-
cialized computational procedures that are still pendently of the Deborah number (Rao and
far from routine, even in the absence of free Finlayson (1991) suggested that the EEME form
surfaces. Computational difficulties arise primarily may not always prevent the catastrophic change
from excessive approximation errors due to (1) oftype and resulting loss ofevolutionary character).
Special topics 339
The EEME formulation was restricted to the on integral constitutive models are still susceptible
upper convected Maxwell model, and is not to convergence difficulties and can be quite costly
easily extended to other differential models for to execute. Hence, they do not necessarily offer
viscoelasticity. The simplest way of regularizing a clear-cut advantage except for liquids that
the momentum equation to be strictly elliptic for require a spectrum of relaxation times (Ra-
a wide variety of differential constitutive models jagopalan, Armstrong and Brown 1993).
is to add the effect of a solvent viscosity. The For viscoelastic free-surface flows, currently
method of choice for incorporating this effect available algorithms are most robust for 'smooth'
into a computational algorithm, in turn, appears flow configurations free of singularities such as
to be the so-called elastic-viscous split stress wetting lines or even sharp corners. Examples
(EVSS) formulation of Rajagopalan and co- most pertinent to industrial coating operations
workers (1990, 1992). It divides the deviatoric include rimming flow of a film in a rotating tube
stress into its elastic and viscous parts and (Rajagopalan et at. 1992), flow and stability of a
thereby assures a dominant, explicitly elliptic film down an incline (Salamon, Armstrong and
operator on velocity in the momentum equations, Brown, 1992), and leveling of films after coating
even without added solvent effect. The EVSS (Keunings 1986). As for viscoelastic free-surface
formulation is also readily adopted to free surface flows with static contact lines, solutions to the
flows (Rajagopalan et al. 1992). For special flow die-swell flow of liquid emanating from a slot,
geometries amenable to spectral or pseudo-spectral tube, or other cross-section are abundant in the
expansions, a third approach capitalizes on the published literature (see Crochet (1989) and
rapid convergence with increasing node number Keunings (1989) for extensive references). Unless
of such expansions to produce highly accurate obtained with one of the recent, numerically
and numerically stable predictions of viscoelastic stable algorithms mentioned above, however,
flows (e.g., Pilitsis and Beris (1989». In conjunction these solutions may not hold up to a test of
with domain decomposition into several large convergence with mesh refinement. Worse yet,
subdomains, spectral collocation methods can unless care is taken to ensure that the singular
mimic the geometric flexibility of conventional stress fields resulting from a particular constitutive
finite-element methods while maintaining the model are consistent with the function spaces
superior accuracy so advantageous for viscoelastic that can be represented by the finite-element
flow computations (Souvaliotis and Beris 1992). approximation, the published solutions suffer
To date, however, this approach has not been from additional numerical artifacts, including
applied to viscoelastic free-surface flow. A fourth spurious limit points (e.g., Rosenberg and Keunings
approach, finally, seeks to take advantage of the (1991), Coates, Armstrong and Brown (1992».
increased realism of integral constitutive models For these reasons, computing viscoelastic coating
in describing complex rheological behavior, and flows with static contact lines is still at the
thereby minimize computational difficulties (e.g., forefront of current research. Viscoelastic flows
Papanastasiou, Scriven and Macosko (1987); with dynamic wetting lines - which involve
Luo and Tanner (1988); Luo and Mitsoulis singularities that are even more severe (see Section
(1990». The limitation to flows without recir- 9.3.1 and also Chapter 3) - have yet to be tackled
culation ofearly algorithms with their streamlined in a satisfactory manner.
elements (Papanastasiou, Scriven and Macosko
1987) can be overcome by decoupling the velocity
9.3.3 SURFACTANT SOLUTIONS -
and pressure calculation from the stress calculation
CONVECTIVE DIFFUSION EFFECTS
in a Picard iteration scheme (e.g., Luo and
Mitsoulis (1990); Rajagopalan, Armstrong and Coating solutions, especially aqueous systems
Brown (1993». Contrary to the belief held by the such as photographic gelatins or pressure-sensitive
early proponents, however, algorithms relying adhesives, often contain a wide variety of surface
active components, commonly called surfactants. discretization technique to fit a finite-element

Small amounts of these materials usually serve mesh that is highly refined near free surfaces and
as wetting agents, solubilizers, sensitizers, color responsible for resolving the convective diffusion
formers and many other formulation purposes. field of the various species into a coarse mesh
Quite often, they are also added to enhance that suffices for the rest of the flow field. From
stability of the coating process (for further details, a rigorous theory of affine deformation, Schunk
see Chapters 4 and lId). Surfactants influence derived a finite deformation dyadic and used it
the coating flow dynamics by concentrating and to compute deformations of the fine mesh in
adsorbing at liquid/gas interfaces, where they terms ofthose of the coarse mesh, the latter being
reduce capillary pressure by depressing surface tracked as part of the free-surface flow problem
tension and create interfacial shear stresses by as described in Section 9.2.2. Interpolation of the
inducing surface-tension gradients. velocity field onto the fine mesh and projection
Accurate computational analysis of coating of the concentration fields onto the coarse mesh
flows affected by surfactants requires not only a are readily accomplished via the finite-element
modified version of the interfacial traction basis function representation of the physical
condition (9.5) to include the dynamic surface domain (i.e. the parametric mapping (9.15c)) and
tension effects, but also interfacial species balances the dependent variables. With properly designed
and a bulk/interface equilibrium model to account interpolation and projection, the efficiency of
for surfactant adsorption and transport along Newton's method can be retained for solving the
interfaces as well as changes in concentration free-surface flow and concentration field simul-
due to interface expansion or contraction. It taneously in a few iterations. Several case studies
furthermore requires bulk transport equations in Chapters ltd and 13 attest to the power and
to describe convection and diffusion of both effectiveness of the two-level discretization tech-
monomer and aggregated micelle species, and nique for complex free-surface flows that are
kinetic models to account for the interchange of strongly influenced by surfactant effects -- and
surfactant monomer between the bulk and micelles also solvent evaporation/condensation. Accurate
of various sizes. Chapter lId, in conjunction with predictions of the latter may require accounting
Chapter 4, lays out in detail all the requisite of the coupled flow and species transport in both
equations and boundary conditions. Fortunately, phases and the complete set ofinterfacial boundary
these equations can be solved together with the conditions, as discussed in detail in Chapter 13.
Navier-Stokes system by augmented standard
finite-element techniques of the sort described in
9.3.4 ELASTOHYDRODYNAMICS COATING
Section 9.2 above.
SYSTEMS
Apart from a much increased size of the
computational problem, the main challenge added Quite a few industrial coating processes employ
by surfactant effects arises from the extremely deformable members such as rubber covered
thin boundary layers in concentration that can rolls, flexible blades, inflated membranes and
develop next to free surfaces. Refining the tensioned webs to control the coated film thickness
discretization near free surfaces to equal level for in a self-metered configuration, or to lessen the
all the unknowns is highly inefficient because sensitivity of the process to the mechanical
solute diffusivities are much smaller than the precision of the coating machine (for further
kinematic viscosity of the liquid and, hence, details, see Chapter t 2c). The nonlinear mechanics
concentration boundary layers are drastically that govern the deformation of many of these
thinner than momentum boundary layers. To members and the elastohydrodynamic coupling
avoid squandering computational work on between solid deformation and liquid flow add
unwarranted resolution of broad boundary layers considerable complexity to the numerical simu-
in the flow field, Schunk (1989) devised a two-level lation of coating processes.
Other methods 341
To predict the effect of deformable roll covers to compute viscoelastic roll deformation coupled
- quite often made of elastomeric rubber - a with lubrication flow, but ran into serious and as
further challenge is to come up with constitutive yet unresolved convergence difficulties.
models and mechanical theories that accurately Computations of coating flows with flexible
describe not only the resilience of the material blades, inflated membranes or tensioned webs are
to local indentation but also the deformations considerably more straightforward, provided the
at one place on the roll surface due to loads flexible members are thin compared with their
acting at other places. The simplest models radius of curvature and their deformation is not
account only for purely elastic roll deformations, altered by creep or other not strictly elastic
but systematic deviations from experimental data behavior. In that case, cylindrical shell theory
cast doubts on whether such models accurately applies (for full details, see Pranckh and Scriven
capture physical reality (for a critical review see (1990a, b, 1989)). It yields a set of differential equa-
Coyle (1992) and also Chapter 12c). Models that tions that describe the position, curvature, and
have been tried in roll coating simulations include tension along deformable members (e.g. equations
a one-dimensional Hookean model which amounts (12c.12) and (12c.13) in Chapter 12c). Pranckh and
to a continuous distribution of linear springs Scriven (1990a, b) recommend Galerkin/finite-
(Coyle 1988; Carvalho and Scriven 1993); a element techniques to discretize these equations
neo-Hookean spring model, also one-dimensional and thereby furnish the distinguished free-boundary
but nonlinear to describe rubber and other equations in finite-element algorithms of the sort
elastomers better (Carvalho and Scriven 1993); described in this chapter. Spurious oscillations,
a one-dimensional integral approximation for however, tend to arise when first-order equations
plane strain of a semi-infinite slab of Hookean such as (12c.12a) are approximated by Galerkin
material, which amounts to a two-dimensional weighted residuals (as they do for the kinematic
and volume-conserving model (Carvalho and boundary condition, as mentioned in Section 9.2.1
Scriven 1993); and a fully two-dimensional finite- above). Here, in the experience of one of us (SFK),
element description of large plane strain of a simple centered differences provide a ready remedy.
curved layer of either Hookean or neo-Hookean Special procedures, described in detail by Pranckh
material (Coyle 1990,1992). These models were and Scriven (1989), are also needed at points where
all solved together with Reynolds' equation for one deformable member separates from another,
lubrication flow (see equation 12c.2)). With the as does the trailing blade from the inflated
exception of the integral representation for membrane in membrane coating.
two-dimensional plane strain, however, the de-
formable-roll models are also readily incorporated
9.4 OTHER METHODS
into algorithms for fully two-dimensional free-
surface flows, especially algorithms with elliptic
9.4.1 ALTERNATIVE FREE-SURFACE
or generalized algebraic mesh generation schemes
PARAMETERIZATION AND
to accommodate the additional free boundaries
MESH-GENERATION SCHEMES: FIXED-GRID
at the rubber-liquid interface(s), and Newton's
ALGORITHMS AND FRONT TRACKING
method to cope with the nonlinearities from the
elastohydrodynamic coupling. The computational Deforming-mesh algorithms of the sort described
challenge is considerably more formidable for in Section 9.2 accurately resolve the position and
finite-strain analysis of nonlinear viscoelastic solid shape of free boundaries. Hence, they are well-
materials, which may be necessary to describe suited to predict steady, two-dimensional flows
adequately elastomeric roll-cover materials used that make up many successful coating operations,
in practice. Bapat and Batra (1984) performed and to assess sensitivity and stability of these
such an analysis for the dry contact between a flows to small two- and three-dimensional
deformable and a rigid roll. Coyle (1990) attempted disturbances. With the algorithms published to
date, however, the topology ofthe flow field must embedded as a steep gradient into the solution
be largely known a priori. Automated parameter- of the governing equations. No additional
continuation or time-stepping procedures can computational tool is introduced to identify the
follow only moderate variations in free boundary position of the fronts. Conventional front-cap-
location. Large alterations in flow geometry re- turing methods have been used with great success
quire cumbersome remeshing procedures, and to locate shocks in compressible flows (e.g., Boris
changes in domain connectivity - as might occur (1989)), for instance. In their simplest form,
during start-up or break-up of a coating bead, for however, they are less well suited to follow
instance - are very difficult if not impossible to material interfaces.
track. Similarly, generally applicable extensions More recent, novel front capturing methods,
to three dimensions of conventional algorithms on the other hand, show great promise for
are far from trivial to program and costly to execute. accurate representation of material interfaces,
Significant changes in flow topolgy, especially and should be tried for coating flow analysis.
break-up or merging of fluid bodies, are more One such method defines the interfaces as a level
readily followed with so called fixed-grid algo- set'P = 0 of a function that is advected according
rithms that keep track of the position of irregularly to a generalized Hamilton-Jacobi equation, viz.
shaped, moving interfaces on a fixed, rectangular o'Pjot + fIV'P! = 0, where f is the propagation
grid that covers both regions occupied by fluid speed normal to the interface (Osher and Sethian
and regions filled with air or another immiscible 1988). Simulations of complicated crystalline
fluid (for extended reviews, see Hirt and Nichols shapes in dendritic solidification (Sethian and
(1981); Hyman (1984); Oran and Boris (1987); Strain 1992) attest to the ability of the method
and Floryan and Rasmussen (1989)). As far back to describe accurately intricate interfacial geo-
as the early I 960s, free-surface flows of remarkable metries. Level-set interface capturing would seem
complexity have been simulated with finite- to be particularly attractive for capillary flows
difference algorithms that solve the time-dependent because it permits computing the local curvature
Navier-Stokes equations on grids fixed in an without the need to reconstruct the interface
Eulerian frame of reference (e.g., Harlow and (Osher and Sethian 1988). To date, however,
Amsden (1971)). Rayleigh-Taylor and Kelvin-Helmholz instabil-
Fixed-grid algorithms are typically quite simple. ities in compressible gas dynamics (Mulder,
Many make use of explicit iterative schemes to Osher and Sethian 1992) are the only free-surface
update the discretized variables at each time step. flows to which the method has been applied.
Nevertheless, fixed-grid methods face several
challenges for viscous free-surface flow analysis.
The key challenge is to devise a scheme that 9.4.1.2 Implicit volume tracking
monitors the position of the interfaces, and to Most common among fixed-grid algorithms are
design an algorithm that advances the interfaces those that rely on interface tracking, i.e. they
as time progresses without spurious numerical introduce additional computational devices to
diffusion. A second major challenge is the precise locate, on the fixed grid, interfaces between
reconstruction of the local slope and curvature immiscible fluids. The earliest schemes employed
of the interface. And a closely related, third implicit volume tracking via some marker quantity
challenge is the accuracy with which the interfacial which monitors the interior motion of the fluids
traction boundary condition (9.5) is satisfied. of interest rather than the interfaces. The position
of the interfaces is not stored explicitly, but is
reconstructed cell-by-cell whenever needed.
9.4.1.1 Front capturing
A prime example is the classical marker-
In so called front capturing methods, the transition and-cell (MAC) finite-difference method (Harlow
in fluid properties across moving fronts is and Welch 1965). It traces distinct fluids with
Other methods 343
mass-less particles that are advected passively track the evolution of ¢(x,Y,t) by solving an
with the computed flow field. Some recent auxiliary advection equation. Pseudo-concentra-
finite-element implementations do the same (e.g., tion functions on fixed finite-element meshes are
Matsushiro et at. (1990». The particles participate gaining popularity in mold-filling simulations
in the dynamics only in determining the position (e.g., Dhatt, Gao and Ben Cheikh (1990); Reddy
of the interface, which must be identified in cells and Reddy (1991», but have not been applied to
that contain tracers of both fluids. The original visco-capillary flows of the sort found in coating
particle-in-cell (PIC) method (Harlow 1964; processes.
Amsden 1966), in contrast, treats the particles as Perhaps the biggest problem with implicit
Lagrangian mass points. It is therefore a Lag- volume tracking is the need repeatedly to
rangian technique (see Section 9.4.2 below), even reconstruct the local position, slope and curvature
though it accounts for momentum and energy ofthe interface. Numerous interface reconstruction
at cell centers of a fixed Eulerian grid. schemes have been devised for a wide variety of
More recent volume tracking schemes abandon applications (e.g., Noh and Woodward (1976);
the computationally costly marker particles in Chorin (1980); Hirt and Nichols (1981); Youngs
favor of a volume-fraction function which is (1982); Barr and Ashurst (1984); Sethian (1984);
taken to be F = 1 in cells occupied by the fluid Chorin (1985); Liang (1991); Fatemi and Kistler
of primary interest, F = 0 in cells outside or in (1992». However, a robust and reliably accurate
another fluid, and 0 < F < 1 in cells that contain scheme for visco-capillary free-surface flows of
the interface. In early implementations, such as the sort encountered in coating processes has
the simple-line interface calculation (SLIC) not been advanced to date (Fatemi and Kistler
algorithm of Noh and Woodward (1976), the 1992).
fraction-of-cell-volume function is transported Many implicit volume-tracking algorithms
directly from cell to cell with a judiciously furthermore rely on questionable ad hoc approxi-
designed donor-acceptor scheme that maintains mations of the dynamics near the interfaces,
F as a step function as the computations advance which makes them even less reliable for visco-
in time. More recent implementations, the most capillary flows. The earliest algorithms merely
prominent being the volume-of-fluid (VOF) set the pressure to its ambient value at the center
method of Hirt and Nichols (1981), advect the of cells adjacent to the free surface, and forced
function F with the flow - and hence the interface the velocities at fictitious nodes outside the fluid
- by solving an auxiliary evolutionary equation, to be equal to those inside (e.g., Harlow and
viz., ofjot + v· VF = o. The interface is recon- Welch (1965». Subsequent improvements sought
structed from F = 0.5. To avoid spurious diffusion to impose the pressure condition at the actual free-
of F that would result from standard finite- surface location (Chan and Street 1970; Nichols
differ-ence discretization, a donor-acceptor flux and Hirt 1971), account for normal stress
approximation is again essential (Hirt and Nichols contributions (Hirt and Shannon 1968; Nichols
1981). A related volume tracking scheme, based and Hirt 1971; Pracht 1971; Hill, Shook, and
on the degree of filling around each vertex in a Esmai11981) as well as capillary pressure due to
triangular mesh, is key in modern fixed-mesh a curved interface (Daly 1967; Nichols, Hirt, and
finite-element simulations of injection molding Hotchkiss 1980, 1985; Liang 1991), and enforce
(e.g., Wang and Lee (1989». Other volume tracking a shear-free interface separately (Nichols and
schemes related to the VOF method employ a Hirt 1971). Boundary conditions for the total
pseudo-concentration function ¢ that fills the normal traction, however, still entered the equation
empty regions with a second, fictitious fluid of set via a modified pressure in boundary cells.
negligible density (as first suggested by Nichols Sicilian and Hirt (1984) tried to impose the
and Hirt (1975» or viscosity (e.g., Thompson traction boundary condition by setting all velocity
(1986». Like the VOF method, such schemes derivatives that involve discretized velocity
components outside the liquid to zero. It is rectangular box with experimental observations.
unclear, however, how this ad hoc procedure can To overcome the problem of excessively small
capture the impact of normal viscous stresses time steps demanded by explicit techniques for
when they are important. Fatemi and Kistler viscous flows, he modified the SOLA-VOF method
(1992) more recently solved for the pressure in of Hirt and Nichols (1981) to account for viscous
the boundary cells, treating the traction condition stresses in an implicit manner as Pracht (1971)
as a boundary condition for the discretized had done much earlier for the MAC method.
momentum equations rather than imposing it Hirt et al. (1990) applied the commercial SOLA-
separately. They validated their formulation for VOF code FLOW-3D to analyze planar spin
an oscillating liquid cylinder in the limit controlled casting of metal melts, a process which resembles
by capillarity, and for a viscous sheet in purely extrusion coating even though it occurs at
extensional flow in the limit dominated by normal Re = 0(1000) and involves solidification. McKib-
viscous stresses. ben and Aidun (1991), evidently unaware of the
Most early applications offixed-grid algorithms work by Hill, Shook and Esmail (1981), computed
with implicit volume tracking were aimed at solutions of a two-dimensional jet emanating
simulating large-scale free-surface flows dominated from a slot in an attempt to demonstrate that
by inertia, such as sloshing in tanks, breaking fixed-grid finite-difference codes with VOF-type
waves, splashing of low-viscosity drops, etc. For volume tracking can yield predictions that
such flows, accurate resolution of the local compare favorably with results from finite-element
interfacial shape is not as critical as for small-scale algorithms of the sort expounded in Section 9.2.
flows that are dominated by capillarity. Besides, Like Hill and his co-workers, however, McKibben
explicit time-stepping procedures are ideally suited and Aidun excluded capillary pressure effects and
to simulate flows at elevated Reynolds numbers. considered only a single value of Re = 300 for
Nevertheless, attempts have been published to which the jet contracts rather than swells as it
apply conventional fixed-grid techniques with does when viscous effects dominate. Miura and
explicit time-stepping and implicit volume tracking Aidun (1992) applied several versions of a VOF
to flows at low and moderate values of Re - algorithm to simulate how pressure variations
some of interest in the context of coating. Hirt induced by flow instabilities in the pond of a
and Shannon (1968) used the teapot effect (cf short-dwell blade coater affect the uniformity of
Chapter llc) as a test case to validate their the coated film. They did not address the accuracy
improved treatment of the normal stress boundary of their results, however. Fatemi and Kistler
condition in a MAC algorithm, but failed to (1992) modeled the impingement region in curtain
account for surface tension which plays a central coating at Ca = 1 and 4 < Re < 10 with their
role in controlling the deflection of a freely falling modified SOLA-VOF algorithm, and obtained
liquid sheet where it is being formed (Kistler and good agreement with predictions from a conven-
Scriven 1994). Harlow and Amsden (1971) also tional finite-element code (Kistler and Scriven
showed a computed prediction of the teapot 1983), especially for the interfacial shape and
effect, but did not comment on the details of the dynamic wetting line position.
free-surface boundary conditions. Hill, Shook Altogther, the simplicity of implicit volume
and Esmail (1981) employed a MAC-type algo- tracking is very attractive, but algorithms currently
rithm to simulate viscous die swell for available are not refined enough to routinely
4.2 < Re < 92.3, and obtained surprisingly good predict coating flows with accuracy comparable
agreement with experiments and standard finite- to that of state-of-the-art finite-element codes
element predictions even though the swelling (Section 9.2). The crux remains the limited
extended over less than one grid size. Golafshani capability of implicit volume tracking to capture
(1988) successfully compared computed results the details of the interfacial shape and to account
for a highly viscous fluid draining from a for the full set of traction conditions. The
Other methods 345
consequences of these shortcomings can be grid line, and approximates the interface by the
particularly grave in the vicinity of apparent new cell edges or diagonals on a locally mildly
singularities such as static separation lines, distorted grid. Glimm and co-workers (Glimm
dynamic wetting lines, or sharply curved interfaces. and McBryan 1985; Glimm et al. 1986, 1988)
Improving interface resolution via grid refinement advanced and refined a similar scheme in which
is typically not a viable alternative with explicit an irregular, triangular finite-element mesh near
time-stepping procedures because the largest the interface connects a lower-dimensional moving
step for which the algorithm is stable is dictated interface grid with an otherwise fixed rectangular
by the smallest grid spacing. mesh. Because the marker points along the
interface advance with the fluid, Glimm's method
can be classified as mixed Eulerian - Lagrangian
9.4.1.3 Explicit interface tracking
(cf Section 9.4.2). Similar schemes that blend
Many of the inherent difficulties of implicit local, deforming meshes near free boundaries
volume tracking can be overcome by explicit with global, fixed finite-element meshes in the
interface tracking, which introduces computational remainder of the computational domain have
elements that expressly describe the position and also been advocated by Couniot and Crochet
shape of interfaces, and store their values at each (1986) and Tabata and Morishta (1989) for
time step. Saving the position of an interface as finite-element algorithms to solve mold filling
a discrete set of heights above a reference line problems. Applications ofGlimm's method include
(Chan and Street 1970; Rirt, Nichols and Romero compressible flows of stratified fluids (Chern et
1975) or plane (Nichols and Rirt 1973) is the at. 1986) and saturation fronts advancing in
most obvious choice but may, of course, become porous media (Daripa et at. 1988). Viscous
multivalued. More powerful procedures track free-surface flows with surface tension have,
the interface by an ordered chain of marker regrettably, not been tackled to date. Fromm
points (Daly 1967; Chan and Street 1970; Nichols (1981, 1986), however, successfully analyzed several
and Rirt 1971; Glimm and McBryan 1985; visco-capillary flows with a related algorithm
Glimm et at. 1986, 1988) or line segments (Miyata that combines a fixed finite-difference grid with
1986), and rely on arc-length to parameterize the local surface coordinates to improve the resolution
local shape of the interface (Ryman 1984). of the dynamics near the interface.
Interfacial marker points are typically advected In comparison to implicit volume tracking,
in a Lagrangian manner with the flow, and hence explicit interface tracking greatly improves the
may have to be added or deleted in regions where accuracy of the local interface representation,
they deplete or accumulate (Nichols and Rirt permitting sub-cell resolution if marker points
1971; Glimm and McBryan 1985; Glimm et al. are packed more densely than the fixed grid (the
1936, 1988). Between the points, the interface is accuracy of the underlying dynamics is, ofcourse,
approximated by interpolants, quite often simple still limited by the cell size of the fixed grid).
straight-line segments (Nichols and Rirt 1971; Advanced methods of explicit interface tracking
Glimm and McBryan 1985; Glimm et at. 1986, (Glimm et at. 1988; Unverdi and Tryggvason
1988), but sometimes also more elaborate piecewise 1992) - and also front capturing of the sort
polynomials (Ryman 1984). introduced in Section 9.4. t.t (Osher and Sethian
While conventional explicit interface tracking 1988) - improve furthermore the treatment of
methods discretize the governing equations on entangling interfaces and associated changes in
a fixed finite-difference grid, Ryman (1984) topology. In particular, they are less likely to
described a local grid-adjustment procedure that cause spurious entrapment of gas pockets within
moves each mesh point within one half spacing the liquid or breaking away of stray particles
from the interface to the nearest horizontal or from the liquid, which both tend to plague
vertical location where the interface intersects a conventional volume tracking schemes. For viscous
free-surface flows, moreover, explicit interface are of practical interest, and should be advanced
tracking algorithms with local grid adjustment for that purpose. While such algorithms may
are capable of incorporating the inter- offer no apparent advantage for the types of
facial traction boundary condition without ad hoc analysis performed so well by standard finite-
approximations. Hence, they ought to match the element methods, they may make amenable to
accuracy of conventional finite-element formula- analysis coating flows which at present are
tions or other special-purpose methods, such as difficult if not impossible to solve. Such flows
boundary-integral techniques, and deserve a more include fully three-dimensional flows, flows for
systematic evaluation for coating-flow analysis. which there is limited a priori knowledge of
Unverdi and Tryggvason (1992) advanced a feasibility, and flows in which the topology
promising hybrid between explicit interface changes dramatically as, for instance, during
tracking and implicit front capturing that may start-up or catastrophic failure ofa coating process.
also be attractive for coating flow calculations.
Their algorithm relies on a special unstructured
9.4.2 ALTERNAnVE FORMULAnONS:
surface grid to monitor the kinematics of the
LAGRANGIAN AND RELATED METHODS
interfaces through a fixed grid that is used to
compute the fluid dynamics. The transition from
9.4.2.1 Strictly Lagrangian methods
one set of density and viscosity values to the
other extends over a finite thickness of the order In an alternative class of numerical methods for
of one cell size of the fixed grid. The tracked simulating free-surface flows, the governing
interface transports the jump in fluid properties, equations are formulated in their Lagrangian
which is approximated from the solution of form and computational entities are employed
Poisson-type equations, and thereby avoids that move with material points in the flow.
spurious diffusion and interface oscillations, Conventional finite-difference grids and finite-
difficulties frequently encountered with conven- element meshes, as well as particle and vortex
tional volume tracking algorithms. The algorithm methods have been used to discretize the governing
implicitly accounts for the interfacial traction equations (see Donea (1983); Leonard (1985);
condition (9.5), yet permits the flow field to be Monaghan (1985); Oran and Boris (1987) and
advanced through the fixed Eulerian grid by a Floryan and Rasmussen (1989) for comprehensive
conventional, MAC-type finite-difference scheme reviews). Because the computational cells or
without any special consideration for the interface particles always follow the same fluid material,
- except for interfacial tension. The latter is Lagrangian methods are ideally suited for free-
incorporated as a body force distributed onto the surface flows. The position ofinterfaces is tracked
fixed grid in a manner first proposed by Peskin in a 'natural' manner without the need for explicit
(1977) to treat elastic boundaries in fixed-grid parameterization, and interfacial boundary con-
flow computations. Brackbill, Kothe and Zemach ditions are readily imposed. Lagrangian methods
(1992) very recently proposed a similar 'pseudo- are also particularly attractive for flows that are
continuum' method to implement capillary dominated by convection. In a frame of reference
pressure in fixed-grid algorithms which, in effect, tied to the material, convective terms are absent
mimics the molecular origin of surface tension. and, hence, the methods avoid the spurious
In summary, sophisticated techniques for diffusion that typically arises from convective
computing the motion of interfaces - such as the transport across grid boundaries in Eulerian
level-set technique (Osher and Sethian 1988) or algorithms. Unfortunately, Lagrangian methods
hybrid interface tracking/front capturing schemes have several drawbacks that have limited their
(Unverdi and Tryggvason 1992) - endow fixed-grid application to viscous free-surface flows.
algorithms with the potential to predict reliably A major drawback arises from the topological
and accurately complicated coating flows that restrictions inherent in Lagrangian grids. The
Other methods 347
earliest schemes relied on strictly Lagrangian completely unstructured Voronoi-Delauny tes-

grids that retain a fixed connectivity structure sellations of the sort first proposed by Dukowicz
and follow specific material elements throughout (1981).
the computation (e.g., Hirt, Cook and Butler Most published applications of Lagrangian
(1970». Such grids become easily distorted, methods for free-surface flows focused on com-
especially in flows with recirculations or large pressible and incompressible flows that are
shear deformations, and are prone to yield large dominated by inertia (see Oran and Boris (1987);
discretization errors. To avoid disastrous grid Floryan and Rasmussen (1989)). In the very first
distortions, several strategies have been developed. application of a strictly Lagrangian method, for
The most common is remeshing, often also instance, Hirt, Cook and Butler (1970) employed
referred to as rezoning, which transforms the a finite-difference technique to analyze the
discretized variables on a distorted grid onto a Rayleigh-Taylor instability and sloshing in tanks,
more regular grid with the same connectivity accounting for surface tension via an excess
structure. In continuous rezoning, the grid is pressure applied along free surface segments.
adjusted in small increments after each time step Lagrangian methods have also been applied to
(e.g., Hirt, Amsden and Cook (1974)); in general visco-capillary free-surface flows of the sort found
rezoning, larger grid adjustments occur intermit- in coating processes. In the most pertinent
tently after many time steps (e.g., Bach and application, Bach and Hassager (1985) advanced
Hassager (1985); Addessio et al. (1986); Rama- a Lagrangian finite-element scheme with general
swamy (1990); Shopov, Minev and Bazhlekov rezoning to solve for the flow field and free
(1992)). Unfortunately, rezoning adds much surface shape in slide coating. The scheme retained
complexity and computational cost to the the viscous and pressure terms in their Eulerian
algorithms. In addition, rezoning requires inter- form and evaluated them iteratively at each time
polation of the discretized variables between two step, but treated the material/mesh coordinates
grids. It thereby reintroduces the spurious as dependent variables as conventional Lagrangian
numerical diffusion in convection-dominated flows schemes do. The results seem to agree with those
that Lagrangian methods seek to avoid in the from conventional finite-element analysis (Kistler
first place. Nevertheless, a considerable amount and Scriven 1983). Bach and Villadsen (1984)
of research has greatly improved the efficiency had earlier used a similar scheme to analyze the
and accuracy of rezoning schemes (e.g., Margolin wave motion in falling liquid films. Fyfe and
and Nichols (1983); Dukowicz (1984); Ramshaw collaborators (1988) successfully analyzed the
(1985); Addessio et al. (1986); and Huh, Golay effects of surface tension on viscous drop
and Manno (1986)). oscillations and capillary waves with a free-
Perhaps the most robust counter to grid Lagrangian finite-difference algorithm. Muttin et
distortion problems in Lagrangian free-surface al. (1991) simulated the filling stage ofa die-casting
flow computations comes from methods that rely process with a free-Lagrangian finite-element
on grid reconnection, often also referred to as scheme. And Shopov and Bazhlekov (1991) used
free-Lagrangian methods. These methods allow a Lagrangian finite-element algorithm with
grid points to acquire new neighbors and thereby rezoning to solve several transient wetting
alter the mesh topology as the algorithm steps problems.
through time. Earlier schemes delete, add and All in all, Lagrangian methods have the potential
reconnect grid points as the need arises (e.g., to accurately predict realistic coating flows.
Crowley (1971); Fritts and Boris (1979); Crowly Because of their inherent ability to track material
(1985); see also Fritts, Crowley and Trease (1985)). fronts, Lagrangian methods are particularly well
Current state-of-the-art schemes (e.g., Fritts, suited to simulate the start-up ofcoating processes,
Crowley and Trease (1985); Fyfe, Oran and Fritts especially when automated grid reconnection
(1988); Muttin et al. (1991)) often make use of schemes accommodate drastic changes in flow
topology. On the other hand, Lagrangian for- Hughes, Liu and Zimmerman (1981); Donea
mulations by necessity produce initial-value (1983); Huerta and Liu (1988); Hayes, Dannelongue
problems. Hence, they are less attractive when and Tanguy (1991); Hurez, Tanguy and Bertrand
the primary interest lays in steady-state solutions (1991); Soulalmani et al. (1991); Tezduyar and
that are so fundamental in systematic coating co-workers (1992)) that seek to resolve moving
flow analyses. The futility of Lagrangian methods interfaces accurately while maintaining acceptable
is most severe when the steady states feature element shapes, even for large deformations of
eddies of enclosed recirculations. From the the flow domain.
applications published to date, it remains unclear Even though many authors derive their ALE
how the computational burden from costly formulations for completely arbitrary referential
rezoning or reconnecting and subsequent inter- kinematics (e.g., Hughes, Liu and Zimmerman
polation schemes stacks up against the cost of (1981); Huerta and Liu (1988); Soulalmani et al.
conventional finite-element algorithms with (1991)) and establish comprehensive strategies
deforming meshes. It is also unclear how to for mesh movement (e.g., Donea (1983); Huerta
handle rapid mass transport across phase bound- and Liu (1988)), most sample computations
aries, which are no longer material surfaces (see published along with the elaborate theories restrict
Chapter 13 in this volume). the Lagrangian motion to free-surface nodes and
move the internal modes with simple ad hoc
procedures. The so called deforming-spatial-
9.4.2.2 Arbitrary Lagrangian-Eulerian
domain/space-time procedure of Tezduyar and
methods
co-workers (1992), which makes use of a space-
To combine the advantages of accurate interface time finite-element discretization but otherwise
tracking and minimal mesh distortion that result is closely related to conventional ALE algorithms,
from purely Lagrangian and Eulerian descriptions also confines the Lagrangian motion to nodes
respectively, hybrid Lagrangian-Eulerian algo- on the moving boundaries, often in the direction
rithms have been developed that let the mesh/grid normal to the boundary. Similarly, most ALE
points move independently of the kinematics of algorithms for simulating materials processing
the coincident flow and formulate the governing flows (e.g., Hayes, Dannelongue and Tanguy
equations in a frame-of-reference tied to the (1991); Ahmed and Alexandrou (1992)) limit the
moving grid/mesh points. The grid/mesh move- Lagrangian description to the normal component
ment is typically arbitrary and user-prescribed, of the velocity at the interface nodes; the other
falling between the two extremes of strictly nodes are held fixed, but are updated after each
Lagrangian formulations in which the grid is time step with an unstructured remeshing scheme
embedded in the fluid and moves with it, and (Hayes, Dannelongue and Tanguy 1991) or a
purely Eulerian formulations in which the grid simple elliptic mesh-generation scheme (Ahmed
remains fixed in a laboratory frame. Such hybrid and Alexandrou 1992). Hurez and collaborators
techniques, commonly referred to as arbitrary (1990,1991) use a time-stepping iterative strategy
Lagrangian-Eulerian (ALE), began nearly twenty based on the kinematic boundary condition to
years ago as an outgrowth of Lagrangian compute steady-state free surface flows; they
finite-difference techniques with continuous invoke an ALE-type description to account for
rezoning (Hirt, Amsden and Cook 1974; Amsden, the displacement of inner-mesh nodes that is
Ruppel and Hirt 1980). The latter are, in effect, needed to avoid severe element distortions near
ALE algorithms themselves, so can be Lagrangian the free surface. Frank and Lazarus (1964) a long
algorithms with general rezoning (e.g., Wang and time ago proposed a technique akin to the ALE
McLay (1986); Ramaswamy (1990)). More recently, concept in which the coordinate system is Eulerian
numerous finite-element based ALE algorithms in one direction and Lagrangian in the other.
have been advanced (e.g., Belytschko (1980); This technique, in turn, is similar to what is
Other methods 349
known as the streamlined finite-element method coating processes, such as wave propagation on
(Papanastasiou, Macsoko and Scriven 1985) in liquid layers (e.g., Hughes, Liu and Zimmerman
which the mesh deforms normal to material lines. (1981); Ramaswamy (1990); Tezduyar et al. (1992»,
As may have become apparent from the die swell with static separation lines (e.g., Hurez,
foregoing discussion, many ALE algorithms are Tanguy and Bertrand (1991», or mold filling flows
quite analogous to 'standard' finite-element with dynamic wetting (e.g., Wang and McLay
algorithms that accommodate changes in free- (1986); Hayes, Dannelongue and Tanguy (1991);
boundary position with deforming meshes, as Ahmed and Alexandrou (1992». Hurez and Tanguy
described in Section 9.2.2. Curiously, researchers (1990) solved for a coating flow proper, namely
devising 'generalized' ALE algorithms have dip coating of Newtonian and Bingham fluids.
uncovered the analogy only recently (e.g., Even though they are potentially quite powerful,
Soulalmani et al. (1991». The reason may be that, ALE-type algorithms share many similarities
for steady-state algorithms with either algebraic with conventional time-dependent finite-element
or elliptic mesh generation, the analogy may be algorithms for viscous free-surface flows, and it
subtle, for these algorithms solve what may be remains unclear what their distinct advantages
called free-boundary problems, whereas ALE are for coating flow analysis. The robustness of
algorithms most often seek to analyze transient ALE algorithms hinges entirely on the ability of
moving-boundary problems. However, transient the user to construct an automated rezoning
algorithms with deforming meshes are in essence scheme that maintains the regularity and or-
ALE methods; the mesh moves with the flow thogonality of the computational mesh yet allows
normal to the interface, yet is adjusted indepen- significant distortions of the fluid boundaries.
dently of the flow kinematics otherwise. In fact, For complicated flow geometries, this task is no
the additional term that arises from the trans- less formidable than the design ofdeforming-mesh
formation of the governing equations into a schemes in conventional formulations (see Section
moving frame of reference (Hughes, Liu and 9.2.2). Nevertheless, the ability of certain ALE
Zimmerman 1981; Huerta and Liu 1988; algorithms to rely on completely unstructured
Soulalmani et al. 1991) is equivalent to the extra meshes (e.g. Sackinger, Schunk and Rao (1996»
term that arises from the time-dependent iso- deserves further examination, especially for
parametric mapping (9.16) in conventional simulating start-up and break-up ofcoating beads.
finite-element algorithms (see Lynch (1982) and
Kheshgi and Scriven (1984». When the Lagrangian
9.4.2.3 Other Lagrangian methods
description is restricted to a single, Cartesian
direction that is nearly normal to the interface, For the sake of completeness, this section briefly
and a simple algebraic procedure is deployed to mentions two additional Lagrangian methods
adjust the mesh underneath - as is so often the which are intriguing but, at present, have no
case (e.g., Hughes, Liu and Zimmerman (1981); practical use for coating flow analysis. Particle
Huerta and Liu (1988); Soulalmani et al. (1991» methods simulate fluid flow with discrete particles
- the much touted ALE approach boils down to whose interactions are designed to mimic, in a
nothing more than a classical Monge represen- collective statistical average, continuum behavior.
tation ofthe sort used in the earliest finite-element While older algorithms - such as the classical
methods for free-surface flows (see Nickell, Tanner particle-in-cell (PIC) method (Harlow 1964;
and Caswell 1974). Amsden 1966), see also Section 9.4.1 - rely on a
The earliest applications of ALE algorithms for fixed Eulerian grid to account for macroscopic
free-surface flows addressed fluid/structure prob- dynamic variables, more recent versions abandon
lems (e.g., Belytschko (1980); Donea (1983». More the mesh concept altogether and use particle
recently, however, free-surface flows have been differencing schemes to simulate mathematical
tackled that are of interest in conjunction with operators and Simpson's rule to perform inte-
grations (Monaghan 1985). Particle methods and Rasmussen (1987); Shokoohi and Elrod
readily track different materials, even for large (1987); Ahn and Ryan (1991)). Most formulations
distortions. However, they are prone to suffer map the deformed flow domain onto a rectangular
from statistical fluctuations, fail to accurately computational domain, modify the governing
describe the interfacial shape, and remain unproven equations to account for the mapping and, in
for incompressible viscous flows (see Monaghan steady-state computations, update the free-
(1985) for a comprehensive review). boundary location by a Picard-type scheme in
Vortex methods follow the evolution oflocalized which the mapping is recomputed after each
vorticity fields in incompressible high-Reynolds- iteration step. Dandy and Leal (1989) abandoned
number flows with a large set of discrete parcels this inefficient successive approximation scheme
of vorticity (see Leonard (1985) for an excellent in favor of Newton's method. They also adopted
overview). These methods are capable of very advanced techniques for computer-aided analysis
precisely tracking moving boundaries with of nonlinear systems to track solution families
interfacial tension (e.g., Tryggvason (1988)). with arc-length continuation (cf Section 9.2.4)
Unfortunately, however, vortex methods cannot and thereby demonstrated that, in many ways,
be made accurate for the viscous flow regime finite-difference methods for viscous free-surface
typically encountered in coating processes. flows can be made to match the capabilities of
finite-element methods.
Still, published applications of finite-difference
9.4.3 ALTERNATIVE DISCRETIZATIONS:
algorithms with mapping of a deforming grid
FINITE-DIFFERENCES AND
onto a fixed rectangular grid have been restricted
SPECTRAL-ELEMENTS
to flow geometries that deviate only moderately
from standard shapes, such as a circle or
9.4.3.1 Finite-difference and finite-volume methods
rectangular strip. The most prominent case studies
Even though finite-element techniques have several include rising bubbles (Christov and Volkov
distinct advantages for coating-flow analysis, as 1984; Ryskin and Leal 1984; Dandy and Leal
outlined in Section 9.1 and documented in Section 1989), die swell (Dutta and Ryan 1982; Ahn and
9.2, finite differences and finite volumes can also Ryan 1991; Liu, Yu and Chen 1991), break-up
be used to discretize the governing equations. of a jet (Shokoohi and Elrod 1987), and flow in
Indeed, finite differences are standard in most cavities with a free surface driven by one of the
fixed-grid algorithms (see Section 9.4.1), and are vertical walls moving downward (Loh and
preferred in many Lagrangian and ALE codes Rasmussen 1987; Esmail and Markov 1989).
(see Section 9.4.2). In the last decade, several Durst et al. (1991) applied a finite-volume scheme
finite-difference and finite-volume formulations to analyze the flow in slot coating (for specific
have also been advanced that seek to emulate results, see Chapter 11 a).
essential aspects of standard finite-element Evidently, finite-difference or finite-volume
formulations, in particular the use of grids that methods can be furbished to the point where
deform in response to changes in free-surface they predict practically relevant coating flows.
location. Some insist that the curvilinear grid be No clear-cut advantages are apparent, however,
orthogonal (e.g., Dutta and Ryan (1982); Ryskin over well-established finite-element methods. A
and Leal (1984); Dandy and Leal (1989)). Others distinct disadvantage offinite-difference methods
employ nearly orthogonal, boundary-fitted grids arises from the need to impose boundary
generated with elliptic methods (e.g., Esmail and conditions at the free surfaces explicitly, which
Markov (1989); Liu, Yu and Chen (1991)), as first requires tedious programming. In addition,
introduced by Thompson and co-workers (1974, incorporating the coordinate transformations
1982), or rely on simple algebraic coordinate into the governing equations is more complicated
stretching (e.g., Christov and Volkov (1984); Loh than the iso-parametric or sub-parametric map-
Other methods 351
ping which finite-element basis functions provide Spectral-element algorithms are potentially
in a straightforward manner. quite powerful for the analysis of realistic coating
flows. They may be particularly attractive for three-
dimensional flows with inherent periodicity in at
9.4.3.2 Spectral-element methods
least one direction (see, for example, Fig. 12a.12
The spectral-element method seeks to combine in Chapter 12). However, in spite ofthe advertised
the geometric flexibility of finite elements with efficiency of preconditioned conjugate-gradient
the potential for improved accuracy that arises iterative schemes (Ho and Patera 1991), the
from the rapid convergence ofdelocalized spectral cost-effectiveness of high-order spectral basis
expansions (Patera 1984; Maday and Patera function remains to be verified by direct com-
1989). The method divides the computational parison with low-order finite-element polynomials,
domain into a few subdomains, or spectral especially for three-dimensional flows. A potential
elements, each supporting a tensor product of problem with spectral-elements is their inherent
polynomial basis functions ofhigh degree (typically inability to capture local singularities without
5 to 15) to approximate the flow variables. As spurious influences on a large part of the
in finite-element methods, iso-parametric mapping computational domain. This difficulty, especially
conveniently describes the shape of distorted critical at wetting lines, can apparently be overcome
spectral elements that are needed to accommodate by confining the size of spectral element that
complex flow geometries, and a variational includes the singularity to a small neighborhood
formulation together with Gauss-type quadrature of the singularity (Grald et al. 1994). However,
generates a discretized equation set. To apply no expedient exists for steep gradients in the flow
the spectral-element method to viscous free-surface field internal to the domain at unknown locations.
flows, Ho and Patera (1990) introduced an
ALE-type description that tracks the evolution
9.4.4 BOUNDARY INTEGRAL TECHNIQUES
of the free-surface with a mesh whose motion is
tied to the fluid velocity normal to the free Discretization of the entire flow-domain ofinterest
surface. They also advanced an iterative solution can be avoided altogether for potential flows
scheme that is implicit in the treatment of the (Longuet-Higgins and Cokelet 1976) and creeping
pressure and viscous elliptic components of the flows (Youngren and Acrivos 1976). In these
Navier-Stokes equations, yet explicit in the limiting regimes, the governing equations are
treatment of the convective operators and the linear and their fundamental singular solutions
coupling of all the operators with the time- - which, for Stokes flow, were first derived by
dependent flow geometry (see also Section 9.2.9). Ladyzhenskaya (1963) - can be used to reduce
Ho and Patera (1990) focused their analysis on a general n-dimensional problem to one or
the stability offalling liquid films in two dimensions. several (n - I)-dimensional integral equations. A
Later, they extended their algorithm to fully solution to a particular flow problem results
three-dimensional free-surface flows (H 0 and Patera when the integral equations are combined with
1991), and illustrated its capabilities with a transient the boundary conditions and solved by a finite-
analysis ofthe stability ofan initially axisymmetric element or other standard collocation technique.
film. Johnson et al. (1991) applied a commercial This approach, known as the boundary-element
version of the algorithm (NEKTON) to verify an method, requires discretized values of the un-
integral film-flow model for the growth ofunduloids knowns, such as velocity or traction forces, along
in an annular film lining a cylindrical tube. the boundaries. At any point within the flow
Recently, the software package has been applied domain, the flow variables can be calculated by
to a variety of coating flows, producing results quadrature in a post-processing step. Even though
that compare favorably with earlier finite-element the method necessitates nontrivial integral evalu-
predictions (Grald et al. 1994). ations and leads to dense coefficient matrices, it
has the potential of reducing the computational dimensional film-flow disturbance generated by
cost because of the comparatively small size of an isolated, small particle arrested on an inclined
the resulting matrix system. slide (Pozrikidis and Thoroddsen 1991); the
Boundary-element methods are particularly profile of a film formed with a slot coater (Yuan
well suited for free-surface flows, for the boundary and Ingham 1991); dip coating, especially the
discretization readily tracks large interface extremely slow and delicate application of
distortions without loss ofaccuracy. The int~rfacial Langmuir- Blodgett films (see Chapter 13); and
velocity is calculated directly and the traction the plunging of a liquid film into a pool as may
condition, including capillary pressure, is easily be found in certain roll-coating configurations
incorporated into the integral equations (Tanzosh, (Hansen 1987; Geonaga and Higgins 1990).
Manga and Stone 1992). Most algorithms for In spite of its many attractive features, the
steady-state computations avoid the nonlinearity boundary-element method is rarely applied to
associated with the free surface by successive realistic coating flows. The primary reason is the
substitution, relying on the kinematic boundary method's ineffectiveness in dealing with non-
condition (e.g., Tran-Cong and Phan-Thien (1988)), linearities within the flow domain. Bush and
the normal-stress condition (e.g., Kelmanson Tanner (1983), for instance, attempted to account
(1983); Lu and Chang (1988)), or even the for inertia effects via a modified body-force term,
shear-stress condition (Yuan and Ingham 1991) but their iterative procedure failed at Reynolds
to update the shape of the boundary. As in numbers larger than 10. Viscoelastic effects in
conventional finite-element algorithms, each two- and three-dimensional extrusion problems
distinguished condition has its own, limited have been handled in a similar manner (Bush,
range of convergence. Pozrikidis (1988) was the Milthorpe and Tanner 1984; Tran-Cong and
first to solve the distinguished boundary condition Phan-Thien 1988). Treating nonlinear effects via
simultaneously with the integral equation(s), using an effective body force requires domain-internal
Newton's method. Boundary-element algorithms discretizations which defeat the main purpose of
for transient free-surface flows typically employ using a boundary-integral technique. The main
an iterative procedure that calculates the surface forte of the boundary-element method remains
velocities for a temporarily fixed shape and therefore its ability to accurately resolve intricate
advances the surface in time via the kinematic interfacial shapes in creeping flows while eco-
boundary condition (e.g., Lee and Leal (1982); nomically concentrating the computational effort
Geller, Lee and Leal (1986)). Yiantsios and where it is most needed. For coating processes,
Higgins (1989) attempted to incorporate the the method is probably most applicable to local
kinematic boundary condition into a fully implicit analysis of small-scale flow features around
scheme using Newton's method. particles and bubbles (e.g., Geller, Lee and Leal
Boundary-element methods have been applied (1986); Pozrikidis and Thoroddsen (1991 I), wetting
to a wide variety of viscous free-surface flow lines (e.g., Goodwin and Homsy (1991)), apparent
problems. Among the most frequently studied singularities in the interface (Goenaga and Higgins
ones are the deformation of drops and bubbles 1990) or the like - possibly matched to conven-
(for a comprehensive overview, see Tanzosh, tional finite-element analyses of the large-scale
Manga and Stone (1992)) and two- and three- flows surrounding the local feature.
dimensional extrusion flows (e.g., Bush and Tanner
(1983); Kelmanson (1983); Tran-Cong and Phan-
9.5 RETROSPECTIVE AND OUTLOOK
Thien (1988)). Case studies relevant to coating
processes have been rather limited, addressing To date, the most comprehensive computational
the effect of isolated and periodic corrugations on analyses of coating flows - and of viscous free-
the two-dimensional film flow down an inclined surface flows in general - rely on sophisticated
slide (Pozrikidis 1988; Hansen 1991); the three- Galerkin/finite-element methods, applied on
Retrospective and outlook 353
moving boundary conforming grids. The last 10 mostly those who have developed the codes. To
to 15 years have seen several significant advances become more generally applicable and more
in the capabilities ofthese methods. Novel elliptic accessible to non-expert users, the methods ought
mesh generation techniques and even some to be improved in several directions. For instance,
generalized algebraic schemes are sufficiently automated mesh generation algorithms are needed
flexible for discretizing very complicated coating- that decide on an optimal sub-division of the
flow configurations. Once set up for a particular flow domain into a few quadrilateral macro-
flow, the resulting meshes are quite robust for elements and, if necessary, adjust their topology
tracking significant changes in position and during a simulation; handle the complex data
shape of the free boundaries with minimal user structures that describe the connectivity between
interference. In combination with Saad's (1989) macro-elements and between finite elements in
iterative Arnoldi algorithm, specialized precon- complex flow geometries; refine the tessellation
ditioning techniques are efficient for computing in regions of steepest and most rapidly changing
the few most dangerous modes of the generalized gradients; and extend all these capabilities to
unsymmetric eigenproblem to which linear, three-dimensional flows with free and unknown
finite-element based stability analysis reduces. boundaries. Much work would also have to be
Critical-point-tracking locates turning and bi- done to provide robust algorithms for stability,
furcation points at which the stability of the operability and sensitivity analysis that are
computed flow changes without the need for trial applicable to a wide range of coating configur-
and error procedures and at a computational ations. More research is also needed in finding
cost only a few times higher than that ofcalculating consistent initial conditions for transient systems
a simple steady operating state. Methods of of index two and higher that are usually
computer-aided sensitivity analysis very effectively encountered in fluid mechanics and control theory;
and systematically probe the frequency response and in more accurately estimating and controlling
of stable coating flows to small-amplitude, oscil- time discretization errors. Finally, coating flow
latory disturbances that are invariably omnipresent computations should be more cleverly combined
in practical coating operations. And state-of-the-art with sophisticated optimization methods.
time integrators for systems ofdifferential-algebraic Existing finite-element methods for coating-
equations are available (e.g., DASSL, DASPK), flow analysis should also be improved in their
should the transient response of a viscous computational efficiency. Even though still under
free-surface flow to large disturbances be ofinterest. development, iterative matrix methods hold great
Altogether, advanced finite-element methods promise to speed up the solution of the large
can predict much of what a coating engineer is linear systems of equations that result from
interested in: they reveal a wealth of detail and Newton's methods in steady-state and transient
insight into the physical mechanisms that control analyses, preconditioning techniques so useful
the two-dimensional flow in successful coater for linear stability analysis, and critical point
operation; they assess the stability ofthe computed tracking and sensitivity analyses which further
flow states, including modes that produce oscil- augment the size of the system. Speed-up of the
latory and three-dimensional flow states; they solution process is especially needed for analysis
systematically map out stability limits that of 3D flows of realistic size, or 2D flows susceptible
delineate operability windows; and they quantify to 3D instabilities, and would also be crucial for
susceptibility ofsuccessful coating flows to external achieving real-time integration in transient analysis
disturbances and thereby identify quality windows. for on-line process control purposes. For the
Unfortunately, many of the advanced methods latter, the benefits of semi-implicit time integration
are highly specialized to a limited class of coating should also be investigated more systematically.
flows (e.g., photographic precision coating) and For certain coating-flow problems, finite-
are accessible only to a selected few experts, element techniques on deforming meshes may
not even be the best choice. Free-surface flows micro-structure, porosity, or surface texture but,
that undergo dramatic variations in free boundary more importantly, ultimate performance of the
shape or changes in topology are much easier coating in its intended function.
handled by fixed-grid methods or free Lagrangian
methods. These methods are also very attractive
when little is known a priori about the flow REFERENCES
geometry. Both methods have seen considerable
advances in the last 10-15 years, and should be Abbott, J. P. 1978. An efficient algorithm for the
determination ofcertain bifurcation points. J. Comput.
revisited for coating flow simulations. Spectral- App!. Math. 4: 19-27.
element techniques may also offer advantages, Addessio, F. L., Carroll, D. E., Dukowicz, J. K.,
especially for 3D periodic flows and deserve Harlow, F. H., Johnson, 1. N., Kashiwa, B. A.,
consideration as well. Maltrud, M. E. and Ruppel, H. M. 1986. CAVEAT:
Currently available methods for computational a computer code for fluid dynamics problems with
coating-flow analysis fall furthermore short in large distortion and internal slip. Los Alamos National
Laboratory Report: LA-I0613-MS.
their ability to account satisfactorily for essential
Ahmed, A. and Alexandrou, A. N. 1992. Compression
physical mechanisms. Most importantly, no molding using a generalized Eulerian-Lagrangian
computational technique advanced to date reliably formulation with automatic remeshing. Adv. Polym.
predicts the important high-speed coating limit Techno!' 11: 203-211.
that arises from excessive air-entrainment, or the Ahn, Y. C. and Ryan, M. E. 1991. A finite-difference
dynamic contact angles observed at lower speeds. analysis of the extrudate swell problem. Int. J.
This shortcoming, however, is not as much Numer. Meth. Fluids. 13: 1289-1310.
Amsden, A. A. 1966. The particle-in-cell method for
computational as it stems from the lack of a
the calculation of the dynamics of compressible
well-developed theory of the sub-microscopic air fluids. Los Alamos Scientific Laboratory Report.
displacement process that constitutes dynamic LA-3466.
wetting. Computing coating flows of highly Amsden, A. A., Ruppel, H. M. and Hirt, C. W. 1980.
non-Newtonian liquids also remains a formidable SALE: A simplified ALE computer program for
challenge, in spite of considerable progress in fluid flow at all speeds. Los Alamos Scientific
modeling viscoelastic flows in general. This Laboratory Report. LA-8095.
precludes an accurate analysis of many industrially Anturkar, N. R., Papanastasiou, T. C. and Wilkes,
1. O. 1991. Stability ofcoextrusion through converging
relevant coating systems. Species transport and dies. J. Non-Newt. Fluid Mech. 41: 1-25.
interfacial mass transfer, electrostatic forces, Aston, B. M. and Thomas, J. W. 1982. An implicit
magnetic forces, or even long-range fluid-solid scheme for water wave problems. Appl. Math.
interaction forces are more readily included in Comput. 10-11: 809-818.
coating flow simulations, but are not as commonly Babuska, B. 1971. Error bounds for finite element
accounted for as they probably should be. method. Numer. Meth. 16: 322-333.
Finally, extending coating process simulations Bach, P. and Hassager, O. 1985. An algorithm for the
use of the Lagrangian specification in Newtonian
to predict essential characteristics of the final fluid mechanics and applications to free-surface
coated product remains largely uncharted territory. flow. J. Fluid Mech. 152: 173-190.
Chapter 13 in this volume leads the way for the Bach, P. and Villadsen, 1. 1984. Simulation of the
special case of sol-gel systems that are dominated vertical flow of a thin, wavy film using a finite-
by evaporation of volatile solvents, but much element method. Int. J. Heat Mass Trans. 37: 815-827.
work remains to be done to compute particle Bapat, C. N. and Batra, R. C. 1984. Finite plane strain
migration and orientation in coating and drying; deformation ofnonlinear viscoelastic rubber-covered
rolls. Int. J. Numer. Meth. Eng. 20: 1911-1927.
diffusion, phase behavior, chemical reactions and Barr, P. K. and Ashurst, W. T. 1984. An interface
stress build-up during drying and/or cure of the scheme for turbulent flame propagation. Sandia
films; and many other phenomena that may National Laboratories Report. 82-8773.
influence not only physical properties such as Bathe, K.-J. 1982. Finite Element Procedures in
References 355
Engineering Analysis. Englewood Cliffs, NJ: Prentice extrusion computations. J. Non-Newt. Fluid Mech.
Hall. 16: 37-51.
Behrens, R. A., Crochet, M. J., Denson, C. D. and Bush, M. B. and Tanner, R. I. 1983. Numerical solution
Metzner, A. B. 1987. Transient free-surface flows: of viscous flows using integral equation methods.
motion of a fluid advancing in a tube. A IChE J. 33: Int. J. Numer. Meth. Fluids. 3: 71-92.
1178-1186. Bussmann, H., Hoffmann, H. and Beck, D. 1992.
Belytschko, T., Kennedy, J. M. and Schoeberle, D. F. Frequency response in slide coaters: FE calculations
1980. Quasi-Eulerian finite-element formulation for and measurements. Paper read at AIChE Spring
fluid-structure interaction. ASME J. Pressure Vessel National Meeting, New Orleans, LA, March 29-
Technol. 102: 62-69. April 2.
Bercovier, M. and Engelman, M. S. 1980. A finite Carey, G. F., Wang, K. C. and Joubert, W. D. 1989.
element method for incompressible non-Newtonian Performance of iterative methods for Newtonian
flow. J. Comput. Phys. 36: 313-326. and generalized Newtonian flows. Int. J. Numer.
Beris, A. N., Tsamopoulos, 1. A., Armstrong, R. C. Meth. Fluids. 9: 127-150.
and Brown, R. A. 1985. Creeping motion of a sphere Carvalho, M. S. and Scriven, L. E. 1993. Effect of
through a Bingham plastic. J. Fluid Mech. 158: deformable roll cover on roll coating. In Proc. 1993
219-244. TAP P1 Polymers, Laminations and Coatings Con-
Bird, R. B. 1976. Useful non-Newtonian models. Ann. ference, Chicago, IL, August 29-September 2, Book
Rev. Fluid Mech. 8: 13-26. 2, pp.451-460. Atlanta, GA: TAPPI.
Bird, R. B., Armstrong, R. C. and Hassager, O. 1987. Chan, R. K.-C. and Street, R. L. 1970. A computer
Dynamics of Polymeric Liquids - Volume 1: Fluid study of finite-amplitude water waves. J. Comput.
Dynamics, 2nd edn. New York: Wiley. Phys. 6: 68-94.
Bixler, N. E. 1982. Stability ofcoating flow. PhD thesis, Chern, I.-L., Glimm, J., McBryan, 0., Plohr, B. and
University of Minnesota, Minneapolis. Yaniv, S. 1986. Front tracking for gas dynamics. J.
Boris, 1. P. 1989. New directions in computational Comput. Phys. 62: 83-110.
fluid dynamics. Ann. Rev. Fluid Mech. 21: Chorin, A. J. 1980. Flame advection and propagation
345-385. algorithms. J. Comput. Phys. 35: 1-11.
Brackbill, 1. u., Kothe, D. B. and Zemach, C. 1992. Chorin, A. J. 1985. Curvature and solidification. J.
A continuum method for modeling surface tension. Comput. Phys. 58: 472-490.
J. Comput. Phys. 100: 335-354. Christodoulou, K. N. 1990. Computational physics of
Brackbill, 1. U. and Saltzman, 1. S. 1982a. Adaptive slide coating. PhD thesis, University of Minnesota,
zoning for singular problems in two dimensions. J. Minneapolis.
Comput. Phys. 46: 342-368. Christodoulou, K. N., Scofield, D. F. and Scriven,
Brackbill, J. U. and Saltzman, 1. S. 1982b. Applications L. E. 1990. Stability of multilayer coating flows to
and generalizations of variational methods for three-dimensional disturbances. Paper read at A IChE
generating adaptive meshes. Appl. Math. Comput. Spring National Meeting, Orlando, FL, March 18-22.
10-11: 865-884. Christodoulou, K. N. and Scriven, L. E. 1984. The
Brandt, A. 1977. Multi-level adaptive solutions to physics of slide coating, dynamic wetting, and air
boundary-value problems. Math. Comput. 31: entrainment. Paper read at A IChE Annual Meeting,
333-390. San Francisco, CA, November 25-30.
Brezzi, F. 1974. On the existence, uniqueness and Christodoulou, K. N. and Scriven, L. E. 1988. Finding
approximation of saddle point problems arising leading modes of a viscous free surface flow: an
with Lagrange multipliers. RAIRO, Serie Rouge. asymmetric generalized eigenproblem. J. Sci. Comput.
8 R-2: 129-151. 3: 355-406.
Brooks, A. N. and Hughes, T. 1. R. 1982. Streamline Christodoulou, K. N. and Scriven, L. E. 1989. The
upwind Petrov-Galerkin formulations for convection fluid mechanics of slide coating. J. Fluid Mech. 208:
dominated flows with particular emphasis on the 321-354.
incompressible Navier-Stokes equations. Comput. Christodoulou, K. N. and Scriven, L. E.1990. Singular
Meth. Appl. Mech. Eng. 32: 199-259. elements at dynamic wetting lines (unpublished).
Burket, R. S., Conaghan, B. F. and Hirshburg, R. I. Christodoulou, K. N. and Scriven, L. E. 1992.
1984. Coating method. US Patent 4,443,504. Discretization offree surface flows and other moving
Bush, M. B., Milthorpe, 1. F. and Tanner, R. I. 1984. boundary problems. J. Comput. Phys. 99: 39-55.
Finite element and boundary element methods for Christov, C. I. and Volkov, P. K. 1984. Numerical
investigation of the steady viscous flow past a viscoelastic flow revisited. J. Non-Newt. Fluid Mech.
stationary deformable bubble. J. Fluid Mech. 158: 42: 283-299.
341-364. Crowley, W. P. 1971. FLAG: a free-Lagrange method
Chung, K. Y. 1985. Shape optimization and free for numerically simulating hydrodynamic flows in
boundary problems with grid adaptation. PhD two dimensions. Lecture Notes in Physics. 8:
thesis, University of Michigan. 37-43.
Cliffe, K. A. 1983. Numerical calculations of two-cell Crowley, W. P. 1985. Free-Lagrange methods for
and single-cell Taylor flows. 1. Fluid Mech. 135: compressible hydrodynamics in two space dimensions.
219-233. Lecture Notes in Physics. 238: 1-21.
Cliffe, K. A. and Spence, A. 1984. The calculation of Cuvelier, C. and Schulkes, R. M. S. 1990. Some
high-order singularities in the finite Taylor problem. numerical methods for the computation of capillary
In Numerical Methods for Birfurcation Problems, free boundaries governed by the Navier-Stokes
IntI. Series Numer. Math., Vol. 70, ed. T. Kuepper, equations. SIAM Review. 32: 355-423.
H. D. Mittelmann and H. Weber, pp.129-144. Dahlquist, G. and Bjork, A. 1974. Numerical Methods.
Basel, and Boston, MA: Birkhaeuser. Englewood Cliffs: Prentice Hall.
Coates, P. J., Armstrong, R. C. and Brown, R. A. 1992. Daly, B. J. 1967. Numerical study of two fluid
Calculation of steady-state viscoelastic flow through Rayleigh-Taylor instability. Phys. Fluids. 10:
axisymmetric contractions with the EEME formu- 297-307.
lation. J. Non-Newt. Fluid Mech. 42: 141-188. Dandy, D. S. and Leal, L. G. 1989. A Newton's scheme
Couniot, A. and Crochet, M. 1. 1986. Finite-elements for solving free-surface flow problems. Int. J. Numer.
for the numerical simulation of injection molding. Meth. Fluids. 9: 1469-1486.
In Proc. 2nd Int!' Con! Numerical Methods in Daripa, P., Glimm, J., Lindquist, B., Maesumi, M. and
Industrial Forming Processes: NUMIFORM 86. McBryan, O. 1988. On the simulation of hetero-
Gothenburg, Sweden, August 25-29, pp.165-170. geneous petroleum reservoirs. In Numerical Simu-
Rotterdam, Netherlands, and Boston, MA: A. A. lation ofOil Recovery, ed. E. M. Wheeler, pp. 89-100.
Balkema. New York: Springer.
Cox, R. G. 1986. The dynamics of the spreading of Davey, A. 1962. The growth of Taylor vortices in flow
liquids on a solid surface. Part 1. Viscous flow. J. between rotating cylinders. J. Fluid Mech. 14:
Fluid Mech. 168: 169. 336-368.
Coyle, D. 1. 1984. The fluid mechanics of roll coating: de Santos,1. M. 1991. Two-phase co-current downflow
steady flows, stability, and rheology. PhD thesis, through constricted passages. PhD thesis, University
University of Minnesota, Minneapolis. of Minnesota, Minneapolis.
Coyle, D. J. 1988. Forward roll coating with Demkowicz, L. and Oden, 1. T. 1986. On a mesh
deformable rolls: a simple one-dimensional elas- optimization method based on a minimization of
tohydrodynamic model. Chem. Eng. Sci. 43: interpolation error. Int. J. Eng. Sci. 24: 55-68.
2673-2684. Dhatt, G., Gao, D. M. and Ben Cheikh, A. 1990. A
Coyle, D. 1. 1990. Nonlinear theory of squeeze- finite-element simulation of metal flow in molds.
roll coating. Paper read at AIChE Spring National Int. J. Numer. Meth. Eng. 30: 821-831.
Meeting, Orlando, FL, March 19-22. Dheur, J. and Crochet, M. 1. 1987. Newtonian stratified
Coyle, D. 1. 1992. Roll coating. In Modern Coating flow through a sudden expansion. Rheol. Acta. 26:
and Drying Technology, eds E. D. Cohen and E. B. 401-413.
Gutoff, pp.63-116. Weinheim, Germany: VCH Dheur,1. and Crochet, M. 1. 1989. Stratified flows of
Publishers. Newtonian and viscoelastic fluids. J. Non-Newt.
Coyle, D. 1., Macosko, C. W. and Scriven, L. E. 1987. Fluid Mech. 32: 1-18.
Film-splitting flows of shear-thinning liquids in DiPrima, R. C. and Rogers, E. H. 1969. Computing
forward roll coating. A IChE J. 33: 741-746. problems in nonlinear hydrodynamic stability. Phys.
Coyle, D. 1., Macosko, C. W. and Scriven, L. E. 1990. Fluids Suppl. II: 165-170.
Stability of film splitting between counter-rotating Do, D. V. and Christodoulou, K. N. 1992. Multilayer
cylinders. J. Fluid Mech. 216: 437-458. curtain coating flows: stability and sensitivity to
Crochet, M. 1. 1989. Numerical simulation ofviscoelastic small 3-D disturbances. Paper read at A IChE Spring
flow: a review. Rubber Chem. Techno!' 62: 426-455. National Meeting, New Orleans, LA, March 29-
Crochet, M. J. and Legat, V. 1992. The consistent April 2.
streamline-upwind/Petrov-Galerkin method for Donea,1. 1983. Arbitrary Lagrangian-Eulerian finite
References 357
element methods. In Computational Methods for Engelman, M. S. and Hasbani, I. 1990. Matrix-
Transient Analysis, eds T. Belytschko and T. R. J. free solution algorithms in a finite-element context.
Hughes, pp.473-516. Amsterdam: North-Holland. Fluid Dynamics International Technical Report 88-1.
Drazin, P. G. and Reed, W. H. 1981. Hydrodynamic Engelman, M. S. and Sani, R. L. 1983. Finite element
Stability. New York: Cambridge University Press. simulation of incompressible fluid flows with a
Duff, I. S. 1984. Design features of a frontal code for free/moving surface. In Numerical Methods in Laminar
solving sparse unsymmetric linear systems out- and Turbulent Flow (Proc. 3rd Inti. Con! Seattle,
of-core. SIAM J. Sci. Stat. Comput. 5: 270-280. WA), eds C. Taylor, J. A. Johnson and W. R. Smith,
Duff, I. S., Erisman, A. M. and Reid, 1. K. 1986. Direct pp.47-74. Swansea, UK: Pineridge Press.
Methods for Sparse Matrices. New York: Oxford Esmail, M. N. and Markov, V. V. 1989. Numerical
University Press. simulation of dynamic contact angles. Trans. Math.
Dukowicz, 1. 1981. Lagrangian fluid dynamics using Inst. Acad. Sci. USSR (English Trans.) 186: 203-208.
Voronoi-Delauny mesh. In Numerical Methodsfor Ettouney, H. M. and Brown, R. A. 1983. Finite-
Coupled Problems, eds E. Hinton, P. Bettes and R. element methods for steady solidification problems.
W. Lewis,pp. 82-104. Swansea, UK: Pineridge Press. J. Comput. Phys. 49: 118-150.
Dukowicz,1. 1984. Conservative rezoning (remapping) Fatemi, E. A. and Kistler, S. F. 1992. Finite-difference
for general quadrilateral meshes. J. Comput. Phys. discretization of viscous free-surface flows. Paper
54: 411-424. read at AIChE Spring National Meeting, New
Dupret, F. 1982. A method for the computation of Orleans, LA, March 29-April 2.
viscous flow by finite elements with free boundaries Finnicum, D. S., Weinstein, S. 1. and Ruschak, K. J.
and surface tension. In Finite Element Flow Analysis, 1993. The effect of applied pressure on the shape of
Proc. 4th Inti. Symp. FEM Flow Problems, Tokyo, a two-dimensional liquid curtain falling under the
26-29 July 1982, ed. T. Kawai, pp. 495-502. Tokyo: influence of gravity. J. Fluid Mech. 255: 647-665.
University of Tokyo Press. Floryan, 1. M. and Rasmussen, H. 1989. Numerical
Durst, F., Haas, R., Peric, M., Scheurer, G. and methods for viscous flows with free/moving bound-
Wagner, H.-G. 1991. Numerische Simulation aries. Applied Mechanics Reviews. 42: 323-341.
zweidimensionaler Schlitzdiisenstromungen bei der Fortin, M. and Fortin, A. 1985. A generalisation of
Fliesserbeschichtung. Coating. 2/91: 52-54, 3/91: Uzawa's algorithm for the solution of the Navier-
90-92,7/91: 259-263, and 10/91: 360-366. Stokes equations. Comm. Appl. Numer. Meth. 1:
Dutta, A. and Ryan, M. E. 1982. Dynamics of a 205-210.
creeping Newtonian jet with gravity and surface Fortin, M. and Glowinski, R. 1983. The Augmented
tension - a finite difference technique for solving Lagrangian Method. Amsterdam: North-Holland.
steady free-surface flows using orthogonal curvilinear Francis, D. c., Christodoulou, K. N., Scofield, D. F.
coordinates. AIChE J. 28: 220-232. and Lang, C. 1990. Fundamental studies of slot
Edwards, W. S., Tuckerman, L. S., Friesner, R. A. and coating. Paper read at AIChE Spring National
Sorensen, D. C. 1994. Krylov methods for the Meeting, Orlando, FL, March 18-22.
incompressible Navier-Stokes equations. J. Comput. Frank, R. M. and Lazarus, R. B. 1964. Mixed
Phys. 110: 82-102. Eulerian-Lagrangian methods. Methods in Com-
Einset, E. and Jensen, K. F. 1992. A finite element putational Physics. 3: 117-179.
solution ofthree-dimensional mixed convection gas Frederiksen, C. S. and Watts, A. M. 1981. Finite-
flows in horizontal channels using preconditioned element method for time-dependent incompressible
iterative matrix methods. Int. J. Numer. Meth. free surface flow. J. Comput. Phys. 39: 282-304.
Fluids. 14: 817-841. Fritts, M. J. and Boris, J. P. 1979. The Lagrangian
Eiseman, P. R. 1982. Orthogonal grid generation. solution of transient problems in hydrodynamics.
Appl. Math. Comput. 10-11: 193-233. J. Comput. Phys. 31: 173-215.
Ellwood, K. R. 1., Papanastasiou, T. C. and Wilkes, Fritts, M. 1., Crowley, W. P. and Trease, H. E. (eds).
J. O. 1992. Three-dimensional streamlined finite 1985. The Free Lagrange Method. Lecture Notes in
elements - design of extrusion dies. Int. J. Numer. Physics, Volume 238. New York: Springer Verlag.
Meth. Fluids. 14: 13-24. Fromm, J. 1981. Finite-difference computation of the
Engelman, M. S.1982. Quasi-Newton methods in fluid capillary jet free surface problem. Lecture Notes in
dynamics. In The Mathematics of Finite Element Physics. 141: 188-193.
Applications, Vol. IV, ed. J. R. Whiteman, Fromm, 1. 1986. Free-surface calculation of capillary
pp.479-487. London: Academic Press. spreading. Lecture Notes in Physics. 264: 283-289.
Fuller, G. G., Cathey, C. A., Hubbard, B. and Zebrowski, finite-element method that also introduces a nearly
B. B. 1987. Extensional viscosity measurements for consistent mass matrix. Part 2: Implementation. Int.
low-viscosity fluids. J. Rheol. 31: 235-249. J. Numer. Meth. Fluids. 11: 621-660.
Fyfe, D. E., Oran, E. S. and Fritts, M. 1. 1988. Surface Gresho, P. M., Lee, R. L. and Sani, R. L. 1980. On
tension and viscosity with Lagrangian hydro- the time-dependent solution of the incompressible
dynamics on a triangular mesh. J. Comput. Phys. Navier-Stokes equations in two and three dimen-
76: 349-384. sions. In Recent Advances in Numerical Methods in
Gartling, D. K. and Phan-Thien, N. 1984. A numerical Fluids, eds C. Taylor and K. Morgan, pp.27-75.
simulation of a plastic fluid in a parallel-plate Swansea, UK: Pineridge Press.
pIastometer. J. non-Newt. Fluid Mech. 14: 347-360. Gresho, P. M. and Sani, R. L. 1987. On pressure
Gaster, M. 1962. A note on the relation between boundary conditions for the incompressible
temporally-increasing and spatially-increasing dis- Navier-Stokes equations. Int. J. Numer. Meth.
turbances in hydrodynamic stability. J. Fluid. Mech. Fluids. 7: 1111-1145.
14: 222-224. Griewank, A. and Corliss, G. F. 1992. Automatic
Geller, A. S., Lee, S. H. and Leal, L. G. 1986. The Differentiation of Algorithms. Philadelphia: SIAM.
creeping motion of a spherical particle normal to Habashi, W. G., Nguyen, V.-N. and Bhat, M. V. 1991.
a deformable interface. J. Fluid Mech. 169: 27-69. Efficient direct solvers for large-scale computational
Glimm, 1., Grove, J., Lindquist, B., McBryan, O. A. fluid dynamics problems. Comput. Meth. Appl.
and Tryggvason, G. 1988. The bifurcation of tracked Mech. Eng. 87: 253-265.
scalar waves. SIAM J. Sci. Stat. Comput. 9: 61-79. Hackler, M. A., Christodoulou, K. N., Hirschburg,
Glimm,1. and McBryan, O. A. 1985. A computational R. I. and Lightfoot, E. J. 1992. Frequency response
model for interfaces. Adv. Appl. Math. 6: 422-435. of coating flows to small three-dimensional disturb-
Glimm, 1., McBryan, 0., Menikolf, R. and Sharp, ances by supercomputer-aided analysis. Paper read
D. H. 1986. Front tracking applied to Rayleigh-Taylor at AIChE Spring National Meeting, New Orleans,
instability. SIAM J. Sci. Stat. Comput. 7: 230-251. LA, March 29-April 2.
Goenaga, A. and Higgins, B. G. 1990. Analysis of free Haftka, T. and Grandhi, R. V. 1986. Structural shape
surface flow problems by the boundary element optimization - a survey. Comput. Meth. Appl.
method. Paper read at AIChE Spring National Mech. Eng. 57: 91-106.
Meeting, Orlando, FL, March 18-22. Hansen, E. B. 1987. Stokes flow down a wall into an
Golafshani, M. 1988. A simple numerical technique infinite pool. J. Fluid Mech. 178: 243-256.
for transient creep flows with free surface. Int. J. Hansen, E. B. 1991. Stokes flow of a fluid layer over
Numer. Meth. Fluids. 8: 897-912. an obstacle on a tilted plane. Math. Comput.
Goldhirsch, I., Orszag, S. A. and Maulik, B. K. 1987. Modeling. 15: 185-193.
An efficient method for computing leading eigenvalues Harlow, F. 1964. The particle-in-cell computing method
oflarge asymmetric matrices. J. Sci. Comput. 2: 33-58. for fluid dynamics. Methods in Computational Physics.
Goodwin, R. and Homsy, G. M. 1991. Viscous flow 3: 313-343.
down a slope in the vicinity of a contact line. Phys. Harlow, F. H. and Amsden, A. A. 1971. Numerical
Fluids A - Fluid Dynamics 3: 515-528. fluid dynamics. In Fluid Dynamics, Los Alamos
Gordon, W. J. and Hall, C. A. 1973. Construction of Scientific Laboratory Report LA-4700.
curvilinear coordinate systems and applications to Harlow, F. H. and Welch, 1. E. 1965. Numerical
mesh generation. Int. J. Numer. Meth. Eng. 7: calculation oftime-dependent viscous incompressible
461-477. flow of fluid with free surface. Phys. Fluids. 8:
Grald, E. W., Chakrabarti, M., Subbiah, S. and 2182-2189.
Gephart, 1. 1994. The spectral element method and Haroutunian, V., Engelman, M. S. and Hasbani, I. 1993.
its application to modeling coating flows. Paper Segregated finite element algorithms for the numerical
read at AIChE Spring National Meeting, Atlanta, solution oflarge-scale incompressible flow problems.
GA, April 17-21. Int. J. Numer. Meth. Fluids. 17: 323-348.
Gresho, P. M. 1991. Some current CFD issues relevant Hasbani, Y. and Engelman, M. 1979. Out-of-core
to the incompressible Navier-Stokes equations. solution of linear equations with non-symmetric
Comput. Meth. Appl. Mech. Eng. 87: 201-252. coefficient matrix. Comput. Fluids. 7: 13-31.
Gresho, P. M. and Chan, S. T. 1990. On the theory Haussling, H. J. 1982. Solution of nonlinear water
of semi-implicit projection methods for viscous wave problems using boundary-fitted coordinate
incompressible flow and its implementation via a systems. Appl. Math. Comput. 10-tt: 385-407.
References 359
Hayes, R. E., Dannelongue, H. H. and Tanguy, P. A. Hughes, T. J. R., Liu, W. K. and Zimmerman, T. K.
1991. Numerical simulation ofmold filling in reaction 1981. Lagrangian-Eulerian finite-element form ulation
injection molding. Polym. Eng. Sci. 31: 842-848. for incompressible viscous flows. Comput. Meth.
Heijer, C. D. and Rheinboldt, W. C. 1981. Onsteplength Appl. Mech. Eng. 29: 329-349.
algorithms for a class of continuation methods. Huh, K. Y., Golay, M. W. and Manno, V. P. 1986. A
SIAM J. Numer. Anal. 18: 925-948. method for reduction of numerical diffusion in the
Higgins, B. G. 1980. Capillary hydrodynamics and donor treatment of convection. J. Comput. Phys.
coating beads. PhD thesis, University of Minnesota, 63: 201-221.
Minneapolis. Hurez, P. and Tanguy, P. A. 1990. Finite element
Hill, G. A., Shook, C. A. and Esmail, M. N. 1981. analysis of dip coating with Bingham fluids. Polym.
Finite-difference simulation of die-swell for a Eng. Sci. 30: 1125-1132.
Newtonian fluid. Canad. J. Chern. Eng. 58: 100-108. Hurez, P., Tanguy, P. A. and Bertrand, F. H. 1991. A
Hirt, C. W., Amsden, A. A. and Cook, J. L. 1974. An finite element analysis ofdie swell with pseudoplastic
arbitrary Lagrangian-Eulerian computing method and viscoplastic fluids. Comput. Meth. Appl. Mech.
for all flow speeds. J. Comput. Phys. 14: 227-253. Eng. 86: 97-103.
Hirt, C. W., Cook, J. L. and Butler, T. D. 1970. A Huyakorn, P. S., Taylor, c., Lee, R. L. and Gresho,
Lagrangian method for calculating the dynamics of P. M. 1978. A comparison of various mixed
an incompressible fluid with free surface. J. Comput. interpolation finite elements in the velocity-pressure
Phys. 5: 103-124. formulation ofthe Navier-Stokes equations. Comput.
Hirt, C. W., Korzekwa, D. R., Rollet, A. D. and Wilde, Fluids. 6: 25-35.
P. 1990. Planar flow casting: modeling and under- Hyman, J. M. 1984. Numerical methods for tracking
standing. Flow Science (Flow Science Inc., Los interfaces. Physica. 12D: 396-407.
Alamos, NM) Report FSI-90-oo-4. Hyun, J. C. and Ballman, R. L. 1978. Isothermal melt
Hirt, C. W. and Nichols, B. D. 1981. Volume of fluid spinning - Lagrangian and Eulerian viewpoints. J.
(VOF) method for the dynamics of free boundaries. Rheol. 22: 349-380.
J. Comput. Phys. 39: 210-225. Isaacson, E. and Keller, H. B. 1966. Analysis of
Hirt, C. w., Nichols, B. D. and Romero, N. C. 1975. Numerical Methods. New York: Wiley.
SOLA - a numerical solution algorithm for transient Jackson, C. P. 1987. A finite-element study ofthe onset
flows. Los Alamos Scientific Laboratory Report of vortex shedding in flow past variously shaped
LA-5852. bodies. J. Fluid Mech. 182: 23-45.
Hirt, C. W. and Shannon, J. P. 1968. Free-surface Jackson, C. P. and Robinson, P. C. 1985. A numerical
stress conditions for incompressible flow calculations. study of various algorithms related to the precon-
J. Comput. Phys. 2: 403-411. ditioned conjugate gradient method. Int. J. Numer.
Ho, L.-W. and Patera, A. T. 1990. A Legendre spectral Meth. Eng. 21: 1315-1338.
element method for simulation of unsteady incom- Jackson, C. P. and Winters, K. H. 1984. A finite-
pressible free-surface flows. Comput. Meth. Appl. element study ofthe Benard problem using parameter-
Mech. Eng. 80: 355-366. stepping and bifurcation search. Int. J. Numer.
Ho, L.-W. and Patera, A. T. 1991. Variational Meth. Fluids. 4: 127-145.
formulation of three-dimensional free-surface flows. Johnson, M., Kamm, R. D., Ho, L.-W., Shapiro, A.
Natural imposition of surface tension boundary and Pedley, T. J. 1991. The nonlinear growth of
conditions. Int. J. Numer. Meth. Fluids. 13: 691-698. surface-tension-driven instabilities of a thin annular
Hood, P. 1976. Frontal solution program for unsym- film. J. Fluid. Mech. 233: 141-156.
metric matrices. Int. J. Numer. Meth. Eng. 10: Joseph, D. D. 1990. Fluid Dynamics of Viscoelastic
379-399. Liquid. New York: Springer.
Hood, P. 1977. Correction. Int. J. Numer. Meth. Eng. Kaneko, N. and Scriven, L. E. 1988. Transient analysis
11: 1055. of a curtain coating flow. Paper read at AIChE
Howard, D., Connolley, W. M. and Rollet, J. S. 1990. Spring National Meeting, New Orleans, LA,
Unsymmetric conjugate gradient methods and sparse March 6-10.
direct methods in finite element flow simulations. Karagiannis, A., Hrymak, A. N. and Vlachopoulos, J.
Int. J. Numer. Meth. Fluids. 10: 925-945. 1988. Three-dimensional extrudate swell of creeping
Huerta, A. and Liu, W. K. 1988. Viscous flows with Newtonian jets. AIChE J. 34: 2088-2094.
large free surface motion. Comput. Meth. Appl. Karagiannis, A., Mavridis, H., Hrymak, A. N. and
Mech. Eng. 69: 277-324. Vlachopoulos,1. 1989. A finite element convergence
study for shear-thinning flow problems. Int. J. Kistler, S. F. and Palmquist, K. E. 1990. A simple,
Numer. Meth. Fluids 8: 121-133. algebraic finite-element discretization offree surface
Katagiri, Y. 1992. Analysis of edge effects in curtain flows. Paper read at AIChE Spring National Meeting,
coating. Paper read at AIChE Spring National Orlando, FL, March 18-22.
Meeting, New Orleans, LA, March 29-April 2. Kistler, S. F. and Scriven, L. E. 1983. Coating flows.
Katagiri, Y. and Scriven, L. E. 1986. Supercomputer- In Computational Analysis of Polymer Processing,
aided analysis of transient response of a coating eds J. R. A. Pearson, and S. M. Richardson,
operation. Paper read at AIChE Annual Meeting, pp. 243-299. London and New York: Applied Science
Miami Beach, FL, November 2-7. Publishers.
Keller, H. B. 1977. Numerical solution of bifurcation Kistler, S. F. and Scriven, L. E. 1984. Coating flow
and nonlinear eigenvalue problems. In Applications theory by finite-element and asymptotic analysis of
ofBifurcation Theory, ed. P. H. Rabinowitz, pp. 45- 52. the Navier-Stokes system. Int. J. Numer. Meth.
New York: Academic Press. Fluids. 4: 207-229.
Kelmanson, M. A. 1983. Boundary integral equation Kistler, S. F. and Scriven, L. E. 1994. The teapot effect:
solution of viscous flows with free surfaces. J. Eng. sheet forming flows with deflection, wetting, and
Math. 17: 329-343. hysteresis. J. Fluid Mech. 263: 19-62.
Keunings, R. 1986. An algorithm for the simulation of Kistler, S. F. and Zvan, G. 1991. Hydrodynamic
transient viscoelastic flows with free boundaries. J. models of forced wetting in coating flows. Paper
Comput. Phys. 62: 199-220. read at 44th IS&T Conference, St Paul, MN, May
Keunings, R. 1989. Simulation of viscoelastic flow. In 12-17.
Fundamentals of Computer Modeling for Polymer Kreis, R. I., Thames, F. C. and Hassan, H. A. 1986.
Processing, ed. C. L. Tucker, pp.403-469. Munich: Application of a variational method for generating
Hanser. adaptive grids. AIAA J. 24: 404-410.
Keunings, R. 1991. Progress and challenges in Kruyt, N. P., Cuvelier, c., Segal, A. and van der
computational rheology. Rheo!' Acta. 29: 556-570. Zanden, 1. 1988. A total linearization method for
Kheshgi, H. S. and Scriven, L. E. 1984. Penalty finite- solving viscous free boundary flow problems by the
element analysis of unsteady free-surface flows. In finite element method. Int. J. Numer. Meth. Fluids.
Finite Elements in Fluids, Volume 5, eds R. H. 8: 351-363.
Gallagher, J. T. Oden, O. C. Zienkiewicz, T. Kawai Ladyzhenskaya, O. A. 1963. The Mathematical Theory
and M. Kawahara, pp.393-434. New York: John of Viscous Incompressible Flow. New York: Gordon
Wiley. and Breach.
Kheshgi, H. S. and Scriven, L. E. 1985. Variable penalty Lee, S. H. and Leal, L. G. 1982. The motion of a sphere
method for finite element analysis of incompressible in the presence of a deformable interface. II. A
flow. Int. J. Numer. Meth. Fluids. 5: 785-803. numerical study of the translation of a sphere
Kheshgi, H. S. and Scriven, L. E. 1988. The evolution normal to an interface. J. Colloid Interface Sci. 87:
of disturbances in horizontal films. Chem. Eng. Sci. 81-106.
43: 793-801. Legat, V. and Marchal, J.-M. 1992. Prediction of
Kikuchi, N., Chung, K. Y., Torigaki, T. and Taylor, three-dimensional general shape extrudates by an
J. E. 1986. Adaptive finite element methods for shape implicit iterative scheme. Int. J. Numer. Meth.
optimization of linearly elastic structures. Comput. Fluids. 14: 609-625.
Meth. Appl. Mech. Eng. 57: 67-89. Leonard, A. 1985. Computing three-dimensional
King, R. c., Apelian, M. N., Armstrong, R. C. and incompressible flows with vortex elements. Ann.
Brown, R. A. 1988. Numerically stable finite element Rev. Fluid Mech. 17: 523-559.
techniques for viscoelastic calculations in smooth Liang, P. Y. 1991. Numerical method for calculation
and singular geometries. J. Non-Newt. Fluid Mech. of surface tension flows in arbitrary grids. AIAA J.
29: 147-216. 29: 161-167.
Kistler, S. F. 1984. The fluid mechanics of curtain Lin, S. P., Lian, Z. W. and Creighton, B. 1. 1990.
coating and other related viscous free surface flows Absolute and convective instability of a liquid sheet.
with contact lines. PhD thesis, University of J. Fluid Mech. 220: 673-689.
Minnesota, Minneapolis. Lin, S. P. and Roberts, G. 1981. Waves in a viscous
Kistler, S. F. 1993. The hydrodynamics of wetting. In liquid curtain. J. Fluid Mech. 112: 443-458.
Wettability, ed. 1. C. Berg, pp. 311-429. New York: Liou, 1. and Tezduyar, T. E. 1991. Computation of
Marcel Dekker. compressible and incompressible flows with the
References 361
clustered element-by-element technique. In Domain moulding process using finite element method with
decomposition methods for partial differential equa- marker particles. Int. J. Numer. Meth. Eng. 30:
tions, eds R. Glowinski et al., pp. 140-150. Philadel- 1569-1575.
phia, P A: SIAM. McKibben, 1. F. and Aidun, C. K.1991. Computational
Liu, J.-J., Yu, T.-A. and Cheng, S.-H. 1991. Finite- visualization ofthree-dimensional free surface flows.
difference solution ofa Newtonian jet swell problem. In Proc. TAPPI Engineering Conference (Nashville,
Int. J. Numer. Meth. Fluids. 12: 125-142. TN, Sept. 30-0ct. 3), Book 2, pp. 711-718. Atlanta,
Loh, C. Y. and Rasmussen, H. 1987. A numerical GA: T APPI Press.
procedure for viscous free surface flows. Appl. Melhem, R. G. and Rheinboldt, W. C. 1982. A
Numer. Math. 3: 479-495. comparison of methods of determining turning
Longuet-Higgins, M. S. and Cokelet, E. D. 1976. The points of nonlinear equations. Computing. 29:
formation of steep surface waves on water. Part 1: 201-226.
a numerical method of computation. Proc. Roy. Meyer, K. A. 1969. Three dimensional study of flow
Soc. Lond., Ser. A. 350: 1-26. between concentric rotating cylinders. Phys. Fluids.
Lowndes, 1. 1980. The numerical simulation of the Suppl. II: 155-165.
steady movement of a fluid meniscus in a capillary Mittal, S. and Tezduyar, T. E. 1992. Space-time finite
tube. J. Fluid Mech. 101: 631-646. element computations of incompressible flows
Lu, W. Q. and Chang, H. C. 1988. An extension of the involving oscillating cylinders. Int. J. Numer. Meth.
biharmonic boundary integral method to free-surface Fluids. 15: 1073-1118.
flow in channels. J. Comput. Phys. 77: 340-360. Miura, H. and Aidun, C. K. 1992. Pressure fluctuations
Luo, x.-L. and Mitsoulis, E. 1990. An efficient algorithm and coat-weight nonuniformities in blade coating.
for strain history tracking in finite-element compu- In Proc. 1992 TAPPI Coating Conference,
tations of non-Newtonian fluids with integral pp.193-216. Atlanta: TAPPI Press.
constitutive equations. Int. J. Numer. Meth. Fluids. Miyamoto, K. 1992. On the mechanism of air
11: 1015-1031. entrainment. Ind. Coating Res. 1: 71-88.
Luo, x.-L. and Tanner, R. I. 1988. Finite-element Miyata, H. 1986. Finite-difference simulation ofbreaking
simulation oflong and short circular die experiments waves. J. Comput. Phys. 65: 179-214.
using integral models. Int. J. Numer. Meth. Eng. Moffat, H. K. 1964. Viscous and resistive eddies near
25: 9-22. a sharp corner. J. Fluid Mech. 18: 1-18.
Lynch, D. R. 1982. Unified approach to simulation Monaghan, 1. 1. 1985. Particle methods for hy-
on deforming elements with application to phase drodynamics. Comput. Phys. Rep. 3: 71-124.
change problems. J. Comput. Phys. 47: 387-411. Morice, P. 1983. Numerical generation of boundary-
Maday, Y. and Patera, A. T. 1989. Spectral element fitted coordinate systems with optimal control of
methods for the Navier-Stokes equations. In State- orthogonality. In Advances in Grid Generation (Proc.
of-the-art Surveys in Computational Mechanics, eds ASME AppliedMechanics, Bioengineering, and Fluids
J. T. Oden, and A. K. Noor. New York: ASME. Engineering Conference; Houston), FED-Vol. 5, eds
Malamataris, N. T. and Papanastasiou, T. C. 1991. K. N. Ghia and U. Ghia, pp. 71-78. New York:
Unsteady free surface flows on truncated domains. ASME.
Ind. Eng. Chern. Res. 30: 2211-2219. Miies, W., Hens, J. and Boiy, L. 1989. Observation of
(Manteuffel, T. A. 1977. The Tchebychev iteration for dynamic wetting process with laser-Doppler
non-symmetric linear systems. Numer. Math. 28: velocimetry. AIChE J. 35: 1521-1526.
307-327. Mulder, W., Osher, S. and Sethian, J. A. 1992. Computing
Marchal, M. and Crochet, M. J. 1987. A new mixed interface motion in compressible gas dynamics. J.
finite-element method for calculating viscoelastic Comput. Phys. 100: 209-228.
flow. J. Non-Newt. Fluid Mech. 26: 77-114. Muttin, F., Coupez, T., Bellet, M. and Chenot, 1. L.
Margolin, L. G. and Nichols, B. D. 1983. Momentum 1991. Newtonian fluid computations in Lagrangian
control volumes for finite-difference codes. In variables with a remeshing technique. Application
Numerical Methods in Laminar and Turbulent Flows to the filling stage of the diecasting process. In Proc.
(Proc. 3rd IntI. Conf, Seattle, WA), eds C. Taylor, 1st IntI. Conf Computational Modelling ofFree and
1. A. Johnson and W. R. Smith, pp.411-421. Moving Boundary Problems (Southampton, England,
Swansea, UK: Pineridge Press. July 2-4), eds L. C. Wrobel and C. A. Brebbia,
Matsuhiro, I., Shiojima, T., Shimazaki, Y. and Daiuji, pp. 207 - 221. Southampton, England: Computational
H. 1990. Numerical analysis of polymer injection Mechanics Publications.
Natarajan, R. 1992. An Arnoldi-based iterative scheme strategies for real matrices. Yale University, Computer
for nonsymmetric matrix pencils arising in finite Science Dept. Tech. Report YALEV/DCS-RR-424.
element stability problems. J. Comput. Phys. 100: Patera, A. T. 1984. A spectral element method for fluid
128-142. flow: laminar flow in a channel expansion. J.
Nelson, N. K., Kistler, S. F. and Olmsted, R. D. 1990. Comput. Phys. 54: 468-488.
A generalized viscous model for the steady shear Pavelle, R. and Wang, P. S. 1985. MACSYMA from
rheology of magnetic dispersions. Paper read at F to G. J. Symbol. Comput. 1: 69-100.
62nd Annual Meeting of the Society of Rheology, Pearson, 1. R. A. 1960. The instability of uniform
Santa Fe, NM, October 21-25. viscous flow under rollers and spreaders. J. Fluid
Nichols, B. D. and Hirt, C. W. 1971. Improved free Mech. 7: 481-500.
surface boundary conditions for numerical incom- Peitgen, H. and Pruefer, M. 1984. Global aspects of
pressible-flow calculations. J. Comput. Phys. 8: Newton's method for nonlinear boundary value
434-448. problems. In Numerical Methods for Bifurcation
Nichols, B. D. and Hirt, C. W. 1973. Calculating Problems, IntI. Series Numer. Math., Vol. 70, eds
three-dimensional free surface flows in the vicinity T. Kuepper, H. D. Mittelmann and H. Weber,
of submerged and exposed structures. J. Comput. pp.352-368. Basel: Birkhauser.
Phys. 12: 234-246. Peskin, C. S. 1977. Numerical analysis of blood flow
Nichols, B. D. and Hirt, C. W. 1975. Methods for in the heart. J. Comput. Phys. 25: 220-252.
calculating multi-dimensional transient free surface Petzold, L. R. 1982. A description of DASSL: A
flows past bodies. In Proc. First Inti. Conf. differential/algebraic system solver. Sandia National
Numer. Methods Ship Hydrodynamics, Gaithersburg, Laboratory Report SAND82-8637.
Maryland, pp. 253-277. Petzold, L. R. 1982. Differential/algebraic equations
Nichols, B. D., Hirt, C. W. and Hotchkiss, R. S. 1980. are not ODE's. SIAM J. Sci. Stat. Comput. 3:
SOLA-VOF - a solution algorithm for transient 367-384.
fluid flow with multiple free boundaries. Los Alamos Pilitsis, S. and Beris, A. N. 1989. Calculation of
Scientific Laboratory Report LA-8355. steady-state viscoelastic flow in an undulating tube.
Nickell, R. E., Tanner, R. I. and Caswell, B. 1974. The J. Non-Newt. Fluid Mech. 31: 231-287.
solution of viscous incompressible jet and free Pitts, E. and Greiller, 1. 1961. The flow of thin liquid
surface flows using finite-element methods. J. Fluid films between rollers. J. Fluid Mech. 11: 33-50.
Mech. 65: 189-206. Poenisch, G. D. and Schwetlick, H. H.1981. Computing
Noh, W. F. and Woodward, P. 1976. SUC (simple turning points of curves implicitly defined by
line interface calculation). Lecture Notes in Physics. nonlinear equations depending on a parameter.
59: 330-340. Computing. 26: 107-121.
Oran, E. S. and Boris, 1. P.1987. Numerical Simulation Potter, D. E. and Tuttle, G. H. 1973. The construction
of Reactive Flow. New York: Elsevier. of discrete orthogonal coordinates. J. Comput.
Ortega, J. M. and Rheinboldt, W. C. 1970. Iterative Phys. 13: 483-501.
Solution ofNonlinear Equations in Several Variables. Pozrikidis, C. 1988. The flow of a liquid film along a
New York: Academic Press. periodic wall. J. Fluid Mech. 188: 275-300.
Osher, S. and Sethian, J. A. 1988. Fronts propagating Pozrikidis, C. and Thoroddsen, S. T. 1991. The
with curvature-dependent speed: algorithms based deformation of a liquid film flowing down an
on Hamiltonian-Jacobi formulation. J. Comput. inclined plane wall over a small particle arrested on
Phys. 79: 12-49. the wall. Phys. Fluids A - Fluid Dynamics. 3:
Papanastasiou, T. C. 1987. Flows of material with 2546-2558.
yield. J. Rheol. 31: 385-404. Pracht, W. E. 1971. A numerical method for calculating
Papanastasiou, T. c., Macosko, C. W. and Scriven, transient creep flows. J. Comput. Phys. 7: 46-60.
L. E. 1985. Streamlined finite-elements and transit Pranckh, F. R. and Scriven, L. E. 1989. Elastohyd-
times. In Finite Elements in Fluids, Vol. 6, eds R. H. rodynamics of membrane coating. University of
Gallagher, G. F. Carey, 1. T. Oden and O. C. Minnesota Supercomputer Institute Report VMSI
Zienkiewicz, pp.263-278. New York: Wiley. 89/83.
Papanastasiou, T. c., Scriven, L. E. and Macosko, Pranckh, F. R. and Scriven, L. E. 1990a. The
C. W. 1987. A finite element method for liquid with elastohydrodynamics of blade coating. AIChE J.
memory. J. Non-Newt. Fluid Mech. 22: 271-288. 36: 587-597.
Parlett, B. and Saad, Y. 1985. Complex shift and invert Pranckh, F. R. and Scriven, L. E. 1990b. The physics
References 363
of blade coating of a deformable substrate. TAPPI Ryskin, G. and Leal, L. G. 1983. Orthogonal mapping.
J. 73(1): 163-173. J. Comput. Phys. 50: 71-100.
Rajagopalan, D., Armstrong, R. C. and Brown, R. A. Ryskin, G. and Leal, L. G. 1984. Numerical solution
1990. Finite element methods for calculation of offree-boundary problems in fluid mechanics. Part
steady, viscoelastic flow using constitutive equations 1. The finite-difference technique. J. Fluid Mech.
with a Newtonian viscosity. J. Non-Newt. Fluid 148: 1-17.
Mech. 36: 159-192. Saad, y. 1980. Variations on Arnoldi's method for
Rajagopalan, D., Armstrong, R. C. and Brown, R. A. computing eigenelements of large unsymmetric
1993. Comparison of computational efficiency of matrices. Linear Algeb. Appl. 34: 269-295.
flow simulations with multimode constitutive equa- Saad, y. 1989. Numerical solution oflarge nonsymmetric
tions: integral and differential models. J. Non-Newt. eigenvalue problems. Comput. Phys. Commun. 53:
Fluid Mech. 46: 243-273. 71-90.
Rajagopalan, D., Philips, R. 1., Armstrong, R. C. and Saad, Y. and Schultz, M. 1986. GMRES: a generalized
Brown, R. A. 1992. The influence of viscoelasticity minimal residual algorithm for solving nonsymmetric
on the existence of steady solutions in two-dimen- linear systems. SIAM J. Sci. Stat. Comput. 7:
sional rimming flow. J. Fluid Mech. 235: 611-642. 856-869.
Ramaswamy, 8.1990. Numerical simulation ofunsteady Sackinger, P. A., Schunk, P. R. and Rao, R. R. 1996.
viscous free-surface flow. J. Comput. Phys. 90: A Newton - Raphson pseudo-solid domain mapping
396-430. technique for free and moving boundary problems:
Ramshaw,1. D. 1985. Conservative rezoning algorithm a finite element implementation. J. Compo Phys.,
for generalized two-dimensional meshes. J. Comput. 125: 83-103.
Phys.59: 193-199. Saeger, R. B., Davis, H. T. and Scriven, L. E. 1990.
Rao, R. R. and Finlayson, 8. A. 1991. Adaptive Adaptive elliptic grid generation for 3-D elemental
refinement of a viscoelastic flow problem with the structures. Univ. ofMinnesota Supercomputer Institute
explicitly elliptic momentum equation. J. Non- Report UMSI 90/73.
Newt. Fluid Mech. 38: 223-246. Saito, H. and Scriven, L. E. 1981. Study of a coating
Reddy, M. P. and Reddy, J. N. 1991. Finite element flow by the finite element method. J. Comput. Phys.
simulation of mold filling processes. In Advances in 42: 53-76.
Finite Deformation Problems in Materials Processing Salamon, T. R., Armstrong, R. C. and Brown, R. A.
and Structures, AMD-Vol. 125, pp.65-74. New 1992. Modification of inertial film instability by
York: ASME. viscoelasticity. In Theoretical and Applied Rheology
Reid, J. K. 1981. Frontal methods for solving finite- (Proc. Xlth IntI. Congr. Rheology, Brussels, Belgium,
element systems of linear equations. In Sparse August 17-21, 1992), eds P. Moldenaers and R.
Matrices and their Uses, ed. I. S. Duff, pp.275-281. Keunings, pp. 213-215. Amsterdam: Elsevier.
New York: Academic Press. Savage, M. D. 1977a. Cavitation in lubrication. Part
Rheinboldt, W. C. and Burkardt, J. V. 1983. A locally 1. On boundary conditions and cavity-fluid inter-
parametrized continuation process. ACM Trans. faces. J. Fluid Mech. 80: 743-755.
Math. Software. 9: 215-235. Savage, M. D. 1977b. Cavitation in lubrication. Part
Rosenberg, J. and Keunings, R. 1991. Numerical 2. Analysis of wavy interfaces. J. Fluid Mech. 80:
integration of differential viscoelastic models. J. 757-767.
Non-Newt. Fluid Mech. 39: 269-290. Scanlan, D. J. 1990. Two-slot coater analysis: inner
Ruschak, K. 1. 1978. Flow of a falling film into a pool. layer separation issues in two-layer coating. MS
AIChE J. 24: 705-709. thesis, University of Minnesota, Minneapolis.
Ruschak, K. 1. 1980. A method for incorporating free Schunk, P. R. 1989. Surfactant and polymer effects in
boundaries with surface tension in finite element coating and related flows. PhD thesis, University
fluid flow simulators. Int. J. Numer. Meth. Eng. 15: of Minnesota, Minneapolis.
639-648. Schunk, P. R. and Scriven, L. E. 1990. Constitutive
Ruschak, K. J. 1983. A three-dimensional linear stability equation for modeling mixed extension and shear
analysis for two-dimensional free boundary flows in polymer solution processing. J. Rheol. 34:
by the finite element method. Comput. Fluids. 11: 1085-1119.
391-401. Schwetlick, H. 1984. Algorithms for finite-dimensional
Ruschak, K. J. 1985. Coating flows. Ann. Rev. Fluid turning point problems from viewpoint to relation-
Mech. 17: 65-89. ships with constrained optimization methods. In
Numerical Methods for Bifurcation Problems, IntI. real matrix. University ofMaryland Computer Science
Series Numer. Math., Vol. 70, eds T. Kuepper, Center (College Park) Technical Report TR-514,
H. D. Mittelmann and H. Weber, pp.459-479. ONR-NOOOI4-76-C-0391.
Basel: Birkhaeuser. Strang, G. and Fix, G. 1973. An Analysis of the Finite
Sethian, 1. 1984. Turbulent combustion in open and Element Method. Englewood Cliffs, NJ: Prentice
closed vessels. J. Comput. Phys. 54: 425-456. Hall.
Sethian, J. A. and Strain, 1. 1992. Crystal growth and Strigberger, J., Baruzzi, G., Habashi, W. and Fortin,
dendritic solidification. J. Comput. Phys. 98: 231-253. M. 1993. Some special purpose preconditioners for
Shanks, S. P. and Thompson, 1. F. 1977. Numerical conjugate gradient-like methods applied to CFD.
solution of the Navier-Stokes equations for 2D Int. J. Numer. Meth. Fluids. 16: 581-596.
hydrofoils in or below a free surface. In Proc. 2nd Subbiah, S., Trafford, D. L. and Gii~eri, S. I. 1989.
Inti. Conf Numerical Ship Hydrodynamics, Non-isothermal flow of polymers into two-dimen-
pp. 202-220. Berkeley: Vniv. of Calif. Extension sional, thin cavity molds: a numerical grid-genera-
Publ. tion approach. Int. J. Heat Mass Transfer. 32:
Shokoohi, F. and Elrod, H. G. 1987. Numerical 415-434.
investigation of the disintegration of liquid jets. J. Tabata, M. and Morishita, A. 1989. Fractional step
Comput. Phys. 71: 324-341. finite element scheme for free boundary problems.
Shopov, P. 1. and Bazhlekov, I. B. 1991. Numerical Theor. Appl. Mech. 39: 251-257.
method for viscous hydrodynamic problems with Tanner, R. I. and Milthorpe, 1. F. 1983. Numerical
dynamic contact lines. Comput. Meth. Appl. Mech. simulation of the flow of fluid with yield stresses.
Eng. 91: 1157-1174. In Numerical Methods in Laminar and Turbulent
Shopov, P. 1., Minev, P. D. and Bazhlekov, I. B. 1992. Flow (Proc. 3rd Inti. Conf), eds C. Taylor, 1. A.
Numerical method for unsteady viscous hy- Johnson and W. R. Smith, pp.680-690. Swansea:
drodynamical problems with free boundaries. Int. Pineridge Press.
J. Numer. Meth. Fluids. 14: 681-705. Tanzosh, 1., Manga, M. and Stone, H. A. 1992.
Sicilian, J. M. and Hirt, C. W. 1984. HYDR-3D: a Boundary integral methods for viscous free-boundary
solution algorithm for transient 3D flows. Flow problems: deformation of single and multiple
Science (Flow Science Inc., Los Alamos, NM) fluid-fluid interfaces. In Proc. 7th IntI. Conf Boundary
Report FSI-81-00-1. Element Technology, Vol. VII, eds C. A. Brebbia
Silliman, W. 1. and Scriven, L. E. 1980. Separating flow and M. S. Ingber, pp. 19-39. Southampton, England:
near a static contact line: slip at a wall and shape Computational Mechanics Publications.
of a free surface. J. Comput. Phys. 34: 287-313. Tezduyar, T. E., Behr, M. and Liou, J. 1992. A new
Smith, R. E. 1982. Algebraic grid generation. Appl. strategy for finite-element computations involving
Math. Comput. 10-11: 137-170. moving boundaries and interfaces - the deforming-
Sonneveld, P. 1989. CGS: a fast, Lanczos-type solver spatial-domainjspace-time procedure: I. The concept
for nonsymmetric linear systems. SIAM J. Sci. Stat. and the preliminary numerical tests. Comput. Meth.
Comput. 10: 36-52. Appl. Mech. Eng. 94: 339-351.
Sonneveld, P., Wesseling, P. and de Zeeuw, P. M. Tezduyar, T. E., Behr, M., Mittal, S. and Liou, J. 1992.
1985. Multigrid and conjugate gradient methods as A new strategy for finite-element computations
convergence acceleration techniques. In Multigrid involving moving boundaries and interfaces - the
Methods for Integral and Differential Equations, eds deforming-spatial-domainjspace-timeprocedure: II.
D.1. Paddon and H. Holstein, pp. 117-167. Oxford: Computation of free-surface flows, two-liquid flows
Clarendon Press. and flows with drifting cylinders. Comput. Meth.
Soulaimani, A., Fortin, M., Dhatt, G. and Quellet, Y. Appl. Mech. Eng. 94: 353-371.
1991. Finite element simulation of two- and three- Thomas, P. D. 1982. Composite three-dimensional
dimensional free-surface flows. Comput. Meth. Appl. grids generated by elliptic systems. AIAA J. 20:
Mech. Eng. 86: 265-296. 1195-1202.
Souvaliotis, A. and Beris, A. N. 1992. Application of Thomas, P. D. and Brown, R. A. 1987. LV decompo-
domain decomposition spectral collocation methods sition of matrices with augmented dense constraints.
in viscoelastic flows through model porous media. Int. J. Numer. Meth. Eng. 24: 1451-1459.
J. Rheol. 36: 1417-1453. Thomas, P. D. and Middlecoff, 1. F. 1980. Direct
Stewart, G. W. 1978. SRRIT - a FORTRAN subroutine control of the grid point distribution in meshes
to calculate the dominant invariant subspaces of a generated by elliptic equations. AIA A J. 18: 652-656.
References 365
Thompson, E. G. 1986. Use of pseudo-concentrations problems arIsmg in prediction of morphological

to follow creeping viscous flow during transient structure. J. Sci. Comput. 3: 77-108.
analysis. Int. J. Numer. Meth. Fluids. 6: 749-761. Unverdi, S. O. and Tryggvason, G. 1992. A front-
Thompson, J. F., Thomas, F. C. and Mastin, C. W. tracking method for viscous, incompressible, multi-
1974. Automatic numerical generation of body- fluid flows. J. Comput. Phys. 100: 25-37.
fitted curvilinear coordinate system for field con- Van Abbenyen, W., Christodoulou, K. N. and Scriven,
taining any number of arbitrary two-dimensional L. E. 1989. Frequency response of a coating flow:
bodies. J. Comput. Phys. 15: 299-319. predictions for slide coating. University ofMinnesota
Thompson, J. F., Warsi, U. A. and Mastin, C. W. 1982. Supercomputer Center Report UMSI 89/40.
Boundary-fitted coordinate systems for numerical Van Abbenyen, W., Goetmaeckers, B., Van de Vijver,
solution of partial differential equations - a review. H., Bussmann, H. and Beck, D. 1990. Experimental
J. Comput. Phys. 47: 1-108. verification of the calculated frequency response of
Thompson, J. F., Warsi, Z. U. A. and Mastin, W. C. a multilayer slide-coating operation. Paper read at
1985. Numerical Grid Generation. New York: AIChE Spring National Meeting, Orlando, FL,
Elsevier. March 19-22.
Tilton, J. N. 1988. The steady motion of an interface Van der Vorst, H. A. 1992. Bi-CGSTAB: a fast and
between two viscous liquids in a capillary tube. smoothly converging variant of Bi-CG for the
Chern. Eng. Sci. 43: 1371-1384. solution of nonsymmetric linear systems. SIAM J.
Torrey, M. D., Cloutman, L. D., Mjolsness, R. C. and Sci. Stat. Comput. 13: 631-644.
Hirt, C. W. 1985. NASA-VOF2D: A computer Visbal, M. and Knight, D. 1982. Generation of
program for incompressible flows with free surfaces. orthogonal and nearly orthogonal coordinates with
Los Alamos National Laboratory Report. LA- grid control near boundaries. AIAA J. 20: 305-306.
10612-MS. Voinov, O. V. 1976. Hydrodynamics of wetting. Fluid
Tran-Cong, T. and Phan-Tien, N. 1988a. Three- Dynamics. 11: 714-721.
dimensional study ofextrusion processes by boundary Wambersie, D. and Crochet, M. J. 1992. Transient
element method. I. An implementation of high- finite element method for calculating steady state
order elements and some Newtonian results. Rheol. three-dimensional free surface flows. Int. J. Numer.
Acta. 27: 21-30. Meth. Fluids. 14: 343-360.
Tran-Cong, T. and Phan-Thien, N. 1988b. Three- Wang, H. P. and Lee, H. S. 1989. Numerical techniques
dimensional study ofextrusion processes by boundary for free and moving boundary problems. In Funda-
element method. II. Extrusion of a viscoelastic fluid. mentals ofComputer Modelingfor Polymer Processing,
Rheol. Acta. 27: 639-648. ed. C. L. Tucker, pp. 369-401. Munich, Vienna and
Tryggvason, G. 1988. Numerical simulation of the Ray- New York: Hanser Publishers.
leigh-Taylor instability. J. Comput. Phys. 75: 253-282. Wang, H. P. and McLay, R. T. 1986. Automatic
Tsiveriotis, K. and Brown, R. A. 1992. Boundary- remeshing scheme for modeling hot forming process.
conforming mapping applied to computations of J. Fluid Eng. (ASME) 108: 465-469.
highly deformed solidification interfaces. Int. J. Weatherburn, C. E. 1927. Differential Geometry of
Numer. Meth. Fluids. 14: 981-1003. Three Dimensions. London: Cambridge Univ. Press.
Tsiveriotis, K. and Brown, R. A. 1993. Solution of Werner, B. and Spence, A. 1984. The computation of
free-boundary problems using finite-element/Newton symmetry-breaking bifurcation points. SIAM J.
methods and locally refined grids: application to Numer. Anal. 2: 388-399.
analysis of solidification microstructure. Int. J. Wilkinson, 1. H. 1965. The Algebraic Eigenvalue
Numer. Meth. Fluids. 16: 827-843. Problem. Oxford: Clarendon Press.
Ungar, L. H. and Brown, R. A. 1982. The dependence Winslow, A. M. 1967. Numerical solution of the
of the shape and stability of rotating captive drops quasilinear Poisson equation in a nonuniform triangle
on multiple parameters. Phi/os. Trans. Roy. Soc. mesh. J. Comput. Phys. 2: 149-172.
Lond. A306: 457 -480. Winters, K. H., Cliffe, K. A. and Jackson, C. P. 1987.
Ungar, L. H. and Brown, R. A. 1985. Cellular interface The prediction of instabilities using bifurcation
morphologies in directional solidification. IV. theory. In Numerical Methods for Transient and
The formation of deep cells. Phys. Rev. B. 31: Coupled Problems, eds R. W. Lewis, E. Hinton, P.
5931-5940. Bettess and B. A. Schrefler, pp.179-198. New York:
Ungar, L. H., Ramprasad, N. and Brown, R. A. 1988. Wiley.
Finite element methods for unsteady solidification Yiantsios, S. G. and Higgins, B. G. 1989.
Rayleigh-Taylor instability in thin viscous films. M. J. Banes, pp.273-285. New York: Academic
Phys. Fluids A - Fluid Dynamics. 1: 1484-1501. Press.
Yokoi, T. and Scriven, L. E. 1989. Profile extrusion Yuan, Y. and Ingham, D. B. 1991. Numerical solution
analyzed and die shape designed by Galerkin's of viscous flows with a free surface. In Computational
method with streamline-adapted basis functions. Modeling of Free and Moving Boundary Problems
University of Minnesota Supercomputer Institute (Proc. 1st IntI. Conf), eds L. C. Wrobel and C. A.
Report UMSI 89/65. Brebbia, pp.325-339. Southampton, UK: Com-
Youngren, G. K. and Acrivos, A. 1976. On the shape putational Mechanics Publ.
of a gas bubble in a viscous extensional flow. J. Zhou, M.-Y. and Sheng, P. 1990. Dynamics of
Fluid Mech. 76: 433-442. immiscible-fluid displacement in a capillary tube.
Youngs, D. L. 1982. Time-dependent multi-material Phys. Rev. Lett. 64: 882-885.
flow with large fluid distortion. In Numerical Methods Zienkiewicz, O. C. 1977. The Finite Element Method.
for Fluid Mechanics, eds K. W. Morton and London: McGraw-Hill.
PART THREE
THEORY AND PRACTICE OF COATING

PROCESSES
ANALYSIS AND DESIGN OF INTERNAL 10
COATING DIE CAVITIES
Robert B. Secor
10.1 INTRODUCTION
from the feed location. Two coathanger manifolds,
The purpose of the internal cavities of coating linearly tapered and curviIinearly tapered are
dies of all types is to distribute the coating liquid often distinguished. The linearly tapered coathan-
in a manner which in conjunction with the rest ger manifold is sometimes referred to as a fishtail
of the coating process produces a liquid film with manifold. The cross-sectional shape of the manifold
uniform dimensions and properties. Whereas the cavities also come in various shapes, (see Fig.
constancy of feed rate and take-away speed play to.2). Circular shapes are very popular for analysis
the biggest role in the down-web uniformity of but not so common in practice. More common
the final film, the liquid passages internal and are the teardrop, triangular, and rectangular
external to the coating die have the biggest shapes.
impact on the cross-web uniformity of the coated Often coating dies are equipped with adjustment
liquid. Although cross-web nonuniformities in mechanisms for tuning the cross-web flow
the liquid passages external to the coating die uniformity produced by the die manifold. Figure
can drastically affect the final coating uniformity, 10.3 shows the two most common mechanisms:
the focus of this chapter is the influence on the the flexible die lip and the choker bar (sometimes
cross-web uniformity of the passages internal to referred to as a straining bar or restrictor bar).
the coating die. The flexible die lip features a weakened die lip
The distribution chambers of most coating with a mechanism for locally moving the lip and
dies have a common general shape. They consist thus changing the slot gap near the die exit.
of a cavity (also referred to as a distribution Traditionally, the die lips are flexed mechanically
chamber or channel) that is oriented substantially with a number of bolts evenly spaced across the
in the cross-web direction which is connected to width of the coating die (cf Kasamatsu (1987);
a slot (or slit or land) on one side, (see Fig. 10.1). Iguchi et al. (1987). Alternatively, the movement
The combination of the cavity and slot is referred of the die lip can be accomplished by thermal
to as the distribution manifold. Two particular expansion (Lowey 1960; Nissel 1976; Anthony
types are often distinguished: the T-manifold and et al. 1988), piezoelectric or magnetostrictive
the coathanger manifold. The T-milllifold, (also translation (Reifenhauser et al. 1986; Erckmann
referred to as an 'infinite cavity' manifold), has 1988). The choker bar is a separate member in
a uniform cross-section across the width of the the die cavity that can be flexed to change the
die. The coathanger manifold features decreasing slot gap locally usually a distance upstream of
slot length and cavity size with cross-web distance the die exit. When either flexible lip or choker
London. ISBN 0 412 06481 2.
370 Analysis and design of internal coating die cavities
~---~) '-'----~
f
(a)
CENTER-FED T-MANIFOLD
IT
ll=====~T END-FED T-MANIFOLD
CIRCULAR TEARDROP
(b)
(c) LINEARLY-TAPERED COATHANGER MANIFOLD

TRIANGULAR RECTANGULAR
Figure 10.2 Common shapes of cavity cross-sections.

The cavity shape near the feed location and near the
distal end of the cavity is shown for tapered cavities.
Circular and teardrop shapes feature a changing
CURVILINEARLY-TAPERED COATHANGER MANIFOLD
curvature while the triangular shape has a constant
curvature along the length of the cavity.
(d)
Figure 10.1 Examples ofcommon coating die manifold flow can be evenly distributed within the range
shapes including: (a) a center-fed T-manifold with a of adjustment of the flexible lip or the choker
single-stage slot; (b) an end-fed T-manifold with a bar. Nevertheless, it is usually wise to design the
multi-stage slot; (c) a center-fed linearly tapered
coathanger manifold, and (d) an end-fed curvilinearly
die manifold for as uniform a distribution as
tapered manifold. The cross-sectional views on the possible as highly contorted flexible die lips or
right-hand side show the variation in manifold choker bars can produce cross-web flow and
cross-section from feed location to distal end. material property differences.
Public literature on the flow analysis of die
bar adjustment is used on a coating die, the role manifolds falls into several broad categories. The
of the distribution manifold is different from that majority use a combination of a one-dimensional
in a die without means of uniformity adjustment. flow model of viscous effects in the cavity and
Ifthere is no means ofadjustment, the distribution the slot to determine flow uniformity. In chro-
provided by the die manifold is what will be nological order, these are:
ultimately observed, and thus, must be quite
uniform. If flexible lips or choker bars are Carley (1954) Newtonian and power-law liquids
present, however, the distribution provided by in T-manifolds;
the die manifold does not need to be so uniform. Pearson (1964) effects of temperature variations
It only needs to be sufficiently uniform that the and die distortion;
Introduction 371
Liu and Hong (1988) flow uniformity and

residence time distribution in coathanger
manifolds with non-circular cavities.
In addition, a number of papers (Miller 1972;
Huang and Yu 1973; Liu 1983; Liu and Hong
1988) specificially address the problem of using
one-dimensional cavity flow models for the
noncircular shapes often encountered. Of these,
Liu and Hong (1988) incorporated a 1i-dimen-
sional model (one velocity component varying
LEXIBLE DIE LIP
in two dimensions) for the flow through the
cavity. Papers that address two- or three-
dimensional flow models in the slot include:
Tadmor, Broyer and Gutfinger(1974) developed
flow analysis network (FAN) method for
two-dimensional slot flow of Newtonian and
non-Newtonian liquids;
Vergnes and Agassant (1986) two-dimensional
flow and thermal analysis of coathanger
manifolds;
Sartor (1990) two-dimensional slot flow analysis
mostly for Newtonian liquids;
CHOKER BAR Vlcek et al. (1990) two-dimensional flow and
(b)
thermal analysis with the control volume
method;
Figure 10.3 Adjustment mechanisms for coating dies: Vrahopoulou (1991) two-dimensional slot flow
(a) flexible die lips and (b) choker bar. ofNewtonian liquids with the boundary integral
method;
McKelvey and Ito (1971) multi-stage slots and Wang (1991) three-dimensional finite element
tapered cavities in T-manifolds; analysis of a coathanger manifold.
Procter (1972) coathanger manifolds and optimal
cavity tapering for fishtail dies; Specific discussions about the thermal analysis of
Klein and Klein (1973) effect of flow rate on manifold analysis can be found in Winter (1975)
uniformity for non-power-Iaw liquids in fishtail and Vergnes et al. (1984).
manifolds; Fairly comprehensive discussions of manifold
Chung and Lohkamp (1976) related size of the analysis can be found in the following:
cavity with the angle of the tapered slot length;
Leonard (1985b) general analysis of the effects
Matsubara (1981) shapes ofcurvilinearly tapered
of viscous, inertial and gravitational effects in
coathanger manifolds in terms of the power-law
T-die manifolds;
index of the liquid and residence time distribu-
Sartor (1990) analysis of viscous and inertial
tion;
effects in coathanger dies;
Booy (1982) electrical resistance network analogy
Michaeli (1992) general analysis of effects
for manifold flow analysis;
encountered in polymer processing.
Winter and Fritz (1986) flow uniformity and
residence time distribution in circular and The intent of this chapter is first to review the
rectangular cavities; analysis in the public literature of the different
flow regimes and then describe how it can be distribution in the cavity and the final flow
best used by coating engineers to design and use uniformity as a result. Although no specific
coating die manifolds. analysis of this appears in the published literature,
analyses of an analogous situation, a liquid jet
impinging on a flat plate, can be found in many
10.2 ANALYSIS OF INTERNAL DIE FLOWS
fluids mechanics textbooks (e.g., Whitaker (1981)
The analysis of the flow in the internal passages pp.254-258; Batchelor (1967) pp.392-396). In
ofcoating die manifolds will be divided according this situation the ratio of the force exerting on
to the four flow sections: (1) feeding of the cavity, the plate to the cross-sectional area of the
(2) flow through the cavity, (3) flow through the impinging jet is related to the kinetic energy
cavity-slot transition, and (4) flow through the arriving in the jet,
slot region. The flow in the immediate vicinity
of the slot exit is the subject of other chapters.
(10.1 )
Figure 10.4 schematically shows these regions.
In the discussion, attention will be paid to the
relationship between the geometry of the die Here p is the liquid density and V is the average
cavities, the rheological characteristics of the velocity of the impinging jet. If this force to area
liquid, and the pressure and flow distribution ratio is analogous to the pressure disruption
that results. caused by inertial forces in the feed section of
coating die manifolds then flow disruption can
be expected when the kinetic energy of the
10.2.1 CAVITY FEED
feedstream (the right-hand side of (10.1» is
Coating die cavities can be fed in a number of significant in comparison with the pressure in
different ways. Center-fed cavities, end-fed cavities, the manifold cavity.
and cavities with multiple feed locations are The method of feeding high viscosity and
possible. Ordinarily, the method of feeding the viscoelastic coating liquids into cavities can also
manifold cavity has little effect on the final flow disrupt the ultimate uniformity. Stafford (1971)
uniformity but there are exceptions. The exceptions discloses that extrusion of a three-layer film from
occur when the flow in the feed section is such a center-fed manifold will yield an uneven
that the process of feeding the cavity significantly distribution of the center layer whereas no
disrupts the pressure distribution in the cavity. disruption occurs in an end-fed manifold. Ac-
For low viscosity liquids, inertial forces in the cording to Cloeren (1993), 'flow divergence' in a
flow of the feed section can disrupt the pressure center-fed cavity leads to a disruption in the
Figure 10.4 Schematic diagram of the general flow pattern in a coating die manifold.
Analysis of internal die flows 373
uniformity of both multilayer and single layer CENTER-FED ANI FOLD

films. The 'flow divergence' results from the '... I.
division of the (feed)stream into substreams that
flow in generally opposite directions to both ends
of the manifold.' Apparently, the response of '" 1.3
melts and other viscoelastic liquids to the stretching
(or extensional) flow that occurs as the feedstream
~
>
1.2
<
splits in a center-fed cavity disrupts the pressure X 1.1
distribution in the manifold cavity and ultimately ~

the flow uniformity (Secor 1995). End-fed cavities '"
b
~ 0.9
have no flow split (or 'flow divergence'), and
hence, have no resulting uniformity disruption. 0.8
Figure 10.5 shows an example of the flow
0.7
nonuniformity caused by the flow split in a
0 4 6 10
center-fed cavity. The upper graph shows the CRO WEB PO ITIO (Xf\\!' 10)
measured flow distribution for a center-fed T-
END-FED MA 'IFOLD
manifold, while the lower graph shows the
1.5
distribution when the same T-manifold is fed
from the end. As will be discussed later, a 1.4
relatively high flow is expected opposite the feed
",1.3
location in aT-manifold from viscous forces
alone, but the nonuniformity would be expected "i i 1.2
>
to be higher in the end-fed manifold. Note that <
xl.I
in this case the flow split causes a dramatically :::>
higher nonuniformity in the center-fed cavity. ~

f-
1
~ 0.9
10.2.2 CAVITY FLOW 0.8
After the feed section, the coating liquid flows 0.7

mainly down the axis of the cavity with a small 0 2 4 6 10
CROSSWEB PO ITION (Xf\\I • IOJ
portion leaking into the slot. Ordinarily, the
cross-section of the manifold cavity is a simply
connected area, but there are exceptions. Michaeli Figure 10.5 Example of the flow nonuniformity
(1992, pp. 205-206) shows a manifold cavity with caused by the flow split in a center-fed cavity. The
a rotating screw inside, while Hiraki et al. (1989) symbols represent experimental data points whereas
describe a rotating rotor in the manifold cavity. the solid lines represent model predictions which
Both are exceptional cases ofattempts to increase ignore any influence of the method offeeding the cavity.
the movement of the liquid in the cavity. In the
cavity flow analysis here, only simply connected
cavity cross-sections will be considered. Figure curved, a (X', ¥', Z') Cartesian coordinate system
10.2 shows a number of the common shapes. is employed where the X' -axis locally points
Because the flow in the manifold cavity is down the cavity axis. The X' -axis is parallel to
mainly down its axis, the interplay of the different the slot exit in T-manifolds, angled to the slot
forces is captured by a balance of momentum in exit in linearly tapered coathanger manifolds,
the direction of the cavity axis (Huang and Yu and curved in curvilinearly tapered coathanger
1973). As the axis of manifold cavity may be manifolds. The use of this coordinate system is
angled slightly to the slot exit and also possibly equivalent to 'unrolling' the curvature, if any, of
the manifold cavity. The momentum conservation

equation for steady flow of liquid in the noncon-
servative form is
(t == ~2 L 2
U dA, Q == L U dA (10.5)
Here the angle 0 is the inclination of the cavity

2
OU o(UV) O(UW)] wall to the axis of the cavity. Because the liquid
[
PoX' + oY' +~ adheres to the walls of the manifold cavity, the
second terms in (10.3) and (10.4) contain nonzero
_ OTx'X' OTx'Y' OTx'z' oP contributions only where the liquid exits the
- oX' + oY' + oZ' - ax' + pGX' (10.2)
cavity and enters the slot. Because the liquid
where (U, V, W) are Cartesian componenets of leaves the cavity substantially perpendicular to
the liquid velocity vector V. T is the stress tensor, the main axis of the cavity, these terms are small
P the hydrodynamic pressure, p the liquid density in magnitude and can be entirely neglected.
and GX' the X'-component of the gravitational The third term on the right side of (10.3)
force vector. The variation of the hydrodynamic represents the role of normal stresses in the cavity
pressure down the axis of the manifold cavity is momentum balance. The normal stresses can be
of primary interest here because it is mostly this viscous or elastic stresses due to extensional flows
that influences the subsequent flow in the slot. or elastic stresses resulting from the shear flow
As the flow in the cavity is mainly down the in the die cavity. The penultimate term in (10.3)
cavity axis, the pressure is nearly uniform across accounts for the shear stresses acting on the
any cavity cross-section. Thus, integration over perimeter of the cavity cross-section. Because the
the cross-sectional area, A, of the cavity yields liquid ordinarily adheres to the cavity there, these
the following for the cavity pressure: stresses are viscous in nature, even for elastic
dP
dX' = -Ap f A
oU
2
pi
oX' dA - A
Sslot
VNU dS
liquids. They constitute the pressure drop in the
cavity when inertial, gravitational and elastic
forces are absent,
+A
1 f A
OTx'X'
ax' dA
(10.6)
+~ Is TNx,dS + pG x' (10.3)

With these modifications, the cavity momentum
balance in its final form is as follows:
Here the subscript N refers to the direction dP dP vis pd ( Q2)

dX' = dX' - AdX' (t A
f
normal to the perimeter of the manifold cavity
but in the Y'Z' -plane.
1 OTx'X'
The first term in (10.3) represents the main +A A oX' dA + pG x' (10.7)
contribution of inertial forces to the cavity
momentum balance, For convenience, it can be It stipulates that the change in the hydrodynamic
rewritten in terms of an inertial shape factor (t pressure down the cavity results from a combi-
(Huang and Yu 1973; Leonard 1985b), "
f
nation of viscous losses, inertial accelerations
J:r
normal stresses and gravitational forces. Becaus~
p
fA
OU2 d
ax' dA = p dX A
2
u dA - p most of the reports in the literature are targeted
~ p~ f
Sslot toward relatively high viscosity melt processing,
often only viscous losses are included in the
U 2 tanOdS U 2 dA cavity momentum balance. Leonard (1985a,
dX' A
1985b), Lee and Liu (1989) and Sartor (1990) are
= P d~' ((t ~2) (10.4)
exceptions tht include non-viscous forces such
as inertia and gravitational forces following the
analysis by Huang and Yu (1973). Sartor (1990) with shear rate. For dispersions and other more
included normal stresses from the extension of complicated materials, the Bingham-Carreau
Newtonian liquids in his cavity momentum - Yasuda representation is sometimes required
balance, even though they are extremely small (Nelson, Kistler and Olmsted 1990),
as shown in (10.37) below. There have been no
accounts in the literature of elastic effects in the
'7(1) = '7 00 + ( '70 - '7 +"y -;
1 -FY)
00
manifold cavity momentum balance.
Practical application of the cavity momentum ~-1
balance requires evaluation of the integrals in [1 + (AWJ-a (10.10)

(10.5) and (10.6) in order to determine the inertial
Figure 10.6 compares the viscosity behavior of
shape factor and the viscous pressure drop. For
two liquids to that ofa power-law liquid. Whereas
simplicity's sake, most analyses assume the
the power-law liquid has a constant slope, both
manifold cavity cross-section is circular even
material A and material B show significant
though most common shapes are very noncircular.
changes in their slope. Material A is representative
Miller (1972) proposed determining the viscous
of many shear-thinning materials that are best
pressure drop in terms of the average wall shear
represented by the Carreau- Yasuda model.
stress, f which was postulated to depend only
Material B is representative of more complicated
on an average shear rate, y,
i
materials that are best represented by the
Bingham-Carreau-Yasuda model. In either the
dP vis _ ~ dS - r(y) - QA (10.8) case of the five parameter Carreau- Yasuda
dX' - A
s
"NX' - R h ' Y = 8R h A
model or the seven parameter Bingham-Car-
reau- Yasuda model, the number of parameters
R h , the hydraulic radius (cf Whitaker (1981)
required to specify the viscosity behavior accurately
p. 159), is the ratio of the area to the perimeter
precludes the efficient use of tabulated values for
of the cavity cross-section and A is a shape factor
manifold flow analysis.
which is different for different cavity shapes. The
The second principal problem is the accuracy
functional dependence of the average wall shear
achieved if values for the shape factor can be
stress on the average shear rate is presumed to
found. Liu and Hong (1988) assessed the accuracy
be independent of cavity shape. Liu (1983) finds
of Miller's procedure by using two numerical
that different shape factors are required depending
methods to discretize the cross-section of the
on the shear-thinning behavior of the liquid
manifold cavity and compute the velocity field
viscosity and tabulates values according to the
U(Y',Z') according to
power-law index of the liquid.
Coating engineers encounter two principal apvis a.. X'Y' a.. X'Z'
problems with this procedure. The first is that (10.11)
ax' = ay' + az'
although the shape factors are tabulated according
to power-law index, common liquids encountered They concluded that for very non-Newtonian
in industrial situations exhibit a variety ofviscosity liquids, Miller's shape factor procedure is very
behavior that is not represented by the power-law inaccurate. As a consequence, the most efficient
viscosity function (Chung and Lohkamp 1976). method for computing the viscous pressure drop
For solutions and polymeric liquids, the Car- in the cavity momentum balance with sufficient
reau-Yasuda viscosity model (cf Bird et al. (1987», accuracy is apparently the Ii-dimensional (one
velocity component varying in two dimensions)
modeling procedure of Liu and Hong. The flow
rate and inertial shape factor in (10.5) can then
exhibits sufficient flexibility to capture almost aU be computed by suitable integration of the
the observed variations in viscosity behavior velocity field.
10
MATERIAL A
,
POWER·LAW LIQUID
WITIl n = 213
~,:~.:,~,~
0.1 III 100 ItX10 10'

SHEAR RATE ( EC ' )
Figure 10.6 Comparison of the viscosity behavior of two materials to that of a power-law liquid. Material
Ashows a general shear-thinning behavior while material B shows yield stress behavior in addition to shear-thinning.
10.2.3 SLOT FLOW Tadmor et al. (1974) was capable of modeling

two-dimensional slot flow of liquids of arbitrary
The flow of the coating liquid in the slot portion
viscosity. Here, a more general treatment is
of coating dies is nearly rectilinear by virtue of
shown with its relationship to the governing
the large aspect ratio of most coating die slots.
equations of the FAN method.
Typically, the length of the slot is 10 to 100 times
Conservation of mass in the two-dimensional
larger than the height. The width of the die slot
lubrication approximation results from the
may in addition be 10 to 100 times the length of
integration of the continuity equation over the
the slot. As a result, the analysis of the flow in
thickness of the slot channel, see Fig. 10.7
the slot of the die is amenable to the lubrication
f
approximation which applies under the following HA(X.Yl
conditions (cf Cameron (1981)); V·YdZ = V,,'Qs = 0,

H.(X.Y)
(~r«l, Re(~r« 1 (10.12) Qs == fHA

H.
Y,,(X, Y)dZ (10.13)
The subscripts A and B refer to the upper and

Here Hand L are the characteristic thickness
and length of the slot flow channel and
Reynolds number.
Re
is the lower surfaces of the slot channel and the subscript
II refers to the plane of the flow channel,
Most analyses of the slot flow assume that the a a
flow is unidirectional from slot inlet to outlet. v" == I ax + J ay' YII == IV + JV (10.14)
Recent work has shown that this assumption can
lead to erroneous results (Sartor 1990; The (X, Y, Z) Cartesian coordinate system used
Yrahopoulou 1991). But even these researches here has the X-axis parallel to the slot exit as
were restricted to Newtonian or power-law shown in Fig. 10.4. The unit vectors I and J point
viscosity models. Apparently, it was not generally in the X-and Y-directions respectively.
realized that the FAN method developed by Conservation of momentum in the lubrication
Figure 10.7 Diagram of the lubrication approximation geometry.
approximation stipulates that the liquid pressure in the lubrication approximation,

is constant across the thickness of the flow
channel. The pressure gradient in the plane of
the flow channel varies in response to the shear
_(I oZ
t llZ -"
OV u/) oVu
oZ (10.18)
stresses in the plane of the flow,

In the lubrication flow of Newtonian liquids,
V P- ih uz or the viscosity is constant throughout the flow
u - oZ channel and the solution of the governing
equations (10.16), (10.17) and (10.18) is straight-
to=tIlZ(Z=O) (10.15) forward. In the case of shear-sensitive liquids,
Because the flow is parallel to the direction of the viscosity varies throughout the channel and,
the pressure gradient, it is convenient to resolve thus, it is convenient to transform the integration
this equation into a component parallel to the variable in the governing equations from the
direction of the pressure gradient, coordinate Z to stress. Rearrangement ofequation
(10.13) using integration by parts followed by
VuP transformation of the integration variable yields
IVuPJ Z = E· (t uz - to), E == IVuPI (10.16)
the following:
The boundary conditions for the surfaces of
the flow channel are sensibly no slip between the Qs = - rJR.RA ZoVudZ
oZ
liquid and its solid boundaries. For the present
purpose, it is convenient to express this in 1 fE'~ oV
integral form, = -IV PI 2 E·(tuz - to) O~I E·dtllz
II E I~Bz
0= JRA oVudZ (10.17)

(10.19)
R. oZ With the application of the no-slip boundary
Because the lubrication flow is nearly rectilinear, conditions (10.17), further transformation of the
only the dominant velocity gradients appear in integration variable from stress to velocity gradient,
the constitutive relation for shear-sensitive liquids and some manipulation, (10.19) is reduced to an
expression for the flow rate in terms of the 10.2.4 CAVITY-SLOT TRANSITION
velocity gradient at the bounding surfaces,
The cavity flow and the slot flow described in
the preceding two sections are joined by a
Q
s
=- (H A - H a)2 LAE
transitional flow between them. Liquid leaves
4
[1 - 112(IL~I)L~ f:
the cavity on one side and enters the manifold
A
2 2 slot. Conservation of mass dictates that whatever
11 (ILI)L dLJ (10.20) liquid leaves the cavity must enter the slot,
LA =E· av"
az (Z = H A) = -E,
av"
az (Z = H a) (10.25)
(10.21) Here NIl is the unit normal vector in the plane

of and pointing outward from the manifold slot.
LA' the velocity gradient at the upper surface, is The balance of momentum over the transitional
determined by applying (10.15) at the upper and flow serves to relate the pressure at the beginning
lower surfaces, of the slot to the pressure in the cavity. All
literature reports of manifold models assume that
IV"PI(H A - H a) = 2110LAI)LA (10.22) the cavity pressure is identical to the pressure at
The combination of (10.22), (10.20) and the the head of the slot. In reality these pressures
conservation of mass principle, (10.13) serve to will differ by an entrance pressure loss which
determine the pressure and flow rate distribution characterizes the viscous losses, inertial forces,
in the slot channel of a die manifold. This extensional viscosity effects and elastic effects in
combination is general in that any viscosity the cavity-slot transitional flow,
model can be incorporated. The only requirement (10.26)
is that the integral in (10.20) has to be evaluated
either analytically or numerically. A related flow, the contraction flow, has been
The FAN method starts with the solution to studied extensively and the roles of the different
the two-dimensional lubrication flow of a New- forces delineated, (Boger 1987; Binding 1988;
tonian liquid with no-slip boundary conditions White, Gotsis and Baird 1987; Binding 1991).
on each boundary, For Newtonian liquids in a circular contraction,
Boger (1987) reports the following for the role
__ (H A -Ha )3 P of inertial forces and viscous losses in the entrance
Q s- V (10.23)
12}l " pressure loss:
For shear-thinning liquids, the Newtonian vis-
A Pent = 0.0725 Re + 0.69 (10.27)
cosity ~ is replaced by an equivalent viscosity 2 'w
defined in terms of the wall shear stress, 'w,
f
The Reynolds number and wall shear stress are
those based on the smallest tube in the contraction
= (10.24)
}lequivalent flow. A similar expression is also reported for
3 : .yd. shear-thinning liquids. Although the entrance
pressure loss in the planar contraction flow that
The combination of(10.23) and (1 0.24) is equivalent occurs in die manifolds would be expected to be
to (10.19) in foregoing development. In either somewhat different, a few points from the circular
case, efficient and accurate two-dimensional contraction flow entrance pressure loss are
lubrication flow analyses ofliquids with arbitrary noteworthy. Because the viscous contribution to
viscosity functions in the slots of die manifolds the dimensionless entrance pressure loss in (10.27)
are possible. is on the order of unity, the viscous loss is not
expected to be significant as long as the slot has analysis of flow in coating dies. They provide the
a relatively large aspect ratio, (e.g., L/H > 10). means of determining expected cross-web flow
Likewise, the inertial contribution to the entrance distribution in a coating die manifold given the
pressure loss is only significant when the Reynolds manifold geometry, the material properties ofthe
number (based on the slot flow) is large. coating liquid and the throughput rate of liquid
Contraction flow analyses also yield information through the manifold.
about the viscoelastic effects in the cavity-slot For example, Fig. to.8 shows the resulting
transitional flow. In abrupt contractions, elastic cross-web uniformity computed for aT-manifold
liquids often show vortices just upstream of the compared with experimental data. The flow
contraction. The vortices are more pronounced uniformity was computed with a combined model
in circular contractions than planar ones and of Ii-dimensional cavity flow, two-dimensional
can be eliminated by a sufficient tapering of the lubrication flow in the slot, and no entrance
entry flow geometry (White, Gotsis and Baird pressure loss in the transitional flow. The gap in
1987). The size of the vortices and likewise the the slot of this manifold varied in the cross-web
size of the entrance pressure loss is largest for direction as is shown in Fig. to.8. As will be
liquids that exhibit large extensional viscosities discussed later, a high flow would ordinarily be
(Luo and Mitsoulis 1990). Binding (1988) provides expected opposite the location of the cavity feed.
an accounting of the extensional viscosity effects Although the T-manifold is fed at a location
on the entrance pressure loss based on consider- which corresponds to the left side of the figure,
ations of the energy dissipation in the contraction due to the nonuniform slot gap the slot efflux is
flow. The entrance pressure loss in a planar highest on the right. This example serves to
contraction for a liquid which shows power-law illustrate the sensitivity of manifold flow to small
viscosity behavior in both extension and shear is changes in the slot gap. According to (10.23) the
flow in the slot is proportional to the cube of
2(2+ ~) the gap for Newtonian liquids. For shear-thinning
( 2)(
2 liquids the sensitivity is higher. As a result, small
3)(1 - P )Re gap variations can lead to large flow nonunifor-
3 3+- 4+- mities.
n n
+ (1 + +t)2
2t 2 (1 n)
Yw
smax
(1 _P2t '!.±.1-)
t+ 1
(10.28)
10.2.6 DIMENSIONAL ANALYSIS
In order to assess the regimes better where the
Here nand t are the power-law indices of the
different forces (i.e. inertial, gravitational and
shear and extensional viscosity and P is the
elastic) play an integral role, dimensional analysis
inverse of the contraction ratio. The first term
is employed. Here we follow closely the procedure
represents the inertial contribution to the entrance
used by Leonard (I985b) for dimensional analysis
pressure loss while the second, which contains
of T-manifolds with some modifications.
the ratio of maximum shear rate to extension
The drop in pressure due to the flow through
rate in the flow, characterizes the viscous and
the slot of the manifold scales according to
extensional viscosity effects.
p* = 2L'1sYs () . 6Qo ( 2
10.2.5 MANIFOLD ANALYSIS H ' '1s == '1 Ys, Ys == WH 2 to. 9)
The combination of mass and momentum where L is the length of the slot, H the gap in
conservation equations for the cavity flow, the the slot channel, Wthe width of the die, and Qo
slot flow and the transitional flow described in the volumetric flow rate of liquid entering the
the last three sections comprise a model for die. The ratio of the viscous pressure drop in the
0.3
0.25
UJ 2.5
,
0 0.2
;;2 z V>
[""
UJ
> 0
/ Q
<:
~ / 0.15
Cl
x ~
~ /
'3
~
uJ
I 0
UJ -6
0
Z-
~ /'
9 0.1
,
/'
/'
'" 0
0 0
0
--G- o 0.05
0.5
o
o
o 50 100 150 200 250 300 350
CROSSWEB PO rno (mm)
Figure 10.8 Comparison of the flow uniformity predicted for a T-manifold with experimental data. The
coating liquid has a viscosity described by (10.10) with '10 = 1.365 Pa s, n = 0.0532, A = 0.0556 s, a = 2,
'1 00 = 0.0057 Pa s, 't"y = 5.96 Pa and F = 511.
manifold cavity to the pressure drop in the die viscous forces in the cavity and is closely associated
slot comprises the viscous number, Nv, to the Reynolds number in the manifold cavity,
Nv = W1]cYc ') , 2Qo Nm Re

"Ie == '1 (Ye' Ye == AR (10.30) (10.32)
RhP*' h
Nv
Here R h is the hydraulic radius of the cavity The influence of gravitational forces on the
cross-section and serves as the characteristic manifold flow is contained in the gravity number
length scale of flow in the manifold cavity. A is which is the ratio of the gravitational force in
the cross-sectional area of the manifold cavity. the cavity to viscous forces in the slot,
Note that as long as the viscosity behavior in
the slot and in the cavity are similar, Nv and (10.33)
hence the relative flow resistance in the cavity
and the slot is independent of the throughput The ratio of the gravity to viscous numbers
rate through the die. This is not the case if inertial reflects the role of gravitational forces in the
forces are significant as measured by the mo- manifold cavity and is closely related to the
mentum number, Nm. It is the ratio ofthe inertial cavity Stokes number,
pressure gradient in the cavity to the slot pressure
drop, Ng St
(10.34)
Nv 32
Nm = PQ~ (10.31) The viscoelasticity of the coating liquid would
A 2 p*
sensibly affect the manifold flow through three
The ratio of the momentum to viscous numbers sources: (1) the 'memory' of viscoelastic liquids;
reflects the relative importance of inertial and (2) elastic normal stresses from shear flow, and
(3) normal stresses from extensional flows. flow. The ratio of extensional to viscous effects
Ordinarily the viscoelastic memory of coating in the cavity is represented by the ratio of
liquids would be expected to influence the manifold extension to viscous numbers which is propor-
flow when the Deborah number, which is the tional to the ratio ofextensional to shear viscosity
ratio of the characteristic memory time of the in the cavity. Here again, the appearance of the
liquid to the characteristic time scale of the flow, cavity aspect ratio in the scaling ofthe extensional
becomes large. But the majority of the distribution effects diminishes their importance. In the
characteristics of a coating die manifold are cavity-slot transition, however, the extensional
determined by flow conditions at or very near properties of the coating liquid are likely to have
the walls of the manifold. At the manifold walls, a bigger influence. Because extensional effects
the coating liquid is stationary or moving very can lead to large entrance pressure losses in the
slowly, and hence, experiences the flow near the flow into the slot according to (10.26), they can
wall for a very long time. As a result, the have a large impact on the hydrodynamic pressure
viscoelastic memory plays an insignificant role in the manifold cavity. However, the extensional
in the manifold flow distribution. effects will have a smaller impact on the flow
Elastic normal stresses from shear flows may distribution of the manifold. Because the cavity
influence the manifold flow in both the cavity momentum balance (10.7) contains only pressure
and the slot. Based on the cavity momentum gradients, only the cross-web gradient of the
balance (10.7), normal stresses in the cavity are entrance pressure loss will affect the flow
expected to influence flow distribution in accord- distribution of the die manifold. That is, if a
ance with the normal stress number, manifold yields a uniform distribution without
'P ·2 entrance pressure losses, it will also yield a
Nns = ~ (10.35) uniform distribution with entrance pressure losses
p*
included. Furthermore, if a manifold has a
where 'P 'e is the first normal stress coefficient nonuniform distribution without entrance pressure
evaluated at the characteristic cavity shear rate. losses, the distribution will be more uniform
The ratio of the normal stress number to the when entrance pressure losses are included. This
viscous number represents the relative importance applies equally to entrance pressure losses from
of normal stress and viscous effects in the cavity inertial forces and extensional viscosity effects.
flow and is proportional to the stress ratio, In addition to the viscous, inertial, gravitational
and viscoelastic forces described above, the flow
Nns SR 'PI Y
Nv (W/R )' SR == -t1- (10.36) distribution of die manifolds can be affected by
h elastic deformation of the coating die itself. This
As it is rare for coating liquids to have stress is characterized by analogy with the deflection
ratios greater than ten (Bird, Armstrong and of a cantilevered beam that has a load that
Hassager 1987) and because the aspect ratio of linearly decreases from the support to the free
industrial manifold cavities are often quite large, end (cf Blodgett (1966». The ratio of the maximum
normal stress effects are not expected to be die deflection to the slot gap comprises the
significant in manifold flows. deflection number,
Normal stresses derived from extensional flows
may affect the manifold flow in both the cavity (j p* L 4
and the cavity-slot transition. In the cavity, their N d == H = 30EIH
influence is represented by the extension number,
Here E is the elastic modulus of the coating die
Nex=t1extQo Nex='lext R~ 2 (l0.37) material and I is the moment of inertia of the
AWP*' Nv 'le 2W die cross-section. For small values ofthe deflection
Here 'lext is the extensional viscosity in the cavity number (approximately N d < 0.01) the manifold
flow distribution is unaffected by the elastic Nahme number (or Nahme-Griffith number)
deformation of the die, but at large deflection indicates when viscous dissipation can be expected
numbers (approximately N d > 0.1) the effect is to influence the flow,
significant.
Na =: '1* V21! d'1l (10.42)
k '1dT
10.2.7 THERMAL ANALYSIS
In the preceding discussion, it has been implicitly The characteristic viscosity, velocity, and thermal
assumed that the coating die and the liquid conductivity are denoted by '1*, V and k. Large
maintain a constant temperature. Occasionally, Nahme numbers (i.e. Na> 1) indicate a strong
temperature gradients are used as a means of influence of viscous dissipation on the flow while
changing the cross-web uniformity or incidentally a small Nahme number represents negligible
introduced when the die manifold is used to influence. Except for very high speed coating of
remove or add heat from a liquid stream before very viscous liquids (e.g. '1* > 103 Pa s, V> 1 m/s),
coating. Less common is the generation of viscous dissipation ordinarily does not play a
temperature gradients due to viscous dissipation significant role in manifold flows.
in the manifold flow. Whether produced by viscous dissipation or
Variations in temperature influence the manifold introduced by temperature differences between
flow through the temperature dependence of the coating die and liquid feed, temperature fields in
liquid viscosity. For polymeric materials, the manifold flows are governed by the interplay
temperature dependence is often represented in between thermal convection with the flow and
terms of a temperature shift factor, ar (cf Bird, thermal conduction mostly perpendicular to the
Armstrong and Hassager (1987)), flow. The Graetz number, which is closely related
to the thermal Pe number, measures the relative
(10.39) importance of thermal convection to conduction
and provides a measure of the degree of thermal
where '1 re r is a reference viscosity at the reference development,
temperature, '1;.f' Two functional forms of the
temperature shift factor are commonly used: (1) VH*2 H
the Arrhenius form: Gz=:--=Pe- (10.43)
rxL* L
ar(T) = exp[AH(~ -
R T
_1)J
'1;er
(10.40) Here H* is a characteristic dimension for
conduction, L * is a characteristic length in the
and (2) the WLF equation: flow direction, rx is the thermal diffusivity and V
is a characteristic velocity. Large Graetz numbers
-Cl (T - T rer) are indicative of convection-dominated tempera-
Iog aT = ~--=--,=-~::.:. (10.41)
C 2 + T - T rer ture field which is still in the infantile stages of
development. Typical Graetz numbers for mani-
Characterization of temperature fields in flow fold flows range from 10 to 104 • The ratio of the
systems can be found in several sources (Winter Graetz number in a manifold cavity to that in
1975; Pearson 1985; Bird, Armstrong and Hassager the slot is approximately equal to the aspect ratio
1987). The salient points are recounted here with of the slot
reference to the flow through coating die manifolds.
If the coating liquid is fed to the manifold with
a homogeneous temperature that is equal to that (10.44)
of the die body, then the only temperature
gradients present would sensibly be those that Thus, thermal conduction plays a bigger role in
arise from viscous dissipation in the flow. The the slot flow than it does in the cavity flow and
Design considerations 383
thermal manipulation of the die slot would have cross-web distribution of coating liquid. The
a larger impact on the flow uniformity. uniformity of the distribution is measured in
Because of the complexity of the developing terms of the uniformity index (Leonard 1985b):
temperature field in manifold flows there have
been relatively few analyses in the public literature. VI=
(-J.Q)
sma -(-J.Q)
/ (10.46)
Qo W
Smin
Vergnes et at. (1984) investigated the effects of
temperature gradients in a coathanger die manifold
which is the ratio of the difference between the
but considered only a mean temperature across
maximum and the minimum efflux from the
the depth of the cavity and slot. Nordberg and
manifold slot to the average slot efflux. A uniform
Winter (1990) used a similar approach in analyzing
distribution and hence, a small uniformity index,
multilayer slot flows. Consideration of only a
results from all flow paths through the manifold
mean temperature across the depth ofthe manifold
have substantially the same resistance to flow.
channels is a little disquieting because the
A number ofdifferent strategies may be undertaken
temperature in the vicinity of the walls of the
depending on the final application of the coating
manifold should influence the flow more than
die. The discussion of these considerations begins
the temperatures elsewhere. Equation (10.20)
shows that the viscosity (and hence, the tempera- with the characteristics of two limiting cases, the
T-manifold and coathanger designs.
ture) in the regions of highest shear rate (i.e. in
the vicinity of the solid boundaries) has a much
greater influence on the slot flow than the to.3.1 T-MANIFOLD DESIGN - NO GRAVITY,
viscosity elsewhere. As a result, the actual influence INERTIA EFFECT
of the temperature field on the flow uniformity
might be lost by a mean temperature approxi- The T-manifold (or infinite cavity) design strives
mation. for equal flow resistance through the die manifold
Bird, Armstrong and Hassager (1987, by incorporating a uniform slot of high flow
pp. 211-215) show a solution to the temperature resistance with a uniform cavity of low flow
field in the flow through channels that might be resistance. In the limiting case of zero flow
quite useful in conjunction with manifold flow resistance in the cavity, a cavity of 'infinite' size,
models. The solution is valid for the initial stages the degree of uniformity is determined only by
ofthermal development, small viscous dissipation the uniformity of the slot gap and length. For
effects, and assumes a linear velocity profile in relatively high viscosity liquids(Nm « 1, Ng « 1),
the vicinity of the channel wall. The resulting the flow uniformity from aT-manifold depends
temperature field is only on the viscous number, Nv, which is the
ratio of viscous losses in the cavity to those in
T- TWall
-----"-=-'=---- = ~
1 IX exp[ - x
3
] dx,
the slot. Leonard (1985b) has presented an
~nitial - TWall r{J-) ° approximate equation for the uniformity index
for power-law liquids in this situation,
y )1/3
X =y ( 9cxx
wall (10.45)
VI = J Nv nO.3909(1-J;')-1.527
where r() is the gamma function, y is the distance
from the walls of the channel and x is the distance (10.47)
in the flow direction.
Here n is the power-law index of the power-law
viscosity function and Nv is defined according to
10.3 DESIGN CONSIDERATIONS
(10.30). Figure 10.9 is a graphical representation
The goal of a manifold design is to define a of (10.47) and shows the values of the viscous
manifold geometry that will produce a uniform number required for acceptable uniformity. Note
I fl ITE CAVITY DESIG • 0 GRAVITY, I ERTIA EffECT
I
I I
• , n~1
I
,'I
0.8
. I
t;S
I . I
Cl : n~0.8
0.6 :
~
~ :
I I
::E , , I
'"g " /
n~0.6
.. I
:z :
::> 0.4 . /
I
n~O.4 . /
. /
0.2 n~0.2
/ '/
I
0.001 0.01 0.1 10

VISCOU UMBER. v
Figure 10.9 The cross-web uniformity of power-law liquids in infinite cavity manifolds according to equation
(10.47). Gravitational and inertial effects are neglected in this case.
that for small viscous numbers (N v < 0.1), a material A in Fig. 10.6. Note the large differences
uniform distribution is predicted regardless of in flow uniformity that result from a relatively
the power-law index, that is, regardless of the small departure from power-law viscosity behavior.
shear-thinning behavior of the liquid. It is this It is noteworthy that the analysis performed
characteristic that in many cases makes the by Leonard (1985b) on which equation (10.47)
T-manifold design an attractive choice. It has is based does not include the possibility of
substantial ability to produce uniform flow entrance pressure losses in the cavity-slot
distributions with a wide variety of liquids. transitional flow. Highly elastic liquids are likely
Another implication of equation (10.47) and to have large extensional viscosities, and hence,
Fig. 10.9 is that the uniformity is dependent on large entrance pressure losses. As a result, equation
the throughput through the die only through the (10.47) represents an upper bound for the
viscous number, Nv. For the special case of uniformity index as the presence of entrance
power-law liquids, the viscous number is entirely pressure losses would improve the uniformity by
independent offlow rate. As a result, the uniformity providing an added flow resistance in the
of a power-law liquid is a constant for all flow cavity-slot transition.
rates. Figure 10.10 shows an example of this for T-manifolds have the disadvantage that for
a power-law liquid with a power-law index equal high viscosity liquids extraordinarily high cavity
to 2/3. Also shown is the expected uniformity pressures or very large cavities may be required
from a liquid with the viscosity behavior of to achieve reasonable uniformity. Rearrangement
I
I
0.8 I
I
:; I
xuJ
I
0 0.6
~ I
~ POWER-LA w L1Q 10. n = 2J3 /
i
~ I
~
Z 0.4 I
::J
/
./
./ MATERIAL A
0.2
0.1 10 100 1000
SLOT SHEAR RATE. 6QJWH2
Figure 10.10 The variation of the cross-web uniformity with flow rate in a power-law liquid and one with
the viscosity behavior of material A in Fig. 10.6. The viscous number, Nv, is 0.39 for the power-law
liquid and varies from 0.17 to 0.52 over the flow rate range for material A.
of(1O.30) expresses the scale of the cavity pressure Matsubara (1980) analyzed the distribution of
in terms of the viscous number: average residence time in a T-manifold with a
circular cavity cross-section. This distribution of
*_ 1 W . (10.48)
residence times is that which arises from the
P - Nv R '1 e Ye different flow paths the liquid takes through the
b
manifold and not the distribution that arises
In order to maintain flow uniformity (i.e. maintain from the roughly parabolic velocity profile through
a small Nv number) while increasing viscosity, the channels of the manifold. The residence time
either the cavity pressure will increase or the distribution, t(X), calculated as the length of the
cavity size must increase to limit the pressure flow path divided by the average velocity through
increase. Large cavity pressures in turn require it, is
very bulky coating dies to resist deflection of the
coating die. The die size required is gauged from t(X) _ 1 nR21n[~J
the die cross-section moment of inertia needed (HQ:L) - + HL W- X
to keep the deflection number defined in (10.38)
small. Large cavities diminish the ability of the nR 2 E
die to resist deflection from hydrodynamic pressure (10.49)
+2HWL
and increase the residence time of the liquid in
the manifold. Large residence times may cause where E is the length of the entrance tube that
thermal degradation problems or other detri- feeds the manifold cavity.Liquid exiting from the
mental effects if the coating liquid is reactive. slot opposite the feed location of the cavity has
the shortest residence time whereas liquid that Gravitational effects can affect the flow uni-
travels through the distal end of the cavity has formity of a T-manifold design in much the same
the longest. The breadth of the residence time way as inertial effects when the cavity axis is not
distribution is clearly influenced by the size of aligned horizontally. Leonard (1985b) shows the
the cavity and the slot. minimum uniformity index for several power-law
liquids appearing between values of 0.5 and 1
for ratio ofthe gravity number to viscous number.
10.3.2 T-MANIFOLD DESIGN - GRAVITY, Negative values of the gravity number always
INERTIA EFFECTS decrease the flow uniformity of the T-manifold.
When gravitational and inertial effects are
negligible in theT-manifold (i.e. Nm « 1, Ng « 1),
the flow exiting from the slot is always highest 10.3.3 COATHANGER DESIGN
adjacent to the cavity feed location. The presence
An alternative to the T-manifold is the coathanger
of gravitational and/or inertial effects can change
manifold design. This design matches the flow
this slot efflux from a maximum to a minimum.
resistance in the cavity and the slot so that each
Acrivos, Babcock and Pigford (1959) describe
path through the die has equal resistance. Pearson
the essential physical mechanisms for the inertial
(1964) describes a coathanger design that tailors
effects.
the flow resistance in the slot using the slot gap.
Both Leonard (1985b) and Sartor (1990) report
McKelvey and Ito (1971) and Cloeren (1993) use
that for Newtonian liquids, the T-manifold flow
a multistage slot in which the gaps in the
distribution can be made perfectly uniform by
individual stages and the total length of the slot
matching the momentum and viscous forces.
are constant across the width of the manifold.
Under this condition, the viscous losses in the
The length of the individual stages have a
cavity flow are exactly offset by the inertial forces.
cross-web variance that changes the slot flow
When the momentum number is much less than
resistance. A more common approach is to use
the viscous number, the slot efflux is still highest
a single-stage slot whose length varies in the cross-
adjacent to the feed location. As the momentum
web direction. That is, the slot length is longer
number approaches and then exceeds the viscous
near the feed port and shorter at the distal end
number, the slot efflux becomes lowest adjacent
of the die so that lower slot flow resistance at
to the feed location. Figure 10.11 shows an
the end is exactly balanced by the resistance from
example of this effect.
the flow through the manifold cavity. The
For shear-thinning liquids, in particular, liquids
difference between the slot length at the feedport
with power-law viscosities, Leonard (l985b)
and the slot length at the end of the die is referred
reports that a perfect flow distribution is not
to as the slot length taper. The amount of taper
possible, but a minimum occurs as the momentum
required in the slot length is given approximately
number is increased in relation to the viscous
by setting the viscous number equal to one in
number. The location of the minimum is shifted
(10.30),
to higher ratios of the momentum and viscous
numbers with lower power-law indices.
It is also noteworthy to add here that since WH'leYe
Ltaper = 2R h "Is
.. y' (10.50)
the analysis by Leonard (1985b) ignores entrance s
pressure losses in the cavity-slot transition, the
predictions of the uniformity index may be Thus larger taper in the slot length is required
over-estimated. An entrance pressure loss due to for more shear-thinning liquids, wider dies, smaller
inertial forces will increase the flow resistance in cavities and larger slot gaps.
relation to the local flow rate, and hence, increase In addition to the taper in the slot, the cavity
the uniformity of the flow distribution. cross-section usually decreases in size from the
CROSSWEll NIFORMITY FOR N,' = 0.16
I I
NII1~O.r)()2
1.115
---
......
II1~O 1117
{ /'
ri NII1=O.031 /
x
::>
--------
ti /
"-
'"
b 0.95
I11=O.O~S
...l
til /
N'n=O.1160
0.9
o 0.2 0,4 0.0 o.

CROSSWEB POSITIO (X/Wj
Figure 10.11 An example of the cross-web uniformity change due to inertial effects. The cavity feed location
is at X = O.
feed location to the distal end in order to reduce the curvilinearly tapered coathanger manifold as
the liquid residence time. Whereas a circular simple to fabricate as the linearly tapered
cavity of constant size is easily fabricated, a coathanger. In addition, because uniformity
circular cavity of decreasing radius is difficult to requirements have generally increased, the superior
machine. It is much easier to fabricate a noncircular resulting flow uniformity obtained from the
cavity (such as the triangular cavity in Fig. 10.2) curvilinearly tapered coathanger manifold makes
using a constant-radius machine tool (Chung it the preferred choice.
and Lohkamp 1976). The size of the cavity is A number of prescriptions are available for
varied by changing the depth of the machine tool designing the appropriate tapers in the slot
cut. length and cavity size in coathanger manifolds.
Two types of coathanger manifolds are often Most are derived under the tenuous assumption
distinguished: (1) the linearly tapered coathanger that the slot flow is everywhere parallel to the
(or fishtail manifold) and (2) the curviiinearly machine direction (opposed to the direction of
tapered coathanger. The linearly tapered coat- the Y-axis in Fig. 10.4). Under this flow condition,
hanger often, but not always, has a linear taper equality of flow resistance for each flow path is
in both the slot length and the cavity size. specified mathematically as a relation between
Traditionally, the linearly tapered coathanger the slot and cavity pressure gradients and the
has been preferred because of its ease of manu- slope of the slot length taper:
facture with acceptable resulting flow uniformity
(Michaeli 1992). The advent of computer-con-
trolled fabrication machinery, however, has made
[:~lol = - [::,JcaVilY Si~~ (10.51)
The angle characterizes the slot length taper,

<jJ
R(X) = R o [ 1 - ~Jn:l,
Ro = [Hsin2n ((3n
dL
- = -tan<jJ (10.52) +1 + 1) w)nJ3n~
dX 1
<jJ(2n + 1) 2n (10.53)
Procter (1972) reports that the optimal uni-
formity from a linearly tapered coathanger is where n is the power-law index of the liquid.
obtained when the cavity radius (or equivalent Figure 10.12 shows the cavity radius taper
radius) at the feed location is 1.5 times bigger according to (10.53). Here and everywhere in this
than that at the distal end. Chung and Lohkamp discussion the cavity is fed at X = 0 and the
(1976), Tadmor and Gogos (1979, pp.545-551), distal end of the cavity is at X = w: Also shown
and Matsubara (1983) report that for perfect is a curve for an infinite power-law index which
uniformity in a coathanger with a linearly tapered corresponds to the cavity radius specified by
slot length, the cavity radius (or equivalent (10.54). Note that especially for shear-thinning
radius) should have the following form: liquids the taper is almost linear over most of
o.
f!!E.0
c::
vi 0.6
::l
is
<
0::
>- --B- n=1
C
>
~ 0.4 - ' S - n=O.6
~n=O.4
-fr- n=O.2
0.2
n=O.1
-n=oo
o 0.2 0.4 0.6 O.

eRO WEB PO [TIO . XIW
Figure 10.12 The taper in the cavity radius (or equivalent radius) for coathanger manifolds according to
equation (10.53). The power-law index of the liquid is given by n. The curve for an infinite power-law index
corresponds to the cavity radius profile dictated by equation (10.54).
the manifold width. In addition, the cavity radius perpendicular to the cross-web direction. Figure
ratio of 1.5 proposed by Procter (1972) is 10.14 shows the relationship between the maxi-
approximately obeyed by (10.53) over a relatively mum slot length to the aspect ratio of the slot
large range of power-law indices. exit. Note that a slot with large aspect ratio is
Obviously, there are a multitude of possible required to maintain a relatively short slot length
coathanger designs that produce uniform cross- in the uniform residence time coathanger manifold
web flow distribution. Other criteria may be design.
invoked to limit the design choices. Matsubara In cases where a uniform residence time
(1979) showed that a uniform residence time distribution is not essential, Matsubara (1981)
distribution is obtained when cavity radius (or has parameterized the cavity and slot profiles in
equivalent radius) has the following profile: terms of a ratio of average residence times, M,
which is the ratio of residence time of liquid that
R(X) = Ro[1- ~]1/3, traverses the entire length of the cavity and exits
the slot at X = W to that of liquid which exits
2 the slot at X = O. The cavity and slot profiles in
R = [H W( 3n + 1 )n~1]1/3
o 1t 2 (2n + 1) (10.54) (10.54) and (10.55) represent a residence time
ratio, M, equal to unity and the profiles for other
The corresponding slot length profile has a more values of the residence time ratio have similar
complicated functional form, forms:
~) = ~p[ (1 _~Y/3 Ro(M) = M1/(3n+3) R~M;l),
L(X,M,P) = VM ;1)(X, ~) (10.58)
A larger cavity than that specified by (10.54)

(I - X/W)1/3 + J(l- X/W)2/3 - P2}] subjects liquid travelling through the cavity to
+ p2 ln { p a larger residence time than that which exits
adjacent to the cavity feed location. The corre-
(10.55) sponding decrease in the required slot length
where the parameter p is the ratio of the area of taper is obtained by replacing P in (10.55) or
the largest cavity cross-section to that of the slot (10.57) with P/M. Figure 10.15 shows how the
exit: maximum cavity radius and slot length vary with
power-law index and residence time ratio.
R
,.,
== 1tR~
HW
= (1tH)1/3[
W
3n + 1 ]3}~3
2(2n + 1)
(056)
1 .
Coincidentally, the cavity defined by (10.54)
or (10.58) has a uniform wall shear rate and shear
Figure 10.13 shows the slot length profile specified stress along its length. The wall shear rate in the
by (10.55) for various values of the area ratio p. slot is also uniform across the manifold width
The maximum slot length, L o' is proportional and is related to the cavity shear rate in the
to the manifold width, following manner;
18101 __ [2(2n + l)M]l/(n+l)
L o =~p • (10.59)
W 2 Ycavily 3n + 1
[.j1 - p2 + p2lnC + .j~)] (10.57)

The slot shear rate is ordinarily larger than the
cavity shear rate and the difference increases with
increasing shear-thinning (i.e., lower power-law
The slot length decreases to zero at the cross-web index) or increasing residence time ratio. Equality
position X = W(l - P3) where the cavity is of cavity and slot wall shear rates necessitates a
SLOT LENGTH PRO ILES
--e--- p = 0.01
-a-p =0.1
0.8
---p =02
----{r- P= 0 3
S.x 0.6
::r --~ =0.4
J:
I- --p =0.5
0
;Z;
UJ
...l
I- 0.4
a
...l
Vl
0.2
-~=....,~
--
o 0.2 OA 0.6 0.8

eRO SWEB POSITION (XIW)
Figure 10.13 The slot length taper for a coathanger manifold with a uniform residence time distribution
according to equation (10.55). The ratio of the maximum cavity cross-sectional area to that of the slot exit is
given by the parameter p.
design with a residence time ratio less than unity profiles of the height and width of the cavity;
(Charbonneaux 1991),
Hc(X) = [~~~r/Tl ~r/3, -
M = 3n + 1 (10.60) ~(X) = aHc(X) (10.61)
2(2n + 1)
a is the aspect ratio of the cavity and f p is the
shape factor which characterizes the effects of
In this case, liquid that traverses the cavity and the finite aspect ratio of the rectangular cavity
exits at X = W has the shortest residence time. (cf Middleman (1977) pp.90-92). The corre-
Winter and Fritz (1986) proposed that rectan- sponding slot length profile is the same as that
gular cavities be used to produce material- given by (10.55) with the following substitution
independent manifold designs by matching the for the parameter {3:
cavity and slot wall shear rates. In cavities with
a constant aspect ratio, equality of cavity and
(3
= [a HfpJ
W
1/3 = H
Hc(X = 0) (10.62)
slot shear rates is achieved with the following
50 o.
:I:
f-
0
Z
UJ
...J
f-
a
...J
'"~
~
~
:2:
10 100 1000
SLOT XIT ASP cr RATIO. WIH
Figure 10.14 The slot length taper for a coathanger manifold with a uniform residence time distribution
according to equations (10.56) and (10.57).
The shear-thinning nature of the liquid (i.e. the taper to 100mm. Reducing the aspect ratio of
power-law index) enters into this manifold design the cavity reduces the slot length taper, but it
only through the shape factor,fp, which depends also increases the sensitivity of the design to the
on both the aspect ratio of the cavity cross- power-law index of the liquid through the shape
section and the power-law index of the liquid. factor. Uniformity of residence time is also
Winter and Fritz (1986) point out that for large sacrificed by reducing the cavity aspect ratio.
aspect ratio cavities, the shape factor, and thus So it is that a material-independent coathanger
the manifold design, is independent of power-law manifold design that provides uniform flow and
index and equal to unity. In addition, uniformity residence time distribution is possible with the
of residence time is achieved if the shape factor use of large aspect ratio rectangular cavities.
is equal to unity. Unfortunately, a large aspect Unfortunately, the application of this design is
ratio cavity also requires either a long slot length often not practical for the wide coating dies
taper or a very small slot gap. For example, if commonly encountered. Either very long slot
the cavity aspect ratio is equal to 10 and the length tapers or very narrow slot gaps are
manifold width is 1m, a slot gap of 2.5 mm required. Long slot tapers often produce unac-
requires a slot length taper of 494 mm and a slot ceptable elastic deformation of the coating die
gap of29 ~m is required to shorten the slot length as portrayed by the deflection number in (10.38).
60 250
---e- n=1
50
---e- n=0.5
~- n=0.5
n=0.1
200
::I:
~
""0
c.: 40
vi
c:
0 150 ~
<
c.: ,
(I)
0
~
>
30
y
,-i
tTl
<
U
Z
0
~ -i
100 .::I:
~
d
~
20
:r
~
50
10
o
10 100 1000
RESiDE CETIMERATIO,M
Figure 10.15 The variation of the coathanger manifold design with residence time ratio and shear-thinning
behavior (power-law index) of the liquid according to equation (10.58). The aspect ratio of the slot exit is held
constant and equal to 1000.
Owing to the sensitivity of the flow distribution Fig. 10.15). Thus, coathanger manifold designs
to slot gap variations (see Fig. 10.8), narrow slot which increase the uniformity of residence time
gaps require prohibitive gap uniformity. The also increase the applicability of the design to
coathanger manifolds described by Matsubara materials of varying shear-thinning behaviour.
(1979,1981) provide flow uniformity with varying This is contrary to the widely-held beliefexpressed
degrees of residence time uniformity, but are by Lippert (1991) that 'Larger flow channels are
dependent on the shear-thinning nature of the less sensitive to rate and viscosity changes than
coating liquid. For many nonreacting coating small channels.' Figure 10.15 suggests that
liquids, uniformity of residence time in the manifold sensitivity to changes in shear-thinning behavior
is not as desirable as insensitivity to the shear- is decreased in coathanger manifold designs by
thinning nature of the liquid. But note that reducing the size of the cavity. Sartor (1990)
according to (10.59), uniformity of residence time showed that increasing the size of the cavities
also decreases the shear rate disparity between and slots in T-manifold designs reduces the
the cavity and the slot which reduces the sensitivity sensitivity to rate and viscosity changes. He also
of the manifold design to material changes (see showed that uniformly increasing the slot length
decreases the sensitivity of coathanger manifold in the slot region of the manifold. Secondary
designs. However, it is still not clear whether cavities provide another chance for flow across
sensitivity is better diminished by uniformly the die if a nonuniform distribution resulted from
increasing the slot length or increasing the slot the flow out of the original cavity. Leonard
length taper and thereby reducing the shear rate (1985a) modeled secondary cavities with a
disparity between the slot and the cavity. procedure very similar to the analysis of primary
Caution is well-advised in the use of the cavities. He reported that there was not a clear
aforementioned coathanger cavity and slot profiles. optimum secondary cavity position and that the
These coathanger designs account only for viscous utility of adding secondary cavities depended on
effects while ignoring the influence of inertia and the viscous number and the shear-thinning
gravitational forces. In addition, very simplified behavior of the liquid. For large viscous numbers
models for the cavity and slot flow were used to and very shear-thinning liquids, it was more
produce the coathanger manifold designs and beneficial to enlarge the primary cavity than it
the liquid was assumed to follow the power-law was to add a secondary one. For small viscous
viscosity model. Although the general features numbers and Newtonian liquids, the opposite
and trends may be correct, the details of the was true.
coathanger designs are likely to be inaccurate. Sartor (1990) analyzed secondary cavities
More accurate coathanger manifold designs can through the lubrication approximation applied
be obtained with the combination of I1-dimen- to the slot flow. He found a clear optimal location
sional cavity flow and two-dimensional slot flow for the secondary cavity at the midpoint of the
models described in the previous sections. Inertia slot. An optimum size of the secondary cavity
and gravitational effects can be incorporated into was also indicated which was half the cross-
coathanger designs computed with these models, sectional area of the primary cavity. In addition,
the accuracy of the cavity and slot flow repre- Sartor (1990) analyzed secondary cavities through
sentation is much improved, and a wide variety an electrical resistor analogy to viscous flow
ofliquid viscosity behavior can be incorporated. based on the work of Booy (1982). The optimal
Sartor (1990) details a procedure that can be secondary cavity location was still at the mid-point
used to compute slot length profiles for coathanger of the slot, but the optimum size was with the
manifold designs in conjunction with a two- secondary cavities slightly smaller than the primary
dimensional slot flow model. The length of the one.
slot at numerous cross-web positions is treated The shape of the secondary cavity can also
as a variable whose value is determined by have a significant impact on the resulting flow
requiring the slot efflux to be uniformly distributed. field. Lee, Wen and Liu (1990) observed closed
It is interesting to note that the slot length recirculations on the upstream side ofa secondary
profiles computed by Sartor (1990) have a shape cavity with glycerine/water solutions. When an
that is very different from that in (10.55) near aqueous polyacrylamide solution was observed,
the distal end of the cavity. The length of the the closed recirculations appeared on the down-
slot tends to approach the end of the cavity stream side of the secondary cavity. In both cases,
substantially perpendicular to the machine reducing the entry or exit angle of the secondary
direction while the profiles from (10.55) shown cavity eliminated the eddy.
in Fig. 10.13 tend to be more parallel to the From the coating engineer's viewpoint, the
machine direction. discussion of the utility and optimal design of
secondary cavities is probably moot. Improve-
ments in the design and fabrication of the
10.3.4 SECONDARY CAVITIES
primary cavity and slot have greatly reduced the
Another option in the design of flow distribution need for secondary cavities. In addition, Lee,
manifolds is the inclusion of additional cavities Wen and Liu (1990) have vividly demonstrated
potential problems that secondary cavities can interdiffusion is an important consideration from
create. As a result, it is often advantageous to this viewpoint.
eliminate secondary cavities. Two-layer coextrusion flow in a slot of
incompatible polymers was analyzed both the-
oretically (Su and Khomami 1992) and experi-
10.3.5 MULTILAYER MANIFOLD DESIGN mentally (Wilson and Khomami 1992). The
When the coating liquid to be distributed by the theoretical analyses depict a rather complicated
die manifold is a multilayer package of a number situation in which the interfacial stability is
of different materials, different design consider- influenced by inertial, viscous and elastic forces,
ations may need to be considered. In some as well as the thickness ratio of the layers.
multilayer coextrusion dies, the layers are Interfacial shear stress does not appear in this
combined externally and the manifold consider- picture. Generally, the studies show that the
ations are essentially identical to single layer dies. interface is stable if the more viscous layer is
It is when the layers are combined internally thick and unstable if the more viscous layer is
through feed block systems or common lands that thin. There is a critical layer thickness ratio, r c :
significant changes occur. Michaeli (1992) provides
a good description of the various externally and r = hmore viscous = '1 more ViSCOUS) (10.63)
internally combined manifold systems. Cloeren hless viscous (
C 'lless viscous
(1979, 1980) also describes die manifolds for
internally combining different layers. defined in terms of the viscosity ratio of the
One of the principal concerns in combining layers. If the layer thickness ratio is greater than
layers in the die manifold is whether or not the r c ' then the elasticity of the liquids has a stabilizing
layer interfaces will remain smooth and stable. effect. Below this critical thickness ratio, stability
Schrenk et al. (1978), Han and Shetty (1978) and can be enhanced by decreasing the elasticity of
Butler (1992) describe interfacial instabilities in the more viscous layer and increasing the elasticity
polymeric melt systems which produce distorted ofthe less viscous one. Inertia generally destabilizes
interfaces. All indicate that the layer interfaces the interface while the effect of any interfacial
become unstable when the interfacial shear stress tension is stabilizing. If the slot has a converging
exceeds a critical value. Schrenk et al. (1978) cross-section, the extensional stresses developed
report that the interfaces remain smooth and in the converging flow have essentially the same
stable until the final die land where the interfacial effect as increasing the elasticity of the layers.
shear stresses are presumably the highest. The The wavelength of the most dangerous mode of
interfacial shear stress can be reduced (and the the interfacial instability scales with the thickness
instability potentially eliminated) by increasing of the more viscous layer:
the thickness or decreasing the viscosity of the A. ~ 27l: hmore viscous (10.64)
layer adjacent to the die walls, reducing the
throughput rate, or increasing the slot gap of the Whether the interfacial instability of multilayer
manifold. Kim and Han (1991) addressed inter- flows is controlled by interfacial shear stress or
facial in~ability in compatible polymeric systems a more complicated combination of viscous,
from tt,'e viewpoint of the interdiffusion and the elastic and inertial forces, the use of stability
subsequent adhesive bond strength that develops criteria in die manifold design requires knowledge
after the interface is formed. High interfacial of the flow conditions in the individual layers
shear stresses inhibit diffusion through molecular and near the interfaces. A number of researchers
orientation in the vicinity of the interface. Weak (Sornberger, Vergnes and Agassant 1986a, 1986b;
adhesive strength and presumably high suscep- Nordberg and Winter 1988, 1990; Puissant et al.
tibility to interfacial instability result. In addition 1992) have developed flow models with this
to interfacial shear stress, the contact time for objective in mind. The models feature approximate
one- and two-dimensional representations of gelation of thermosetting liquids, it would be

multilayer flow in slots that provide information detrimental if the coating liquid is a dispersion
about the interface locations and conditions. The susceptible to shear-induced coagulation. The
common practice of adjusting layer viscosity by rate of coagulation in the absence of repulsive
changing its incoming temperature necessitates forces is proportional to the shear rate (Russel,
the inclusion of thermal effects in the flow models Saville and Schowalter 1989). Thus, in order to
with the accompanying difficulties discussed above. minimize the total coagulation during the flow
Another concern in the design of multilayer through the die manifold, both shear rate (or
die manifolds is the tendency of low viscosity wall shear stress) and residence time should be
liquids to encapsulate those of higher viscosity. kept small.
In the manifold flow of wide dies, this results in As was mentioned in previous sections, the
a preferential flow of the lower viscosity liquids slot efflux varies according to, at least, the cube
towards ends of the manifold (Butler 1992). of the slot gap. For shear-thinning liquids the
Cloeren (1993) states that manifold cavities of dependence of flow on the slot gap is higher. As
rectangular cross-section are superior to those a result, maintaining uniformity of the slot gap
ofcircular shape for distributing multilayer flows. across the width of the coating die manifold is
It may be that encapsulation may be responsible a high priority. Machining tolerances and elastic
for some of the rationale. deformation of the die due to the hydrodynamic
pressure from the liquid flow through the manifold
are common causes of gap nonuniformities. The
10.3.6 OTHER CONSIDERAnONS
deflection number defined in (10.38) can be used
There are a number of other considerations that to gauge when elastic die deflections can be
may impact the design of a coating die manifold. expected to be excessive. In addition, the deflection
These include the effects of elastic deformation number shows that the parameters which have
of the die, temperature nonuniformities and the largest influence on the deflection of the die
implications of reactive liquids. Charbonneaux are the length and gap of the slot. Shorter slot
(1991) points out that die designs for reactive lengths and larger slot gaps will decrease the
thermosetting materials should maintain the amount of deflection.
shear stress at the walls of the cavity and slot The slot efflux also varies in proportion to the
above a critical value to avoid plugging of the reciprocal of the liquid viscosity. Maintaining
die glue to gell formation. Schweizer (1992) also uniform viscosity through temperature uniformity
reports that removal ofbubbles and contaminants across the width of the manifold is also a priority.
from the walls of the die manifold is facilitated Equation (10.39) describes the expected viscosity
by maintaining the wall shear stress above a variation due to temperature differences. Tem-
critical value which depends on the contaminant perature nonuniformities may arise from variations
size and its interfacial properties. The coathanger in the liquid feed, differences between the
designs described in (10.58) and (10.61) are well temperatures of the liquid feed and the die, or
suited for this task as they maintain a constant temperature variations in the die itself.
shear stress along the length of the cavity. Proper design ofcoating dies to insure minimal
T-manifold designs do not fulfill this requirement elastic die deflection and sufficient heat transfer
and are likely to plug near the ends of the cavity. to control the temperature uniformity is a subject
However, Sartor (1990) notes that even in that largely falls outside the scope of this writing.
T-manifolds the wall shear stress can be increased A number of the significant mechanical design
by drawing a recycle liquid stream from the distal considerations as summarized by Michaeli (1992)
end of the cavity. are as follows: (1) the number of parts should be
Although increasing the wall shear stress may kept to a minimum; (2) sealing surfaces should
be beneficial in preventing plugging due to be small and flat; (3) a few, large die bolts are
better than many, small die bolts to hold the die feature better uniformity of residence time and
together; (4) the die dimensions must be sufficient lower cavity pressures, but the flow uniformity
to keep elastic deformations acceptable; (5) the tends to be more sensitive to liquid material
flow channel surface should be polished to a properties. Detailed understanding of the best
surface roughness less than 0.2 J,lm, and (6) the design strategy for a particular application requires
die material should be selected based on corrosion analysis ofthe interplay between the various fluid
resistance, machinability, toughness, surface mechanical forces. This is provided by the solution
hardness and thermal conductivity. of the appropriate mass and momentum conser-
vation principles. In addition, mechanical con-
siderations such as elastic deflection due to
10.4 SUMMARY hydrodynamic pressure in the manifold cavity
The flow distribution that results from the flow and appropriate heat transfer for temperature
through internal cavities of a coating die is a uniformity are also important considerations.
result of the interplay between the manifold Overall, the science of die manifold design and
geometry and the viscous, inertial, gravitational, analysis is fairly well-developed. Exceptions,
and elastic forces in the coating liquid. In some which should be the focus of future research,
cases, the elastic deformation of the die itself include understanding of the effects of temperature
plays an important role. In the manifold cavity nonuniformities, design of the best cavity
flow, the dominant forces are viscous losses, ends and design of manifolds for multilayer
inertia and gravity. In the slot flow, viscous forces flows.
dominate. In the cavity-slot transitional flow,
inertial forces as well as elastic forces in the form
of extensional viscosity are the key components. REFERENCES
A detailed understanding of the effect of each Acrivos, A., Babcock, B. D. and Pigford, R. L. 1959.
requires the simultaneous solution ofthe combined Flow distribution in manifolds. Chemical Engineering
mass and momentum conservation principles in Science. 10: 112-124.
all three regions. Anthony, 1. D., Jr., Leffew, K. W. and Trentacosta,
A rudimentary understanding of the role of J. D. 1988. Thickness control system for an extrusion
each of the forces is gained from a set of coating apparatus. US Patent 4,765,941.
Batchelor, G. K. 1967. An Introduction to Fluid
dimensionless numbers. The viscous, momentum, Dynamics. Cambridge: Cambridge University Press.
and gravity numbers reflect the importance of Binding, B. M. 1988. An approximate analysis for
the viscous, inertial and gravitational forces in contraction and converging flows. J. Non-Newt.
the manifold cavity to the viscous forces in the Fluid Mech. 27: 173-189.
slot. The importance of elastic forces in the form Binding, 8. M. 1991. Further considerations of
of normal stresses either from shear or extensional axisymmetric contraction flows. J. Non-Newt. Fluid
Mech. 41: 27-42.
flow in the cavity is captured in the normal stress
Bird, R. 8., Armstrong, R. C. and Hassager, O. 1987.
and extension numbers. Finally, the character- Dynamics of Polymeric Liquids. New York: John
ization of elastic deformation of the die due to Wiley.
hydrodynamic forces in the liquid flow is captured Blodgett, O. W. 1966. Design of Welded Structures.
in the deflection number. Cleveland, Ohio: James F. Lincoln Arc Welding
The design strategies of coating die manifolds Foundation.
vary depending on the intended die application. Boger, D. V. 1987. Viscoelastic flows through contrac-
T-manifold cavity designs are suited for dis- tions. Ann. Rev. Fluid Mech. 19: 157-182.
Booy, M. L. 1982. A network flow analysis ofextrusion
tributing a variety of liquids of varying material dies and other flow systems. Polymer Eng. Sci. 22(7):
properties, but may result in excessive liquid 432-437.
residence time or excessive die deformation due Butler, T. I. 1992. Effects offlow instability in coextruded
to high cavity pressures. Coathanger designs films. TAPPI Journal, September: 205-211.
References 397
Cameron, A. 1981. Basic Lubrication Theory. Chichester, Liu, T. 1983. Fully developed flow of power law fluids
West Sussex, England: Ellis Horwood. in ducts. Ind. Eng. Chem. Fundam. 23: 183.
Carley, 1. F. 1954. Flow of melts in 'crosshead'-slit Liu, T. and Hong, C. 1988. The pressure drop/flow rate
dies; criteria for die design. J. Appl. Phys. 9: equation for non-Newtonian flow in channels or
1118-1123. irregular cross-section. Polymer Eng. Sci. 28(23):
Charbonneaux, T. G. 1991. Design of sheet dies for 1559-1564.
minimum residence time distribution: a review. Lowey, R. E., Jr. 1960. Plastic extrusion die. US Patent
Polym.-Plast. Technol. Eng. 30(7): 665-684. 2,938,231.
Chung, C. I. and Lohkamp, D. T. 1976. Designing Luo, x.-L. and Mitsoulis, E. 1990. A numerical study
coat-hanger dies by power-law approximation. Modem of the effect of elongational viscosity on vortex
Plastics. March: 52-55. growth in contraction flows of polyethylene melts.
Cloeren, P. F. 1979. Method for forming multi- J. Rheology. 34(3): 309-342.
layer laminates. US Patent 4,152,387. Matsubara, Y. 1979. Geometry design of a coat-
Cloeren, P. F. 1980. Variable thickness extrusion die. hanger die with uniform flow rate and residence time
US Patent 4,197,069. across the die width. Polymer Eng. Sci. 19(3): 169-172.
Cloeren, P. F. 1993. End feed extrusion. US Patent Matsubara, Y. 1980. Residence time distribution of
5,234,649. polymer melt in the T-die. Polymer Eng. Sci. 20(3):
Erckmann, 8. 1988. Shaping tool for an extruder or 212-214.
calibrator for thermoplastic material. US Patent Matsubara, Y. 1981. Coat hanger die. US Patent
4,721,447. 4,285,655.
Han, C. D. and Shetty, R. 1978. Studies on multilayer Matsubara, Y. 1983. Residence time distribution of
film coextrusion II.Interfacial instability in flat film polymer melts in the linearly tapered coat-hanger
coextrusion. Polymer Eng. Sci. 18(3): 180-186. die. Polymer Eng. Sci. 23(1): 17-19.
Hiraki, Y., Tanaka, Y. and Noda, S. 1989. Coating McKelvey, J. M. and Ito, K. 1971. Uniformity of flow
method. US Patent 4,828,779. from sheeting dies. Polymer Eng. Sci. 11(3): 258-263.
Huang, 1. C. P. and Yu, H. S.1973. Pressure distributions Michaeli, W. 1992. Extrusion Dies for Plastics and
in porous ducts of arbitrary cross section. J. Fluids Rubber. New York: Oxford University Press.
Eng. Trans. ASME. 95: 342-348. Middleman, S. 1977. Fundamentals ofPolymer Processing.
Iguchi, K., Nitta, S., Wada, H. and Sano, T. 1987. T-dies New York: McGraw-Hill.
adapted for extrusion molding. US Patent 4,704,083. Miller, C.1972. Predicting non-Newtonian flow behavior
Kasamatsu, T. 1987. Film-forming T die for low in ducts of unusual cross section. Ind. Eng. Chem.
viscosity resin. US Patent 4,708,629. Fundam. 11(4): 524-528.
Kim, J. K. and Han, C. D. 1991. Polymer-polymer Nelson, N. K., Kistler, S. F. and Olmsted, R. D. 1990.
interdiffusion during coextrusion. Polymer Eng. Sci. A generalized viscous model for the steady shear
31(4): 258-269. rheology of magnetic dispersions. Paper read at 62nd
Klein, I. and Klein, R. 1973. Computer modelling of Annual Society of Rheology meeting, 21-25 October
coat hanger dies may be cheaper for the long run. 1990, Santa Fe, New Mexico.
SPE Journal. 29: 33-37. Nissel, F. R. 1976. Thickness control system for an
Lee, K. and Liu, T. 1989. Design and analysis of a extrusion die. US Patent 3,940,221.
dual-cavity coat-hanger die. Polymer Eng. Sci. 29(15): Nordberg III, M. E. and Winter, H. H. 1988. Fully
1066-1075. developed multilayer polymer flows in slits and
Lee, K., Wen, S. and Liu, T. 1990. Vortex formation in annuli. Polymer Eng. Sci. 28(7): 444-452.
a dual-cavity coat-hanger die. Polymer Eng. Sci. Nordberg III, M. E. and Winter, H. H. 1990. A simple
30(19): 1220-1227. model of nonisothermal coextrusion. Polymer Eng.
Leonard, W. K. 1985a. Effects of secondary cavities, Sci. 30(7): 408-415.
inertia and gravity on extrusion dies. ANTEC '85, Pearson, 1. R. A. 1964. Non-Newtonian flow and die
pp.I44-148. design. Trans. J. Plastics Institute. 32(99): 239-244.
Leonard, W. K. 1985b. Inertia and gravitational effects Pearson, 1. R. A. 1985. Mechanics ofPolymer Processing.
in extrusion dies for non-Newtonian fluids. Polymer New York: Elsevier Science Publishing Co.
Eng. Sci. 25(9): 570-576. Procter, B. 1972. Flow analysis in extrusion dies. SPE
Lippert, H. G. 1991. Slot die coating for low viscosity Journal. 28: 34-41.
fluids. In Coatings Technology Handbook, ed. D. Puissant, S., Vergnes, 8., Demay, Y. and Agassant,
Satas. New York: Marcel Dekker. 1. F. 1992. A general non-isothermal model for one-
dimensional multilayer coextrusion flows. Polymer Tadmor, Z. and Gogos, C. G. 1979. Principles of
Eng. Sci. 32(3): 213-220. Polymer Processing. New York: John Wiley.
Reifenhiiuser, H., Beisemann, H., Reitemeyer, P. and Vergnes, B. and Agassant, J. F. 1986. Die flow
Grabowski, R. 1986. Extruder or calibrating die. US computations: A method to solve industrial problems
Patent 4,594,063. in polymer processing. Advances in Polymer Tech-
Russel, W. B., Saville, D. A. and Schowalter, W. R. nology. 6(4): 441-455.
1989. Colloidal Dispersions. New York: Press Syndicate Vergnes, B., Saillard, P. and Agassant, J. F. 1984.
of the University of Cambridge. Non-isothermal flow of a molten polymer in a
Sartor, L. 1990. Slot coating: fluid mechanics and die coat-hanger die. Polymer Eng. Sci. 24(12): 980-987.
design. PhD thesis, University of Minnesota. Vlcek, J., Mailvaganam, G. N., Vlachopoulos, J. and
Schrenk, W. 1., Bradley, N. L., Alfrey, Jr., T. and Maack, Perdikoulias, 1. 1990. Computer simulation and
H. 1978. Interfacial flow instability in multilayer experiments of flow distribution in flat sheet dies.
coextrusion. Polymer Eng. Sci. 18(8): 620-623. Advances in Polymer Technology. 10(4): 309-322.
Schweizer, P. F. 1992. Fluid handling and preparation. Vrahopoulou, E. P. 1991. A model for fluid flow in
In Modern Coating and Drying Technology, eds E. dies. Chem. Eng. Sci. 46(2): 629-636.
D. Cohen and E. B. Gutoff. New York: VCH Wang, Y. 1991. Extrusion of rubber compounds and
Publishers. highly filled thermoplastics through coathanger dies.
Secor, R. B. 1995. Viscoelastic effects in die manifold Intern. Polymer Processing. 6(4): 311-317.
flows. Paper read at 67th Annual Society of Rheology Whitaker, S. 1981. Introduction to Fluid Mechanics.
meeting, 8-12 October 1995, Sacremento, CA. Malabar, Florida: Robert E. Krieger Publishing
Somberger, G., Vergnes, B. and Agassant, J. F. 1986a. Company.
Two directional coextrusion flow of two molten White, S. A., Gotsis, A. D. and Baird, D. G. 1987.
polymers in flat dies. Polymer Eng. Sci. 26(7): 455-461. Review of the entry flow problem: experimental and
Somberger, G., Vergnes, B. and Agassant, J. F. 1986b. numerical. J. Non-Newt. Fluid Mech. 24: 121-160.
Coextrusion flow of two molten polymers between Wilson, G. M. and Khomami, B. 1992. An experimental
parallel plates: non-isothermal computation and investigation of interfacial instabilities in multilayer
experimental study. Polymer Eng. Sci. 26(10): 682-689. flow of viscoelastic fluids. Part I. Incompatible
Stafford, L. 0. 1971. Apparatus for casting multi- polymer systems. J. Non-Newt. Fluid Mech. 45:
layer composite film. US Patent 3,583,032. 355-384.
Su, Y. and Khomami, B. 1992. Interfacial stability of Winter, H. H. 1975. Temperature fields in extruder dies
multilayer viscoelastic fluids in slit and converging with circular, annular, or slit cross-section. Polymer
channel die geometries. J. Rheology. 36(2): 357-387. Eng. Sci. 15(2): 84-89.
Tadrnor, Z., Broyer, E. and Gutfinger, C. 1974. Flow Winter, H. H. and Fritz, H. G. 1986. Design of dies for
Analysis Network (FAN) - a method for solving the extrusion of sheets and annular parisons: the
flow problems in polymer processing. Polymer Eng. distribution problem. Polymer Eng. Sci. 26(8): 543-553.
Sci. 14(9): 660-665.
SLOT COATING Iia
Franz Durst and Hans-Gunter Wagner
lla.l INTRODUCTION
Chapter 11 presents pre-metered coating methods Downstreom Distribution Upstream
Die Lip Chamber Die Lip
for which all the liquid fed into the coating die
by a metering device, e.g., a displacement pump,
is deposited on the substrate. Therefore, the
average wet film thickness H co is predetermined
for a given feed flow rate V·, coating width in
the cross-web direction, J¥, and substrate speed,
U, and is independent of rheological properties of
the coating liquid.
The purpose of this chapter is an explanation
of the physical background of slot coating as a
basis for understanding the advantages, the Coo ted Cooting Wetting Subs rote
~Im Bead Meniscus
drawbacks and the limits of coatability. It is
shown how information on these limits is obtained Figure l1a.l Slot coating.
and how it is used to predict coating windows.
Internal and external design go hand in hand for
a stable coating process and aspects of the
external design are the topic of this chapter. Downstream Distribution Upstream
As the terms slot and extrusion coating suggest, Die Lip Chamber Die Lip
in this process (Figs I 1a.! and 11a.2), the coating

liquid is forced through a coating die on to the
substrate. Within the coating die, a distribution
system, consisting of a distribution chamber and
a feed slot, provides a uniform liquid flow rate
Gap
over the coating width J¥, as explained in detail Clearance
I--~~~HG
in Chapter 10. The feed slot separates the
downstream and upstream die lips. The domain
bounded by the die lips and the substrate is the
coating gap; the width of this domain is known Coated Film Forming Wetting Substrate
as the gap clearance, H G • The hydrodynamic Film Meniscus Meniscus
process of formation of the coated layer occurs

Figure lla.2 Extrusion coating.
within the liquid bridge (also called the coating
London. ISBN 0 412 06481 2.
402 Slot coating
bead) between the coating die and the moving observed in slide coating, which is described in
substrate. Depending on the forces acting on the Chapter 11 b. In knife coating, which is discussed
liquid, different shapes of the liquid bridge can in Chapter 12a, the flow is similar to that within
be observed. The two-dimensional representation the coating gap under the downstream die lip.
of the bead shows two free surfaces: the film The control of the wet film thickness by an
forming meniscus is located downstream of the independent metering device is of advantage in
feed slot and forms the wet film surface; the many applications, in particular for high precision
wetting meniscus is upstream of the feed slot and coating. However, the wet film thickness cannot
wets the substrate. be chosen arbitrarily, because for all coating
In the following we distinguish between slot processes there exists a parameter space of stable
coating and extrusion coating operation. In slot operating conditions outside the limits of which
coating operation, viscous and capillary forces coating defects occur. In slot coating, flow
are dominant. The gap clearance is small, from instabilities can be found which lead to typical
less than twice up to more than ten times the variations of the coating thicknesses as shown
final wet film thickness. The coating gap under in Fig. lla.3. The boundary ofsuccessful operating
the downstream die lip is filled by the liquid and conditions is reached when these instabilities
the upstream coating gap can be partially or exceed the tolerable limit of product quality. The
completely filled. In extrusion coating operation, coating window includes the parameter space of
the liquid bridge has the shape of a short liquid liquid properties, operating conditions and die
curtain. The die lips are usually not wetted by geometry, within which defect-free coatings can
the coating liquid, as shown in Fig. lla.2. The be produced.
gap clearance can be more than 100 times the Slot and extrusion coating are very versatile
wet film thickness. Stable bridging of the coating operations for coating single layers on to a web.
gap is possible owing to the rheological properties With a single external design, a very wide range
of the coating liquid, e.g. viscoelasticity, or to of applications with the viscosity of the coating
external forces, such as magnetic fields which can liquids varying between less than 1 mPa sand
accelerate magnetic liquids in the liquid curtain several thousand Pa s can be handled. Coating
and increase the liquid inertia (Ankenbrand, liquids with even higher viscosity are applied
Herrmann and Wilke 1969) and electrostatic with extrusion coaters. The web velocities can
force fields which can assist in positioning the be less than 0.1 mls or exceed 10 m/s. The minimum
coating bead (Nadeau 1960; Miller and Wheeler wet film thickness is in the order of 10 ~m with
1965). an achievable variation of less than ± 2.5%.
We shall concentrate on slot coating operation Some examples of materials coated with slot
and provide additional references on extrusion coaters are photosensitive materials, e.g., photo-
coating operation as appropriate. Flow processes resist, magnetic suspensions, wax, inks, hot melt
in extrusion coating, e.g. 'heel formation' and the adhesives, silicon, rubber and foams (Aurin 1985;
'teapot' effect, are very similar to those in curtain Campbell 1980; Lippert 1987). Low viscosity
coating, which are explained in Chapter llc, but melts of alloys, metals and organic materials can
because the gap clearance is small, acceleration also be applied (Carpenter and Steen 1990). In
of the coating liquid by gravitational forces, highly viscous melts of plastics and adhesives,
which is typical of curtain coating, is negligible. viscoelastic effects will dominate and these are
Increasing the gap clearance in extrusion coating, typically extrusion coated. Slot coaters can be
when the direction of the falling liquid curtain used for the simultaneous application of multiple
coincides with the earth's gravitational vector, layers, and many patents describe multilayer slot
will lead to a curtain coating operation. coating. However, multilayer slot coating is not
Flow processes similar to those within the very common in industrial applications.
coating gap under the upstream lip can be First we give a selected overview on equipment
Description of equipment 403
~o' I I~
~
-
Swelling and Weeping
al Upstream Meniscus ~ Air Enlralnmenl
~
Too Much Vacuum: -
Swelling and Weeping I
Rivulets I
at Upstream Meniscus Too Fast·
---..... Air En~rainmenl
Low Flow Umil --.....:
IA;~
Barring!
Cross-web
Ribbing
Web Speed
Figure lla.3 Coating window and flow instabilities.
used for slot coating in Section 11a.2. In Section posItioning of the coating die relative to the
11a.3 the physics of slot coating is explained and substrate, is varied to find an ideal design for
Section l1aA focuses on stable operating condi- any application. Fig. 11a.5 shows different
tions. The influence of small disturbances on the geometries of the coating die. The lip geometry
stability of the coating flow is discussed in influences the liquid flow and the pressure
Section 11a.5. Section 11a.6 provides information distribution in the coating gap. In the case of
on practical issues of slot coating and a summary highly viscous coating liquids, the pressure within
in Section l1a.7 closes this chapter. the coating bead can become substantial. The
maximum pressure can be reduced by inclined
lla.2 DESCRIPTION OF EQUIPMENT
The basic elements of slot and extrusion coating
Pulsation
processes are shown in Fig. l1aA. The coating Damper
liquid is pumped from the liquid preparation to Filter
the coating die. The pulsation damper reduces
oscillations induced by the feed system, which
can cause flow instabilities. The filter system
keeps back impurities and large particles to
prevent coating defects such as streaks. Details
on liquid preparation and transport to the coating
die were given in an overview on fluid handling
by Schweizer (1992).
The center piece is the coating die. Many
Wet Film Coating Feed Slot
different die designs are shown in the patent Bead
literature, publications and brochures from die
manufacturers. The external design, in particular Figure lla.4 Basic elements of slot and extrusion
the shape, the inclination of the die lips and the coating.
404 Slot coating
-u -u (a)
(a) (b)
(b)
-u -u
(c) (d)
Vacuum
Box
(c) (d)
Figure tla.5 Various die designs of slot coater.
Figure lla.6 Devices for the pressure control and
multilayer slot coating.
lips (O'Brien 1984) as indicated in Figs lla.5a
and 5b, or by special lip designs as shown by
Kageyama and Yoshida (1986), Figs lla.5c and downstream meniscus as indicated in Fig. 11a.6d
5d, which create sinusoidal pressure variations (Browatzki et al. 1975). This device has to be
within the coating bead. A high manufacturing designed very carefully as the air flow, which is
quality of the die and the application roll is necessary to sustain the over-pressure, can have
necessary as the gap clearance can be very small. a damaging influence on surface uniformityoflow-
Die lips are in general planar, mainly owing to viscosity wet films. A further method to apply
ease of manufacturing. The deviations in flatness the pressure difference was shown by Cameron,
of the die lips and roundness of the coating roller Alfred and Hills (1968), where the pressure of the
should be less than ± 1....m for gap clearances coating liquid under the upstream lip can be
less than 50 ....m. According to Fahrni and controlled by circulating the liquid as shown in
Zimmermann (1978), the curvature of the edges Fig. l1a.6a. Similar stabilizing effects on the
should be less than 50 ....m where the meniscus coating bead can be obtained by an electrostatic
should pin. assist or by a magnetic assist in the case of a
Beguin's (1954) suggestion to apply a pressure magnetic fluid.
difference between the upstream and downstream Several examples of the many patents on
menisci was a very important contribution to multilayer slot coating are shown in Figs lla.6a-6c
increasing the coating speed in an operation and lla.7b. The liquid layers can be combined
mode where the gap clearance is greater than within the feed slot as Fig. lla.6b demonstrates
twice the wet film thickness. It has revolutionized (Russel 1956; Jackson 1976) or within the coating
an entire industry. The necessary pressure gap (Cameron, Alfred and Hills 1986), see Figs
difference, which can reach values of 3000 Pa, lla.6a and lla.7b. Combinations of slide and
can be achieved with the help of a vacuum box slot coating principles were proposed as indicated
located at the upstream meniscus. An example in Fig. lla.6c by Russel (1956) and Jackson (1976).
is given in Fig. lla.6c (Woodworth, Winkler and In general, the web will be supported by a
Jackson 1982). The same effect is produced by a rigid or rubber-covered roll, see Figs lla.6a-6d,
box which creates an over-pressure at the or by a flat surface. It is also possible to push
Successful steady state operation: forces, principles 405
the coating head towards a tensioned web. The mechanisms for modifying the clearance of the
coating liquid should act as a lubricating layer feed slot to provide the desired fluid distribution
and prevent the contact of web and coating die in the cross-web direction. The control of the
as shown in Fig.lla.7a (Tanaka and Noda 1988). cross-web flow uniformity is a topic of internal
Such elasto-hydrodynamic coating processes are design. Neverthless, variations of the clearance
described in Chapter 12d. Slot coaters can be of the feed slot have to be considered as they
used to coat both surfaces of the web simulta- can change the flow field within the coating bead.
neously as shown in Fig. lla.7b (Ishiwata and Owing to the narrow coating gap in slot
Nagai 1970), and in this way one coating operation coating, great care has to be taken in properly
provides the support for the other. positioning the coating die relative to the
Many commercially available slot coaters have application roll. Exact repeatability of the
positioning is required. The support must allow
an in and out movement of the die normal to
the surface of the roll, with two adjustments.
First, a coarse in/out adjustment with a movement
over a distance up to 200 mm provides fast access
for cleaning of the die lips and prevents damage
while a splice of the substrate passes the coating
station. Second, a fine in/out adjustment for
perfect alignment provides independent adjust-
ment of the gap clearance at both ends of the
die, monitored by micrometer gauges. Two
concepts are used for further fine adjustment of
the coating gap, namely a straight movement
normal to the in/out system and an adjustment
of the angle of attack of the die lips. A favorable
center of the angle of attack movement is the
downstream edge of the downstream die lip
where the film forming meniscus pins.
11a.3 SUCCESSFUL STEADY STATE

OPERATION: FORCES, PRINCIPLES
tta.3.t QUALITATIVE DESCRIPTION OF THE

FLOW
To maintain a stable coating process, the
interaction of inertial, viscous, capillary and
external forces acting on the coating bead must
be properly managed. Various parameters which
determine the position and shape of the coating
bead and the flow processes within it are listed
below.
Fluid and material properties

• The wetting behavior of the coating fluid on
Figure lla.7 Tensioned web slot coating. the web and the coating die. The static and
406 Slot coating
dynamic contact angles, () and D, of the fluid • External forces, e.g. direction ofgravity, electric
with the die lips and the substrate, respectively, or magnetic fields.
are shown in Fig. 11a.8. Detailed information • Gas pressure at the upstream and downstream
on the complexity of the wetting behavior and menisci, P u and Po, respectively.
contact angles is given in Chapter 3.
• The rheological properties of the coating The dimensionless numbers of relevance for
fluid, which very often show non-Newtonian the slot coating process are in general the Reynolds
behavior. The Newtonian viscosity is 11. number, Re, the capillary number, Ca, the Bond
• The surface tension, a, of the coating fluid. number, Bo, and the dimensionless pressures, Pi:
• The density, p, of the coating fluid.
pUH oo JJ.U
• Properties of the substrate, e.g. deformability, Re=--· Ca=-;
dewatering of the coating fluid or electrostatic JJ. ' a
surface charges. (11a.t)
Geometric parameters
These clarify the importance of the various
• Design variables of the coating die, which forces. Chapter 8 provides useful information on
define shape, dimensions, inclination and these dimensionless quantities. All geometric
position of the die lips and feed slot. values are made dimensionless by the wet film
• Design variables, which define shape of the thickness, H 00.
web, gap clearance and radius of the application In nearly all practical applications of slot
roll. coating, the Bond number is much smaller than
unity, implying that the influence of gravity is
negligible and the coating die can be operated
Operating parameters
in any position relative to the gravity field. The
• Web speed, U. capillary number is often greater than unity at
• Wet film thickness, H 00' given by web speed, high coating speeds and hence the effect of
U, feed flow rate, V*, and coating width, J¥. capillary forces on the coating flow is minor. Fig.
Downstream Upstream
Die Lip Die Lip
L Lu
H.
- - - lJ
Figure 11a.8 Geometric parameters of slot coater.

IIa.8 sketches the coating bead and shows the slot, see Fig. 11.9c. The capillary pressure drop
geometric parameters used. across the free surfaces and the pressure drop
A qualitative description of the change in the caused by friction at the die lip are in equilibrium
coating bead due to variations of the feed flow with the pressure drop caused by the viscous
rate is shown in Fig. Ila.9. The speed of the forces of the moving substrate. Increasing the
substrate is constant. If the flow rate is too low, flow rate further increases the pressure drop
the liquid forced out of the feed slot cannot within the downstream coating gap. Hence the
maintain a stable liquid bridge across the web pressure level at the feed slot increases. The liquid
and single drops or rivulets will form on the web, is pushed into the upstream coating gap against
as shown in Figs lla.9a and 9b. The pulling the direction of the web motion. The wetting
action of the moving web on the liquid is stronger meniscus moves into the rear gap and the
than the stabilizing effects of liquid inertia and upstream die lip will be wetted as indicated in
capillary forces. Disturbances, present in any Figs lla.9d and ge. Figure lla.9f shows the
coating process and propagating faster than the coating bead when the limit of stable coating
fluid velocity, lead to a rupture of the liquid conditions is reached after a further increase in
bridge. Therefore, the liquid forms several bridges the flow rate. The pressure drop within the
in the cross-web direction, each being stable downstream coating gap also increases. The
owing to the higher flow rate within it. A detailed pressure at the feed slot is so high that it cannot
treatment of the stability of curtains is given in be reduced sufficiently by the viscous forces
Chapter 11 c. Increasing the flow rate will lead within the upstream coating gap. The capillary
to a stable situation. In extrusion coating, a forces at the upstream meniscus cannot compen-
coating bead will form as indicated in Fig. Ila.2. sate for the pressure difference between the liquid
The cohesive strength of the liquid or external and the gas. The upstream meniscus swells and
magnetic or electrostatic forces compensate for liquid wets the upstream die shoulder of the
the viscous force exerted by the pulling action upstream die lip. The wetting meniscus can no
of the web. In slot coating, the coating liquid longer bridge the gap and the coating liquid spills
will fill the gap under the downstream lip, the on to the web. If the downstream meniscus does
free surface of the wet film is formed at the not pin at the edge of the downstream die lip,
downstream edge of the lip and the wetting the film-forming meniscus climbs up the die
meniscus bridges the gap at the exit of the feed shoulder.
(a) (b) (e)
(e)
~(f)
-u
i
Figure lla.9 Influence of the flow rate on the shape ofthe coating bead. The flow rate is increased from (a) to (I).
408 Slot coating
lla.3.2 ANALYTICAL DESCRIPTION OF SLOT and Pu, at the downstream and upstream menisci.
COATING The pressure balance is given by:
The qualitative picture of the interplay of forces Apou = ApDl + APlZ + ApZ3 + Ap3U (t ta.2)
within the coating bead can be quantified by
Ruschak (1976) analyzed the limiting flow
analytical expressions. In this section these
conditions within the coating bead of a slot
expressions will be used to explain the pressure
coater when capillary forces are dominant and
distribution, the location of the menisci and the
Higgins and Scriven (1980) extended this work
occurrence of recirculation zones in a slot coater
to include viscous and capillary forces. In the
whereas the analysis of the same equation~
following, their considerations are used to
prov!~es invaluable information about the op-
determine the parameters which influence the
erabIlIty of a slot coater presented in Section
fr~e ~urfaces an~ the flow of a Newtonian liquid
11 a.4. The flow processes within the coating bead
wlthm the coatmg bead in such a way that a
are divided into zones with different flow regimes.
stable and steady coating process is possible.
The flow within the first zone at the film-
It is assumed that the Bond number is much
forming meniscus rearranges from the flow within
smaller than unity and inertial effects are negligible.
the coating gap into the fluid at rest within the
The. fre~ surfaces will then be influenced only by
wet film. The pressure difference, AP Ol ' between
cap~llanty and viscosity. If the capillary number
the pressure within the fluid, Pl' and the gas
Ca IS small, no dynamic forces are present and
pressure, Po' is the capillary pressure.
the capillary pressure at the free surfaces can be
Figure l1a.l indicates the location of the relevant
calculated as for a static drop. The wetting and
pressures. The flow processes within the down-
film forming menisci are curved and their shape
stream coating gap, the second zone, are caused
can be approximated in the simplest form as
by the d~a~ of the moving substrate and possibly
cylinders. The curvature of the surfaces induces
by a dnvmg pressure gradient. The pressure
a capillary pressure, which is calculated from the
gradient is given by the pressure difference , AP 12'
Young- Laplace equation. This is explained in
between the pressure at the film forming zone,
Chapter 2.
Pl' and the pressure at the feed slot, pz. At the
. The capillary pressure has to equal the pressure
feed slot the fluid turns and viscous shear, normal
dIfference, Ap3U' between the pressure within the
stresses and sometimes inertia determine the flow
fluid; P3' and the gas pressure, Pu, at the wetting
processes. If the upstream coating gap is filled,
memscus. The curvature of the free surface can
then a further zone exists, within which a pressure
be calculated from geometrical considerations
gradient driven flow opposes the drag flow of
for given static and dynamic contact angles {}
the web so that the net flow rate within the zone
and v, respectively. The actual value of these
is zero. The driving gradient is the pressure
contact angles is strongly related to the history
difference, ApZ3' between the pressure at the feed
ofwetting, as discussed in Chapter 3. The position
slot, pz, and the pressure at the upstream meniscus.
of the wetting meniscus depends on the pressure
At the wetting meniscus, the last zone, the fluid
distribution in the coating bead. All possible
flowing against the direction of the web has to
shapes of the wetting meniscus are given by the
turn at the free surface and is then accelerated
following equation:
by the web in the direction of the web. The
capillary pressure at the upstream meniscus is A cos{}* + cos V
the pressure difference, Ap3U' between the pressure LlP3U = (11a.3)
hG •u
within the fluid, P3' and the gas pressure, pu. The
so called equilibrium state requires that the The angle {}*, included by the free surface and
pressure differences across the zones balance the the horizontal at the static contact line, is indicated
pressure difference, Apou, of the gas pressure, Po in Fig. 1ta.8, where {}* = {} + f3 within the
Iii
upstream coating gap. The resulting pressure shown in Fig. 11 a.8.
difference is positive when the meniscus is curved
into the coating bead and negative when the 1 t
meniscus bulges out.
D2 j = t; 0 h?(x) dx;
t; I'i h~(x)
An equation for the pressure difference, Ap01' (tta.7)
at the film forming meniscus can be derived by t t
D3 j = 0 dx
the method of matched expansions (Wilson 1982).
The resulting equation is valid for small capillary The influence of the flow at the feed slot and
numbers (Ca« 1). The influence of inertia is the contributions from the areas where the flow
neglected (Re « 1) and the terms ofthe first-order is not fully one dimensional can be neglected if
approximation do not reflect the influence of the length of the die lips is sufficiently greater
gravity (Bo « 1). For capillary numbers greater than the gap clearance. The numerical calculation
than unity viscous forces will dominate within presented in the following section demonstrates
the coating bead and the pressure difference, that it takes less than one gap clearance in the
Ap01' can be set to zero: flow direction to reach the fully developed flow
condition.
1.3376. Ca2f3 - 0.1551· Ca; Ca« 1 In the limiting case of small capillary numbers,
Ap01 = {0 Ca> 1 the pressure differences Ap12 and Ap23 are
negligible in comparison with the pressure
(1Ia.4) differences at the menisci. The slot coater operates
in the capillary mode. The pressure difference at
The flow processes within the downstream and the upstream meniscus is always positive. Hence,
upstream coating gap are influenced by the if the gas pressure difference between upstream
geometry of the coating gap. If the gap clearance and downstream menisci is zero, then the surface
is small compared with the length of the die lips, of the wetting meniscus must curve into the
then the flow is one-dimensional and the influence coating bead to satisfy equation (lta.2). Reducing
of inlet and boundary effects on the pressure the gas pressure at the wetting meniscus by a
difference can be neglected. The flow can be vacuum box has a pulling action on the wetting
expressed by the Reynolds lubrication approxi- meniscus and it may bulge out.
mation as a superposition ofCouette and Poiseuille A common situation in slot coating is the
flows. Details on the lubrication approximation second limiting case, where the capillary numbers
are given in Chapter 8. The net flow rate within are much greater than unity. The slot coater
the downstream gap is equal to the premetered operates in the viscous mode and the contributions
flow of the coating liquid. The viscous pressure of the pressure differences at the menisci are
drop is negligible. The pressure difference Ap23 within
the upstream coating gap is always positive and
Ap12 = Ca· 6 ·lo (D2 o - 2· D3 o ) (1Ia.5) independent of the flow rate. It increases with
increasing web speed, increasing viscosity and
The contributions of Couette and Poiseuille decreasing gap clearance. If the pressure difference
flows under the upstream die lip are identical Apou is zero, then equation (1Ia.2) can only be
and the net flow rate is zero. The pressure drop satisfied if the pressure difference Ap 12 within the
follows as downstream coating bead is negative. From
equation (1Ia.5), it can be concluded that the
Ap23 = Ca· 6· lu . D2 u (1Ia.6) pressure gradient in the flow direction is always
negative if the gap clearance is less than twice
The terms D2 and D3 take into account the the wet film thickness. It might be assumed now
variable clearance of the coating gap, hi (X), as that a sufficiently high pressure difference in the
410 Slot coating
gas pressures at the free surfaces allows arbitrarily 300 I!m can reach values of up to 200 Pa, and at
high values of the gap clearance. However, that a gap clearance of 100 I!m it is up to 600 Pa. It
is not true, as there exists a 'low flow limit', which is therefore reasonable to neglect the influence
is explained in Section lla.4. of the gas boundary in the balance of forces.
In many applications, viscous and capillary A similar analytical approach can be used for
forces have to be considered and operating the design of coaters for axisymmetric substrates
conditions are possible where a positive pressure such as cables, wires and fibers. Analytical
difference !1P12 within the downstream coating expressions for the axisymmetrical case were
gap occurs. The border between positive and presented by Lange et al. (1991). The extension
negative pressure differences within the down- to the case oftwo-layer slot coating was described
stream coating gap is the pure Couette flow by Scanlan and Scriven (1990) and Cohen and
condition, where the gap clearance is twice the Scriven (1992). The resulting equation system
wet film thickness and no pressure gradient provides information on the existence of recir-
exists. If the dimensionless gap clearance is culation zones and the position of the interface
greater than 2, then the Poiseuille flow opposes between the two coating liquids, which is an
the Couette flow and the pressure in the flow important issue in this type of coating operation.
direction increases. Secor (1991) and Strenger and Secor (1991)
The onset of recirculating flow within the included the influence ofthe viscoelastic behavior
downstream coating gap is given by a velocity of the coating liquid in the simplified equations.
profile which has a zero velocity gradient in the They considered the flow within the downstream
wall normal direction at the wall of the die lip. coating gap, which is operated at less than twice
By superposing the velocity profiles of Couette the wet film thickness. They found that the elastic
and Poiseuille flows, the velocity gradients of stress counteracts the negative pressure gradient
which at the wall are of the same magnitude with in the flow direction at high shear rates, which
opposite sign, it can be deduced that there is no increases the possibility of ribbing instability.
recirculating flow under the downstream die lip
if the gap clearance is smaller than three times
the wet film thickness. If the gap clearance is 11a.3.3 NUMERICAL SIMULAnONS OF SLOT
greater than this a recirculation zone can be found. COATING
These equations reflect the physics of slot The analytical description of the flow within slot
coating well and provide a very good qualitative coaters has its limits and numerical calculations
picture. One might wonder why the effects of the are a powerful tool for simulating the flow
gas boundary layer at the wetting meniscus have processes of slot coating. Chapter 9 elaborates
not been included in the balance of forces. The on the details of numerical simulations. Detailed
maximum pressure which can be exerted on the information on flow processes within regions
wetting meniscus by the gas flow is the stagnation which are experimentally accessible only with
pressure, which is calculated as great difficulty were provided, for example, by
Sartor (1990) and Durst et al. (1988, 1991).
(Ila.8) Velocities, pressure, shape and location of the
menisci and size and location of recirculation
zones within the coating bead can be predicted.
This pressure is 65 Pa at a very high web speed Such programs are further useful for studying
of 10m/s and reduces to 16.25 Pa at a high web the influence of parameter variations on the
speed of 5 m/s. The density p of the gas is coating flow and its stability. In principle, all
1.3 kg/m 3 . The capillary pressure at the wetting acting forces, properties of the fluid and the
meniscus of a fluid with a surface tension of wetting behavior of the coater materials can be
0.03 N/m and at a very high gap clearance of included in these calculations. However, all
calculations published in the literature contain upstream die lip. The size of the coating bead is
some simplifications. Major shortcomings are determined by the pressure drop within the
the treatment of the wetting behavior and upstream and downstream coating gap. Increasing
simulations of three-dimensional flows with free the upstream gap clearance reduces the pressure
surfaces, which are still difficult and laborious. drop and the wetted length of the upstream die
Figure lla.!0 depicts a numerical grid employed lip increases. Within the coating bead the
in the calculations, profiles of the velocity distribution ofthe streamlines changes according
component parallel to the web and a streamline to the gap clearance.
plot of the coating bead. The flow within the The pressure distribution within the coating
downstream and upstream coating gap is one- bead is given by Fig. 11 a.12 for four of the slot
dimensional and fully developed, as shown by configurations. The pressure at the end of the
the unchanged velocity profiles and the parallel calculation domain in the downstream direction
streamlines. At the entrance and exit zones of is identical to the ambient pressure at the
the coating gap the fluid needs only about one downstream free surface (p = 0). The pressure
gap clearance to assume the fully developed state. decreases in the upstream direction owing to the
The arrows point to the stagnation points, where curvature of the free meniscus. Within the
the fluid is at rest, which indicate the end-points downstream coating gap the pressure gradient
of a streamline which separates the recirculation is constant, which indicates the fully developed
zone from the main flow. flow regime. Across the exit of the feed slot the
Figure lla.ll shows various geometries of the pressure is nearly constant, although there are
coating gap under the same flow conditions. The strong changes in the flow direction. Within the
upstream mensicus is free to move along the upstream coating gap the pressure gradient is
Figure 11a.l0 Numerical simulation of slot coating: numerical grid, velocities and streamlines. Re = 9.3,
Ca = 0.05, lipov = 1.6, e = pinned, v = 110°, hG •o = hG •v = 5, hF = 10, 10 = Iv = 20.
412 Slot coating
0.
(a) 8
J
"t:l
'"
---
v 6
..0
g' 4
I
iiio 2
(b~
u
v 0
.c
... - 2
c
5 -4
~ -6
(c) ...::>v -8
---
Cfl
Cfl
'" -10
... ~b
-p~!'ZZ*!'ZZ~~m:z:;t~~w~=zZi~
AIL
c.. _I 2
---Jz
-20 0,0 20 40 60 80
(d)
X - Position
---J
(e)
J &§W Figure 11a.12 Pressure profiles within the coating
lib
bead; for parameters see caption to Fig. lla.!1.
(f)
I
---
& 5??1 Without vacuum the upstream meniscus is located
close to the feed slot. The coating liquid is pulled
(g)
--- k 4 further upstream by an increase in the vacuum.

The meniscus will pin when the edge of the
L
upstream die lip is reached and changes its
curvature to compensate for the pressure difference.
(h)
J The limiting curve of the stable positions of the
upstream meniscus can be determined with this
kind of numerical simulation (Sartor 1990). The
Figure 11a.11 Numerical simulation of slot coating:

various die designs. Re = 10, Ca = 0.05, I D = Iu = 25,
hF = 10, e = 60°, v = 110°, ~PUD = 10. (a) h G.D =
h G•u = 1.75; (b) h G•D = h G.u = 5; (c) hG,D = hG.u = 7.5;
(d) hG,D = hG•u = 1.75, f3D = 6°; (e) hG,D = 1.75, hG•u =
4.37, f3D = 6°, f3u = _6°; (f) hG •D = 1.75, hG •u = 4.37, • l\\
I
f3u = - 6°; (g) hG.D = hG,u = 1.75, f3D = f3u = 6°; A 0) l\~ ,,~
(h) hG.D = 2.5, hG,u = 4.
7 / I!hj
u
again constant. The upstream meniscus has to

~
• • increasing
•
take a position where the vacuum pressure
applied at the free surface of the upstream
/~A;;; vacuum
meniscus is in equilibrium with the capillary

pressure of the free surface and the pressure Figure 11a.13 Change in shape of the coating bead
within the coating bead. due to vacuum, ~PDU increases from 3 to 9 in
Different vacuum levels change the location increments of 1. Re = 10, Ca = 0.1, ID = lu = 10,
ofthe upstream meniscus as shown in Fig. lla.!3. e
hG,D = hG.u = hF = 5, = 60°, v = 110°.
same method can be used to find the limiting displace the liquid flowing within the coating
positions ofthe downstream meniscus to determine gap. A vortex will form within the feed slot which
the low flow limit, which is explained in Section blocks the exit of the latter and provides a
lla.4. higher momentum to the liquid leaving the feed
Figure lla.14 shows the flow field at the exit slot. This vortex will be completely within the
of the feed slot where the upstream coating gap feed slot if the clearance of the coating gap is
is filled by the coating liquid. The Reynolds hG < 2. For higher values of hG the vortex
number is constant and the clearance of the will intrude into the coating gap and combine
coating gap and the feed slot are varied. From with the recirculation zone within the down-
analytical expressions in the previous section it stream coating gap. The geometric con-
could be deduced that recirculations within the figurations which are believed to be optimum
downstream coating gap can only be observed show no additional regions of recirculating
if the dimensionless coating gap clearance is liquid.
greater than 3. The numerical simulation in Fig. There is still a definite need to improve the
lla.14 points out that if the dimensionless numerical predictions. More powerful computer
clearance, hF , of the feed slot is smaller than the programs are needed which can handle efficiently
dimensionless clearance, hG , of the coating gap, complex three-dimensional coating flows. A
then additional vortices will appear at the die drawback of numerical simulations is that often
lips upstream and downstream of the feed slot realistic boundary conditions, e.g. contact angles,
even at clearances of the coating gap hG < 3. are unknown, and values have to be assumed.
These vortices are caused by the high momen- Sensitivity studies are required to determine the
tum of the coating liquid exiting the feed slot. influence of the specific boundary condition on
The flow processes are similar to a jet emanating the flow processes. Nevertheless, the accuracy of
from a nozzle. If the clearance of the feed slot the calculations provided so far is sufficient in
is greater than about twice the clearance of most cases for engineering purposes and provides
the coating gap, then the mean momentum of a deeper understanding of the physics underlying
the liquid exiting the feed slot is too small to thin film coating processes.
Figure lla.14 Flow processes within the coating bead at the feed slot. Re = 10.
414 Slot coating
11a.3.4 EXPERIMENTAL STUDIES OF SLOT
AND EXTRUSION COATING
The easiest way to obtain information on the

coating process is by visual observation of the
film quality, i.e. the occurrence of undesired
surface structures. As an indirect method, it
provides information only on the effects of the
flow processes within the coating bead. The
experimental methods utilized to study the flow
within the coating bead range from these simple (a)
investigations to very sophisticated and time
consuming flow visualization experiments (Sartor
1990). The benefits of sophisticated experimental
investigations and suitable experimental methods
are discussed in Chapter 7.
Durst et al. (1988) monitored the coating bead
of a slot coater in side view with a video camera.
They used a scaled-up slot coater to provide easy
optical access to the coating bead. Fluids of
higher viscosity permitted operating conditions
similar to those of the real size slot coater, where
the coating gap is very narrow. The flow was
visualized with aluminum flakes as tracers, and
dynamic shape changes of the coating bead due
to variations of the web speed, flow rate and gap
clearance were recorded. The qualitative com- Figure lla.15 Flow visualization ofextrusion coating
parison of the flow processes with numerical in the presence of magnetic forces (Gilbert and Eckel,
simulations showed good agreement, even though 1992). Re = 0.5, Ca = 3.3, St mBg = 2. (a) Magnet is
opposite to die, (b) magnet is upstream of die.
the capillary numbers differed between the
scaled-up experiment and the computations of
the real-size slot coater. Grader (1985) investigated in the direction normal to the web and assists
the flow processes within the upstream coating in positioning the wetting line. The pulling action
gap. He monitored the movement of the wetting of the magnetic force is demonstrated in Fig.
meniscus within the gap, and found good 11 a.15b, where the magnet is positioned further
agreement between experimental data and nu- upstream. A 'heel' is formed as the magnetic fluid
merical simulation. is pulled in a direction opposite to the web motion.
An example of flow visualization of extrusion Colored dyes and hydrogen bubbles were used
coating is given in Fig. lla.15. Gilbert and Eckel to visualize the flow patterns in slot coating
(1992) looked into the shape change of the liquid (Sartor 1990) and in two-layer slot coating (Cohen,
bridge of a magnetic fluid within the force field Suszynski and Scriven 1992). The experiments
of a permanent magnet. Fig. 11 a.15a shows the reveal interesting information on eddy motions
flow when the magnet is positioned underneath and yielded vortex maps as shown in Fig. 11 a.16.
the web immediately below the exit of the feed The figures indicate the different flow processes
slot. The shape of the liquid bridge is similar to with recirculation zones within the coating bead.
that for curtain coating presented in Chapter Flow structures which can be found in single-layer
Hc. The permanent magnet accelerates the fluid slot coating are given by combinations of the
DOWNSTREAM INTERLAYER UPSTREAM
DIE EDGE PINNING LOCATED DOWNSTREAM MENISCUS FREE TO MOVE
~~
/1.. \\71.11
I~ ~ I
PINNING WITH RECIRCULATION FREE TO MOVE ON LIP AIR ENTRAINMENT LIMIT
WETTING OF DIE SHOULDER LOCATED UPSTREAM WETTING OF DIE SHOULDER
WETTING WITH RECIRCULATION LOCATED UPSTREAM WEEPING

WITH RECIRCULAT10N
Figure lla.16 Microvortex map of two-layer slot coating (Cohen, Suszynski and Scriven, 1992).
flow charts in the left- and right-hand columns effects change the pressure field within the coating
of Fig. lla.16. Flow processes in two-layer slot gap. Owing to this, the ribbing instability occurs
coating are given by combinations of flow charts and shrinks the size of the coating window
in the left-hand, middle, and right-hand columns. dramatically. Tanaka and Yasui (1992) investigated
The various combinations show how complex the influence of viscoelasticity in slot and extrusion
flow conditions can exist within the coating bead. coating. They observed that in extrusion coating
It is of practical interest to know the optimum the spacing between the thickness variations of
flow conditions and flows with no vortices are the liquid layer is increased by increasing the
desired. These are depicted by the figures in the shear rate within the feed slot and is decreased
topmost row. The flow charts show that there by increasing the elongation rate within the
are several admissible locations of the interface liquid bridge. The minimum spacing of the
between the two layers. In general, mixing of the variations was the clearance of the feed slot, hF •
two layers has to be avoided and therefore stable The wetting meniscus in slot coating showed
positioning of the separation line at corners is similar curvature variations in the cross-web
desired. direction. By increasing the shear rate within the
Many coating fluids show a non-Newtonian feed slot, the spacing increased. The spacing
rheological behavior; viscoelastic fluids in par- decreased by a combination of smaller curvature
ticular can be difficult to handle. Strenger and variations when the meniscus moved further
Secor (1991) found that in slot coating viscoelastic upstream within the upstream coating gap. All
416 Slot coating
these effects were amplified by increasing the curvature of the free surface, which opposes the
viscoelasticity of the fluid. change in the pressure difference across the
The complexity ofthe flow studies may require memscus.
specially designed coating equipment with glass The limiting values of the curvature at the
and Plexiglas rollers, through which flow visual- downstream meniscus are derived from the
ization and optical velocity measurements can following considerations. The minimum value of
be performed. Steps in that direction were taken r m at the downstream meniscus is determined by
by Cohen, Suszinski and Scriven (1992). the smallest radius ofcurvature which can bridge
the coating gap for a given static wetting angle
(). The smallest radius is found for a static wetting
lla.4 STABLE COATING OPERATIONS angle of zero. There exists no limit for the
Successful coating requires that the operating maximum value of the radius r m' The stability
parameters lie within the parameter space of condition can be fulfilled for all values of () as
stable coating conditions. This area of study, long as the meniscus is not pinned at the
two-dimensional flow processes is called the downstream edge of the die lip and there is no
coating or operability window. The operating restriction on the height, he' of the wetting line
conditions have to be chosen in such a way that climbing up the die shoulder (see Fig. l1a.8 for
fluctuations in the operating parameters, which details). The limiting cases at the film forming
are present in any process, will not cause the meniscus can therefore be given by
operating point to leave the stable operation
area. It is useful to define a further and more hG,D - 1 hG,D - 1 + he 1t
----'=-....,...,-~r
1 + cos ()*
~
1 - sin ()*
, o ~ ()* <"2
restrictive quality window, which is limited by m
operating parameters where the quality standard (lla.9)

h - 1 1t
of the specific product is not achieved owing to G,D ~ r ~ 00' - ~ ()* < 1t
1+ cos ()* "" m "" , 2 ""
undesired eddies, potentially nonuniform wetting
lines or high sensitivity to external disturbances. If the wetting line climbs up the downstream die
The theoretical description of the flow with shoulder, then flow situations will be observed
asymptotic approximation equations presented as depicted in Fig. lla.16, where a vortex forms
in Section lla.3 is useful for efficient calculation at the static wetting line of the downstream
of coating windows. This analysis is restricted, meniscus. This is an undesired flow situation as
as it allows one only to exclude the parameter the wetting line which is not pinned may no
space where the coating flow is unstable and longer be straight and this increases the possibility
stable coating conditions definitely cannot be ofa nonuniform wet film in the cross-web direction.
achieved. This does not necessarily imply that At the upstream meniscus the same static
within the predicted coating window any stable stability consideration as for the film forming
coating condition exists. In this section, the meniscus can be made. The radius of curvature
limiting parameters of the coating window are of the upstream meniscus is given by
determined using equations (1Ia.l) to (1Ia.?) and
the inequalities contained in these equations are hGu+he
r = ---=-,=-----"-- (lla.l0)
discussed with respect to static equilibrium m cos ()* + cos u
conditions at the menisci.
A stable static equilibrium condition at the The wetting line, as shown in Fig. 11 a.8, may
meniscus of the free surfaces requires that the climb up the die shoulder by an amount he. Three
capillary pressure compensates the pressure cases of the meniscus location are distinguished,
difference at the meniscus (.::\PDl and .::\P3D) and as follows.
that any change in this pressure difference leads At the downstream edge of the upstream die
to a new stable equilibrium by changing the lip, the static contact angle can take the following
Stable coating operations 417
values:
() - n + p + Y1 < ()* < () + P (11 a.ll) 20 ...-.:---.;---.;-----..;:-.........,.........,---.--~--_____,
Stability requires that a positive pressure change

is compensated by a decrease in the radius of ~
curvature and this is equivalent to an increase ~15

<l
in the angle ()*. The minimum value of () is zero.
The value of ()* can increase and the meniscus
will be pinned at the edge until ()* = () + n - Yl'
If ()* becomes larger, then the meniscus will climb
up the die shoulder. Stable conditions are possible
for values of ()* between 0 and n and he = O. Cl 5
Within the coating gap, the angle ()* can take (l)
s....
on only one value, which is given by ()* = () + p. ::l
rn
If the die lips are inclined, stable conditions can ~ 0
be achieved for I-
0..
n
p< 0 and 0 ~ ()* < 2"
0.3 0.5 0.8 1.0 1.3 1.5
(lla.12) Capillary Number Ca
n
p> 0 and 2" < ()* ~ n (a)
If the inclination angle p of the die lip is zero,

no stable position exists within the coating gap.
The wetting meniscus moves freely within the
gap and has to jump to an edge of the die lip in
order to find a new stable position when the :::>
pressure difference changes. In reality, viscous ci 15

<l
forces within the coating bead will compensate
for the pressure fluctuations and stabilize the
position of the wetting meniscus. A positive
pressure difference, ~.p3V' fluctuation increases
the wetting length under the die lip and the shape
of the meniscus is unchanged. A negative value Cl 5
of ~P3V decreases the wetting length. (l)

I-
::l
At the upstream edge, the angle ()* can take rn
rn
the values (l) a
I-
0..
() + P< ()* < () + P+ n - Y2 (lla.13)
and stable conditions are possible only for values - 5 -f-T>.,....,..;.,....,.~;;:;::;=;:::TT-r-+-r-rt=in-..,..,-"'-.--r-r1
of ()* between zero and n and he = O. 0.0 0.3 0.5 0.8 1.0 1.3 1.5
Figure lla.17 shows two typical examples of Capillary Number Ca
(b)
coating windows with vacuum at the upstream
meniscus calculated from the inequalities contained
in equations (lla.l) to (lla.7) and equations Figure lla.17 Coating windows: influence of gap
(lla.9) to (lla.13). The abscissa gives the coating clearance. ID = Iv = 20, v = equation lla.16. (a)
speed expressed by the capillary number. The hO •D = ho .u = 2.5; (b) ho .D = ho .u = 1.75.
418 Slot coating
ordinate shows the difference between upstream the coating window periodic variation of the film
and downstream gas pressures. The static wetting thickness in the web direction can be observed.
angle, e, is zero and the dynamic wetting angle, This kind of coating defect is also called barring
v, is calculated from equation (lla.14) (Guthoff and is caused by periodic changes of the location
and Kendric 1982): of the upstream menisci. In this manner, the
l) = 541· UO. 22 .(TO. I I • J.l0 . 18 (11a.14)
volume of the coating bead fluctuates and,
assuming a constant feed flow rate, the film
The values are to be input in the dimensions thickness will change according to the volume
of kg, m and s. Three limiting conditions are variation.
calculated for each capillary number. They are At the lower or starving limit, the pressure
distinguished by the location of the upstream drop within the coating bead due to viscous
meniscus. The lowest line in the graph represents forces is too high to be balanced by the capillary
the flow where the upstream die lip is not wetted forces of the menisci and ribbing of the wet film
and the upstream meniscus pins at the edge of thickness occurs (cf Ruschak (1985». These
the die lip. This is also called the starving periodic thickness variations in the cross-web
condition. The broken line shows the case where direction are due to a nonuniform curvature of
the upstream meniscus pins at the upstream edge the menisci in the cross-web direction, by which
and has its maximum value of curvature into the the coating bead tries to compensate for the
gap. The upper line shows the condition where viscous pressure drop. These ribbing patterns
the meniscus bulges out and assumes its maximum will transform into rivulets, with uncoated streaks
curvature. The pressure difference between the between them, if the pressure inequality increases.
broken line and the upper line describes the At the upper border of the coating window,
influence of the capillary action at the upstream swelling and weeping of the meniscus can be
meniscus on the coating window. The difference found. If the limit point given by equation
between the broken line and minimum is the (lla.13) is exceeded, swelling of the meniscus
pressure drop due to viscous forces within the occurs, which can lead to weeping, i.e. the
upstream coating gap. With increasing capillary discharge ofcoating liquid from the coating bead.
number (web speed) the pressure drop within the This causes thickness variations in the cross-web
upstream coating gap increases and therefore the direction as not all liquid supplied by the die is
height of the coating window increases. coated on to the web at the same time.
Figure lla.17b demonstrates an important The dotted line in Fig. 11a.17a indicates a limit
characteristic of the operating behavior of a slot of stable operation, the low flow limit, which can
coater. If the coating gap is smaller than twice be derived from equation (1Ia.3). Again it is
the wet film thickness, then the lowest curve assumed that the meniscus is ofcylindrical shape.
showing the lower stability limit has a negative The maximum pressure difference ApDI is found
slope. No vacuum is necessary at all web speeds, for the smallest meniscus radius (compare equation
whereas in Fig. lla.17a vacuum is necessary if (11a.9» which bridges the minimum gap clearance
the capillary number Ca exceeds a certain value. of the downstream coating bead. For reasons of
With increasing web speed (Ca) the vacuum level simplicity the second term equation (lla.3) is
necessary to maintain a stable coating condition e
omitted and the static contact angle is set to
increases. The size of the coating window, zero. The capillary number, which indicates the
admissible pressure differences and range of low flow limit, follows as
(h 2_1Y/2,
coating speeds depicted in Fig. 11a.17b are much
greater than those in Fig. lla.17a. Ca max = 0.65 ,D Ca« 1 (11a.15)
G
Outside the domain bounded by the minimum
and maximum pressure difference no stable coating If Ca exceeds the critical limit, the meniscus
conditions can be found. At the lower border of will recede into the gap or break. At this limit
Stable coating operations 419
coating defects due to ribbing or rivulets can be
found. With increasing Re, liquid inertia decreases 100
the receding action and the limiting capillary
number is raised by a certain amount, as shown
in Fig. 11 a.18. The marked points in Fig. 11 a.18 E 80
are taken from numerical flow simulations of the ::::>...
downstream meniscus and show the limiting c

::r:
condition. The low flow limit exists only for Q)
60
values of hO,D greater than twice the wet film u
C
thickness. <ll
I-
<ll 40
In practical applications, defects caused by Q)
air-entrainment along the wetting line of the U
upstream meniscus will define a further coating a..

<ll
speed limit. Equations to predict the onset of CJ 20
wet film thickness
air-entrainment are given in Chapter 3.
The analytical expressions for the prediction
of coating windows are very helpful for the 0
0.0 0.5 1.0 1.5 2.0
selection of the optimum operation parameters. Web Velocity U (m/s)
Figure 11 a.19 gives an example which shows the
admissible variations of the coating gap as a Figure lla.19 Coating window: gap clearance versus
function of the coating speed for the coating of coating speed. L D = L u = I mm, H D = H u , (J =
20 Ilm thick wet film. At low coating speeds and 0.03N/m.
in the case of a low viscosity coating liquid, high
gap clearances can be selected. On the other viscous liquid, the gap clearance has to be less
hand, at high speeds and in the case of a highly than twice the film thickness. With increasing
viscosity and web speed, the capillary number
increases. This implies that the influence of
surface tension becomes negligible compared
til 10.0
U
with the viscous forces and stable coating
~ f....-(O conditions can be achieved only for gap clearances
Q)
,D 8.0 \ less than twice the wet film thickness.
S \
The analytical expressions can be compiled
;:1
z 6.0
*\
(e)J I
into a PC program which predicts operational
windows for slot coaters (Lange and Wagner
>..
~ I I 1992). The influence of non-Newtonian rheology
~
.....
1\
of the coating fluid can be approximated by using
I I
0.. 4.0
ttl (d) 1\ the viscosity of the current shear rate within the
U \ 1\ gap. Normal stress components can also be
~ II unstable considered as was done by Secor (1991) and
:>< 2.0 (C~\ I
ttl I
\ \ Strenger and Secor (1991). PC programs of this
S (a) I
(bY" ,~\..
kind visualize the influence ofparameter variations
0.0 , on the coating window and demonstrate that the
0.0 2.0 4.0 6.0 8.0 10.0
Gap Clearance h G following parameter changes decrease the size of
the coating window:
Figure lla.18 Low flow limit: numerical and analytical
data. (a) equation lla.15; (b) Re = 0.1; (c) Re = 1; (d) • increasing viscosity of the coating fluid;
Re = 10; (e) Re = 100; (f) Re = 1000. • decreasing surface tension of the coating fluid;
420 Slot coating
• increasing downstream lip length of the slot 4.0 , . . - - - - - - - - - - - - - - - - - - - - - ,

coater;
• increasing the coating gap; experiment
2:
• increasing web speed. 0-
<13.0 00000 swelling
00000 ribbing
Comparison of experimental and predicted Q)
o
I::
coating windows shows good agreement and this Q)
I..
justifies the presented approach to obtain par- ~2.0
ameter combinations for stable conditions. Figure is

l1.a.20 shows Sartor's experiments (1990) and in Q)
I..
::l
Fig. lla.21 the authors' own experimental results gJ 1.0
Q)
are depicted, which were compared with predicted I..
0.. /'
/'
/'
coating windows from equations (11 a.1 ) to (11 a.7) /'
/'
0
with the limiting conditions, equations (1Ia.9) to ,/
0.02 0.04 0.06 0.06 0.10

(1Ia.15), given above. Capillary Number Ca
(a)
lla.5 SMALL DISTURBANCES AND FILM

STABILITY
5.0 , . . - - - - - - - - - - - - - - - - - . ,
The considerations provided in the previous
section do not permit any conclusions regarding experiment
::>
the growth of disturbances with time. Linear and ~4.0
<1
00000 swelling
nonlinear stability analysis of the flow processes 00000 ribbing
can provide such information. However, only a Q)
few linear stability analyses of flow processes of

g 3.0
Q)
I..
slot coaters could be found in the literature. It ~
.....
might be that the confined flow field of the slot is 2.0
coater which allows a relative easy and fairly Q)
I..
::l
reliable prediction ofcoating windows still satisfies (/l
gj 1.0
the needs of industry. This situation contrasts I..
0..
with multilayer slide and curtain coating where
this kind of analysis is common, as Chapters 11 b
and Hc show. 0.10
Capillary Number Ca
0.20 0.30 0.40
Using lubrication approximations, a stability (b)

analysis of the flow at the downstream meniscus
under small disturbances will fail to predict the Figure 11a.20 Coating window: comparison of
onset of instability, as was shown by Bixler and experiment with theoretical predictions (Sartor, 1990).
Scriven (1982). They used a finite element (a) ID = Iv = 20.3, hO •D= ho .v = 3, PD = 22S,
simulation instead to calculate the base flow in v = equation lla.16; (b) ID = Iv = 12, hO •D = 2.5,
a two-dimensional flow domain and studied the ho .v = 3.19, v = equation lla.l6.
influence of three-dimensional disturbances on
the stability of their two-dimensional flow. The
meniscus was not pinned and could move along • increasing the capillary number;
the inclined die face. Bixler and Scriven (1982) • decreasing the flow rate;
could show for the limited parameter range of • decreasing the channel divergence angle;
their study that the following will induce ribbing • decreasing the Reynolds number.
instability: An analysis of the sensitivity of the upstream
Guide to practitioners, problems in practice 421
1.0..,....----------------
global energy balance of the flow processes
within the coating bead including the viscous
experiment
::>
8. 0 .6 forces. This balance leads to a nonlinear differential
<><><><><> swelling
<l 00000 defect free film <> equation. Analytical solutions were presented for
<l> two limiting cases. At low capillary numbers
g 0.6 <> capillary forces will dominate. The simplified
-
<l>
s-. equation reflects the coupling of the two menisci.
<l>
is 0.4 This is not considered in equations (11 a.l) to

<l> (l1a.15) and instabilities will occur if
s-.
;::l /
'"~ 0.2 ~(_1_ + _1_) > 0

/
s-. <> /
/ (l1a.16)
p., / dlw rm,D rm,U
/
/
where lw denotes the wetted length of the upstream
0.01 0.02 0.03 0.04 0.05 die lip and r m,D and r m,V are the radii of curvature
Capillary Number Ca
of the downstream and upstream meniscus,
(a)
respectively. For the second limiting case purely
viscous behavior can be assumed and a condition
for the minimum wet film thickness is derived.
1.0 . , - - - - - - - - - - - - - - - - - - ,
It is expected that disturbances will be damped if
experiment hG,u (lD + lw) h
2:0.6
0<><><><> swelling
121 < (1Ia.17)
0- D
<l 00000 ribbing
<l>
o No further predictions for the dynamic behavior
aQ) 0.6 of the coating bead and the growth rate of
s-.
Q)
:::: instabilities have been published even though the
is 0.4 numerical methods are available today.
Q)
s-.
;::l
til
til 11a.6 GUIDE TO PRACTITIONERS, PROBLEMS
~ 0.2
p., IN PRACTICE
The proper design and positioning of the die are
0.02 0.04 0.06 0.06 0.10 only one important part ofthe stable and reliable
Capillary Number Ca slot coating process. There exist many sources
(b) of disturbances which can amplify instabilities
and create coating defects. All elements of the
process need to be considered. In general, the
Figure lla.21 Coating window: comparison of
gap clearance in slot coating is small and therefore
experiment with theoretical predictions. (a) 10 =
Iv = 14.8,hG ,o = 6.5,h G ,v = 2,8, v = equation (lla.16);
very sensitive to vibrations. Even small amplitudes
(b) 10 = Iv = 25, hG,o = hG.v = 6,25, v = equation in the range of a few micrometers can create the
(lla.16). barring defect. This variation of the wet film
thickness in the downstream direction can be
meniscus to three-dimensional disturbances was caused only by pulsations of the flow rate or
made by Higgins and Brown (1984). It confirmed changes of the web speed, e.g., due to stick and
the results ofthe static analysis given by equations slip effects. Coating defects such as different wet
(lla.10) to (lla.13). film thicknesses in the cross-web direction, which
Higgins and Scriven (1979) suggested predicting are due to a misaligned coating gap, can be
the dynamic behavior of the coating bead by a prevented by a high-precision system for posi-
422 Slot coating
tioning the die relative to the application roll. The methods for preventing air-entrainment
Special measures, e.g., automated control of gap in slot coating are similar to other coating
clearance, can prevent damage of the die lips operations. However, as the free surfaces exposed
while a splice passes through. to the ambient gas are very small compared with
The narrow gap clearance requires that much other coating methods, degassing of the coating
effort has to be put into preparation and delivery liquid can increase the potential of the coating
of the coating liquid. Experience shows that the liquid to dissolve bubbles entrained at the dynamic
coating liquid should be degassed and filtered to wetting line, and by this means the occurrence of
remove all contaminations, e.g., bubbles and defects due to air entrainment can be reduced
particles. Cleaning of the web is also essential as (Miyamoto 1991).
dirt will accumulate at the die lips, creating The operating conditions should be chosen
streaks of uncoated web and in the worst case such that the pressure drop in the coating gap
these accumulations will lead to a rupture ofthe web. is much less than that in the feed slot. In this
At small Bond numbers a slot coater can be manner, the uniform distribution of the coating
operated in any position relative to earth's liquid is not influenced by the flow within the
gravitational field. However a position should coating bead. Fixed die lips are preferred over
be chosen such that no bubbles are entrapped adjustable ones. Nevertheless, for applications
within the liquid distribution system. Bubbles where the viscosity of the liquids coated with the
transported by the liquid can stick to the edges same die differs widely, a uniform liquid distribu-
of the die lips or block the coating gap. This tion might be achieved only by changing the
leads to a wavy wet film surface in the cross-web pressure drop within the feed slot in the cross-
direction, and streaks and uncoated areas may web direction. This can be done by automatic
appear on the substrate. positioning systems, e.g. piezo positioners, which
The edges of the die lips should be very sharp, control the gap clearance of the feed slot in
as mentioned in Section lla.2, and the curvature combination with an in-line measurement of the
should be uniform in the cross-web direction. wet film thickness (Lippert 1987). A high pressure
Cleaning and handling requirements set limits at the exit of the feed slot spreads the coating
to the radius of curvature. If the radius is too liquid in the cross-web direction. In some
small the lips are difficult to handle and the edges applications it is possible to block the coating
are easily damaged. The angle YDl included gap by soft inserts at the edges to limit the coating
between die lip and shoulder (see Fig. lla.8) width.
should be small (about 90°) to prevent the By blocking of the feed slot it is possible to
wetting of the die shoulder. If the wetting line reduce the coating width or coat stripes. Blocking
does not pin at the edge, vortices can occur, as the feed slot will change the liquid distribution
shown by Cohen, Suszinski and Scriven (1992), and special care has to be taken not to damage
and the wetting line may become distorted, which the lips.
can lead to wavy wet film surfaces. In the case of long die lips and application
Drying processes at the static wetting lines rolls of small diameter, the pressure level under
create depositions of non-volatile components the downstream die lip can become so small that
which distort the wetting line so that wavy films dissolved gas and solvent will leave the coating
occur. Especially if the coating liquid contains liquid and bubbles will be created. Figure lla.22
highly volatile solvents, encapsulation of the die gives an example of pressure distributions within
can prevent the rapid evaporation ofthe solvents. the coating gap. The die is translated horizontally,
For removing the sources of streaks, the coating parallel to the x-axis. Even at ambient pressure
gap can be cleaned with a soft, flat tool. In this at the downstream and upstream menisci, the
way contaminants sticking to the die lip will be pressure within the coating bead can decrease
removed. well below the ambient pressure. It can further
Summary and outlook 423
300.0
240.0
180.0
120.0
p.,
(l) 60.0
.t-.
::l
[fJ
[fJ
0.0
6.x
(l)
.t-.
P-. rRoll
-60.0
-50.0 50.0 150.0 250.0 350.0 450.0
Figure 11a.22 Variable gap clearance: pressure profiles within the coating bead. Re = 0.66, Ca = 2.78,
I D = Iv = 200, h G •min = 1.5, hF = 5, rroll = 8000. (a) ~x = 150; (b) ~x = 100; (c) ~x = 50.
be seen in this configuration that the coating gap the downstream die shoulder during the start-up
is diverging in the downstream direction and that of the system is fairly high and starving conditions
the pressure increases in the downstream direction, should be chosen.
which can enhance the ribbing instability.
It is commonly believed that a coating process
without vortices is the most stable. Further
11a.7 SUMMARY AND OUTLOOK
reasons for avoiding vortices within the coating
bead are that particles deposit at the walls in Slot coating is a very versatile operation which
areas of low shear rate, namely vortices and has a broad range of applications. It is premetered
recirculation zones, and change the flow field. and therefore allows the control of the film
Vortices promote separation processes and an thickness independent of liquid properties. High
uncontrollable residence time of the liquid, which coating speeds can be achieved and very low-
endangers the stable operation of the process. and high-viscosity coating liquids can be applied.
The asymptotic equations given in this chapter The maximum gap clearance in slot coating is
can be implemented into PC programs (Lange often very small, thereby requiring high precision
and Wagner 1992) which are useful for monitoring of the die and the applicator roll and in their
and controlling the instantaneous operating relative positioning. Dirt and dust particles can
conditions. It is possible to indicate the actual very easily create serious coating defects if the
point of operation within the coating window to coating gap is narrow. Hence the goal has always
assist the operating personnel. These coating been to coat at high speeds with a high gap
windows can also be used to find a stable path clearance. Under certain conditions, the gap
for the start-up of the process. The risk of wetting clearance can be increased at low and moderate
424 Slot coating
web speeds with a vacuum box located at the REFERENCES
upstream meniscus. The flow within the coating Ankenbrand, F., Herrmann, K. and Wilke, W. 1969.
bead is confined by the die lips and the web. Verfahren zum Auftragen magnetisierbarer binde-
Only a very small surface area of the coating mittelhaltiger Schichten auf einer Triigerfolie mit
liquid is exposed to the ambient gas during the einem ExtrudergieBer. German Patent 1,907,212.
process oflayer formation. Therefore, even coating Aurin, W. 1985. Hot melt pressure sensitive adhesive
liquids with a high content of extremely volatile coating equipment. In Proceedings of the 1985
solvents can be applied with a small risk of Polymers, Laminations and Coating CO'1lerence,
pp. 43-49. Atlanta: TAPPI.
defects by early evaporation ofthe solvent. Many Beguin, A. E. 1954. Method of coating strip material.
patents describe multilayer slot coating processes; US Patent 2,681,294.
however, the number of industrial applications Bixler, N. E. 1982. Stability of a coating flow. PhD
today is small compared with slide and curtain thesis, University of Minnesota, Minneapolis.
coating. Bixler, N. E. and Scriven, L. E. 1982. Stability of slot
In extrusion coating the positioning and and knife coating by finite element analysis. Paper
precision of roll and die are less sensitive to errors read at AIChE 1982 Spring National Meeting, at
Orlando, FL.
as the admissible gap clearance is much higher Browatzki, K., Maus, F., Platz, S. and Schweicher, W.
than in slot coating. The advantages of a higher 1975. Verfahren und Vorrichtung zum BegieBen
gap clearance are obtained at the expense of bandformiger Unterlagen. German Patent 1,921,044.
greater disturbances by edge effects. The flow Cameron, E., Alfred, P. and Hills, R. 1968. High speed
processes in extrusion coating are closely related coating apparatus. US Patent 3,413,143.
to curtain coating. Campbell, D. P. 1980. Slot orifice coating in 1980. In
The confined coating bead of a slot coater Proceedings of the Paper Synthetics Conference,
Cincinnati, Ohio, pp. 309-313, Atlanta: TAPPI.
permits the description of the flow processes by Carpenter, J. K. and Steen, P. H. 1990. Inertial effects
approximate equations. These equations can be in the fluid mechanics of planar-flow melt-spinning.
used to calculate coating windows, which are a Paper read at AIChE 1990 Spring National Meeting,
valuable tool for designing and controlling slot at Orlando, FL.
coating processes. In the case of fluids with Cohen, D. and Scriven, L. E. 1992. ViscocapiIlary
simple rheology, the predicted areas of stable model of two-layer slot coater. Paper read at AIChE
coating conditions agree very well with experi- 1992 Spring National Meeting, at New Orleans, LA.
Cohen, D., Suszinski, W. and Scriven, L. E. 1992. Flow
mental findings. visualization of two-layer slot coating. Paper read
Steadily increasing requirements on product at AIChE 1992 Spring Meeting, at New Orleans, LA.
quality and optimization of the coating process Demirdzic, I., Durst, F., Peric, M. and Wagner, H.-G.
will drive researchers to study the flow phenomena 1990. A finite volume numerical method for free
within the coating gap in more detail in order surface flow analysis. Paper read at AIChE 1990
to obtain a better prediction of coating windows. Spring National Meeting, at Orlando, FL.
Still, the influence of strongly viscoelastic coating Durst, F., Haas, R., Peric, M. and Scheurer, G. 1988.
Computational study of the flow in slot coaters.
liquids cannot be adequately simulated by Paper read at AIChE 1988 Spring National Meeting,
numerical programs. The avoidance of undesired at New Orleans, LA.
edge effects and three-dimensional flow structures Durst, F., Haas, R., Peric, M., Scheurer, G. and
will also draw more attention. Future research Wagner, H.-G. 1991. Numerische Simulation
will concentrate even more on combined experi- zweidimensionaler Schlitzdiisenstromungen bei der
mental and numerical studies, with a stronger F1iesserbeschichtung. Coating 2/91, 3/91, 7/91,10/91.
emphasis on numerical simulations than today; Fahrni, F. and Zimmermann, A. 1978. Coating device.
US Patent 4,109,611.
experimental studies tend to be expensive and Gilbert, N. and Eckel, A. 1992. Analysis of extrusion
numerical simulation programs are becoming coating in the presence of external forces. Paper
more and more powerful while simultaneously read at AIChE 1992 Spring National Meeting, at
the cost of computing is decreasing. New Orleans, LA.
Notation 425
Grader, L.1985. Untersuchungen zum BenetzungsprozeB, Sartor, L. 1990. Slot coating: Fluid mechanics and die
dargestellt am Beispiel der Beschichtung von Infor- design. PhD thesis, University of Minnesota,
mationsaufzeichnungsmaterialien. Dissertation, Tech- Minneapolis.
nische Hoschschule 'Carl Schorlemmer', Leuna- Scanlan, D. J. and Scriven, L. E. 1990. Two-slot coater
Merseburg, Germany. analysis: Separation line issues in two-layer coating.
Guthoff, E. B. and Kendrik, C. E. 1982. Dynamic Paper read at AIChE 1990 Spring National Meeting,
coating angles. AIChE J. 28: pp. 456-466. at Orlando, FL.
Higgins, B. G. and Brown, R. A. 1984. Multiple Schweizer, P. 1992. Fluid handling and preparation.
equilibrium shapes of partially constrained menisci: In Modern Coating and Drying Technology, eds
a quasi-static mechanism for instability of a coating E. D. Cohen and E. B. Gutoff, pp. 23-62. New York:
bead. Chem. Eng. Sci. 39(9): 1339-1345. VCH Publishers.
Higgins, B. G. and Scriven, L. E. 1979. Coating beads Secor, R. B. 1991. Process modeling ofcontact extrusion
behave as nonlinear oscillators that have limited and coextrusion. Tappi Journal. Nov.: 168-174.
stability. Paper read at AIChE 72nd Annual Meeting, Strenger, M. R. and Secor, R. B. 1991. Behavior of
at San Francisco, CA. knife coaters: Operating bounds, thickness control,
Higgins, B. G. and Scriven, L. E. 1980. Capillary and stress characteristics. Paper read at IS&T 44th
pressure and viscous pressure drop set bounds on Annual Conference, St. Paul, MN.
coating bead operability. Chem. Eng. Sci. 35: Tanaka, T. and Yasui, Y. 1992. Anomalous phenomenon
673-682. in die coating with viscoelastic fluids. Paper read
Ishiwata, M. and Nagai, Y. 1970. Uberzugsverfahren at AIChE 1992 Spring National Meeting, at New
und Uberzugsvorrichtung. German Patent Appli- Orleans, LA.
cation, 1,939,038. Tanaka, Y. and Noda, S. 1988. Beschichtungsvorrich-
Jackson, B. W. 1976. US Patent, 3,996,885. tung. German Patent 3,144,655.
Kageyama, T. and Yoshida, M. 1986. Extrusion Wilson, S. D. R. 1982. The drag-out problem in film
coating apparatus. UK Patent 2,120,132. coating theory. J. Eng. Math. 16: 209-221.
Lange, u., Durst, F., Raszillier, H. and Wagner, H.-G. Woodworth, C. B., Winkler, D. E. and Jackson, B. W.
1991. Axisymmetric slot coater for cables, wires and 1982. German Patent 2,403,313.
fibers: Theoretical treatment and layout consider-
ations. Paper read at IS&T 44th Annual Conference,
St. Paul, MN.
Lange, U. and Wagner, H.-G. 1992. DESC: A computer NOTATION
program for the design of slot coating devices. Bo Bond number
LSTM-Erlangen, U niversity of Erlangen-N urn berg,
Ca capillary number
Germany.
Lippert, H. G. 1987. Slot die coating - A technical D2 averaged square of inverse dimensionless
update on recent techniques and design developments. gap clearance
In 1987 Polymers, Laminations and Coating Conference D3 averaged third power of inverse dimen-
Book 1. Atlanta: T APPI. sionless gap clearance
Miller, F. D. and Wheeler, 1. J. 1965. Coating high He height of meniscus climbing up the die
viscosity liquids. US Patent 3,206,323. shoulder
Miyamoto, K. 1991. On the mechanism of air
HF clearance of feed slot
entrainment. Ind. Coating Res. 1: 71-88.
Nadeau, G. F. 1960. Method of coating a liquid hG,D dimensionless clearance of downstream
photographic emulsion on the surface of a support. coating gap
US Patent 2,952,559. HG,D downstream clearance of coating gap
O'Brien, W. G. 1984. Beveled edge metered bead hG,u dimensionless upstream clearance ofcoat-
extrusion coating apparatus. US Patent 4,445,458. ing gap
Ruschak, K. J. 1976. Limiting flow in a pre-metered HG,u upstream clearance of coating gap
coating device. Chem. Eng. Sci. 31: 1057-1060.
h oo dimensionless final wet thickness ofcoated
Ruschak, K. 1. 1985. Coating flows. Ann. Rev. Fluid
Mech. 17: 65-89. film
Russel, T. A. 1956. Multiple coating apparatus. US H 00 final wet thickness of coated film
Patents 2,761,417 and 2,761,418. LD length of downstream die lip
426 Slot coating
length of upstream die lip (J static wetting angle

dimensionless pressure at locations i = D, (J* static wetting angle measured between
U, 1,2 and 3 free surface and parallel to the web
Pi pressure at locations i = D, U, t, 2 and 3 f.l Newtonian viscosity
R radius of application roll p density
Re Reynolds number (J surface tension
Rm mean radius of curvature v dynamic wetting angle
U coating/web speed
v* feed flow rate
W width of coating in cross-web direction
Indices
D downstream
Greek letters
U upstream
f3 inclination angle of die lip m mean
y angle included by two die lip faces C climbing
!J.Pij dimensionless pressure difference between G gap
locations i, j = D, U, 1, 2 and 3 F feed
SLIDE eGATING llb
Jules Hens and Willy Van Abbenyen
llb.l INTRODUCTION
In Chapter 11a a first premetered coating system at a small distance (0.2 to 0.4 mm) from a moving
- the slot coater - was discussed. An important web, supported by a backing roll. Coating liquids
limitation of the slot coater is the difficulty of are fed through distribution cavities and slots,
building dies for coating more than two layers forming a multilayer stack on the incline without
simultaneously. In the photographic industry the convective mixing. The liquids are bounded
production ofcolor films requires the superposition sideways by edge guides.
of many (up to twenty) layers; so it is not At the end of the inclined plane, the liquid
surprising that slide coating - allowing an bridges the narrow gap between coater and web,
unlimited number of premetered layers to be wets the web and gets entrained by it. The liquid
coated simultaneously - was invented and puddle, extending from the coater to the web
developed in the photographic industry (Mercier, and bounded by an upper and a lower free
Torpey and Russell 1956). surface is commonly called the bead. If the flow
A slide coater (Fig. 11 b.1) basically consists of in this bead is stable, steady, uniform and without
a multilayer die with an inclined plane, placed vortices, the slide coating operation is satisfactory.
- - \ - - - UPSTREAM BLOCK
SLOT EXIT
V\---~r---------}- SLOT
\---7'"~---1- MA IFOLD
DIE LIP BASE PLA TE DIE BLOCK
Figure llb.l Perspective view on a slide coater.
London. ISBN 0 412 06481 2.
428 Slide coating
Generally a slightly sub-ambient pressure P v publications, including PhD theses, although

is applied below the bead, at most 1000 Pa less much information is also drawn from presentations
than the ambient pressure PA above. The difference at the coating symposia held every two years as
~P = P A - P v is often referred to as the 'bead part of the AIChE meetings.
vacuum' or 'backpresure' (Beguin 1954). As will
be shown, such a 'bead vacuum' is useful in order
llb.2 GEOMETRY OF A SLIDE COATER
to broaden the operability window, i.e. the range
ARRANGEMENT
of parameters within which a satisfactory coating
can be made.
I1b.2.1 DIE ASSEMBLY - NUMBER OF LAYERS
Alternatively the bead can be stabilized by
means of an electrical potential difference (of the A slide die comprises a number of blocks, in
order of 1000 V) between the backing roll and principle one per layer, mounted on an inclined
the coating liquid (Nakai and Okada 1968). base plate. It is convenient to use rectangular
Slide coating can be performed at web speeds blocks that are (nearly) identical, but they may
ranging from very low (say 0.1 m/s) up to rather also be designed individually according to the
high (say 4 m/s). Typical viscosities in slide coating rheology and the flow rate of each specific layer
are between 0.5 mPa s and a few hundred mPa s. (Paton 1968).
The total wet coated thickness is typically of the The number of layers that can be coated
order of 100 flm; it can be substantially thicker, simultaneously is in principle unlimited. As early
but it is difficult to coat thinner than about as 1977 an example with seven layers was described
20 flm. The total thickness, however, may be (Dittmann and Rozzi 1977); today, some films
distributed over several layers, each of which can for color photography have twenty and it is
be as thin as a few microns. desirable to manufacture such films in one pass
It should be clear at this point that not every at one coating station.
combination of the parameter values cited is In practice, the stability of the multilayer flow
possible; on the other hand, in special cases the on the inclined plane may become a problem
parameter range attainable can even be ex- upon increasing the length of the die by adding
tended. We shall explore the possibilities and the more blocks (see also Section 11b.4.3). In addition,
limitations of slide coating in more detail in the the reliability of any system is severely decreased
following sections. upon increasing the number of components.
In Section 11 b.2 we shall discuss important These and other potential problems have to be
details of slide coater geometry, such as slot exits, taken into consideration when deciding upon the
inclination angle, lip geometry, etc. In Section number of layers to be coated simultaneously.
11 b.3 we shall attempt to give an insight into
slide-coating flow, based upon principles of fluid
11b.2.2 SLOT EXIT
mechanics, flow visualization experiments and
numerical calculations. This will enable us to The coating solutions, making up the different
understand some of the limitations of slide layers, are uniformly distributed across the width
coating. More limitations are covered in Section of the die by a manifold/slot system, one for each
11 b.4. In Section 11 b.5 we discuss how the layer; this topic is covered thoroughly in Chapter
quality of the coated product may be affected by 10.
external disturbances and in Section 11 b.6 the Additionally, the geometry of the slot exit (at
main defects of slide coating will be reviewed. which a newly formed layer joins the stack on
Finally in Section 11 b.7 we summarize the the inclined plane) is crucial in slide coating (and
advantages and drawbacks of slide coating. in slide-fed curtain coating, see Chapter llc).
The contents of this chapter are mainly based Some problems and some proposed solutions are
upon information from patents and scientific discussed below and are illustrated in Fig. llb.2.
Geometry of a slide coater arrangement 429
~/
(a) (b)
(e) (g) (e)
(d) (h) (f)
Figure llb.2 Slot exit details.
In the most simple geometric arrangement the the top layer tends to flow back, wetting the
slot is formed by two rectangular blocks mounted upstream slide surface (Fig. 11 b.2b). Unfortunately,
in such a way that their top sides making up the there is little chance that the thus formed static
slide are in one plane (Fig. 11 b.2a). The fluid wetting line is straight, resulting in 'bandedness',
arriving from this slot flows underneath the other another nonuniformity across the web, but less
layer(s), but due to its momentum it is likely that sharp than streaklines (see also Section 11 b.6.1).
separation will occur, giving rise to a vortex. This The separation vortex can be avoided by
is undesirable, because in practice such a vortex increasing the width of the slot near the exit.
is generally neither two-dimensional nor steady Padday (1961) proposed the geometry shown in
(see Schweizer (1988), see also Chapter 7), making Fig. 11 b.2c. Such a design has one serious
the coating nonuniform. Furthermore a particle drawback, however; a stagnant zone (with very
can become trapped in such a vortex and a sharp, slow recirculation) in the widened part of the slot
prolonged nonuniformity (a streakline, Section creating the possibility that suspended particles
11b.6.1) will result. Additionally, in such a geometry settle causing streaklines. The same idea, however,
430 Slide coating
can be employed in other and better ways, e.g. wetting process (see also Sections 11 b.3 and
by chamfering as in Fig. 11b.2d (Russell 1969; 11 bA.l).
Ade 1976). Such a geometry is impossible, however, in
Backflow and upstream wetting at the upper- coating machines in which the web is conveyed
most slot can be avoided by elevating the upstream (nearly) vertically upwards into a chilling or
block (Fig. 11b.2e); Ade (1976) combined this drying zone. Furthermore, the web is preferably
with a chamfered downstream block (Fig. 11 b.2f). kept in contact with the backing roll over an
The same concept can be used for the other angle of at least 5" to 10· beyond the coating
slots too (Figs 11 b.2g and 2h). The step height position in order to make sure that web vibrations
should be chosen according to the flow parameters cannot perturb the coating bead. Hence the die
in such a way that the flow of the upper layers is often mounted with its leading edge under the
is nearly rectilinear. A drawback of such a horizontal plane through the axis of the backing
'stepped' geometry is the necessity for more roll (Fig. 11b.3b: IX < 0).
complicated edge guides: designs based upon Even when the web leaves the coating roll
simple single plates cannot be used. (nearly) horizontally, it is more convenient to
An optimal slot exit design is based upon the have the die positioned as in Fig. 11 b.3b, because
rheology and flow rates of the actual layers; of the start-up procedure. Indeed, immediately
visualization experiments and/or numerical before the start of the coating operation, the
simulations are indispensable (see Chapter 7 and liquid coming from the slide is not yet taken
also Kistler (1983), Scanlan (1990) and Berghezan away by the web, and hence has to be evacuated
(1991)). somehow. With a geometry such as shown in
Fig. 11 b.3b everything flows into the vacuum
box (see Section 11 b.2.8), and can be drained
llb.2.3 APPLICATION ANGLE IX
easily; a design such as that shown in Fig. 11 b.3a
Intuitively, it might seem logical to mount the requires a starting plate or the like, as with
die and the backing roll in such a way that the curtain coating (see Chapter lIe). This extra
axis of the roll lies in the plane of the slide (Fig. complication probably explains why nearly all
11 b.3a: IX = P); in such a geometry the momentum publications mention a geometry as shown in
of the arriving liquid will optimally assist the Fig. 11 b.3b.
(a) (b)
Figure llb.3 The application angle IX.

Geometry of a slide coater arrangement 431
llb.2.4 SLIDE INCLINATION ANGLE f3 exponentially with increasing slide length (see
It should be possible to find an optimal value Section 11 bA.3). In practice, manufacturing issues
for the inclination of the slide (the angle /3 with impose a minimum length per block (ofthe order
the horizontal plane) because both extremes of 3 to 6 cm) to maintain mechanical stability.
(/3 : : : 0° and /3 : : : 90°) have their own problems, The larger the coating width, the more difficult
as explained below. it is to obtain a good die with a slide that is as
The smaller the inclination, the thicker the short as possible.
layer on the slide H s will be:
H= (
3JiQ )1/3 (l1b.l)
11 b.2.6 GAP WIDTH H G
s pg sin /3 The die is placed in close proximity to the web,
at a gap of between 0.1 and 0.5 mm, preferably
where Q is the total flow rate per unit width, Ji between 0.2 and 004 mm. For completeness an
is the viscosity, p is the fluid density, and 9 is alternative geometry should also be mentioned
the gravitational acceleration. (Timson 1973; 1976) in which a gap as small as
This expression holds for the one layer case, 25 ~m can be obtained by using a pressure
but can easily be extended to a multilayer system. chamber instead of a backing roll, but this
The momentum of the flow on the slide is geometry does not appear to be widely used.
proportional to the flow rate squared, divided A small gap produces a coating bead which
by the thickness on the slide; so the momentum, is less prone to disturbances (because the surface,
which makes up the hydrodynamic assist of the on which e.g. pressure variations can act, is
wetting process (see Section 11b.4.1), tends to smaller) and hence the coating window (or more
zero with decreasing inclination. precisely the quality window, see Sections 11 bAA
On the other hand, increasing the inclination and 11 b.5) is larger. From this point of view a
makes a flow less stable, as the growth rate of gap of the order of 0.1 mm is preferred.
waves on an inclined plane (see Section IlbA.3.1) On the other hand, splices (and other web
increases with sin /3 (Yih 1963; Kao 1968). unevennesses) must pass the coating die. With
A steep incline positioned in such a way that too narrow a gap there is a serious risk of web
the momentum optimally assists the wetting rupture; and if the web does not break, but merely
process has been described (Burket, Conaghan touches the die face, the static wetting line may
and Hirshburg 1984). The drawbacks ofinstability become distorted, yielding streaklines (see Section
on the slide and start-up complication are, in 11b.6.1). When these types of problem occur, an
our opinion, not worth the possible gain in increase of the gap width is recommended.
maximum coating speed.
As far as we know, a complete study of the
optimalization of the inclination angle of a llb.2.7 DIE FACE
multilayer slide die has never been published. In
The detailed geometry of the die lip is important
most publications and studies an angle between
in view of the positioning of the static wetting
15° and 30° is mentioned. It is the authors'
line. It should be straight in order to obtain a
experience that in this range a few degrees are
uniform coating (a well known feature common
not crucial.
to all die-coating operations). This can be achieved
by 'pinning' the contact line at a sharp edge,
according to Gibb's principle (see Chapters 3
llb.2.5 LENGTH OF THE DIE
and 9). Fahrni and Zimmermann (1977) defined
The length of the slide should be as small as a sharp edge at which a wetting line will pin as
possible, because waves, if present, may grow a physical edge with a radius of curvature of less
432 Slide coating
than 100 Jlm, preferably less than 50 Jlm. On the llb.3 SATISFACTORY STEADY STATE
other hand of course, the sharper the edge, the OPERATION
greater the risk of damaging it, and of the
operator getting injured when manipulating the In slide coating different flow regions can be
die. distinguished: the zone near the slot exit where
Whether the static contact line of the lower the slide flow is formed, the flow on the slide
meniscus is pinned at the slide edge or wets the itself, a transitional zone between slide and bead,
slide face lip depends upon both geometry and the bead zone and another transitional zone in
flow parameters. The larger the angle y between which the flow rearranges towards plug flow to
the die face and the web (see Fig. 11 b.3), the more be finally convected away by the web (see Fig.
the flow will tend to pin at the slide edge. 11 b.4, and examples of flow visualization in
Increasing web speed or viscosity or decreasing Chapter 7).
the gap width or the pressure difference over the The forming flow at the slot exit delivering
bead also forces a 'wetting' contact line to move one fluid underneath the other(s), has been
upward, and eventually pin at the edge. thoroughly studied by Kistler (1983) and Scanlan
Wetting should be avoided, not only because (1990), by numerically solving the Navier-Stokes
the contact line may become crooked, but also equations.
because of the possibility of recirculation in the On the slide the flow eventually reaches a
lowermost part of the bead between die lip and one-dimensional steady state regime which can
web (see Section 11b.3.4). Furthermore, numerical be easily calculated in the case of Newtonian
calculations show that a pinned contact line liquids, the flow profile being a (multiple) parabola
is qualitatively advantageous, due to reduction (Aktharuzzaman, Wang and Lin 1978). Other
of the sensitivity to disturbances (see Van liquids yield a modified, yet always parabola-like
Abbenyen, Christodoulou and Scriven (1988); see profile on the inclined plane.
also Section 11 b.5). Where the slide flow ends (because the liquid
eventually impinges on the web) the free surface
transforms into a standing wave. This typical
transitional flow pattern was first described by
11b.2.8 VACUUM BOX
Ruschak (1978), further by Kheshgi, Kistler and
In order to create a pressure difference over the Scriven (1992) and studied numerically in the
bead a so called vacuum box is built between case of slide coating by Kobayashi, Saito and
the die and the backing roll. Precision labyrinth Scriven (1982).
seals are mostly used to minimize the leakage of Higgins and Scriven (1979) showed that the
air (inevitable due to the moving web) into the flow in the transitional zone out of the bead
vacuum box. Maintaining the pressure difference exponentially rearranges into a plug flow, or
requires continuous evacuation of air out of the more precisely a solid body translation with the
vacuum box. This should be done carefully and web, at least if one disregards the effect of gravity.
uniformly, because the coating quality can be Galehouse and Colt (1984) re-examined both
impaired by fluctuations in the applied pressure the flow at the end of the slide and the flow
difference, especially when a relatively large gap leaving the bead, attempting to study the slide
is used (see Section 11 b.5). coater analytically. Christodoulou and Scriven
Multiple chambers have been described (Krussig (1989) used these analytical approximations as
1981), even with adjustable walls (Bassa 1984). boundary conditions for their finite element
In general, standing acoustic waves in the vacuum calculations of the flow field in a slide coater.
box should be avoided and noise sources (such We shall now focus on the flow in the bead
as the fan generating the pressure difference) need region, which is really the heart of the slide
to be damped carefully. coater. Here the liquid arriving from the incline
Satisfactory steady state operation 433
Hoo/O•9
u F
v.g
TAKE AWAY ZO E I
I I "
'I
I
I
"
,
l _ _ : _,_
Figure llb.4 Definition sketch of a slide coating bead: the control volume and the forces acting upon it.
is deflected, wets the moving web and becomes line the coating liquid displaces the air that
entrained by it. Two important processes take previously adhered to the web. This is called the wet-
place in the bead. First, at the dynamic wetting ting of the web and it is the subject of Chapter 3.
434 Slide coating
We shall return to this subject in the next section neglected. In part these simplifications are
(Section 11 bA.l). Secondly, the two free liquid-air physically clear or intuitively reasonable; others
interfaces allow the bead to change its geometry rely on results of flow visualization or numerical
in such a way that the rate at which the liquid flow calculations; and some are only justified a
is entrained by the web exactly balances the rate posteriori because the results seem to correlate
at which liquid is delivered to the coater, making qualitatively with practical coating experience.
it possible to use satisfactorily a premetered
coating system such as slide coating with a wide
lib.3.l VERTICAL MOMENTUM BALANCE -
range of parameters.
FORCES AT WORK
Many years ago Levi, Cvetkov and Babcin
(1966) investigated slide coating in the limiting The liquid delivered to the bead is accelerated
case where capillary forces dominate over both by the viscous action of the moving web, as
viscous and inertial effects. They came to the described in classical boundary layer theory
conclusion that the determining phenomenon (Sakiadis 1961). Hens and Boiy (1986) found a
was the change in the mean radius of curvature good correlation between Sakiadis' boundary
Rm of the upper free surface. Hens and Boiy layer length L l1 and the height Lb of the slide
(1986) investigated the inertia-dominated regime coating bead when inertia dominates. Here we
and found that the contact length of the boundary shall also consider, or at least mention, capillary
layer was the determining parameter. effects, gravity, pressure difference over the bead,
In the remainder ofthis section the geometrical electrostatic forces, incoming momentum, and
adaptivity of the bead is discussed further. The effects of the air entrained by the web.
flow in the bead is entirely two-dimensional (if A given control volume (abcdefg) is defined
satisfactory, otherwise it can be three-dimensional (Fig. 11 bA) which corresponds to the bead
and/or unsteady - see below). It can be calculated region. It is bounded by the two free surfaces
numerically by means offinite element techniques (bc) and (fg), the moving web (ab), the slide
as shown by a number of investigators (e.g., Beck surface (de), an entrance section (ef) somewhere
and Mues (1988); Christodoulou and Scriven under the trough of the standing wave near the
(1989); Schunk (1989); Chen (1992); see also end of the incline, at a distance Ldc from the edge
Chapter 9). (d) (how the point (e) is chosen will be explained
However, here we want to emphasize the later), and an exit section (ag) located where the
physics involved by using an analytical approach; mean velocity ofthe entrained liquid has reached
unfortunately, an accurate quantitative analysis 90% of the web speed U; the layer thickness at
of this kind has not been published to this day. this position is H 00/0.9, H 00 denoting the final
Therefore, we shall follow the, inevitably ap- wet layer thickness. The distance between the
proximate, qualitative approach as proposed by exit section (ag) and the dynamic wetting line (b)
Hens and Van Abbenyen (1994): they estimated we call the bead height Lb. For the sake of
the different forces acting upon the coating bead, simplicity, we assume a vertically moving web,
and developed an expression showing the order a vertical die lip face (cd), a Newtonian liquid
of magnitude of the dimension of the bead and with constant surface tension (1 and the lower
the qualitative effect ofdifferent coating parameters free surface located at (c) (at a small distance Lcd
on it; assuming that this dimension can vary from the edge (d), with an apparent static contact
between a lower and an upper bound, the angle eo and an apparent dynamic contact angle
approximate expression suggests how some of eD as defined in Fig. 11 b.4.
the limits of the operability range ofa slide coater The rate of change of momentum in this
may be estimated. control volume is brought about by the sum of
This analysis is approximate, because some all the external forces (both bulk and surface
terms are only roughly estimated, and others are forces) acting upon the liquid. The horizontal
component of momentum plays a role in the Drag force of the die lip (cd)
wetting process, but is not considered here. The
F ~ /l(oU/OX) Lcd (lIb.?)
vertical component controls the entrainment of V• 1
the liquid by the web. With velocities, momentum

and forces taken to be positive in the upward Drag force of the slide (de)
direction, and with Us denoting the mean velocity
of the flow far upstream on the slide and liP the F v • s = /l(oU/oy)L de (11 b.8)
pressure difference over the bead, the terms of
this vertical momentum balance (per unit width)
can be estimated as follows. Drag force of the air (bc)
Fv • a ~ /lair (oUair/OX) H O cot g(OD) (l1b.9)
Momentum leaving (ag) by convection
M out ~ 0.9 pU 2 H 00 (11 b.2) Dynamic pressure of air, entrained by the web,
at (bc)
Momentum entering (ef) by convection (l1b.l0)
The developed flow far upstream at the slide

carries a momentum of Surface tension at (g)
(lIb.11)
At the entrance section, chosen under the trough

of the standing wave, the liquid is accelerated Surface tension at (f)
(by the pressure gradient caused by the curved
We may choose the inflow section (ef) of our
meniscus, see also further) from Us to a value
control in such a way that the free surface at the
usually not more than U, and consequently the
point (f) makes an angle (- p) with the horizontal
film thickness is smaller than H s ' but not smaller
plane. The choice may seem arbitrary, but by
than H 00' Hence we may estimate:
doing so, the net effect of capillarity on the
upstream part of the flow (on the incline) is
vertically upward; it causes a decrease in pressure
under the trough of the standing wave, and
Drag force of the web (ab) accelerates the flow; this effect has been taken
into account by the momentum at the inflow
Sakiadis (1961) calculated the total viscous friction
section in equation (11 b.4). The surface tension
force over the boundary layer:
effect on the control volume in (f) then becomes:
(11 b.5)
(11 b.12)
Here, however, we not only consider the boundary
layer length L~, but the bead height Lb , which
also includes the zone in which the boundary Surface tension at (b)
layer reorganizes towards a plug flow. In this Fa•b = -Cos(OD)a (lIb.13)
reorganization zone the viscous effect is less than
in the zone where the boundary layer forms.
Therefore it seems reasonable to estimate: Surface tension at (c)
(l1b.6) F a • c = -cos(Oo)a (1Ib.14)
436 Slide coating
Pressure difference over the bead met in satisfactory steady state slide coating.
However, this restriction also implies that the
F!J.p = -HGI:::t.P (1Ib.15)
limiting case where 00 approaches 180 cannot 0
be addressed (see Section 11b.4.1).

Gravity Finally, sin(p) is typically of the order of 0.5.
Thus MaUl - Min becomes:
The weight of (unsupported) liquid can be
estimated roughly as: MaUl - Min ~ 1.5pU 2 H oo (l1b.18)
With equations 11b.5 through 11 b.18, the mo-
mentum balance then becomes:
Reactive force of the die
1.5pU 2 H oo ~ fJ.ULb/Hoo + 0.5 a - HGI:::t.P
F0 L de cos(P) (11 b.17)
'"
- pgLb(H oo + H G )/2 - Fe. (l1b.19)
Electrostatic force Fe.

llb.3.2 THE BEAD HEIGHT 4
This force is hard to quantify, but it is retained
here in order to show its influence qualitatively. From the approximate momentum balance
(11 b.19) we can derive an expression for the
(dimensionless) bead height Lb/H 00:
11 b.3.1.1 Approximations - simplifications
Lb 1.5pU 2 H oo -0.5a+HGI:::t.P+F e•
It is reasonable to suppose that the reactive force H oo ~ fJ.U-pgH oo (H G +H oo )/2
of the die F o and its viscous force F v ,. approxi-
mately balance out that part of the gravitational (l1b.20)
force F g that is proportional to slide length (and
that was not mentioned in equation (l1b.16)). or
The dynamic pressure of the air F p • a can be
neglected, because PairlPliquid is of the order of ~b = (inertia - capillary forces )
10- 3. Furthermore, the capillary forces at the 00 + vacuum forces
lower free surface F<1, band F<1. approximately
C + electrostatic forces
cancel each other if the lower meniscus is not
too strongly curved, a situation often encountered viscous forces - gravity (l1b.21)
in practice (Faust 1975). We further neglect the We can write (11 b.20) in terms of dimensionless
viscous action F 1 of the die lip because Lcd/L b
V
•
numbers as follows:
has been assumed to be small. Hence (aU lox) is
small because the flow rate downwards in the Lb 1.5 We - 0.5 + N!J.p + N e•
gap between the web and the die lip is small. H ~ Ca(1 - St*) (l1b.22)
oo
This assumption is consistent with flow visual-
ization and laser-Doppler measurements of where
satisfactory steady state flows. In the next sections,
We = pU 2 H oola (Weber number) (11 b.23)
however, we shall also encounter cases where
this simplification probably is not fully valid N!J.p = HGI:::t.Pla (dimensionless pressure
anymore; so caution is needed here. Furthermore, number) (11 b.24)
the viscous action Fv,a of the air acting on the
(dimensionless electrostatic
lower free surface is also neglected. This is
force) (11 b.25)
certainly permissible as long as cot g(Oo) remains
bounded, i.e. if 00 < e.g. 16<t, a situation commonly Ca = fJ.Ula (capillary number) (l1b.26)
St· = St(H G + H 00)/2H 00 (modified Stokes value of L b above a lower critical value Lb,min:
number) (11 b.27)
(llb.32)
St = pgH~/J.lU (Stokes number)
Neglecting gravity and electrostatic effects, we
(11 b.28) can rewrite (11 b.22) as
It is interesting to consider the limiting case Lb/H oo :::: (1.5 We + NlJ.p - O.5)/Ca (llb.33)
studied by Hens and Boiy (1986), for a Weber
With assumption (llb.32) this becomes:
number that is very large (We» 1), indicating
that capillary forces are small compared with 1.5 Re + (N IJ.P - O.5)/Ca > Lb,min/H 00 (llb.34)
inertial effects. In most cases St· is much smaller
than unity, and may be neglected. Further, or otherwise, from (llb.32) and (llb.20)
assuming a case without pressure difference and (1.5pU2H~ + (HGAP - 0.5 (j)Hoo)/J.lU > Lb,min
electrostatic forces (N IJ.P = N es = 0), equation
(llb.22) reduces to (llb.35)
(llb.29) Expression (11 b.34) suggests that a minimum

flowrate is needed for slide coating to be feasible.
where An experimental study by Gutoffand Kendrick
(1987) clearly shows the existence of such a 'low
Re = We/Ca = pUHoo/J.l flow limit of coatability'. This is the minimum
(Reynolds number) (llb.30) layer thickness H oo,min that can be coated at a
given speed. Eventually, at very small H 00 values
and hence the bead becomes unstable and breaks up. This
break-up is similar to the collapse of a liquid
(llb.31) curtain at small We numbers (Brown 1961). Often
this phenomenon starts at the edges; it mayor
It should be no surprise that this relationship may not stabilize: sometimes the bead completely
is close to the one given by Sakiadis (1961) using collapses, but often a new stable but undesired
boundary layer theory, except that we find a flow can persist, with very thick edges due to
coefficient of 0.82 instead of 1.616; this is because necking-in, or reorganized into 'rivulets' (a steady
we deliberately defined our bead height longer regular pattern of wet 'ropes', separated by
than the boundary layer length, and because we completely dry paths). The study of Gutoff and
took the incoming momentum into account. Kendrick (1987) shows that at higher viscosities
Hens and Boiy found a coefficient of about 1.28. the minimum layer thickness increases (Fig.
In the following sections we shall study the 11 b.5) and presents regression formulae for the
cases in which, according to expression (llb.22), experimentally determined limits.
the bead height L b tends to zero and to infinity, The qualitative agreement between the low
because intuitively it is clear that in practice there flow limit as described by Gutoff and Kendrick
must be limits to L b , and hence both very small (1987) and inequality (11 b.35) is illustrated in
and very large values of L b are likely to set Fig. 11 b.6, by plotting the ratio of the actual
operability limits. coating thickness H 00 over the minimum thickness
. as found experimentally by Gutoff and
H co, min
Kendrick (1987) versus the bead height L b for a
llbJJ SMALL VALUES OF L b
number of cases, described in Table 11 b.1. The
It is clear that the previous analysis is meaningless apparent correlation suggests that the lower
when L b becomes negative or zero. Let us assume coatability limit may correspond to small values
therefore that satisfactory coating requires a of Lb' Expression (llb.35) then indicates how
438 Slide coating
10000
2 5
~ 1000
en
en
(])
c:
~
u
£
m
~ /l. mPa.s
E 1. 100% glycerin 1110
:J
E 2. 95% glycerin 520
'c 100 3. Mineral oil 170
~
4. 88% glycerin 150
5. 70% corn syrup 120
6. Corn oil 59
7. 70% glycerin 25
8. 54% corn syrup 21
10 '--_-'-----'--'-.............J....L........_----L._'---L.-I.......................I ' - - _ - ' - - ' - - - ' -.....I.....I.....I.....I~I!

0.01 0.1 10
Web velocity (m/s)
Figure 11 b.5 Minimum wet thickness vs. web velocity for slide coating various Newtonian liquids - 500 Pa bead
vacuum. From: 'Low flow limits of coatability on a slide coater.' E.B. Gutoff and C.E. Kendrick (1987) A1ChE J.
33 (1): 141-145. (Reproduced by permission of the American Institute of Chemical Engineers © 1987 AlChE.)
this operability limit might be overcome by H oo,min varies with (J.l./U)1 /2: i.e., at low speed one
increasing the coating thickness H 00 and/or has to coat more thickly in order to coat faster,
decreasing the viscosity J.l. and the surface tension whereas at high speed the minimum layer thickness
(T. Also the positive effect of a pressure difference decreases with increasingwebspeed. This behavior
on the bead may be observed. was observed by Gutoff and Kendrick (1987),
Interestingly, a consequence of assumption but only in the case of a polymeric liquid. With
(11 b.35) is that H 00. min does not vary monotonically Newtonian liquids one cannot reach high enough
with the coating speed U. Indeed, if we rewrite coating speeds to detect this phenomenon, because
(11 b.32) with (11 b.20) and then solve for H 00' we of another defect (air entrainment, see Section
find the following relationships: 11 b.4.1).
at small Re: H oo,min '" J.l.U (l1b.36)

llb.3.4 LARGE VALUES OF L b
at large Re: H oo,min '" (J.l./U)1 /2 (11 b.37)
In the last section we looked for a coatability
At very low speed, H oo,min is found to be limit due to a bead becoming too small. Anal-
proportional to (J.l.U). At very high speed, however, ogously, we may suppose that for a satisfactory
10.,.--------..,.---------,---------,.-----------,
:Q'
o
~
c
~
Q)
I
.............................................................................................~..
u
c
ro
:t::
o
:;
~
(j)
Ql
:;
c
g 0.1
8 o
I
.....
8
I
0.01 + - - - - - - - 1 - - - - - - - - - - 1 - - - - - - - + - - - - - - - - 1
0.01 0.10 1.00 10.00 100.00
Bead height Lb (mm)
Figure llb.6 Correlation between the bead height L b calculated with (11 b.20) and the ratio of layer thickness
H <Xl over the minimum wet layer thickness H <Xl.min as found by Gutoff and Kendrick (1987) at 500 Pa bead
vacuum.
coating a bead cannot be infintely large either. or otherwise, from (llb.38) and (llb.20)
Let us therefore assume that there is an upper
critical value Lb. max in satisfactory slide coating: (1.5pU2H~ + (HGAP - O.5a)H oo )/j1.U < Lb,max
(llb.40)
(11 b.38)
Expression (11 b.39) suggests that for satisfactory
Large values of L b can, according to equation slide coating there is a maximum Reynolds
(llb.22), arise in three ways: number for any given capillary number, or
equivalently, that there is a minimum capillary
(1) for We» 1 and Ca « 1, i.e. for Re » 1; number for any given Reynolds number. There
(2) with very large values of N dpand/or N es ' or is also an effect of the gap and the pressure
(3) with (1 - St*) close to zero. difference over the bead, with the pressure force
apparently reducing the useful parameter range.
In the remainder of this section we shall treat We suggest that the defect arising at this
these three cases in turn. For simplicity we supposed limit may be vortex formation. Indeed,
neglect electrostatic forces, i.e. N es = O. For both flow visualizations (Schweizer 1988; Chen 1992)
cases (1) and (2) gravity can be neglected. Using and finite element calculations (Christodoulou
(1Ib.33) assumption (llb.38) then becomes: and Scriven 1989; Chen 1992) show that in
certain coating conditions vortices may develop
1.5Re + (NdP - O.5)/Ca < Lb.max/Hoo (llb.39) in the bead, as illustrated in Fig. lIb.? As stated
440 Slide coating
Figure llb.7 Vortices in the bead of a two-layer slide coating. Left: visualized flow field; right: computed
flow field. Parameter values: webspeed U = 0.25 m/s; gapwidth H G = 250 ~m; vacuum liP = 257 Pa; slide
inclination p = 20·. Bottom layer: thickness H oo.B = 90 ~m; viscosity f.lB = 15 mPa s; surface tension lTB = 69 mN/m;
density PB = 1170kg/m 3 . Top layer: thickness Hoo,T = 91 ~m; viscosity f.lT = 15mPas; surface tension
3
lTT = 69 mN/m; density PT = 1170 kg/m . From K. S. A. Chen: Studies of multilayer slide coating and related
processes; PhD thesis, University of Minnesota (1992); available from University Microfilms International,
Ann Harbor, MI 48106 USA.
above (see Section 11 b.2.2) vortices are undesirable with Po being the property number, defined by
because they are neither steady nor two-dimen- Schweizer as
sional and because particles and bubbles may
Po = g/14/ pq 3 = Ca 3 St/Re 2 (llb.42)
get trapped, causing defects such as, e.g. streaklines
(see Section 11 b.6.1). In the next sections we shall (It should be noted here that this definition
discuss these vortices, and show some correlation deviates from the one used throughout this book,
with the bead height Lb' by the (-1/3)th power). He used a constant slide
inclination of 15° and found no significant influence
of gap width or pressure difference on the upper
llb.3.4.1 Case 1: Upper vortex vortex.
From his visualization experiments Schweizer Chen (1992) deduced from his experiments that
(1988) deduced that a vortex near the upper free vortices occur when the drawdown ratio Hs/HOC)
(which is proportional to St- 1 / 3 ) becomes smaller
surface appears when the capillary number is
than some limit, depending upon the capillary
smaller than Cacril' which is given by the
expression: number. For small Ca, the maximum value of
St decreases with decreasing Ca; for large Ca,
Ca = 0.5406 POO.1441 the upper limit of St becomes independent of Ca.
crit
It is remarkable that in this work no effect of
exp(0.8434RePoO.2170) (llb.41) the Reynolds number was reported, although the
existence of a maximum Reynolds number for lower part of the meniscus. Chen (1992) derived
the onset of a vortex near the upper free surface from his experiments a correlation between the
was clearly established by Schweizer (1988), see Ca number and the critical pressure difference
above, and also suggested by Christodoulou and for die face wetting, but his expression does not
Scriven (1989). include the gap width, although this parameter
obviously plays an important role.
11 b.3.4.2 Case 2: Lower vortex

Correlations, comments and conclusion on vortices
A very large value of N I1P (resulting from a high
in the bead zone
pressure difference and/or a large gap) also leads
to a large value of Lb' Intuitively it is clear that In Fig. 11 b.8 we plotted the ratio of the actual
in such cases the bead will tend to fill the gap capillary number over the critical value Ca crit
between the web and the die face. Severe wetting (equation (lIb.41)) versus the calculated bead
of the die will occur, and, apart from a possible height L b , for the cases defined in Table lIb.I.
crooked static contact line, another vortex may The apparent correlation that is illustrated in
appear between the die lip and the web, as seen this figure, and the qualitative agreement between
in visualizations and numerical calculations expressions (11 b.38 to 11 bAO) and the published
(Schweizer 1988; Chen 1992). evidence on vortices makes the proposition tenable
To the authors' knowledge no systematic that in practice vortices are likely to occur when
study has been published about the vortex in the the bead heigth L b exceeds some critical value.
100 .-------~------__._------.__-----___,
.... <>8
Q) 10 ...................................................................................................................
N
Q) <>
~ <>
.r.
0
(/)
.-
0
co
U
(;j
U
()
()
()
0.14-------+-------+-------1----------l
0.01 0.10 1.00 10.00 100.00
Bead height Lb (mm)
Figure llb.8 Correlation between the bead height Lb calculated with (llb.20) and the ratio of the capillary
number over the critical capillary number Ca crit for the onset of an upper vortex, as found by Schweizer (1988).
442 Slide coating
Table llb.l argued that this occurs when the amount ofliquid
entrained by the viscous boundary layer equals
Inclination angle, P(degrees) 30 the flow rate supplied from the slide before the
Gap width, H G (11m) 350
Pressure difference, tJ.P (Pa) 500
boundary layer reaches the free surface. Beyond
Layer thickness, H 00 (11m) 30-70-110 that point, the web will entrain more liquid, but
Viscosity, JJ. (mPa s) 1-3-10-30-100 this has to flow down again near the upper free
Web speed, U (m/s) 1-2-3 surface in order to maintain the flow rate balance.
Surface tension, (1 (mN/m) 15-40-65 A lower vortex (as in Fig. llb.ge and 9f and in
Fig. llb.7) forms when the gap between the web
and the die lip is filled with liquid and the amount
This is illustrated in Fig. llb.9. Starting with a of liquid entrained by the viscous action of the
base case without vortices (Fig. 11 b.9a) the value web is smaller than that allowed by regular
Lb is increased by varying V, H 00' Jl, H a and tiP Couette-Poiseuille flow, resulting in a vortex that
respectively, as specified in Table llb.2. In each partially fills up the gap width. Such vortices are
of these cases a vortex is observed. well known in slot coating (see Chapter lla). It
Criterion (1Ib.38) seems to cover not only the is intuitively clear that this will be the case when
vortex at the upper free surface, as predicted by both Lb and the gap width H a are large.
Schweizer's correlation, but also the vortex formed Vortices may also appear underneath the
between the die face and the web in the case of standing wave at the end of the slide (as weakly
a large gap width or pressure difference. Expression visible in Fig. Ilb.9c). According to Schweizer
(llbAO) indicates how this operability limit may (1988) this is due to the large adverse pressure
be overcome by decreasing the coating thickness gradient caused by the impingement on the web
H 00' the gap width H a and the pressure difference ofthe liquid arriving from the slide. Such vortices
tiP and/or increasing the web speed V, the appear at high Weber numbers.
viscosity Jl and the surface tension (1. Finally, a weak vortex may form near a static
Figure 11b.9 also illustrates that at least four contact line when the die lip is wetted (as
types of vortices are possible, probably caused noticeable in Figs 9d, ge and 9f). This is a well
by different physical mechanisms. known feature of flow in a sharp corner, as
An upper vortex (as in Fig. 11b.9b, 9c, 9d and ge analyzed by Moffat (1964).
and in Fig. llb.7) will appear when the liquid at The mechanisms causing vortices have been
the free surface decelerates towards a stagnation elucidated more in general by Goenaga and
point and the flow reverses. Schweizer (1988) Higgins (1992).
Table Ub.2
U Hoo JJ. HG tJ.P Lb

Figure (m/s) (JJ.m) (mPas) (p.m) (Pa) (mm) Ca/Caorit
9a 0.6 100 8 200 200 1.6 1.31

9b 0.2 100 8 200 200 1.8 0.53
9c 0.6 275 8 200 200 ILl 0.79
9d 0.6 100 5 200 200 2.6 1.05
ge 0.6 100 8 500 200 3.0 1.31
9f 0.6 100 8 200 500 2.9 1.31
Constant: Inclination angle p: 30'

Surface tension u: 4OmN/m
Density: lOOOkg/m 3 .
Figure llb.9 Flow field (calculated by B. Goetmaeckers) for the parameter sets given in Table llb.1. (a) is
the base case, without vortices. (b)-(f) all have an increased bead height as compared with the base case, due
to the change in one parameter; they all suffer from vortices.
444 Slide coating
Figure llb.9 Continued.
Ilb.3.4.3 Case 3: Bleeding expressions, but intuitively obvious, is that

'bleeding' also occurs when the die lip is either
It is worth commenting, that formula (llb.22) very short or is sharply cut away.
allows L b to become infinite, when the denomi-
nator of the expression tends to zero, i.e. when
Ilb.3.5 CONCLUSIONS ON THE BEAD HEIGHT
St* approaches unity. Such a high value of St*
- APPLICABILITY
is, however, only reached with parameter settings
seldom met in industrial coating practice. St* is In the previous sections qualitative correlations
usually ofthe order of 10- 2 orless. From (llb.24) were noted between the vertical momentum
and (llb.25) it can be immediately deduced that balance and some operability limits of slide
this limit will occur only when coating a very coating. We proposed that the lower coatability
thick, very low viscosity layer at very low speed limit might be related to small values of the bead
using a large gap. height Lb and that vortices are likely to occur
Under such conditions it is physically impossible when Lb is larger than some higher critical value.
for the web to entrain all the liquid delivered by 'Bleeding' also appears to be observed at high
the slide. In coating experiments 'bleeding' (also values of Lb'
known as 'pull through' or 'weeping') occurs: We therefore conclude that the bead height L b
part of the liquid is not entrained by the web, is an important feature in slide coating, and that,
but leaks into the vacuum box, or flows out of in order to achieve a satisfactory coating, it
the bead at the edges. This results in a deposited should remain between some maximum and
layer that is much thinner than expected, far from some minimum value: too large beads suffer from
uniform and of ill-defined thickness. vortices or bleeding, too small beads become
Again there is a qualitative correlation between unstable and will break up.
an operability limit ofslide coating and expressions In practice, the layer thickness H 00 and the
(llb.22) and (llb.38). Not apparent from these viscosity !J. have the largest influence on L b (see
Limits of operability 445
(11b.20». These are the most important parameters llb.3.5.2 Non-Newtonian fluids
to control in slide coating. A liquid with very
Many of the liquids used in practice are not
high viscosity cannot be coated very thinly
Newtonian. At the high shear rates (> 10 5 S -1)
because of bead break-up and an extremely low
close to the dynamic wetting line (Mues, Hens
viscosity will give rise to vortices or 'bleeding' in
and Boiy 1989) many polymeric liquids display
the case of very large layer thicknesses.
shear-thinning behavior as shown by Blake,
It should be emphasized, however, that equation
Clarke and Ruschak (1992) for gelatin-water
(11 b.22) and the criteria (11 b.32) and (11 b.38)
solutions. This makes the viscosity dependent
should not be considered as more than rules of
forces more difficult to estimate, but analytical
thumb. They merely indicate the parameters that
approximations can still be useful. When, however,
may influence the operation of the slide coater.
viscoelasticity is involved, these approximations
Any specific flow can be analyzed much more
become useless, because elastic effects have not
precisely by flow visualization and/or by numerical
been taken into account and viscoelasticity effects
simulation (as, for example, Fig. 11 b.9jTable
can be important, as we shall briefly discuss in
11 b.2, which also illustrates that values of L b
Section 11 bAA.
must be interpreted in a relative way).
llb.3.5.1 Multilayer coating Ilb.3.5.3 Surfactants

Relations (llb.l) to (llbAO) were derived for It is common practice to add surfactants to
single layer coating. In practice, however, multi- coating liquids (see also Chapters 4 and lId). As
layer coating is much more important. When the a result tangential forces may develop at the free
liquid properties (p, J1.) do not differ too much, surfaces due to surface tension gradients caused
approximate calculations can be performed by by the expansion and compression of the free
using some weighted viscosity as proposed by surfaces. Such gradients may not change the
Chen (1992): overall force balance, but they do influence the
exact shape of the free surfaces (Schunk 1989)
J1.- 1 = J1.;1QB/Q + J1.~1 QT/Q (l1bA3) and may therefore affect the incidence of vortices.
where Surfactant effects have not been taken into
consideration in the overall force balance approach
J1.B' J1.T;
Viscosity of the bottom and top used in this chapter so we cannot even think of
layers respectively finding a correlation whatsoever in this way.
QB' QT; Flow rate of the bottom and top Numerical simulation may be valuable in gaining
layers respectively an insight into their effect, but it is difficult to
Q = QB + QT collect all the experimental data necessary for
correct modeling of surfactant behavior. Experi-
When the differences in liquid characteristics ments, including flow visualization, therefore
are appreciable, analytical calculations become remain indispensable to this day.
too crude and confusing and flow visualization
and/or numerical calculations are the best
approach. It is the authors' experience that the
llb.4 LIMITS OF OPERABILITY
single layer analysis presented above can be
useful in planning multilayer experiments and Our intention in the previous chapter was
calculations and in interpreting the results. In to elucidate the flow field of a slide coating
general, multilayer coating with too large a bead bead for steady state two-dimensional flow with
height leads to vortices and with too small a the liquid satisfactorily wetting the web. The
bead height leads to break-up. low flow limit, the onset of vortices and
446 Slide coating
weepingjbleeding were discussed as limitations Table llb.3

of satisfactory operation. However, other restric-
tions to satisfactory operation also exist. a b
If the air adhering to the arriving web cannot Burley and Kennedy (1976) 0.335 0.67
be removed and replaced by the liquid, chaotic Gutoff and Kendrick (1982) 0.67
two-phase flow results. This phenomenon is Bracke, De Voeght and J008 (1989) 1.0 1.0
known as air entrainment, and although related
to the low flow limit discussed in the previous
section, it is really a different physical phenomenon. 1989). They all deduced a relationship of the
It will be covered in Section II b.4.1. Coating form:
flows can also rearrange spontaneously into a
(11b.44)
steady but three-dimensional pattern, known as
ribbing (Section 11 b.4.2). Finally, they can also Table 11 b.3 gives the values of a and b disclosed
oscillate spontaneously, producing a two-dimen- by the different authors.
sional but unsteady (periodic) flow (Section A dependency of UAE on the viscosity is clearly
IlbA.3). established by all the authors. There is, however,
All these limitations together determine the so no agreement between different investigators
called operability window, being the range of concerning the influence of surface tension on
parameters that permit satisfactory coating, i.e. UAE' The dependency on surface tension is,
steady two-dimensional coating without vortices. however, much less important because the (1-
This is addressed in Section 11 bAA. Moreover, it range used in practice is rather small (ca. 20 to
is desirable that small external disturbances do 72 mN/m). Such a narrow range of (1 makes it
not give rise to unacceptably severe coating difficult to determine its dependence precisely,
defects. The required coating quality will, of which could explain the lack ofagreement between
course, be dependent upon the specifications the different investigators (Kistler 1993). The
required by the user. The susceptibility to minor influence of other parameters such as the surface
variations in coating conditions and the resulting properties of the web have been studied, but their
quality window will be covered in Section 11b.5. influence is not clearly established.
The air entrainment limit in a slide coater is
close to the air entrainment limit in a plunging
tape geometry. Gutoff and Kendrick (1987)
11b.4.1 AIR ENTRAINMENT AND LOW FLOW
investigated UAE in slide coating with Newtonian
LIMIT
liquids whose viscosities ranged from 21 to
As mentioned above, wetting of the web, which 21I0mPas. At layer thicknesses larger than
is common to all coating systems, is one of the approximately 100 Jlm the air-entrainment limits
main functions of the bead. Wetting process in both geometries are reported to be almost
failure delineates one of the operational limits of identical, and the limit in slide coating is hardly
the slide coater. Plunging tape experiments (see influenced by the pressure difference !1P.
Chapter 3) show that the wetting process fails When coating thinner layers, however, the
when the dynamic contact angle eo reaches 180· bead collapsed at much lower coating velocities
and air entrained by the web is not displaced by (Fig. 11 b.5). This phenomenon, known as low
the liquid. This occurs at a critical velocity UAE • flow limit, has already been discussed in Section
As the physics of the wetting process is still lIb.3.3. In this case a pressure difference over
poorly understood, the critical velocity UAE has the bead allows substantially thinner layers to
been studied experimentally by several investi- be coated. A possible physical explanation might
gators (Burley and Kennedy 1976; Gutoff and be that the pressure difference, apart from
Kendrick 1982; Bracke, De Voeght and Joos increasing the bead height as mentioned before,
affects the position of the dynamic wetting line stable flow on the slide and satisfactory high
in such a way that more of the momentum of speed wetting can be met.
the liquid arriving from the slide becomes available
to assist the dynamic wetting. (Schweizer 1993:
llb.4.2 RIBBING - A STEADY
personal communication). The momentum ofthe
THREE-DIMENSIONAL FLOW
liquid impinging on the web is also significant
in curtain coating (see Chapter 11c), where it is In Sections 11 b.3 and 11 b.4.1 we have already
much more important than in slide coating (Van mentioned that the limit of coatability can be
Abbenyen, Mues and Goetmaeckers 1992; Blake, extended to thinner and more viscous layers by
Clarke and Ruschak 1992). applying a small pressure difference over the
Electrostatic forces can also be used to assist bead (0 to 1000 Pal. However, at sufficiently high
the wetting of the web by the liquid and to pressure differences a steady regular pattern of
increase the critical velocity UAE (De Geest and lengthwise streaklines, known as 'ribbing', is
Verkinderen 1967; Nakai and Okada 1968). This formed. Other names used to describe this defect
can be achieved by giving the substrate a surface include 'vacuum streaks', 'corduroy', 'rakelines'
charge, or by connecting the backing roll to a and 'phonography', The wavelength of the ribs
high voltage power supply. Feng and Scriven is typically a few millimeters. Although it is
(1992) illustrated these effects by numerical difficult to detect precisely the onset of ribbing,
modeling. Unfortunately, the obtaining of a it is clear that, in practice, a maximum pressure
perfectly homogeneous distribution ofelectrostatic difference limit exists in the operability window
charges appears to be difficult, as illustrated by beyond which the ribbing pattern becomes
the large amount of patents claiming improve- unacceptable (Saito, Ishizuka and Fuchigami
ments (e.g. Nakajima and Miyamoto (1993); 1982).
Hartman (1989)). It may thus impair coating When surfactants are present, extension and
uniformity, making electrostatic assist unusable compression of the free surfaces causes local
whenever high coating quality is required. changes in the surface concentration (adsorption)
In multilayer coating UAE is determined by the of the surfactants. This effect is highly dependent
local viscosity of the layer that wets the web. The upon the time scale. It is quantified by the surface
operability limits of slide coating can therefore elasticity, being defined as the derivative of the
be significantly widened by introducing an surface tension with respect to the relative change
additional layer underneath the other layers. of the surface. Valentini et al. (1991,1992) have
This layer, often called the carrier layer, should studied the influence ofsurfactants on the ribbing
be relatively thin and should have a low viscosity, phenomenon. Their work indicates that surface
easily wet the web and yield high UAE values elasticity suppresses the onset of ribbing.
(Dittman and Rozzi 1977). In this way liquid The influence of other coating parameters is
layers with high solid contents can be coated at not clearly documented in the literature on slide
high viscosities and at high speed without air coating.
entrainment problems. A low viscous liquid layer The spontaneous appearance of ribbing is not
underneath can, however, endanger the stability limited to slide coating; it is well known in roll
of the multilayer flow on the inclined plane (see coating, blade coating and other flows with free
Chapter 8 and Section 11 b.4.3.1). Therefore, surfaces (see Chapters lla and 12a). The flow,
Choinski (1978) recommended the use of a expected to be steady and two-dimensional, loses
shear-thinning lowermost layer, having a high stability to three-dimensional disturbances and
viscosity at the low shear rate on the inclined cross-web waves or ribs start growing. At some
plane and a lower viscosity at the very high shear finite amplitude of the ribs the flow regains
rate near the wetting line (Mues, Hens and Boiy stability and becomes steady, but remains three-
1989). In this way, conflicting requirements for dimensional (Savage 1992).
448 Slide coating
Numerical simulation can predict whether or permitting the use of a higher pressure difference
not ribbing will occur, as well as the wavelength without the occurrence of ribbing (Isayama and
that is likely to appear at the onset of ribbing. Takehara 1980; Choinski 1981). According to the
Such calculations were carried out and experi- authors' experience, however, the manufacture
mentally verified for roll coating by Coyle, of such dies is complicated and they are hard for
Macosko and Scriven (1990). For slide coating the operators to handle. Only in very special
the onset of ribbing was calculated by Chris- cases might the benefits balance the drawbacks.
todoulou (1990), but experimental verification is
not yet forthcoming.
llb.4,3 BARRING-A TWO-DIMENSIONAL BUT
The wavelength that can be calculated is the
UNSTEADY FLOW
one that appears at the onset of ribbing. This
wavelength may change as the flow parameters Barring (also called chatter, chatter marks or
exceed their critical values at the onset of ribbing crosslines) is a coating defect caused by a two-
(Mill and South 1967). The amplitude of the ribs dimensional but unsteady flow. Oscillations occur,
that appear in the steady state flow cannot yet either spontaneously or excited by external
be predicted to this day. disturbances, resulting in a coated product showing
The exact cause of the instability in the case periodic cross-web thickness undulations. Section
of slide coating is not yet completely clear. A llb.S deals with barring as a result of external
Taylor-Gortler mechanism (basically centrifugal disturbances such as mechanical vibrations or
instability) was suggested, but immediately rejected pressure difference variations and in Section
again as unlikely (Schweizer and Scriven 1983). llb.6.2 we shall discuss the effect of periodic
Hens and Boiy (1986) contended that a higher pointwise disturbances. In the remainder of this
pressure difference resulted in a greater adverse section we shall focus on spontaneously unsteady
pressure gradient in the upper bead region due flow.
to capillarity, this in analogy with the adverse The instability may originate on the inclined
pressure gradient hypothesis put forward for the plane, in a non-horizontal web section after
roll coating case (Coyle, Macosko and Scriven coating or in the bead itself. The defect will have
1990). The most recent hypothesis (Christodoulou a different appearance depending on the location
and Scriven 1989) attributes the 3-D instability of the instability.
to an adverse normal stress gradient in the bead.
Although none of these theories has been
Ilb.4.3.1 Flow instabilities on the inclined plane
confirmed so far, there is evidence supporting
the most recent one. Schunk (1989) reviewed the The flow of one or more layers on the inclined
scarce data in the literature on the effect of plane is prone to instabilities: an infinitesimal
polymeric additives and concluded that these are perturbation of the free surface profile may grow
mostly destabilizing with respect to ribbing. He while traveling downwards. Wave speed and rate
also calculated numerically a slide coating flow of damping or amplification can be calculated
with extensional thickening fluid, which suggested from the Orr-Sommerfeld equations, as described
that the adverse normal stress gradient increases in Chapter 8. These equations describe an infinite
due to the extensional thickening effect. This is number of wave modes, most of them fortunately
not at variance with the hypothesis of Chris- highly damped. Nevertheless with N layers, N
todoulou and Scriven (1989), but it does not con- leading modes can be associated with the free
firm it either: more evidence in support is required. surface and the interfaces. For each mode there
Special coater geometries with 'upswept lips' can be a most amplified wavelength, eventually
have been proposed, delivering the liquid from the most amplified wavelength of the dominating
the slide with a velocity component in the (i.e. most amplified) mode, if any, will be visible
direction of the movement of the web and in the coated product.
Analytical solutions are available for the limiting the coating unevenness became unacceptable
case of long waves, i.e. with a wavelength much beyond a threshold value. In his experiments
larger than the layer thickness on the slide. Both with photographic material this threshold value
viscosity and density stratification can be de- Amax was approximately 2000.
stabilizing, as shown by Kao (1968) and Wang, Figure 11 b.l0 shows the dimensionless growth
Seaborg and Lin (1978). The effect of density is distance Le,min/Hs,bottom for a two-layer coating
less important, however, because the possible as a function of the viscosity ratio m =
jump in density at an interface is very much f-ltop/ f-l bottom' as obtained by solving the
smaller than the possible jump in viscosity. Orr-Sommerfeld equations numerically
Weinstein (1990) extended the calculations to (Kobayashi 1992). The two modes, the surface
shear-thinning fluids, and showed that the surface mode with a characteristic wavelength of the
mode is governed by some mean value of the order of 10 cm and the interface mode with a
viscosity, while for the interface modes the jump typical wavelength of the order of 1 to lOmm,
of the local viscosities at the interface matters. are represented. When m is close to unity, the
Weinstein and Kurz (1991) clearly showed that surface mode dominates. When m> 1.4 or
a three-layer case is destabilized by a middle m < 0.73 the interface mode is the most dangerous
layer that has a lower viscosity or lower density one, the smallest value of Le,min occurring when
than the adjacent layers. the top layer viscosity is substantially lower than
Smith (1990) elucidated the physical mechanisms the bottom layer viscosity. Generally, an interface
that give rise to wave amplification on an inclined mode is more problematical than a surface mode
plane (see Chapter 8). because there is no interfacial tension acting as
The long-wave condition, however, is not a damping force (on the assumption that miscible
always met in slide coating and consequently fluids are concerned).
numerical methods have to be used to solve the When more than two layers are involved,
problem. Kobayashi (1992) investigated the Kobayashi (1992) recommended viscosities de-
behavior of multilayer flows down an inclined creasing from top to bottom with the viscosity
plane and calculated that an interface mode is ratio of adjacent layers as close to unit as
always unstable, as soon as the viscosity of the possible: smaller than 2, preferably smaller than
adjacent layers becomes different. 1.5. This seems to be logical. When two layers
In practice, however, it is perfectly possible to have the same characteristics, they will act as a
apply a number of layers, stratified in viscosity single layer and interfacial modes will be absent.
and density, with a very good coating quality. In practice, however, it is perfectly possible to
Consequently, slide flows that are theoretically deviate from this rule of thumb, as illustrated by
unstable can be 'practically stable': if it takes a several publications (Dittmann and Rozzi 1977;
very long distance (compared with the length of Koepke et al. 1984), although according to
the slide) for a small wave to grow up to a Weinstein and Kurz (1991) one has to avoid a
significant amplitude, the instability will not be middle layer that has a lower viscosity or a lower
reflected in the final product. Kobayashi (1992) density than the adjacent layers.
therefore defined the growth length L e as the When dealing with fluids used in practice the
length required for a perturbation to grow by a situation becomes even more complicated, even
factor e. Le,min is then the growth distance of the for the very simple one-dimensional flow on an
most amplified wavelength and an amplification inclined plane. Surface elasticity (due to surfactants)
factor Amax relates L e , min to the length L s of the for instance brings additional damping (Schunk
inclined plane: 1989; Baumlin 1990; Gutoff 1993; see also Chapter
lId). Viscoelasticity is destabilizing at small
(11b.45)
Reynolds numbers and stabilizing at moderate
Kobayashi (1992) determined empirically that Reynolds numbers magnifying both magnitude
450 Slide coating
103 ~---------'",-------r----, 10 ~-----~--------,
···········0....
"0
·······0
Qi
C\ .c
E
"''0.
!
:J
<:
..Q.... .' OJ
>
til
'0'''''0'...0''
~
e'"
/
:J
OJ
Cl
<:
til ...... Surlace mode
"C
V; 10-1 ·0· Interlace mode
.p' o
~
~/....
....•..~ ...... Surlace mode
..... ·0· Interlace mode
1 '--_.J..---'--'-.L..J....L..l.J..J....._-'---'--L....L....l--L.J..1J 10-2 '--_.L--'--'--L.L.L.U..L-_..J.....-'---'-.L...L~

0.1 1 10 0.1 1 10
Viscosity ratio (m) Viscosity ratio (m)
Figure llb.l0 Dimensionless growth length (left) and most dangerous wave number (right) of a two-layer
flow on an inclined plane as a function of the viscosity ratio. From: Kobayashi, c. 1992. Stability analysis of
film flow on an inclined plane. I. One layer, two layer flow. Ind.Coat.Res. 2: 65-88. (Reproduced by permission
of the editor.)
and wavespeed of nonlinear waves (Salamon, plane, only with the layer thickness being an
Armstrong and Brown 1992). Furthermore, Van order of magnitude smaller (100 ~m rather than
Abbenyen and Mues (1992) suggested that diffusion 1 mm). If the flow is unstable a wavy pattern
might play a destabilizing role by creating a results with typically a sub-millimeter wavelength.
watery layer between two layers of different The waves are across the web, but not completely
chemical composition. straight. This defect resembles the structure of a
We conclude that, despite the study on an beach at low tide and is known as runback or
inclined plane by many authors, it is still, at least sagging (see Chapter 6).
in multilayer coating, incompletely understood.
Nevertheless the available theories are helpful
and provide a reasonable approximation, some
11b.4.3.3 Bead
researchers indeed reporting qualitative agreement
between experiment and prediction (Baumlin In the previous sections we discussed instabilities
and Pasquet 1988; Kobayashi 1992). giving rise to waves on the inclined plane and
on the web, both fairly straight zones ofthe flow.
It is also possible that the bead itself oscillates
11b.4.3.2 Non-horizontal web section
as a whole. The coated web then exhibits waviness
Flow in a non-horizontal web section after a with a fairly straight pattern of cross-web 'bars'
coating station is similar to flow on an inclined with a wavelength of the order of centimeters.
Whether this oscillation occurs spontaneously

or results from external perturbations is often Air Entrainment
difficult to determine in practice. When the defect
occurs at once under manufacturing conditions,
it is likely that an external disturbance will be
the cause. Katagiri and Scriven (1986), however,
predicted the existence of an oscillatory state. Low
Katagiri (1988) experimentally confirmed his Flow
Limit
predictions and by applying linear stability analysis
Christodoulou (1990) illustrated the physics Vortices
involved (see also Chapter 9). He showed that
an oscillatory state is characterized by a pair of
Reynolds Number (linear)
complex conjugate eigenvalues with a vanishing
real part, the corresponding eigenmode then Figure llb.ll Possiblefonn ofan operability window
being a periodic vibration that is neither damped of slide coating in dimensionless representation.
nor amplified. Nearby states, with their leading
eigenvalues having a small but negative real part, 11 b.ll. In such a diagram Schweizer (1988)
will be stable but very sensitive to external represented his upper vortex correlation (11 b.41),
disturbances (see Section 11 b.5). only using the property number Po (a function
Fruhner, Kragel and Kretzschmar (1989) of Re, Ca and St, see (llb.42)) as a parameter.
showed that the bead is likely to start oscillating To the authors' knowledge, such dimensionless
when the coated layer thickness Hex:> is small and diagrams have not yet been published.
the viscosity JJ. is large, i.e. near the low flow limit. This may be due to the development and
practical use ofslide coating long before theoretical
analyses were published, with practitioners
Ilb.4.4 OPERABILITY WINDOWS
preferring dimensional variables to dimensionless
The operability window is the region in the numbers.
multidimensional parameter space in which a The coating parameters include liquid properties
stable, steady, two-dimensional flow is possible. such as p, (J and JJ. (or the rheology in general),
Depending on the requirements of the coated geometrical parameters such as gap width H G ,
product, parts of the operability window may be application angle iX, inclination of the slide p,
unusable due to certain combinations of par- length of the slide L s' and process parameters
ameters producing a coating quality, which is such as web speed U, wet layer thickness Hex:>
too sensitive to external disturbances (see Section and pressure difference AP.
11 b.5). The rest of the operability window can Most of these parameters cannot be chosen
be termed the quality window. In this region the freely. For an existing coating station iX, p and
flow is steady, stable, two-dimensional, free of L s are fixed. Possible variation in (J and p is very
vortices and sufficiently unresponsive to external limited and the same is true for the coating gap
disturbances from the surroundings. H G • Viscosity JJ. and wet layer thickness Hex:> can
In principle, one could represent the operability usually be varied, but within limits and not
limits in a multidimensional space, e.g. in terms independently of each other (because the dry
of the dimensionless numbers We (or Re), Ca, layer thickness is normally fixed, and viscosity
St*, N L1P and N es ' suggested by equation (1Ib.22). depends upon concentration). Because for econ-
Even when neglecting N es ' a four-dimensional omical reasons one wants to coat as fast as
space remains. A two-dimensional section rep- possible, provided the quality remains acceptable,
resenting Ca as a function of Re with St* and the degrees of freedom in viscosity and layer
N L1P as parameters, would probably look like Fig. thickness are usually used to maximize the
452 Slide coating
coating speed and quality according to the

principles discussed in Sections 11 b.3 and II b.4
so far. Hence only one parameter remains really
free: the pressure difference over the bead AP. Vortices
There are, in practice two types of two-

dimensional sections of the operability diagram
in use. In the first H 00 versus U and H 00 versus
fl. diagrams are used to illustrate the practical Low
implications of the physics of slide coating. They Flow
Limit
show the operability limits as described above,
with air-entrainment at U > UAE , low flow limit
of coatability at small H 00 and large fl., and the
occurrence of vortices at very large H 00 and small (a)
VIScosity iJ (Log)
fl.. Figure llb.12 is a qualitative example. In

practice, it is well known that slide coating is
unsuitable for coating very thin layers or highly
viscous liquids and certainly not for a combination
of both. Equation (llb.35), despite its being a Vortices
very rough approximation, gives an indication
of why this is the case. It is equally well known Air
Entrainment
that thick layers of low viscosity cannot be
applied with great uniformity, due to the appear-
ance of vortices, as discussed in Section llb.3.4
and indicated by equation (llb.40).
In the second, AP versus U diagrams are used
to determine the coatability range, the maximum Low Flow Limit
and minimum allowable pressure difference with
a given set of parameters indicating the proximity
to the operating limits of coatability. This type (b)
Web speed U (Log)
of diagram seems to be very popular in practice,

probably because it is also quite easy to obtain Figure llb.12 Typical form of two coating windows
experimentally. of slide coating: (a) layer thickness H co versus viscosity
An example of what such a AP versus U Jl on a semilogarithmic scale; (b) layer thickness H co
diagram with constant values of H 00' fl. and H a versus web speed U on a logarithmic scale.
might look like is represented in Fig. 11b.13. At
high AP values, ribbing (see Section 11 b.4.2) and
bleeding (see Section 11b.3.4) delineate the coating ribbing is deferred to much higher AP values.
window. In practice it is known that at low AP Gutoff and Kendrick (1987), on the other hand,
the sensitivity to disturbances increases, setting showed that a polyvinyl alcohol solution can
a limit to the quality window. At even lower AP have a much larger 'low flow limit of coatability'
values, the bead edges break loose and the bead than a 'corresponding' Newtonian liquid. It was
will break up (low flow limit), either completely also mentioned above that ribbing is promoted
or in rivulet form. by viscoelasticity.
This figure 11b.B is typicaifor a gelatine-water In general we conclude that it is usually wise,
coating, but these diagrams may be rather different if not necessary, to check the operability window
when working with other fluids. The diagrams experimentally with the coating liquids that will
given by Chen (1992) were for glycerine, in which be used in practice, because the rheology of the
Sensitivity to disturbances 453
900.-----------------------------,
800
lil
e:, 700
a..
<I 600
Q)
() 500
c:
Q)
Qj 400
;:
"'0
~ 300
::J
(/)
~ 200 Edge Failure

Air
a.. Entrainl"(1ent
100
Low Flow
o Limit
o 25 50 75 100 125 150
Web Speed U (m/min)
Figure llb.I3 Example of a practical coating window of a slide coater: pressure difference !:l.P versus web
speed U on a linear scale. Data are from 10% gelatin in water, viscosity 20 mPa s, surface tension 32 mN/m.
fluids used may considerably influence the coating process, do not produce a nonuniform product.
flow. Whatever the precautions taken, such disturbances
cannot be avoided completely, just minimized.
There are many sources ofexternal disturbances.
llb.5 SENSITIVITY TO DISTURBANCES
The coating die may oscillate if it is not properly
It is obvious that, in order to be ofany importance isolated from its environment, in which all kinds
in practice, a coating flow has to be stable (or of mechanical vibrations can be present. The
practically stable): a small disturbance may cause backing roll may have a certain runout. In both
a defect, but the flow should recover to its cases the gap width is not perfectly constant, but
original state when the disturbance is over (or fluctuates periodically.
the small disturbance may become amplified, but Acoustic noise, e.g. originating from the fan
to such a minor extent that the quality of the generating the pressure difference, is nothing but
final product remains within tolerances). Several a periodic disturbance of the pressure difference
unstable states cannot even be realized experi- across the bead, causing forced oscillations of
mentally because any disturbance will eventually the bead by exciting the lower free surface.
cause the bead to break up. Numerically, however, Web speed and flow rates usually are auto-
steady states, either stable or unstable, can be matically controlled and maintained between
calculated. Eigenvalue calculations or transient very narrow limits, but small periodic fluctuations
behavior will reveal the (in)stability of the flow about the set-point can always occur.
(Katagiri and Scriven 1986; Christodoulou 1990; Furthermore, air currents can disturb the flow
see also Chapter 9). on the inclined plane and in the bead and the
In practice, only that part of the operability web to be coated is also likely to have small
window can be used, in which small external variations in wettability that may influence the
disturbances, inevitably present in any industrial coating quality.
454 Slide coating
The degree of resulting nonuniformity depends multilayer slide flow (i.e. the transfer function
upon engineering design parameters such as the between the periodic disturbance ofsome external
mechanical robustness of the coating equipment, parameter (e.g., flow rate) and the resulting
the quality of process control (determining the variation in some flow characteristics (mostly the
amplitude of the possible disturbances) and the layer thickness at the exit of the bead) as a
susceptibility of the flow to small perturbations. function of frequency) is strongly dependent
To obtain the optimal coating quality in a given upon the type of perturbation (Fig. llb.14). In
coating facility, one should choose the parameters the two-layer slide-coating flow they modeled,
of the flow in such a way that susceptibility to web speed variations caused defects at low
disturbances is minimized. frequencies with a resonance peak at about
Several investigators have established that the 20 Hz, the flow field indicating that this resonance
flow in slide coating becomes extremely sensitive mode was associated with the upper free surface.
to disturbances near the low flow limit (Hackler Horizontal vibrations of the backing roll on the
et ai. 1992; Fruhner, Kragel and Kretzschmar other hand were found to be particularly dangerous
1989). Indeed, at small L b values, the liquid mass at very high frequencies, because the lower free
in the bead is relatively small, which may intuitively surface seemed to show a resonance mode at
result in a greater susceptibility to disturbances several hundred hertz. They also found evidence
such as vibrations of the die or the backing roll, of so-called 'exchange modes': the reflection, at
variations in vacuum, unevenness of the web, certain distinct frequencies, of perturbations in
fluctuations in the premetered flow rate, etc. the flow rate of one layer in thickness variations
Neither theoretical analysis nor experimental of another layer.
study of the dynamic behavior of a flow is easy. In our own experience, the shape of such
Weinstein, Baumlin and Servant (1993) recently calculated frequency responses is rather typical:
studied the effect of external perturbations on amplitudes may change significantly and the
the flow on an (oscillating) inclined plane and peak frequencies may shift a bit as coating
showed both theoretically and experimentally parameters (layer thicknesses, viscosities, web
that the entrance region is responsible for exciting speed, pressure difference, etc.) are varied, but
the various wave frequencies which are observed the appearance of the different response curves
downstream. remains largely unchanged.
The dynamic behavior of a coating flow as a Remarkably, the dominant eigenfrequencies of
whole is even more complicated. There are the upper and the lower free surfaces can be
several eigenfrequencies and the response to estimated from a rather crude approximation.
perturbations cannot be predicted without An undamped second-order system is known to
numerical calculations. Such calculations were have an eigenfrequency
first made by Katagiri and Scriven (1986) and
later more efficiently by Van Abbenyen, Chris-
no = (lj2n) * (kjm)1/2 (llb.46)
todoulou and Scriven (1988), assuming that the where k is the stiffness or the spring constant,
disturbances were small enough for the response and m is the mass of the system. The mass of the
to remain linear, an assumption that was found coating bead per unit width is estimated to be
to hold quite well for disturbances of web speed
m ~ pLbHG (llb.47)
and pressure difference with amplitudes of up to
10%, as shown by comparison with nonlinear A rough estimate of the spring constant k may
time-dependent calculations. Hackler et ai. (1992) be obtained from the ratio of the surface tension
extended this type of calculation to three- to the length over which it is active, which for
dimensional periodic disturbances. the upper free surface is of the order of L b and
Van Abbenyen, Christodoulou and Scriven for the lower one is H G . It can be easily verified
(1988) showed that the frequency response of a that, for the coating flows used in practice, these
Sensitivity to disturbances 455
WEB SPEED FLOWRATE LOWER LAYER
- -
2.4 <> <> Lower 2.4 <> <> Lower
{;---{; Upper {,---{;
Upper
~ 2.0 ~ 2.0
_0
cu-
_0
cu- ••
e (/) .!: ell
u::: ell
C1l
1.6 LLeIl
C1l 1.6
- e .' , -e
O.:II.
e .~ 1.2
."'"1\. 0..:.::
1.2 ,
.
<I'
o .... 11
'II. .• t e.~ ,
o..c o..c
00 0 /'
;: .... \'/\ ....

....... -3.i_Z· X-Ilo - Ilo - Il' 0000. 0 °&
;: d:
cu • ...
/lo
~ 0.8 .~ ~ 0.8
'0: C1l "
~C1l
cu >- cu>- ,
. .'.. .•
·e·... '
> ...J
cu 0.4
'. >cu 0.4 ,
...J .' ,
'
... tlf4
"
0.0 0.0 0" • ~''''

10° 10 1 102 103 10° 10 1 10 2 10 3
(a)
Frequency (Hz) Frequency (Hz)
(b)
GAPWIDTH (WEB VIBRATION) VACUUM
--
2.4 <> <> Lower 0.30 <> Lower
{;---{; Upper .:.---.:. Upper
~ 2.0 _?j 0.25
ea
. ..
0
e cu-
- ....
.' , "
u:: (/) ' .!: ell
(/) 1.6 LL ell 0.20 .. ,:
C1l C1l
0 e -0..:.::
e
.
.:II.
e 1.2 '.
.~ e.~ 0,15
;: .c
ea ....
0 O..c 00,,," :-'.
i: C1l
ea
~
0.8
. ....,,::
'.
.-
:; .... 0.10'
~
> ea>-
~C1l
0.4 . cu>-
...J >~ 0.05
o. 0 +..,,-..........,....,...,........,~.,........,.."T'T"'~;::--r--,-.,.-.-r..,.,.j
1 2
100 10 10 10 3
Frequency (Hz)
(c)
Frequency (Hz) (d)
Figure llb.14 Some frequency responses of a two-layer slide coater. (Amplitude A% means: 1% variation
in the considered parameter causes A% variation in the thickness of the (upper or lower) layer). From: Van
Abbenyen, W., Christodoulou, K. and Scriven, L. E. 1988. Frequency response of a coating flow: predictions
for slide coating. University of Minnesota Supercomputer Institute, Report UMSI 89/54, March 1989.
approximations yield eigenfrequencies of the agreement with numerically calculated frequency

order of 10 to 50 Hz for the upper free surface, responses. The resonance frequencies do agree,
and of the order of 50 to 200 Hz for the lower one. but the amplitudes are not quantitatively correct
Slide coating experiments show only limited (Katagiri 1988; 1990; Van Abbenyen et al. 1990;
456 Slide coating
Bussmann, Hofmann and Beck 1992). This is or slowly moving liquid (Russell 1969), deposition
probably due to the complexity of this kind of of particles can be prevented (see also Chapter
experiment. 15, where this matter is discussed in the context
As yet frequency response calculations are not of die design optimization). Furthermore, the
entirely verified, but may be helpful because liquids should be adequately filtered before coating.
frequency responses yield interesting dynamic Discontinuous three-phase contact lines
information for the coating engineer. Such (air-liquid-solid or liquid I-liquid 2-solid) are
information, however, should be used with care: another important cause of streaks. Pinning the
interpretation by an expert 'skilled in the art', as free liquid surfaces or interfaces on a sharp edge
well as some experimental work remain indis- (see Section llb.2.7) is a way of ensuring that a
pensable. contact line remains straight. Encapsulating the
coating equipment in an enclosure filled with
saturated gas (Ishiwata et ai. 1980) has been
llb.6 COATING DEFECTS IN SLIDE COATING proposed to avoid local drying, another source
of streaks that is especially relevant when dealing
A coating defect always results from either
with volatile liquids.
unstable coating conditions or, when operating
Scratches in the underlying layers may also
under stable conditions, by unwanted but identi-
show up as severe streaklines, even if the defect
fiable external causes. The former type of defects
on the web before coating may seem insignificant,
have been dealt with in Sections 11 b.3 and 11 bA,
as in the case of a scratched 'subbing layer' (i.e.
the latter will be addressed in this section. We
the very thin layer often applied to the base to
shall successively consider lengthwise streaks,
promote adhesion).
chatter, fat edges, spreading problems, blotchiness
Splices crossing the bead may disturb the static
and point defects, which are probably the most
contact line at the die lip, particularly when this
frequently encountered defects in slide coating,
contact line is not pinned and the splice quality
but the list is by no means complete. A more
is poor. Furthermore, air may get trapped at the
complete survey of coating defects, not limited
trailing edge of the splicing tape, preventing
to slide coating, has been presented by Gutoff
rewetting of the web surface. Steps can be taken
(1993).
to facilitate rewetting, e.g. temporarily increasing
the pressure difference over the bead (Fowble
1975; Finnicum, Pegoni and Story 1991), locally
Ilb.6.1 LENGTHWISE STREAKLINES AND
roughening the web surface (Takagi et at. 1975),
'BANDEDNESS'
or hydrophobizing the tape zone (Bourns and
Lengthwise streaklines, not appearing in a regular McDonald 1969). Splicing the web at its backside
pattern, such as ribbing, but rather randomly, (Verkinderen et al. 1979) prevents air getting
are the most encountered defects in slide coating. trapped at the trailing edge of the splicing tape.
'Bandedness' can be considered to be a weak, The problems related to the passage of splices
smoother form of streaklines. 'Real' streaklines can of course be avoided by a die withdrawal.
can be very sharp, and disastrous for coating Unfortunately the start-up of coating after
quality. They can be caused in many ways, the interruption brings its own difficulties. Indeed,
most important of which will be discussed below. at the first contact between the multilayer liquid
Particles can become trapped, somewhere on and the web, the layers are deposited upside
the inside of the die (Vrahopoulou 1992), at the down. Normally the flow reorganizes itself
exit of the slots on the inclined plane, or even at spontaneously in less than one second, but a
the die face lip or on the inclined plane (Pozrikides thick zone is generally formed immediately
and Thoroddsen 1991). By correctly designing downstream from the line where the liquid starts
the coater geometry to avoid zones with stagnant wetting the web. As the drier will generally be
Coating defects in slide coating 457
incapable of handling such a thick coating, the web is moving, resulting in thickness variations,
result is a dirty machine and a sticky web at the at an angle with the web direction.
winding station. The thick starting zone can be The sensitivity of the flow to disturbances is
thinned by increasing the pressure difference over addressed in Section 11b.5. How such disturbances
the bead, thereby reducing the thickening of the can be eliminated will not be discussed here. We
layer near the end of the slide when the die just refer the reader to the article by Gutoff
approaches the web during start-up (Johnson (1993), because this problem is common to
1965). Local roughening of the web surface can several coating systems.
also be advantageous (Takeda and Oyama 1976).
An alternative start-up procedure relies on a
llb.6.3 FAT EDGES
combination of electrostatic assist, a temporarily
narrower coating gap and deliberately altered Thick edges at both sides of the coating width
flow rates of the different layers during start- occur in nearly all coating systems (see Chapter
up (Willemsens, Van Abbenyen and Criel 1989). 3). The patent literature abounds in tips to
Apart from patent literature, there is little work minimize their effects once they have been formed.
published on start-up in slide coating. Owing to In slide coating thick edges can already originate
the complexity of the process, scientific research in the die (Vrahopoulou 1991). They can also
has only recently been started (Fenoglio et al. arise on the inclined plane, when the edge guides
1990; Palmquist et al. 1992). are improperly designed, i.e. if their height differs
from the thickness of the liquid layer on the slide,
the liquid-air interface is curved and a pressure
llb.6.2 CHATTER AND CHEVRONS
gradient along the interface arises. Under such
When a coating suffers from chatter, the cause conditions liquid gets transported sideways,
is not immediately clear. Either there is hydro- creating a thinner and a thicker zone on the slide.
dynamic instability (barring, see Section 11 b.4.3), Furthermore, the thinner zone may give rise to
in which case the problem can be solved by 'necking-in' of the coating bead ('edge failure').
changing the coating parameters, or some external The coating edges thus formed are, of course,
disturbance may induce the defect, which requires extremely thick: up to twice the nominal layer
the identification of the disturbance and its thickness or even more. During the drying process,
elimination or at least minimization (see Section thick edges can develop due to locally lower
11 b.5). Externally induced chatter or crosslines surface tension as a result of faster evaporation
can have many causes, which can be identified there - provided that the liquid viscosity is still
by measurements. The most common are mech- low enough to allow flow.
anical vibrations (affecting the gap between the The formation of thick edges on the inclined
coating die and the backing roll) and pressure plane can be prevented by edge guides with the
variations (affecting the upper or the lower free height of the surfaces facing the liquid flowing
surface of the bead). Other causes of chatter are down the incline equaling the thickness of the
flow rate variations in one of the liquids and web liquid layer (Barstow 1966). As slide coaters are
speed fluctuations. generally used for many products with different
A particular form ofchatter is called 'chevrons': layer thicknesses on the incline, it has been
thickness undulations that are not cross-web, but proposed to use edge guides with surfaces sloping
at a diagonal angle. The cause of this defect is a from the inclined plane towards the outside
periodic disturbance that is not uniform over the (Koepke et al. 1982). Alternatively it is possible
width of the coating but rather located at one to avoid the thick edges in certain cases by
fixed point, often the edge of the coating meniscus. choosing the width of all but the uppermost slot
Such a disturbance initiates a wave traveling smaller than the desired coating width (Willemsens
from the edge to the center of the web, while the 1980).
458 Slide coating
11b.6.4 SPREADING PROBLEMS bead and cause defects, which can affect areas
ofthe web surface that are larger than the particle
Multilayer coating can be faulty due to surface
dimensions. Indeed, at constant flow rate a
tension effects. On the inclined plane the uppermost
thinner coated area must necessarily be followed
layer(s) should not neck in, but spread evenly over
by a thicker one, after the dust particle has left
the underlying ones. To achieve this it is sufficient
the bead (Do and Scriven 1991,1992). When a
that the surface tension of the layers decreases
particle sticks to the backing roll, the defect is
from bottom to top by one or more mN/m per
repeated every time the particle deforms the bead
layer. This condition is not necessary, however,
zone.
as only the uppermost layer (in contact with the
Particles, gel slugs or bubbles in a coating
air) has to have a lower surface tension than the
solution can create unwanted defects, the size of
other layers (Bacon and Fogg 1969; Do and
which are much larger than the particles them-
Scriven 1991). It may be difficult, however, to start
selves, due to flow effects. Particularly in multilayer
the coating or to recover from temporary defects
coating such defects can be devastating because
when the surface tension is not properly stratified.
the interface becomes disturbed and layers may
Defects due to spreading problems, such as
become partly interchanged. Spreading phenom-
craters, repellencies, compass, comets, fish eyes,
ena can also occur, as explained in Section 11b.6.1.
etc. can also arise after coating. Such defects are
Coating defects caused by particles cannot be
also encountered in other coating systems and
eliminated by improving the coating technique.
therefore will not be treated here. They have been
Steps must be taken to remove particles from
discussed by Gutoff (1993) and by Do and
the web and the backing roll (scraper, airknife,
Scriven (1991). For a discussion of spreading
vacuum brush system, etc.) and from the coating
problems the reader is referred to Chapters 4 and 6.
solutions (filtering) before coating.
11b.6.5 BLOTCHINESS OR MOTTLE

llb.7 CONCLUSIONS
Blotchiness or mottle is a coating defect which
is exhibited as random thickness fluctuations Slide coating is a versatile method of premetering
without any regular pattern or direction. The and assembling multiple layers and applying
most obvious cause of this defect is an uneven them simultaneously to a web, with an excellent
air flow over the coated layer before the coating uniformity and within an industrially interesting
is chilled or sufficiently dried down. operability range. Some of these layers may even
Mottle can, however, also be caused by be so thin or so viscous that they cannot be
inhomogeneous properties of the substrate, such coated separately.
as nonuniform surface energy or electrostatic In obtaining high quality products, with other
charges, which affect the dynamic contact angle (single layer) coating systems, the production
at the wetting line. Blotchiness is also observed speed is rather limited. Multilayer slide coating
in other coating systems, but a slide coater is can meet the same or even higher quality standards
particularly sensitive to it, as the relatively small at much higher speeds. In the next chapter it will
bead height means that contact angle fluctuations be shown that curtain coating can reach even
can temporarily disrupt the balance between higher speeds with multilayers and comparable
delivered and entrained flow rate. The reader is quality.
again referred to Gutoff (1993). Blade coating and elastohydrodynamic systems
can coat at much higher speeds, but with a
limited uniformity.
11 b.6.6 POINT DEFECTS
To obtain a coated product with an excellent
Dust particles adhering either to the front side quality it is necessary, however, to design carefully
or to the back side of the web can disturb the the coating installation (control of vibrations,
References 459
noise, dust etc.) and to acquire a lot of operating read at AIChE Spring Meeting, March 30-April 2,
skills (start-up, spreading, edges, streaklines, 1992, New Orleans, LA, paper 43c.
optimum pressure difference, etc.). Bourns, R. T. and McDonald, L. G. 1969. Verfahren
zum kontinuierlichen Auftragen einer Schicht auf
Drawbacks of slide coating are the rather large
ein laufendes Band. Deutsche Offenlegungsschrift
minimum wet layer thickness (not less than about 1,805,734.
20 J.lm) and the relatively limited solid content Bracke, M., De Voeght, F. and 100s, P. 1989. The
(to achieve low thicknesses at high coating kinetics of wetting: the dynamic contact angle.
speeds), certainly when compared with roll or Progr. Colloid Polyrn. Sci. 79: 142-149.
blade coating. Brown, D. R. 1961. A study of the behavior of a thin
That multilayer slide coating is neither a mere sheet of moving liquid. J. Fluid Meeh. 10: 297-305.
Burket, R. S., Conaghan, B. F. and Hirshburg, R. I.
curiosity nor an old and obsolete coating system
1984. Coating method. US Patent 4,443,504.
is demonstrated by the continuous stream of Burley, R. and Kennedy, B. S. 1976. An experimental
patents disclosing small improvements and study of air entrainment at a solid/liquid/gas interface.
scientific articles aimed at explaining its operation. Chern. Eng. Sci. 31: 901-911.
Bussmann, H., Hofmann, H. and Beck, D. 1992.
Frequency response in slide coaters: F. E. calculations
REFERENCES
and measurements. Paper read at AIChE Spring
Ade, F. 1976. GleitfHichenbeschichtungsvorrichtung, Meeting, March 30-April 2, 1992, New Orleans,
Deutsche Offenlegungsschrift 2521995; US Patent LA, paper 43h.
4,041,897. Chen, K. S. A. 1992. Studies of multilayer slide coating
Aktharuzzaman, A. F. M., Wang, C. K. and Lin, and related processes. PhD thesis, University of
S. P. 1978. Wave motion in multilayered liquid films. Minnesota, Minneapolis.
J. Applied Meeh. 45: 25-31. Choinski, E. 1. 1978. Method of multilayer coating.
Bacon, R. E. and Fogg, D. B. 1969. Method of making US Patent 4,113,903.
multilayer coatings containing a water resistant Choinski, E. 1. 1981. Method and apparatus for
layer. US Patent 3,425,857. coating webs. US Patent 4,283,443.
Barstow, F. C. 1966. Cascade coating apparatus for Christodoulou, K. N. 1990. Computational physics of
applying plural layers of coating material to a slide-coating flow. PhD thesis, University of Min-
moving web, US Patent 3,289,632. nesota, Minneapolis.
Bassa, A. 1984. Layer casting apparatus. European Christodoulou, K. N. and Scriven, L. E. 1989. The
Patent 0,119,850. fluid mechanics of slide coating. J. Fluid Meeh. 208:
Baumlin, 1. M. 1990. Surfactant effects on surface 321-354.
waves offlowing liquid. Paper read at AIChE Spring Coyle, D. 1., Macosko, C. W. and Scriven, L. E. 1990.
Meeting, March 18-22, 1990, Orlando, FL,paper 19h. Stability ofsymmetric film-splitting between counter-
Baumlin,l. M. and Pasquet, A. 1988. Flow stability rotating cylinders. J. Fluid Meeh. 216: 437-458.
on an inclined plane: modeling and experimental De Geest, W. and Verkinderen, P. 1967. Improvements
results. Paper read at AIChE Spring Meeting, relating to the coating of webs with liquid compo-
March 6-10, 1988, New Orleans, LA, paper 5b. sitions. GB Patent 1,061,431.
Beck, D. and Mues, W. 1988. Analysis of flow field in Dittmann, D. A. and Rozzi, A. 1977. Method of
slide coaters by FE-calculations and comparison multi-layer coating. US Patent 4,001,024.
with LD-measurements. Paper read at AIChE Spring Do, D. V. and Scriven, L. E. 1991. Analysis of coating
Meeting, March 6-10, 1988, New Orleans, LA, defects, especially those caused by particles and
paper 2b. bubbles in slide coating. Paper read at IS&T 44th
Beguin, A. E. 1954. Method of coating strip material. Annual Conference, May 12-17, 1991, St. Paul, MN.
US Patent 2,681,294. Do, D. V. and Scriven, L. E. 1992. Dynamics of
Berghezan, D. 1991. Simulation numerique d'ecoule- particles and bubbles in coating flows. Paper read
ments en couches intervenant dans l'industrie at AIChE Spring Meeting, March 30-April2, 1992,
photographique. PhD thesis, Universite Catholique New Orleans, LA, paper 40t.
de Louvain-la-Neuve, Belgium. Fahrni, F. and Zimmermann, A. 1977. Beschichtun-
Blake, T. D., Clarke, A. and Ruschak, K. 1. 1992. gsvorrichtung. Deutsche Offenlegungsschrift
Hydrodynamic assist of dynamic wetting. Paper 2,614,068.
460 Slide coating
Faust, H. L. 1975. Menisci studies in bead coating. Isayama, S. and Takehara, N. 1980. Beschichtungsvor-
MS thesis, Drexel University. richtung. Deutsche Offenlegungsschrift 2,951,515.
Feng, J. and Scriven, L. E. 1992. Electrostatic effects Ishiwata, M. et al. 1980. Beschichtungsverfahren und
in continuous liquid coating. Paper read at AIChE Vorrichtung zur Durchfiihrung dieses Verfahren.
Spring Meeting, March 30-April 2, 1992, New Deutsche Offenlegungsschrift 2,948,962.
Orleans, LA, paper 38d. Johnson, H. R. 1965. Method of coating strip material.
Fenoglio, A. et al. 1990. On the physics of establishing US Patent 3,220,877.
coherent coating beads and continuous liquid Kao, T. W. 1968. Role of viscosity stratification in the
coatings. Paper read at AIChE Spring Meeting, stability of two-layer flow down an incline. J. Fluid
March 18-22, 1990, Orlando, FL, paper 81n. Mech. 33: 561.
Finnicum, D., Pegoni, J. and Story, T. 1991. Method Katagiri, Y. 1988. Natural oscillation in slide-coating
and apparatus for use in bead coating a web with flow. Paper read at AIChE Spring Meeting, March
liquid composition. International Patent Application, 6-10, 1988, New Orleans, LA, paper 2c.
Publication Number WO 91/014969. Katagiri, Y. 1990. Comparative study of unsteady
Fowble, W. F. 1975. Method of coating a spliced web. behavior in slide and curtain coating. Paper read
US Patent 3,916,043. at AIChE Spring Meeting, March 18-22, 1990,
Fruhner, H., Kragel, 1. and Kretzschmar, G. 1989. Orlando, FL, paper 96c.
Untersuchungen zur Stabilitat der F1iissigkeitsbriicke Katagiri, Y. and Scriven, L. E. 1986. Supercomputer-
beim Wultstbeschichtungsverfahren. J. In! Rec. aided analysis of transient response of a coating
Mater. 17: 107-115. operation. Paper read at AIChE '86 Annual Meeting,
Galehouse, D. and Colt, J. 1984. Simplified analytical November 2-7, 1986, Miami Beach, FL.
solutions of the free fluid surfaces associated with Kheshgi, H. S., Kistler, S. F. and Scriven, L. E. 1992.
slide coating. Paper read at AIChE Winter Meeting, Rising and falling film flows: viewed from a first-
Atlanta, 1984, paper 16c. order approximation. Chern. Eng. Sci. 47(3): 683-694.
Goenaga, A. S. and Higgins, B. G. 1992. Mechanisms Kistler, S. F. 1983. The fluid mechanics of curtain
for flow reversal in coating operations. Paper read coating and related viscous free surface flows with
at AIChE Spring Meeting, March 30-ApriI2, 1992, contact lines. PhD thesis, University of Minnesota,
New Orleans, LA, paper 49i. Minneapolis.
Gutoff, E. B. 1993. Avoid coating and drying defects. Kistler, S. F. 1993. Personal communication.
Chern. Eng. Proc. 93(1): 49-55. Kobayashi, C. 1992. Stability analysis of film flow on
Gutoff, E. B. and Kendrick, C. E. 1982. Dynamic an inclined plane. I. One layer, two layer flow. Ind.
contact angles. AIChEJ. 28(3): 459-466. Coat. Res. 2: 65-88.
Gutoff, E. B. and Kendrick, C. E. 1987. Low flow Kobayashi, c., Saito, H. and Scriven, L. E. 1982. Study
limits of coatability on a slide coater. AIChEJ. 33(1): of slide coating by finite-element method. Paper
141-145. read at AIChE Winter Meeting, Feb. 28-March 3,
Hackler, M. A. et al. 1992. Frequency response of 1982, Orlando, FL, paper 45e.
coating flows to small three-dimensional disturbances Koepke, G. et al. 1982. Randleisten an Gleitflachen
by supercomputer aided analysis. Paper read at von Kaskaden- und Vorhanggiessern. Deutsche
AIChE Spring National Meeting, March 30-April Offenlegungsschrift 3,037,612.
2, 1992, New Orleans, LA, paper 43d. Koepke, G. et al. 1984. Verfahren zur Mehrfachbes-
Hartman, R. L. 1989. High speed curtain coating chichtung von bewegten Gegenstanden oder Bahnen.
process and apparatus. International Patent Appli- Deutsche Offenlegungsschrift 3,238,905 A1.
cation, Publication Number WO 89/05477. Krussig, K. 1981. Vorrichtung zum begieBen von
Hens, 1. and Boiy, L. 1986. Operation of the bead of Bahnen mit viskosen GieBlosungen. Deutsche
a pre-metered coating device. Chern. Eng. Sci. 41(7): Offenlegungsschrift 3,014,816.
1827-1831. Levi, S. M., Cvetkov, G. M. and Babcin, A. I. 1966.
Hens, J. and Van Abbenyen, W. 1994. Explaining the Physical principles of the extrusion method for
limits of operability of slide coating. Paper read at coating ofphotographic emulsions. Uspechi N aoetch.
AIChE Spring Meeting, April 17-21, 1994, Atlanta, Photogr. (11): 74-84.
GA, paper 7h. Mercier, 1. A., Torpey, W. A. and Russell, T. A. 1956.
Higgins, B. G. and Scriven, L. E. 1979. Interfacial shape Multiple coating apparatus. US Patent 2,761,419.
and evolution equations for liquid films and other visco- Mill, C. C. and South, G. R. 1967. Formation of ribs
capillary flows. Ind. Eng. Chern. Fundarn. 18(3): 208. on rotating rolls. J. Fluid Mech. 28(3): 523-528.
References 461
Moffat, H. K. 1964. Viscous and resistive eddies near instability in thin liquid films. J. Fluid Mech. 217:
a sharp corner. J. Fluid Mech. 18: 1-18. 469-485.
Mues, W., Hens, J. and Boiy, L. 1989. Observation of Takagi, A. et al. 1975. Oberzugsverfahren Deutsche
a dynamic wetting process using laser-Doppler Offenlegungsschrift 2,440,280; US Patent 4,024,302.
velocimetry. AIChEJ. 35(9): 1521-1526. Takeda, H. and Oyama, N. 1976. Coating method.
Nakai, S. and Okada, T. 1968. Proci:de d'enduction. US Patent 3,959,528.
French Patent 1,517,840. Timson, W. 1. 1973. Coating apparatus. US Patent
Nakajima, K. and Miyamoto, K. 1993. Coating method. 3,749,053.
European Patent 0,530,752 AI. Timson, W. J. 1976. Coating apparatus. US Patent
Padday, 1. F. 1961. Multiple coating apparatus. US 3,958,532.
Patent 3,005,440. Valentini, 1. et al. 1991. Role of dynamic surface
Palmquist, K. et al. 1992. Visualization of a falling tension in slide coating. Ind. Eng. Chern. Res. 30(3):
liquid film in air carried by a roller. Paper read at 453-461.
AIChE Spring Meeting, March 30-April 2, 1992, Valentini, 1. et al. 1996. Surface elastic effects in
New Orleans, LA, paper 40hh. premetered coating techniques. Ind. Eng. Chern. Res.
Paton, 1968. Kaskadenbeschichtungsvorrichtung. 35: 434-449.
Deutsche Offenlegungsschrift 1,752,885. Van Abbenyen, W., Christodoulou, K. and Scriven,
Pozrikides, C. and Thoroddsen, S. 1991, The defor- L. E. 1988. Frequency response of a coating flow:
mation of a liquid film flowing down an inclined Predictions for slide coating. Paper read at AIChE
plane wall over a small particle arrested on the wall. Spring Meeting, March 6-10, 1988, New Orleans,
Phys. Fluids A. 3(11): 2546-2558. LA, paper 2d.
Ruschak, K. J. 1978. Flow of a falling film into a pool. Van Abbenyen, W. et al. 1990. Experimental verification
AIChEJ. 24(4): 705-709. of the calculated frequency response of a multilayer
Russell, T. A. 1969. Vorrichtung zum Zufiihren fliissigen slide coating operation. Paper read at AIChE Spring
Beschichtungsmassen. Deutsche Offenlegungsschrift Meeting, March 18-22, 1990, Orlando, FL, paper 96a.
1806887; US Patent 3,474,758. Van Abbenyen, W., Mues, W. and Goetmaeckers, B.
Saito, H., Ishizuka, S. and Fuchigami, S. 1982. Instability 1992. Explaining limits of operability in curtain
of the slide-coating flow. Paper read at AIChE coating. Paper read at AIChE Spring Meeting,
Winter Meeting, Feb 28-March 3, 1982, Orlando, March 30-ApriI2, 1992, New Orleans, LA, paper 43a.
FL, paper 47c. Van Abbenyen, W. and Mues, W. 1992. Comments
Sakiadis, B. C. 1961. Boundary-layer behavior on on 'Stability analysis of multilayer stratified flow
continuous solid surfaces. AIChEJ. 7(2): on an inclined plane'. Ind. Coat. Res. 2: 89-93.
21-25. Verkinderen, P. A. et al. 1979. Method for the continuous
Salamon, T. R., Armstrong, R. C. and Brown, R. A. coating of webs having spliced joints. European
1992. Modification of inertial film instability by Patent 0,001,465.
viscoelasticity. In Theoretical and Applied Rheology, Vrahopoulou, E. P. 1991. A model for fluid flow in
eds P. Moldenaers and R. Keunings, Elsevier Science dies. Chern. Eng. Sci. 46(2): 629-636.
Publishers B.V. pp. 213-215. Vrahopoulou, E. P. 1992. Flow distortions around
Savage, M. D. 1992. Meniscus instability and ribbing. particles between parallel walls with application to
Industrial Coating Research. 2: 47-58. streak formation in slide-coating methods. Chern.
Scanlan, D. J. 1990. Two-slot coater analysis: Inner Eng. Sci. 47(5): 1027-1037.
layer separation issues in two-layer coating. MS Wang, C. K., Seaborg, J. J. and Lin, S. P. 1978.
thesis, University of Minnesota, Minneapolis. Instability of multi-layered liquid films. Phys. Fluids.
Schunk, P. R. 1989. Surfactant and polymer additives 21: 1669-1673.
in coating and related flows. PhD thesis, University Weinstein, S. 1. 1990. Wave propagation in the flow
of Minnesota, Minneapolis. of shear-thinning fluids down an incline. AIChEJ.
Schweizer, P. M. 1988. Visualization of coating flows. 36(12): 1873-1889.
J. Fluid Mech. 193: 285-302. Weinstein, S. 1. and Kurz, M. R. 1991. Long-
Schweizer, P. M. 1993. Personal communication. wavelength instabilities in three-layer flow down an
Schweizer, P. M. and Scriven, L. E. 1983. Evidence of incline. Phys. Fluids. A3(11): 2680-2687.
Gortler-type vortices in curved films. Phys. Fluids. Weinstein, S. J., Baumlin, 1. M. and Servant, J. 1993.
26(3): 619-623. The propagation of surface waves in flow down an
Smith, M. K. 1990. The mechanism for the long-wave oscillating inclined plane. AIChEJ. 39: 1113-1123.
462 Slide coating
Willemsens, K. S. 1980. Method and device for slide m mass of the system (the coating bead) per
hopper multilayer coating. European Patent unit width (kg/m)
0,018,029; GB Patent 7,913,627. M momentum per unit width (kg/s 2)
Willemsens, K. S., Van Abbenyen, W. N. and Criel, momentum entering the control volume by
F. B. 1989. Coating method. European Patent convection, per unit width (kg/s 2)
0,300,098.
M OUI momentum leaving the control volume by
Yih, C. 1963. Stability ofliquid flow down an inclined
plane. Phys. Fluids. 6(3): 321. convection, per unit width (kg/s 2)
Yoshida, K. and Ishiwata, M. 1970. Verfahren und momentum of the flow on the slide, per unit
Vorrichtung zum Regulieren der Beschichtungs- width (kg/s 2)
oder Uberzugsdicke. Deutsche Offenlegungsschrift number of layers (-)
1,929,176. ambient pressure (kg/ms 2)
pressure under the bead (kg/ms 2)
total flow rate per unit width (m 2/s)
flow rate of bottom, resp. top layer (m 2/s)
NOTATION web speed (m/s)
exponent (ofsurface tension) in correlation [ - ] critical velocity for the onset ofair entrainment
amplification factor of the most amplified (m/s)
wavelength [ - ] Us mean velocity of the flow far upstream on
b exponent (of viscosity) in correlation [ - ] the slide (m/s)
FD reactive force of the die per unit width (kg/s 2) x horizontal component
Fe. electrostatic force per unit width (kg/s 2) y vertical component
Fg gravity force per unit width (kg/s 2)
F p •a dynamic pressure ofair per unit width (kg/s 2) Greek letters
FliP force of pressure difference over the bead per IX application angle (degrees)
unit width (kgfs2) f3 slide inclination angle (degrees)
Fv viscous friction force per unit width (kg/s 2) y angle between die face and web (degrees)
F•. a drag force of the air per unit width (kg/s 2) !1P pressure difference over the bead (kg/ms 2)
F•.• drag force of the die lip per unit width (kg/s 2) J.l viscosity (kg/ms)
F•. s drag force of the slide per unit width (kgfs2) J.lB' J1.T viscosity of bottom, resp. top layer (kgfms)
F •. sat drag force of the boundary layer per unit P density (kg/m 3 )
width (kg/s 2) Pair density of the air (kgfm 3 )
drag force of the web per unit width (kg/s 2) a surface tension (kg/s 2)
surface tension force (in point f) per unit 80 apparent static contact angle (degrees)
width (kg/s 2) 8D apparent dynamic contact angle (degrees)
g gravitational acceleration (m/s 2 ) no eigenfrequency (=(1/2n)·(k/m)1/2) (S-I)
HG gap width (m)
H. layer thickness on the slide (m)
H", eventual wet layer thickness on the web (m) Dimensionless numbers
H",.min minimum coating thickness (m) Ca capillary number ( = J,lUla)
k stiffness of the bead per unit width (kg/ms 2) Caerit critical capillary number for the onset of
Lb bead height (m) vortices
L b •max upper limit value of bead height (m) m viscosity ratio ( = J,ltoplJ,lbotlOm)
4.min lower limit value of bead height (m) N es dimensionless electrostatic force (=Fe.la)
Lcd' Lde length scales of control volume (m) N liP dimensionless pressure number ( = HG!1Pla)
Le distance for a perturbation to grow by a Po property number (=gJ,l4/ya 3 )
factor e (m) St Stokes number (= pgH~/J,lU)
idem, for most amplified wave (m) St· modified Stokes number
length ofthe inclined plane ofa slide coater (m) (=St(H G + H",)/2H",)
Sakiadis' boundary layer length (m) We Weber number ( = pU 2 H ",Ia)
CURTAIN COATING llc
Kimiaki Miyamoto and Yoshinobu Katagiri
Hc.I INTRODUCTION
Curtain coating is one ofthe three most prominent freely falling liquid sheet. One difficulty arises
pre-metered coating methods. The other two, slot from a lower limit in flow rate below which the
coating and slide coating, are discussed in the falling liquid sheet suddenly breaks apart. Another
preceding Chapters lla and llb. Even though difficulty stems from the susceptibility to disturb-
curtain coating can be performed with coating ances of the unsupported liquid sheet which
dies similar to those used in slot or slide coating, results in unevenness of the coated film. Tech-
it is quite different. The most distinguishing nological measures for dealing with both difficulties
feature is a liquid sheet that falls freely over a are explained in Section llc.6.
considerable height before it impinges onto the Recent applications of curtain coating include
substrate to be coated. Curtain coating has two multilayer precision coating, such as that used
outstanding advantages: it can apply thin liquid in manufacturing ofphotographic film. A historical
layers to irregular surfaces, and it can coat at review, however, shows that the sophisticated
very high speeds. The first capability is made technology of modern curtain coating evolved
possible by the large distance between the coating from much simpler origins. The first industrial
die and the moving surface, and also the impinging use of the method can be found in a German
velocity of the falling liquids sheet which is patent by Taylor (1903) for a machine that
usually much higher than the vertical velocity manufactured chocolate candies. A falling curtain
component of the substrate motion. Together, of molten chocolate emerged from a slit to coat
the two effects attentuate the consequences of candy cores in a uniform manner. Several decades
up-and-down movement of the substrate. The later, curtain coating became important in the
second capability may be attributed to the localized furniture industry. Up until the 19605, in fact,
pressure that results from the high impinging curtain coating was used primarily for coating
velocity. That pressure presumably squeezes out pieces and sheets because of its tolerance for
the air film that may be entrained between the irregular shapes and the possibility of recovering
liquid and the substrate, and thereby prevents unused coating liquid by a collection system
shear-induced instability ofthe air/liquid interface placed underneath the parts being conveyed
which would otherwise cause catastrophic air through the falling curtain (e.g., Wandtke (1964».
entrainment. This and other operating limits are Applications for continuous webs, such as
treated in full detail in Section llc.5. aluminum foil and corrugated paperboard, became
Although curtain coating is suitable and common in the 1960s because coating operations
attractive for modern mass production, it poses could be performed at speeds as high as, say,
some inherent difficulties, mostly related to the 8 mls (Poirier 1966; Booth 1970). Cox (1968)
464 Curtain coating
described two types of sheet-forming equipment.
The first, shown in Fig. lIe. 1, forms a free-falling
sheet from an overflowing weir. The second type,
illustrated in Fig. 11c.2, makes use ofa slot-shaped
orifice at the bottom of a pressurized coating
head. Both types are limited to coating single layers.
The prototype of modern multilayer curtain
coaters was invented by Hughes (1970). Figure
11c.3 shows an example of a particular embodi-
ment with a multilayer slide-fed die. Hughes not
only disclosed a device and method for multilayer
precision coating, but also illustrated the wide
range of curtain coating with fairly viscous
liquids at high speeds. Hughes' invention fur-
thermore includes specific configurations for air
shields and edge guides. The detailed design of
these parts is essential for demanding applications,
especially multilayer precision coating of photo-
graphic and other imaging products. Further
details on all the important parts of a multilayer Figure lIc.2 Orifice type curtain die.
curtain-coater die, including some of the latest
innovations, are explained in Section 11c.2 below.
Section 11c.2 also introduces typical coating
Figure lIc.3 Slide-fed type curtain die (after Hughes

1970) with feeding nozzle (1); cavity (2); slot (3); slide (4);
elongated slide (5); edge guide (6); lip (7); underside of
Figure 1Ic.l Weir type curtain die. lip (8); air shield (9); and catch pan for start-up (10).
Methods and equipment 465
conditions. Section Ilc.3 explains the phenomena distribute the liquid uniformly in a lateral direction;
and steady flow fields involved in successful slots that discharge the liquid onto an included
curtain coating, as well as the effects of coater slide surface where one layer is stacked on top
designs. Section IlcA addresses the susceptibility of the next; an extended curved slide portion
of curtain coating to external disturbances, along which the multilayered liquid film accel-
including both the stability of a falling liquid erates; a lip that forms a uniform liquid sheet;
curtain (Section l1c.4.l) and the response of vertical edge guides that maintain a desired width
curtain coating to ongoing disturbances (Section of the falling sheet; and air shields that protect
IlcA.2). Section 11c.5 discusses instabilities of the liquid curtain from air currents and thereby
curtain-coating flows in the dynamic wetting enhance its stability. The curtain height can be
region and related operability windows. Section changed by adjusting the vertical position of the
11c.6 provides a summary of important patents die with respect to the substrate to be coated. A
and other practical concerns. second embodiment ofa multilayer curtain coating
die, the slot-fed type (Fig. l1cA), merges liquid
layers within the die manifold and extrudes them
llc.2 METHODS AND EQUIPMENT
together from a slot. The resulting flow field in
Figures 1Ic.3 and IlcA show the essential parts the sheet forming zone is quite different from
of curtain coating devices according to the that in a slide-fed type die.
particular embodiments described in Hughes' Another type of a curtain-coating die was
(1970) patent. A first embodiment, the slide-fed disclosed by Miyake (1973). It is illustrated in
curtain-coating die (Fig. Ilc.3), consists of liquid Fig. 11c.5. The die has two slide surfaces, both
feeder nozzles that are supplied at a constant with negative inclinations. As a result, there is
rate by precision metering pumps; cavities that no static contact line where the falling curtain
forms. The patent claims that separating the
Figure llc.4 Slot-fed type curtain die (after Hughes Figure I Ic.S Reversed slide type curtain die (Miyake,
(1970». 1973).
466 Curtain coating
static contact lines from the sheet-forming region maintain a stable curtain. This can limit the range
prevents coating defects caused by the appearance of applicability of curtain coating operations.
of dried coating solution in the vicinity of the Greiller nevertheless highlighted the capability
static contact line. Another variation yet is a of curtain coating to apply thin layers at very
combination of two slide-fed dies, shown in Fig. high speeds by comparing the operable range of
llc.6 (Fahrni 1976). This configuration features flow rate in curtain coating with that in slide
two static contact lines in the sheet forming zone, coating. Further details on the operability of
as well as two static contact lines away from the curtain coating are given in Section llc.5 below.
lips where the liquid layers emerge from the feed
slots. Fahrni's design produces a multilayered,
llc.3 STEADY-STATE OPERATIONS
symmetrically falling liquid sheet.
As is evident from the above descriptions, the To organize the discussion of curtain coating,
key distinction between the four types of curtain and also to facilitate a theoretical analysis of the
dies is the number of static contact lines in the process, the flow in curtain coating is best divided
sheet forming zone. The slot-fed type and the into several zones, namely a sheet forming zone,
Fahrni type have two, the slide-fed type has one, a curtain flow zone, and an impingement zone
and the inverted-slide type has none. Practical (Kistler 1983). These zones, shown in Fig. llc.7,
implications of the different designs result from are common to all types of curtain coating
differences in the velocity profile in the sheet devices. In curtain coating devices of the slide
forming zone and differences in the adsorption type, there is also a film-forming zone at the slot
time of surface active agents when present. exit and a falling-film-flow zone along the inclined
Hughes (1970) claimed a wide range ofsuccessful slide. The flows in these zones are the same as
curtain coating conditions, including speeds from in slide coating, and are discussed in Chapter
0.75m/s to 10m/s; total coated weights from llb. For that reason, they are not covered here
13.8 g/m 2 to 233.5 g/m 2 , viscosities of the lower- again. Single- and multilayer falling-film flows
most layer ranging from 4 to 80 mPa s; and
curtain heights from 0.05 m to 0.2 m. Brown
(1961), and later Greiller (1972), pointed out that
a minimum flow rate is necessary in order to
Figure llc.6 Two merging, slide-fed curtain coating Figure llc.7 Flow zones in curtain coating (Kistler
dies (Fahrni 1976). 1983).
Steady-state operations 467
and their stability are also addressed in depth in sometimes referred to as the 'teapot effect' (Kistler
Chapters 8 and lld. 1983; Kistler and Scriven 1994). A second
mechanism is wetting of the underside of the lip.
It can give rise to hydrodynamic hysteresis
He.3.1 SHEET FORMING ZONE
which, in turn, stems from the nonlinearities
An important part of curtain coating is the inherent to viscous free-surface flows and permits
formation of a freely falling liquid sheet. The flow multiple stable curtain shapes and flow fields at
fields in the sheet forming zone of symmetrical, the same flow parameters. A third mechanism is
slot-fed curtains have been calculated by Kistler conventional wetting hysteresis that arises from
(1983). Contours of shear stress and transverse differences in advancing and receding contact
velocity, shown in Fig. llc.8, reveal that the angles (Chapter 3). It has its physical origins in
transition from a shear flow to an extensional surface roughness and inhomogeneities, and is
flow is confined to a rather small range of confounded by the presence of a sharp corner.
- 1 < x < 3, where x is the distance from the This set of phenomena is not only of theoretical
die exit in units of half the slot clearance. Kistler interest but also of considerable practical im-
(1983) analyzed not only the formation of a portance in curtain coating with a slide-fed die
falling curtain of low or moderate viscosity, but - where formation of a straight wetting line
also the extrusion of a highly viscous liquid for without wetting past the sharp corner is key to
a sheet casting process, which makes use of a maintaining a uniform curtain. Kistler and Scriven
very similar slot-fed die. In extrusion, of course, (1994) demonstrated all three mechanisms in
there can be considerable die swell. In curtain experiments, and analyzed them theoretically by
coating, in contrast, there is very little or no die means of the finite element method.
swell at all. Kistler found that a small amount As for the deflection ofthe falling liquid curtain
of swelling can be promoted by very narrow slot when the static wetting line remains pinned at
clearances. His computed results show, however, the sharp corner, the flow rate per unit width,
that even when there is no die swell the separation Q, or equivalently the Reynolds number,
angles of the liquid-air interface can be fairly Re == pQ/Jl, is the key variable (p is the liquid
high, ranging from 1300 to 1800 when measured density and Jl the viscosity). The amount of
from the die-internal slot surface. Unless the deflection is not a monotonous function of flow
liquid wets the solid very poorly, such angles are rate, but exhibits a distinct maximum at an
likely to violate Gibbs' inequality (see Chapter intermediate flow rate, as illustrated in Fig.
3), and the liquid is prone to wet the die lip past llc.10. Kistler and Scriven (1994) found that the
the sharp corner at the slot exit. angle of inclination of the slide, p, has a fairly
The flow fields of sheet formation from a weak influence. The effect of the property
slide-fed die have also been analyzed by Kistler parameter Po == (J(p/Jl4g)1 /3 is quite significant,
(1983) (see also Kistler and Scriven (1994)), as buts its influence is mostly confined to shifting
shown in Fig. llc.9. The trajectory of the liquid the Reynolds number at which the maximum
sheet where it leaves the slide lip is controlled deflection occurs. In the definition of Po, (J is the
by gravity, inertia, viscous and capillary forces. surface tension and g = 9.81 m/s 2 is the gravi-
The computed pressure and stress contours tational acceleration. Kistler and Scriven (1994)
indicate the existence of a singularity at the static emphasized that the deflection comes from purely
contact line, where sudden acceleration takes hydrodynamic effects and not from wetting and
place. The resulting maldistributions of stress spreading phenomena associated with the static
near the lip cause the liquid sheet to deflect contact line.
toward the underside of the lip. When the Gibbs inequality condition is violated,
The deflection of the falling curtain is one of the static contact line can move past the sharp
the mechanisms at the root of a set of phenomena corner and cause the liquid to wet the underside
0.21 ~~...
-, .J ']I
0,3'6
U '_~ ' 0.1

'~
::::Jl-l~\\J- ',.~ ) t~ ::' ~- ~:. - ~" :-111 \ U,
X ~
0.7'
0.8715
1111/1111 IAI-I ~.'jjJ

-0 OS -
II
-0.015
(a) (b) (c) (d) (e) (f)
Figure llc.8 Flow fields in sheet forming zone for slot-fed type curtain coating die (Kistler 1983; see also Kistler and Scriven 1994): (a) stream
lines; (b) particle speed; (c) transverse velocity component; (d) pressure; (e) normal stress difference; (t) shear stress. Re = pQ/J-l = 10, St = pgW 3 /(J-lQ) = 0.5,
Ca = J-lQ/(WI1) = 2, where g = 9.81 m/s 2 , Q is the volumetric flow rate per unit width, Wis half the slot clearance, J-l is the liquid viscosity, p the
liquid density, and 11 the surface tension. The horizontal scale is exaggerated by a factor of 2.
aJ b) c) d) e) f)
Figure llc.9 Flow fields in sheet forming zone for slide-fed type curtain coating die (Kistler 1983; see also
Kistler and Scriven 1994): (a) stream function; (b) particle speed; (c) pressure; (d) normal stress difference; (e)
shear stress; (I) vorticity. Re = pQ/Jl = 5, Po = (J/(Jl4 g/p)1/3 = 7.30, fJ = 60·, where fJ is the angle of slide
inclination measured from horizontal.
of the lip. According to the inequality, wetting below. Further confounding whether and, if so,
depends on the static contact angle of the how far the liquid wets the lip past the sharp
liquid-solid system (}e' the cut-back angle y corner is contact angle hysteresis. This hysteresis,
between the inclined slide and the lip underside, as well as the hydrodynamic hysteresis, makes
and the separation angle of the interface that the contact line position strongly dependent on
results from the trajectory of the deflected curtain the history of the preceding back-and-forth
and is a function of Re, Po and /3. Kistler's finite movement of the line.
element analysis with free movement of the The hydrodynamic hysteresis arises because
contact line revealed how the equilibrium position the dependence of the contact line position on
of the static contact line is affected by the various parameters, especially the Reynolds
above-mentioned five parameters. In addition to number, is not a single-valued function but
the flow parameters that control the deflection, exhibits multiple steady states at the same
(}e and y have a strong influence on the contact parameter value. Even if a set of parameters
line position, a phenomenon that is generally corresponds to a pinned condition, operating at
known and incorporated in patents explained those parameters does not assure that the static
470 Curtain coating
6,--------------------------------,
w
(,) 3
Z
«
l-
(/')
0
0
...J
«
(,)
-3
i=
a:
w
>
-6
~.
-9 +-------,----f------,--fL---,-=----,,----....,-------,.--,----,---"''--,-------,--'-!
-4 o 4 a 12 16 20
HORIZONTAL DISTANCE
Figure 1Ic.10 Hydrodynamic deflection of the falling sheet at different flow rates (Kistler 1983; see also
Kistler and Scriven 1994). Po = a/(Jl4 g/ p)1/3 = 6.35, f3 = 60°.
contact line is indeed pinned. Figure llc.ll a lower critical flow rate, the contact line advances
illustrates how the contact line migrates with past the lip but, at first, does so very gradually
changes in Reynolds number, and how hysteresis until a critical Reynolds number is reached at a
appears as a consequence of drastically different lower turning point. Below the lower critical
contact line positions for increasing and decreasing Reynolds number, the contact line jumps past
flow rates. At low rates, the contact line remains the lip to wet the underside again. Kistler and
near the sharp corner but Gibbs' inequality is Scriven (1994) confirmed this characteristic
not satisfied and the lip underside is slightly wet. behavior of the contact line position as calculated
As the Reynolds number increases, the contact by the finite element method with experimental
line advances further past the lip, until a critical measurements.
Reynolds number is reached at an upper turning In Figure llc.11, for purposes of simplicity,
point. Beyond that Reynolds number, the contact the effects of contact angle hysteresis are ignored.
line recedes in a jump-like manner back to the Using different contact angles for advancing and
sharp corner and stays pinned with further receding situations, Kistler and Scriven (1994)
increases in Re. When, on the other hand, the were able to match their computed predictions
flow rate is decreased from high values that with their experiments. Nonetheless, Kistler and
correspond to a pinned condition, the contact Scriven (1994) stress that much of the hysteresis
line remains pinned down to flow rates well seen as part of the teapot effect is purely
below those at the upper turning point. When hydrodynamic in origin, even though it is
Gibbs' inequality can no longer be satisfied past influenced by a wetting parameter that exhibits
8r--,----,--,---,-----r------,---r---,---.----,---.---,
x*
6
z
0
i=
en
0
a..
w 4
z
:J
I-
0
«
I-
Z 2
0
0
O'--_....l.-_--L_ _.J....:._-L----"''''''''''......_ ......_-.1.._....;;....~_--L..._--L

_ __'___~
2 4 6 8 10 12
REYNOLDS NUMBER Re=pQ/1l
Figure llc.ll Teapot effect: migration of the contact line and its hysteresis (Kistler 1983; see also Kistler
and Scriven 1994). Po = (J/(JJ4 g/ p )I/3 = 6.35, P= 60·, 8e = 67", y = 60·. The symbols are explained in the text.
hysteresis itself, namely the contact angle. Kistler the lip and a wavy wetting line, both giving rise
and Scriven (1994) also examined the effect of to down-web striations and streaks in the coated
contact angle on the hydrodynamic hysteresis of layers. Kistler and Scriven (1994) showed, by
the contact line position. Their finite-element means offinite-element based design studies, that
computations reveal that the contact line position decreasing the cut-back angle l' between the
becomes less and less sensitive to flow rate as inclined slide and lip underside is very effective
the contact angle increases, but that an isolated in eliminating wetting past the sharp corner.
hysteresis loop may separate from the main loop Other lip designs aimed at preventing uneven
shown in Fig. llc.ll before, above a critical wetting and related streaking are discussed in
angle, wetting no longer occurs over the entire Section llc.6.1 below.
range of flow rates because Gibbs' inequality is Another practical issue intimately linked to
always satisfied. the teapot effect relates to the curved trajectory
These results are very relevant to practical of a deflected liquid sheet - which may deviate
aspects of successful operation of a slide-type substantially from the vertical trajectory imposed
curtain coating die because wetting of the lip by straight edge guides. When the curtain
underside can lead to degradation of the coating deflection is prominent, a 'crease' may form
material trapped in the dead flow zones beneath where the more or less uniformly deflected curtain
472 Curtain coating
in the central portion of the coating width meets In this formula, Vo is the average velocity at the
the standing wave that emanates from the point slot exit. The distance X 0 is taken to be
of initial contact of the curtain with the edge
guide and delimits the portion of the falling sheet (llc.2)
that is strongly influenced by the edge guide.
This 'crease', in tum, causes a 'kink' in the Brown's data points show agreement between
dynamic wetting line that can become a major the measured and the calculated velocity only
source for down-web striations or streaks. Ridley when the distance X is larger than two or three
(1979) suggested that the use of a vertical edge times X o. In addition, a comprehensive finite-
guide in the form of a plane perpendicular to the element analysis of the sheet-forming region
curtain instead of the commonly used rod solves (Kistler 1983) reveals that the value of X 0 may
this problem. However, this solution is not not be a simple function of the liquid properties
always suitable. In fact, sometimes it may aggravate like 2(4J1./p)2f3 g- 1 / 3 , as conjectured by Brown
the unevenness of the coated layers near the edge (1961).
guide, as mentioned below. For curtains with curved trajectories that
Curtain deflection from the teapot is not result from hydrodynamic deflection at a slide
always harmful, however. Suga et al. (1991), for lip, Kistler (1983) gave a generalized analysis to
example, claimed that when the deflection is obtain a set of integrodifferential equations in
small and the flow rate is high, the three- terms of curvilinear, orthogonal coordinates.
dimensional flow resulting from the bowed shape Those equations, applicable to all sorts of two-
of the dynamic contact line drives small bubbles dimensional liquid sheets, are important as starting
toward the edges, and thereby prevents a special points for various approximations. However,
kind of air entrainment caused by bubbles that they are also very complicated and can provide
would otherwise get trapped in the impingement exact solutions only in special cases.
flow, especially when there is a heel (see Section In the thin film approximation, the integrodif-
llc.3.3 below). ferential equations reduce to simple differential
equations for the thickness profile and angle of
inclination of a falling liquid sheet (Kistler 1983).
llc.3.2 CURTAIN FLOW ZONE This approximation assumes that the characteristic
length scale over which significant changes in
In successful curtain coating operation, the most thickness or angle occur in the streamwise direction
significant influence of the flow in the freely is much greater than that of the thickness. For
falling curtain is the final velocity at which the a vertically falling sheet, for instance, the
liquid impinges onto the substrate. As discussed streamwise momentum balance reduces to
further in Section llc.5 below, that velocity has
a significant effect on the operating window.
Brown (1961) measured the velocity within a
curtain falling from a slot by means of a rotating
mirror that projected the trajectories of bubbles
in the falling liquid sheet. He found that the
curtain reaches a free-fall regime a short distance
below the slot exit and proposed an empirical Here, h is half the curtain thickness, measured
formula for the curtain velocity V as a function in units of (Re/8)(4J1./p)2 /3g-1 /3; x is the distance
of the vertical coordinate X in the form of a along the symmetry axis of the curtain, measured
free-fall parabola that starts from the point X 0 in units of(4J1./p)2 /3 g -1 / 3; and hx and hxx are the
below the slot: first and second derivatives of h with respect to
x. The Reynolds number is Re = pQ/J1., the
Stokes number is St = pgW 3 / J1.Q, and the capillary
number is Ca = JlQIWa, where 2W is the slit rapidly evolves into a parallel, fully developed
clearance. Kistler (1983) used equation (lIc.3), flow. The rapid development ofthe flow is caused
and similar ones for deflected sheets, to provide by the strong shear stress imparted by the
outflow boundary conditions for finite-element substrate, which gives rise to a boundary layer
analyses confined to sheet-forming zones. that emanates from the vicinity of the dynamic
Thus far, the formation and free fall of the wetting line. For cases with larger heels than
curtain has been discussed independently of the those in Fig. llc.12, Ogawa, Kashiwabara and
impingement flow downstream. In certain flow Scriven (1992) demonstrated agreement between
regimes, however, capillarity can transmit a computed stream lines and experimental flow
significant amount of upstream influence from visualization employing hydrogen bubbles and
the impingement region, and 'pull' the curtain in fluorescent dyes, as illustrated in Fig. lIc.14.
the direction of the substrate motion, resulting Heel formation is closely associated with an
in a bent trajectory toward the substrate. This so- operating bound because, as the heel grows large,
called 'pulled-film' phenomenon is observed when it typically incorporates one or several recirculating
the flow rate is low or the curtain height is small. eddies. These eddies may trap bubbles or degraded
Because the phenomenon can cause film-breakage particles of the coating solution, which in turn
and hence limits operability, the exact range of can cause coating streaks. Large heels may also
conditions and countermeasures are discussed in be susceptible to various instabilities, including
Section 11c.5 on the coatability window. instabilities that result in periodic flow structures
in the cross-web direction, and a special kind of
air-entrainment (Kistler 1983). Kistler (1983)
llc.3.3 IMPINGEMENT ZONE
explained how the position of the dynamic
The flow in the impingement zone is, of course, wetting line and the closely related size of the
at the heart of curtain coating, for it largely heel can be estimated from a boundary-layer
controls the operating behavior of the process. theory for moving flat surfaces (Sakiadis 1961).
Kistler (1983) was the first to investigate the Details of the model and specific results are
evolution of free-surface shape and associated described in Section 11c.5 below on the coating
flow field in the impingement region. He did so window.
by performing finite-element calculations in the The phenomenon of the dynamic wetting line
limit of very high capillary numbers where the getting pulled underneath the impinging curtain
Reynolds number Re = pQIJl and the ratio U IV is also closely linked to an operating limit,
of the substrate speed U to the impingement namely massive, visible air entrainment. Not
speed V become the dominant parameters. Figure only this limit, however, but also dynamic wetting
lIc.12 shows that high Reynolds numbers or, below that limit and its exact influence on
equivalently, high flow rates, and also low speed successful curtain coating is associated with
ratios promote the formation of a pronounced considerable quantitative uncertainty, especially
'heel' that pushes the dynamic wetting line as far as theoretical analysis is concerned. The
upstream. At small values of Re and large values reason is that the physics of the local liquid-air
of U IV, on the other hand, the dynamic wetting displacement process are not completely under-
line tends to get pulled underneath the impinging stood (for an in-depth discussion, see Chapter
curtain. Kistler (1983) generated the contours in 3). Miyamoto and Scriven (1982) advanced the
Fig. lIc.13 of various flow variables in the hypothesis that 'a thin air film is always entrained
impingement zone. The contours of pressure and and subject to breakage and dissolution'. This
stress clearly signal a singularity at the dynamic hypothesis is supported by experimental evidence
contact line. The pressure contours also reveal from works by Mues, Hens and Boiy (1989) and
the buildup ofan impact pressure right underneath Miyamoto (1991). However, quantitative sub-
the impinging curtain. Downstream, the flow microscopic models are still lacking that can
474 Curtain coating
(a)
UIV=O.5 2
(b)
Figure He.12 Heel formation and pulled film for Ca = 1oo (Kistler 1983). The symbols are explained in the
text. (a) UfV = 1; (b) Re = 2.5.
predict both the apparent dynamic contact angle understood, as evidenced by the results cited
under successful coating conditions and the rate above (cf Figs llc.12 to llc.14). The air
of momentum transfer through an entrained and entrainment that presumably occurs during
collapsing air film. successful coating is very local and does not affect
In spite of the quantitative uncertainty at the the global flow as long as the air film is thin
dynamic wetting line, the steady, macroscopic enough so as not to cause instability in the liquid
flow field in the impingement zone is fairly well layer. Nevertheless, in coating flow analyses at
Sensitivity to small disturbances 475
(d)
(e)
-0.2
-0.6
(b)
0.2 .2
-1.0 .6
Figure llc.13 Details of flow field in impingement zone (Kistler 1983): (a) particle speed; (b) pressure; (c)
normal stress difference; (d) shear stress. Re = 2.5, UIV = 0.5, Ca = 100.
macroscopic length scales, the above-mentioned 11c.4 SENSITIVITY TO SMALL DISTURBANCES

uncertainty manifests itself in two unsettled
difficulties: one is the need to impose an apparent llcA.1 STABILITY OF A LIQUID CURTAIN
dynamic contact angle which remains an arbitrary
input parameter unless taken from some experi- The formation of a stable liquid curtain is the
ment or empirical correlation; the other is the first step in a successful curtain coating operation.
choice of an approparite slip coefficient in an ad At low flow rates, the liquid curtain tends to
hoc boundary condition that removes the singu- detach from the edge guides or break into
larity at the dynamic contact line. Kistler (1983) individual streams. As the flow rate is reduced
avoided the first difficulty by computing in the further, the liquid dribbles off in the form of
limit of high capillary numbers and specifying individual droplets from various places at the lip
no contact angle at all. This is feasible because of the coating die. Pritchard (1986) observed a
a specified contact angle affects the computed number of unstable flow patterns in liquid films
free surface profile only in the immediate proximity that fall from the end of a plate at low Reynolds
of the contact line. Ogawa and Scriven (1994) numbers.
examined the effects of specified dynamic contact The film thickness at the foot of the curtain,
angles by comparing four cases, namely, Kistler's which generally is at most several times the final
formulation, one specifying 140°, one specifying coating thickness, is less than the thickness of
170°, and a hybrid of Kistler's and specifying the coating beads in other die coating methods.
140°. The computed dynamic-contact-line posi- Consequently, the capability of curtain coating
tions showed insensitivity to both a specified in depositing a thin film at low coating speeds
contact angle and the slip coefficient. falls well behind slide coating, which is also used
476 Curtain coating
Figure llc.14 Impingement flow: comparison of theory and experiment (Ogawa, Kashiwabara and Scriven
1992). Re = pQ/J.L = 10, Ca = J.LU/(J = 0.08, U/V= 0.29.
as a precision multilayer coating method. Greiller limit. On the other hand, slide coating can be
(1972) suggested that the minimum flow rate to operated at a flow rate less than 0.1 cm 2 Is if the
form a stable curtain with an aqueous gelatin web runs slowly enough. Therefore, curtain coating
solution containing surfactants is around 0.5 cm 2Is is most suitable for high speed coating when
(per unit width of the curtain). Our experience moderate coating thicknesses are desired. A long
has been that this value is close to the practical curtain, which is advantageous in delaying the
onset of air entrainment (see Section llc.5 below),
"~
is more difficult to form in practice than a short
one. Curtain heights up to 250 mm are recom- /8 K NO SIJRFAC'-ANT
o SURFACTANT ;"
mended, based on both stability and operability • SUR'ACTANT '/1'
considerations. ~ /6
Greiller's (1972) experimental results further-

more suggest that there is an optimum range for
,
.., /4
U
~--.. 0- _0"''' ""
U
liquid viscosity, and that surfactants play an ~

important role in stabilizing the curtain (Fig. i
- ._-.-
1.0
ltc.15). Greiller did not offer an explanation for "-:

08
the optimum in viscosity. A possible scenario 6CPS J$CPS tooeps
might be that increasing viscosity enhances
damping of disturbances and thereby stabilizes (a)
the curtain but, at the same time, accelerates the

20 x
boundary layer development along the edge o
NO SURFACTANT
SURFACTANT ~ •
• SIlRF~cr~Nr 's'
guide and thereby promotes breakage of the /8
curtain. The effect ofsurfactants may be attributed ~

~ 16
to the reduced surface tension and to surface
elasticity, as is often the case in the stability of ~
~ ''I
......
U
film flow down an inclined plane (see Chapters
.
.-.---. .
4 and lId). At the edge of the curtain, surfactants ~
/
act either to prevent contraction of the curtain "...... /0

____ .----T-..--..--
by enhancing the wettability of the edge guide,
or promote contraction by the Marangoni effect. 08
J$CPS /OOCPS
6CPS
,
U"
This point is discussed further in Section llc.6. (b) 6 8 10 12 /4 /6 18 ZO
Several authors have reported theoretical X GELATIN
stability analyses of falling liquid curtains in

attempts to describe the operating bound that
arises from the low-flow limit. Brown (1961)
_
:":f o SuRFACTANT
NO SePIFAcrANr
;:.
• SURFACTANT '8'
followed Taylor's (1959) analysis of the stability ~ /6 b~
~ "0
of a water bell. Taylor's stability argument is ~ 14 ' .... , ~ ........
based on a simple momentum balance on a free
edge that would result from the formation of a
~
~/2
....0 . . . . ,
_ • .:.:::.-~-:..:e:.-_. __._
.. a""'''
hole in the liquid sheet (Fig. ltc.16). The inertia :k y

/;'
~
force pV 2 H (or pQ V) must be greater than the /0
/
/
surface tension force 20" for such a hole not to 08 -:
grow and be carried away with the liquid stream. 6CPS J$CPS looeps
,
6
Here, H is the thickness of the curtain, and Q is (c)
2 18 20
again the volumetric flow rate per unit width.

The resulting condition for stability is
Figure He.IS Minimum flow rates for curtain stabil-
pV 2 H > 20" or pQV> 20" (ltcA) ity (Greiller 1972): (a) flow rates to which a falling
In dimensionless form, (1lcA) becomes curtain, first established at a high flow rate, can be
reduced without breaking; (b) flow rates at which the
We>2 (1Ic.5) free-falling curtain is resistant to breaking upon intro-
ducing a solid object into the curtain and moving it
where We = pQV/O" is the Weber number which about; (c) flow rates at which the free-falling curtain
compares inertial forces with surface tension will reform on its own after being deliberately broken.
478 Curtain coating
the stability of a falling curtain in experiments.
Liquid 2a They applied different pressures to each side of
a falling curtain, and found that practically stable
Air
curtains can exist over wide ranges of flow
conditions, including conditions that violate the
Figure llc.16 Forces acting at the free edge that
simple stability criterion We > 2.
would result from breakage of a falling liquid sheet. Altogether, recent theories and experiments
concur in supporting the conclusion that falling
curtains are linearly stable regardless of the
forces. Brown (1961) suggested that (llc.S) must Weber number. So how does a falling curtain
hold right above the impingement zone where disintegrate? Edge effects may be the key: they
large disturbances may exist. Of course, this cause local thinning of the curtain near edge
simple analysis cannot explain the whole story guides which could trigger break-up. Many
of curtain stability. In fact, Brown's experiments solutions have been proposed to prevent such
themselves showed that the region near the slot local thinning of the curtain. They include injection
exit is apparently stable even though the simple of an additional liquid stream to the edge region,
stability criterion is not satisfied because of low or optimization of the edge guide designs (see
speeds. Section 11c.6 for more details). Another likely
Lin (1981) carried out a linear stability analysis cause for loss of curtain integrity comes from
of falling liquid sheets, employing a first order nonlinear effects such as those arising from
approximation that takes into consideration contaminants and bubbles in the liquid, which
both spatial and temporal disturbances. He may become nucleation sites for holes growing
predicted that only those spatially growing sinuous in the curtain (Do and Christodoulou 1992).
disturbances are unstable whose group velocity
is in the upstream direction. In other words, they
llcA.2 RESPONSE TO SMALL DISTURBANCES
are unstable ifthe velocity ofthe wave propagation
in the upstream direction is greater than the The study of coating sensitivity to small, ongoing
velocity of the fluid particle falling downstream. disturbances is ofgreat importance from a practical
This turns out to be the case exactly when standpoint. Several computational analyses of
We> 2. However, a later analysis by Lin, Lian curtain coating have been carried out, all based
and Creighton (1990) shows that, when We> 2, on the finite-element method (consult Chapter 9
the curtain is stable in a vacuum and is convectively for details on the various methodologies). Kaneko
unstable (i.e. asymptotically stable) in the presence and Scriven (1988) analyzed the case of short
of surrounding gas. When We < 2, the large-time curtains with an applied pressure difference.
asymptotic amplitudes of sinuous disturbances They used a linearized transient analysis, i.e. time
is nonvanishing but bounded. In conclusion, a integration of the governing Navier-Stokes system
falling curtain of finite length is stable to small linearized about the steady state of interest.
disturbances in all cases, which contradicts Lin's Katagiri (1990) compared slide and curtain coating
earlier results. in regard to their frequency responses with
Do and Christodoulou (1992) presented results various kinds of disturbances by solving the fully
from a full-fledged finite-element stability analysis nonlinear, time-dependent equations directly.
(cf Chapter 9) and found that curtain-coating Ogawa and Scriven (1990) reported results from
flows are linearly stable for both two-dimensional a linearized frequency response analysis for single-
and three-dimensional disturbances. Their finding and two-layer curtain coating. Do and Chris-
supports the conclusion of the latest work of Lin todoulou (1992) performed a linear stability
and co-workers. analysis of curtain coating with respect to two-
Finnicum, Weinstein and Ruschak (1993) tested and three-dimensional disturbances.
All the analyses suggest in essence the same evidently being much more sensitive to small
general result, namely that curtain coating is pressure variations than slide coating.
highly susceptible to small disturbances acting In reality, air currents disturb a curtain
on the unsupported liquid sheet. This is not randomly rather than pushing the curtain over
surprising, as a very small difference in pressure its entire width simultaneously. As a result, the
between the two sides of a curtain will drastically coated layer exhibits wide bands of streak-like
alter the trajectory of the curtain (Finnicum, unevenness. When the coating liquid contains a
Weinstein and Ruschak 1993). If the applied volatile solvent, the Marangoni effect originated
pressure difference oscillates sinusoidally in time, by the irregular solvent evaporation magnifies
the curtain undergoes a swinging motion like a the unevenness of the final coated layer. Therefore,
pendulum. As a consequence, the coated film will the amount of volatile solvent in the lowermost
suffer a large-amplitude unevenness in the down- and uppermost layers should be minimized. To
web direction (i.e. 'barring'). Figure IIc.I7 shows protect the curtain from the air flow carried by
the frequency response of the coating thickness the moving web, an air shield is essential. The
in curtain coating to sinusoidally varying dis- influence of ambient air currents, such as the air
turbance in pressure at the upstream free surface flow used to condition the entire coating room,
ofthe curtain, and compares it with the response should also be minimized. Systems designed to
of slide coating to an oscillating disturbance in contain the deleterious influence of air flows are
applied vacuum. The difference in amplitude reviewed in more detail in Section I Ic.6.4.
reaches two orders of magnitude, curtain coating Whereas the falling curtain is highly susceptible
to ongoing external disturbances, the coating
;fti
108 .--------,-----,-----, bead in the impingement zone is more tolerant
Co.
Vlo
VI=-
of disturbances than the coating bead in other
Ww
z a: _ _, / - \ CURTAIN COATING pre-metered coating methods. Examples of the
"'::> ~ "".. second type of disturbances include fluctuations
\ ".\
UVI
J: VI
:s'"" g:w 10 in coating speed (Fig. lIc.I8), heterogeneous
u:
~ <{
o V ~~.
zm • • \J
u: ?;; ~.\. 10
u. Z
00
z
O<{
1=
'. ~-
cnC I
I==> VIC
SLIDE COATING Ww CURTAIN COATING
.
<{'""
Z W
- U
"'0.
~
> u.
3 I/"f"
'"
UVI
:Em .0=:----:-'-0 I \
z" '""w
;;i~ '\f'~,
:it
Cl 0.1 :;.<::.----'------'---------'
z z
u: Z
i';2
/1'" .\.
1 10 100 1000 SLIDE COATING
FREQUENCY(Hzl
z~
o=> .,
~ tl 0.1
0:3
<tu.
Figure ltc.17 Computed frequency responses of >
slide and curtain coating to perturbations in surround- z"
:it
ing pressure (Katagiri 1990). The vacuum is perturbed Cl
by ± 40 Pa in slide coating; the pressure at the 0.01

upstream free surface of the curtain is perturbed by 1 10 100 1000
±IPa. p= 103kg/m 3; I"/=20mPas; a=40mN/m; FREQUENC'f(Hzl
HOC) (final film thickness) = 1.05 x 1O-4 m;
U = 1.67 m/s; as (slide angle in slide coating) = ISO, L Figure Ilc.IS Computed frequency responses of
(gap) = 2.5 x 1O-4 m; t:,.p (vacuum in slide coat- slide and curtain coating when the web speed is
ing) = 150 Pa; ac (slide angle in curtain coating) = 45°; perturbed sinusoidally by ± 5% (Katagiri 1990). The
He (curtain height) = 0.02m. parameter values are the same as in Fig. ltc.17.
480 Curtain coating
surface energy of substrates (Fig. llc.19), wavy movement of free surfaces in curtain coating
substrates, and splice passages. Such disturbances when the web speed fluctuates by ± 5% (Katagiri
affect only a localized region around the dynamic 1990). Clearly, a localized area, particularly inside
contact line and immediately downstream. The the boundary layer along the web, is disturbed
curtain remains intact because the disturbances whereas the liquid curtain remains steady. This
do not propagate upstream (see Lin (1981)). resilience to perturbations in the coating bead is
Figure llc.20 shows the computed maximum the reason why curtain coating can be used to
coat discontinuous materials or air-supported
webs (see Fig. llc.21, taken from Yoshida et al.
(1978)). For the same reason, electrostatic assist
works very well in combination with curtain
coating: a heterogeneous electrostatic charge
distribution on the web causes less unevenness
in curtain coating than in slide or slot
11. 1 1----"7"!:....-----..t.e:....-.::...--~.-__i coating.
\
Ribbing, a corduroy-type pattern of regularly
/
/\ spaced lines in the down-web direction which is
lUll '--- ----l

CURT~N COATING
---L
•
often seen in slide or slot coating (see Chapters
11 a and 11 b), is not normally observed in curtain
coating. Ribbing results from an instability of a
~
I 13 100 1000
FREQ.8IC(Hz) two-dimensional viscous free-surface flow to
three-dimensional disturbances. Finite-element
stability analysis corroborates that curtain
Figure 11c.19 Computed frequency reponses of slide coating is stable to such disturbances (Do and
and curtain coating when the dynamic contact angle is Christodoulou 1992). Rib-like streaks, however,
perturbed sinusoidally by ± 5° (Katagiri 1990). The are sometimes encountered in curtain coating
parameter values are the same as in Fig. lie.! 7.
of non-Newtonian liquids. A detailed analysis
of this phenomenon has not yet been per-
formed.
\
BOUNDARY LAYER THICKNESS
(SAKIADIS)
Figure lle.20 Calculated maximum movement of

free surafces in curtain coating when the web speed is
perturbed sinusoidally ± 5% with a frequency of 30 Hz
(Katagiri 1990). The parameter values are the same as Figure He.21 Curtain coating on air supported
in Fig. lIc.17. substrate (Yoshida et al. 1978).
Operability 481
llc.5 OPERABILITY • heel formation at high flow rates and low
coating speeds (see Fig. llc.23c);
lIc.5.! COATING WINDOW • air entrainment with heel formation at high
The dimensionless groups which govern the flow flow rates and high coating speeds (Fig.
field in the impingement region ofcurtain coating llc.23d).
are the Reynolds number Re = pQlj1, the web
speed normalized by impinging velocity, UIV, llc.5.2 PULLED FILM
and the capillary number Ca = j1Vla. Kistler
(1983) showed, by means offinite-element analysis, A highly bent curtain without a freely falling part
that the limiting regime of high capillary numbers tends to form at low flow rates and low coating
provides a good approximation to curtain coating speeds (Fig. llc.23a). The mechanism can be
at high speeds, for it captures many of the explained as follows. As the flow rate is reduced,
essential characteristics of impingement flow. In the viscous drag imparted by the substrate shifts
the limit of Ca - 00, the coating window can be the dynamic contact line downstream and forces
drawn in a plane of Re versus UIV, for it depends the curtain to bend near the impingement region.
only on those two dimensionless groups. Figure At the bent upstream meniscus, capillary forces
llc.22 shows a schematic of such a coating balance the inertia force of the falling liquid.
window; Fig. llc.23 shows photographs of Below a certain flow rate, the capillary pressure
unsuccessful curtain coating past several of the dominates over the inertia force and the curtain
limits. The phenomena limiting successful curtain- starts to be pulled. The Weber number, which is
coating operation are: the ratio of inertia forces to surface tension forces
(cf Section llc.4.1), is the appropriate dimen-
• curtain disintegration at low flow rates (already sionless index for this phenomenon. The pulled
discussed in Section 11 c.4 and then taken up film is typically close to instability and a slight
further in Section llc.6); reduction of the flow rate causes the curtain to
• pulled film phenomenon at low flow rates and break up.
low coating speeds (see Fig. llc.23a below); Ogawa, Kashiwabara and Scriven (1992)
• air entrainment at high coating speeds and predicted the film pulling phenomenon from a
relatively low flow rates (see Fig. llc.23b); finite-element analysis of one-phase flow, that is,
they neglected air flow. Examples of their analysis
are given in Fig. l1c.24. These authors also
performed flow visualization, and confirmed that
L~
heel rormauon
3ir tnlrainmenl
& heel
the free surface shapes are well predicted by the
computational analysis. Their numerical predic-
tions reveal that high surface tension and low
curtain heights promote the pulled-curtain
coatable phenomenon (Fig. llc.25), in accord with the
explanation given above. In Fig. l1c.25, the
l~
air enU3lnmem
criterion used to delineate the pulled-film regime
puliedtilm is whether or not the dynamic contact line is
~ ~
located past the vertical extension of the down-
stream free surface of the curtain which, at the
onset of the pulled-film regime, still exhibits a
log IV region of near-vertical fall. Experimental obser-
vations, corroborated by finite-element predictions,
Figure llc.22 Schematic of coating window in cur- indicate that liquid viscosity and coating speed
tain coating. have a relatively small effect on preventing film
482 Curtain coating
Figure llc.23 Photographs of limiting phenomena in curtain coating: (a) pulled film; (b) air entrainment;
(c) heel; (d) air entrainment with heel.
pulling. Likewise, air flow, which might be speeds as high as 6 mls with an impingement
suspected to be highly influential, turns out to velocity of 1.4 m/s. The maximum coating speed
be insignificant in the pulled-curtain phenomenon. claimed by Hughes (1970) is 10 m/s. These speeds
The way to avoid the pulled-curtain effect and are much higher than the speeds of dynamic
thereby widen the coating window is to increase wetting without deleterious air entrainment in
curtain height if the curtain is sufficiently stable, plunging-tape experiments (cf. Chapter 3), and
or to lower the dynamic surface tension of the also the maximum operating speeds typically
lowermost layer. obtained with other die coating methods.
Nevertheless, air entrainment is still a key
phenomenon in curtain coating, and can limit
IlcS3 AIR ENTRAINMENT
the maximum coating speed. Past the limit, V-
The large pressures created by the impingement shaped air pockets followed by streaks ofentrained
of the freely falling liquid can suppress the air bubbles are often seen, as shown in Fig.
intrusion of an incipient thin air film and thereby Ilc.23(b). These features closely resemble those
delay the onset of visible air entrainment in associated with air entrainment ofa tape plunging
curtain coating. Brown (1961) reported that into a pool (see Chapter 3). It is generally
stable curtain coating can be accomplished at presumed that air entrainment sets in when the
Operability 483
(a) (b)
(c)
Figure llc.24 Calculated effect of Re and Ca on pulled film phenomena (Ogawa, Kashiwabara and Scriven
1992): (a) not pulled (Re = 7.4, Ca = 0.25); (b) pulled (lower Q, viz. Re = 6.7, Ca = 0.25); (c) pulled (higher u,
viz. Re = 7.4, Ca = 0.22).
dynamic contact angle reaches 180 (see again

0
suspected that the conjoining-force-driven insta-
Chapter 3). bility breaks the air film into small bubbles.
Miyamoto (1986) observed two types of air When the bubble number and size become
entrainment with curtain coating: in the first, excessive, visible air entrainment appears. Craters
illustrated in Fig. l1c.26a, trails of bubbles and found in line together with bubbles were considered
craters are visible in an otherwise well-coated to be the outcome of the dissolution of small
film; in the second, on the other hand, the liquid bubbles. The second type of air entrainment was
layer is completely disrupted by the entrained obtained by coating a gelatin solution of a lower
air, as in Fig. llc.26b. The first sample was viscosity. Miyamoto (1986) attributed the resulting
obtained by coating a 122 mPa s aqueous gelatin defect to a Kelvin-Helmholtz instability caused
solution on to a gelatin-subbed substrate. by the velocity difference between the liquid and
Miyamoto (1986) attributed the first type of air the entrained air. However, all these hypotheses,
entrainment to a conjoining-pressure-driven which rationalize the two types ofair entrainment
collapse or a Rayleigh-Taylor type instability of by proposing two distinct mechanisms, have not
a thin film of entrained air. According to the yet been completely verified, and identification
hypothesis by Teletzke, Davis and Scriven (1988), of the various mechanisms of air entrainment are
such a thin air film is always entrained between still subject of ongoing research. A more detailed
an advancing liquid and a substrate above a discussion can be found in Chapter 3.
critical speed. Miyamoto and Scriven (1982) Kistler (1983) obtained an operating window
484 Curtain coating
Us = 8.12[Vcosex]0.81/'100. 19 that summarizes
E 100
their experimental data for bone gelatin and a
Z PULLED FILM
.s 80 polyethylene terephthalate base. Here, Us denotes
z the maximum coating speed, V is the impinging
0
U5 60 velocity, ex is the inclination of the web from the
i
Z
W
I-
FREE FALL horizontal plane at the point of impingement (ex
w 40 is positive when the web runs downward) and
0
<l: '10 is the viscosity at zero shear rate (A. and n in
u..
a:
::::>
wj the Carreau model are expressed as functions of
(/)
0
'10)' Illustrative data are shown in Fig. l1c.28.
I) 20 40 60 The second ingredient in the explanation of
(a) CURTAIN HEIGHT [mm) Blake, Clarke and Ruschak (1994) is a simplified
model that predicts the position of the dynamic
80 contact line from boundary layer theory (see also
E Chapter 3). A comparison of experimental
Z
.s air-entrainment data and predicted dynamic-
z 60 PULLED FILM
wetting-line positions revealed that the maximum
0
U5 coating speed without catastrophic air entrainment
Z
W 40 is highest when the wetting line is located beneath
I- FREE FALL
w the impinging curtain. This finding suggests that
0
<l: 20 the hydrodynamic forces resulting from the impact
u..
a: of the curtain delay the establishment of an air
::::>
(/)
0, bearing between the liquid and the solid, and
0.5 l.0 1.5 2.0 2.5 thereby 'assist' dynamic wetting.
(b) FLOW RATE [cm 2 /s) The analyses of Kistler (1983) and Blake,
Clarke and Ruschak (1994) make plain that long
Figure l1c.25 Operating bounds arising from pulled curtains and low-viscosity liquids are effective in
film phenomenon (Ogawa, Kashiwabara and Scriven delaying the onset of air entrainment. The rapid
1992): (a) effects of curtain height and surface flow development near the dynamic wetting line
tension; (b) effects of flow rate and surface tension. suggests a positive effect of a shear-thinning
Base condition: p = 103 kg/m3, (J = 60 mN/m, Q = coating liquid on delaying the onset of air
1.9 cm 2 /s, " = 30 mPa s, He (curtain height) = 0.02 m, entrainment (Kistler 1983; Ogawa and Scriven
U = 2m/s. 1994). Another important parameter is the
impingement angle of the curtain with respect to
for curtain coating from his finite-element analysis the substrate, as is also evident from the correlation
in the high-capillary-number regime and compared obtained by Blake et al. Sone, Hanawa and
his predictions with experiments, as shown in Kageyama (1989) suggested that the preferred
Fig. 1Ic.27. He took computed solutions with angle between the falling liquid sheet and the
an apparent dynamic contact angle of 180· to moving web ranges from 120· to 150· (the angle
identify a hypothetical limit of air entrainment. is measured on the upstream side of a curtain).
This limit qualitatively captures the effect of the The mechanism behind the impingement angle
Reynolds number on air-entrainment velocity in effect can be rationalized with the concept of
experiments, as seen in the figure. 'hydrodynamic assist' but is not fully understood.
Blake, Clark and Ruschak (1994) rationalized The influence of surface energy of the solid
the capability of curtain coating to coat at high substrate is weaker in curtain coating than in
coating speeds with the concept of'hydrodynamic slide coating because hydrodynamic effects are
assist'. Their starting point is the correlation so dominant in the impingement flow.
Operability 485
(a) (b)
Figure llc.26 Photographs of coated samples with air entrainment (Miyamoto 1986): (a) bubbles and craters
on mostly well coated film (conjoining-foree-driven instability); (b) completely disrupted layer (shear-driven
instability).
llc.5.4 HEEL FORMATION for moving flat surfaces (Sakiadis 1961). According
to the model, the position ofthe dynamic wetting
When the flow rate is increased, the dynamic line is dictated by the development length needed
contact line migrates upstream in proportion to for the displacement thickness of the boundary
the increase of the distance needed for the layer to reach the final film thickness. That
boundary layer to develop (see Fig. llc.12). The development length is placed in the flow so that
resulting heel with its eddy often traps bubbles, the corresponding nominal (99%) thickness of
agglomerates, gels, or other particles resulting the boundary layer is reached at the vertical
from a deteriorated coating solution. Furthermore, extension of the top free surface of the falling
in the presence of a heel, the dynamic contact curtain. The final result for the dynamic contact
line tends to not be straight, causing various line position x* is
types of down-web striations.
Kistler (1983) and Blake, Clarke and Ruschak (1 - x*) uIV = 0.383 Re (1Ic.6)
(1994) explained heel formation by means of a where x* is normalized by the thickness of the
simple model based on boundary layer theory falling sheet and measured from the vertical
486 Curtain coating
100,- -.--- --, ----,

I I
k It
HEEL
AND
AIR ENTRAINMENT
~
a0- EEL
Il
Q)
a: -
a:
10..- HEEL ~ "-
_~
UJ
~ •.=..=. ~~/
l5'R
(])
~
::J
'" --• ··000 ENTRAIN-
, • •0 0 ~ MENT
Z
(/) "'- 0 /= /
o
oz
~
1f-
;00 /'~H-~/Y_
>- /?" i'/
UJ
a: (0)' ~
~O~ 0/
+y ,,'0:1
*~I
Q)O
0.1 L - - ...l....' ---1I -------J
0.01 0.1 1 10
SPEED RATIO u/v
Figure llc.27 Finite-element predictions of air entrainment and onset of heel formation, as well as comparison
with operating limits measured in experiments (Kistler 1983). Ca'U/V= 3; for definition of x*, see Section
llc.5.4 below.
portion of the rear free surface. Figure Ilc.29 Measures to avoid large heels include increasing
compares the contact line position estimated UIV and reducing Reynolds number. A higher
from equation (llc.6) with results obtained from surface tension shortens the length of the boundary
finite-element calculations. The simple model layer through an increased pressure gradient in
suitably explains the mechanism of heel formation the bead and also makes the heel smaller.
when the flow rate is increased. In Fig. llc.27,
experiments are compared with finite-element
llc.5.5 AIR ENTRAINMENT WITH HEEL
predictions. In spite of significant quantitative
FORMATION
differences, the predictions approximately replicate
the experimental trends. Altogether, the findings At high flow rates, the maximum coating speed
confirm that the dynamic wetting line position is bounded by massive air entrainment in the
and associated heel formation are controlled by presence of a heel (Fig. Ilc.23d). This regime of
a balance between macroscopic, hydrodynamic air entrainment is quite distinct from air entrain-
mechanisms, namely the inertia of the impinging ment at a low flow rate (Section Ilc.5.3), and
liquid and the momentum transfer that accelerates exhibits different flow characteristics. In particular,
the liquid of the given volumetric flow to the air entrainment in this regime seems to cause
final coating speed. At moderate capillary numbers, slip between the liquid and the substrate, which
of course, surface tension also comes into play. results in the formation of very large heels. A
Operability 487
5
......
Ii)
E 3
o
~
M
E
.£ 2
o
0
0 180 36(! 540 720 900
Us [em/5]
• 6.0,0 • 10.2.0 • 25.4.0
• 10.2.45
• 10.2.0 • 25.4.0
Figure llc.28 Maximum coating speed of curtain coating (Blake, Clarke and Ruschak 1994). The data points
indicate the conditions for clearing of air entrainment as coating speed is reduced. The liquids were aqueous
solutions of bone gelatin with viscosities of 63 mPa s (black symbols) and 220 mPa s (gray symbols). The legend
gives the curtain height He in centimeters, followed by the impingement angle in degrees (measured between
the direction of web motion and the horizontal).
complete understanding of the mechanisms Kistler (1983) also noted bistable pairs of flow
involved has yet to be reached. It is plausible, states with and without air entrainment, and
however, that air entrainment becomes massive plotted experimental limits of air entrainment in
in the presence of a heel because the heel lessens the presence of heels. Figure llc.27 shows that,
the hydrodynamic assist of wetting. The transition in that regime, the maximum coating speed
between flow states with and without air entrain- decreases slightly when the flow rate is increased.
ment is very abrupt, and is accompanied by a At the time, finite element analysis was not
dramatic change in heel size (in the regime developed enough to predict this phenomenon,
described in Section 1Ic.5.4, in contrast, the heel which still awaits theoretical prediction.
changes size slowly and continuously as a function Degrauwe and Hens (1986) coated a polyester
offlow parameters). There seems to be a nonlinear base with aqueous gelatin solutions of various
jump in the flow states, as is manifested by the viscosities using a slide-fed curtain coating die.
existence of a hysteresis in air entrainment speed, They found that the maximum coating speed is
reported by Blake, Clarke and Ruschak (1994). almost independent of the Reynolds number
These authors found that the clearing speed of (which includes viscosity) and also curtain length
air entrainment when they decreased the flow (Fig. 11 c.30). These findings suggest that, in the
rate is much lower than the speed at the onset of heel-flow regime, microscopic wetting mechanisms
air entrainment when they increased the flow rate. dominate over the macroscopic hydrodynamics.
488 Curtain coating
100 .,--------,--.....,..---..,...-.--...,...-..,.-,-"""'-,-'
1 1 -r if J-
+-
~.
.... 3 +------i-- - - - - H=108mm...L
-l-----'-- . H=7Omrn
:x:
I-
2 +------i-- ------- H=30mm J-
!
~
Z
w
t----
oJ
I-
Z
W
:;
C. 2
o
..J
LU o -~-"'--- UN =4
>
LU
C -1 _ I I! I I
0.1 1 10
REYNOLDS NUMBER Re
~ 10 +----...;---l..-;--U-.\-+-~I--+-l
Figure llc.29 Effect of Reynolds number and speed

ratio on wetting line position: comparison of FEM
calculation and a simple model based on Sakiadis'
boundary layer theory (Kistler 1983).
Therefore, microscopic parameters such as surface

1/ I
energy of the substrate are important for maxi-
mizing coating speed. Surface roughness will
have an effect as well, there being an optimal 7j...····
choice for surface roughness that is most effective
.../-"j
1...•• i
in breaking up the entrained air. Hartman (1989), 1 I
! !
1-'-,----4---;....----;:---+----+--+---+-...;.......I
for example, showed that the maximum curtain
coating speed of photographic materials on matt .1
Vmax [m/s]
paper (rough) is higher than on glossy paper
(smooth), in contrast to slide coating. Electrostatic
assist is also effective to assist air displacement Figure llc.30 Maximum coating speed of curtain
in the presence of a heel. coating (Degrauwe and Hens 1986). v (kinematic
viscosity) = 13, 25 and 38 x 10- 6 m 2 /s, 0' = 34mN/m,
He (curtain height) = 0.03, 0.07 and 0.108 m, Q = 1,
llc.6 PRACTICAL PROBLEMS 1.5,2, and 2.5cm 2/s.
llc.6.1 LIP DESIGN

suitable radii of curvature for the elongated slide
Curtain-coating dies of the slide-fed type usually surface, ranging from 20 to 40 mm. They claim
incorporate an elongated, downward bent slide that these radii reduce the minimum flow rate
block along which the surface velocity of the needed to maintain a stable curtain. Accelerating
falling film increases. This configuration involves the liquid flow decreases the surface density of
numerous design factors, such as the inclination surfactants and increases the local surface tension,
of the plane slide upstream, the radius ofcurvature thereby reducing the difference in the surface
of the elongated slide, the inclination of the slide tensions between the top and the bottom free
at the terminating lip, and the angle at which surfaces. That way, the curtain is stabilized.
the underside of the lip is cut back. Fahrni and The inclination of the slide at the lip affects
Zimmermann (1978), in their patent, describe the teapot effect as mentioned in Section llc.3.
Practical problems 489
So does the inclination of the cut-back of the lip
underside, as is also explained in that section. In
practice, of course, that inclination has an upper
limit dictated by the shape of the lip and fabrication
capabilities. Greiller (1975) showed an example
of 30· from horizontal. Dittman et al. (1976)
disclosed lip designs with a chamfered, overhanging
slide and a cut-back on the underside, some of
which Kistler and Scriven (1994) also analyzed.
Fahrni and Zimmermann (1978) chamfered the
lip to form a sharp edge.
llc.6.2 EDGE GUIDES

Edge guides are essential for maintaining the
coating width of a curtain: without them, the
curtain tends to contract into a triangular sheet
and, sufficiently far downstream, forms a liquid
cylinder. Edge guides can be a problem, however,
because by necessity they involve three-dimensional
flow fields that are prone to cause local variations
in coating thickness. Edge guides can also
significantly influence the stability of the curtain.
Several patents suggest the use of different
liquids in the edge regions from those in the
central region in order to obtain a stable curtain. Figure llc.31 Partial curtain at edge region (Ridley
Ridley (1977) proposed a system with three 1977).
distinct partial curtains that together constitute
an integrated curtain, as shown in Fig. Hc.31.
In this system, the flow rate of the edge region
can be increased independently to make the 23
whole curtain stable, even when the central
portion requires a lower flow rate. Improvements
to this patent address the thick edge regions that
are unattractive because they may delay drying. I
Watanabe, Tanaka and Yoshimura (1980), for 21; 1
instance, installed a jet of liquid adjacent to the ;1:
'I:
.R
edge guide as shown in Fig. llc.32. They claim !
that a small amount of added liquid flowing in \
the direction of the curtain is sufficient to prevent 36'
36
III
shrinkage or breakage of the curtain. Recently, II
Reiter (1989) proposed an edge configuration
that flushes the top part of the edge guides with
a low-viscosity liquid, and then extracts most of 21 22
that liquid by means of a suction device near the
lower end of the edge guides. This improvement Figure llc.32 Jet adjacent to guide bar (Watanabe et
minimizes the drying load at the edges. al. 1980).
490 Curtain coating
A suitable choice of edge-guide material is also 100
important for maintaining a stable curtain. Greiller
(1975) used stainless steel rods which, since, have
80
become standard. Other materials such as
polyvinyl chloride and Teflon appear in examples
~
~
E Boundary layer
of later patent embodiments. As a general rule ~ 60
~
of thumb, wettable materials assure a stable ·u0
curtain but produce a thicker coated edge. Katagiri
(1992) illustrated the effect of contact angle by
means of fully three-dimensional finite-element
..
0;
>
u
·c
40
calculations. His results suggest not only a strong >" 20

influence on curtain thickness near the edge
guide but also a danger ofcurtain break-up when 0
the contact angle is large. The mechanism of the u.o 0.5 1,0 1.5 2.0
edge stability is quite complicated. Nishida (1990) (8) Distanee from edge guide (em)
measured the vertical and horizontal velocities
in a curtain with a laser-Doppler velocimeter.
He uncovered the development of a Blasius type 4,0
boundary-layer along the edge guide, and a
resulting flow toward the center of the curtain
(Fig. Ilc.33). This lateral flow creates a thin film 3.0
near the edge guide. Katagiri (1992) analyzed the ~~
same flow field using the three-dimensional ~
'g
finite-element method mentioned already. His 0;
>
2.0
predictions (Fig. llc.34) reveal the same flow
characteristics as Nishida's measurements. The "E2
·c
flow profile near the edges strongly depends on 0
:I: 1.0
the capillary forces that arise from wetting of the
edge guides by the liquid, and the forces coming
from the surface tension gradient in the curtain 0.0
caused by a difference ofadsorption time between 0.0 0,5 1.0 1.5 2.0
the central region and the edge region. The (b) Distanee from edge guide (em)
capillary force attracts the liquid toward the edge
guides, and the surface tension gradient creates Figure llc.33 Vertical and horizontal velocities ob-
a flow toward the center of the curtain. Creating tained by laser-Doppler measurements in the curtain:
(a) vertical velocities; (b) horizontal velocities (Nishida
a balance of these effects is essential for preventing 1990).
thinning of the curtain near the edge guides
(Nishida, Katagiri and Suga 1992). move the curtain die from a holding station to
the running web, as shown in Fig. 11c.35
(Kawahara and Yamagishi 1979). Several design
llc.6.3 START-UP
improvements have appeared in patents. Reduc-
Two methods for starting up a curtain coating tion of the thickness at the very beginning of
operation have appeared in patents. One method coating, where it is usually thicker than the
is to remove a catch pan that receives the falling subsequent normal position, is vital for high
liquid near the moving substrate during prep- speed coating. Solutions have been proposed for
aration or interruption of the coating machine, improving the shape of the pan edge (Kozak
as shown in Fig. llc.3. The other method is to 1989; Morikawa, Kobayashi and Miyamoto 1989).
T (t4E l6: II • 040 TINE 16:U·0(
Slot Exit
CASE NO. I CASE NO. I
11.75E-l
1. 4E-l
I 1.25E-l
1.2E-l ~:1 qi~-.. I .7.S0E-2

Q)
:0 Jllld,t~.~
~Wtl,:'JSSt~ I U2.S0E-2
1. OE-l ::J
(9 iiijli-UIIIIII 1-J-2.S0E-2
~ II 11'1...1
r r I~ I I
Q)
0>
8.0 -0 . .IIIIINII IN I I U-7.S0E-2
W
III ..... l-l.2SE-l
5.0
III~-'III
,II
~ : I -1.7SE-l
4.0
-2.7SE-l
1
2.0
r"<-'
-3.2SE-l
.-3.7SE-l
0.0
~
~ Flow
[ CONTOUR) ( CONTOUR 1
VELOCI TY (X. y. Z) YELOCITYOC. Y.Z)
NAX , 1.3S22f:-1 MAX: 2.aS6BE-2
NIN , 0.0000 tl1N : -9. 3.492E-.L
<a) (b)
Figure llc.34 Velocity contours obtained by three-dimensional finite-element analysis (Katagiri 1992): (a) vertical velocity contour; (b) horizontal
velocity contour. p = 10 3 kg/m 3 , " = 50mPa s, (J = 4OmN/m, 2W (slot clearance) = 0.04 m, Q = 4cm 2 /s, (J«contact angle at edge guide) = 90·,
Hdcurtain height) = 0.1 m.
492 Curtain coating
liquid sheet and the dynamic contact line. Another
r+--.,..,- -- --,, approach to preventing disturbances to the curtain
I is reducing the ambient air flow. O'Connor (1981)
suggested shrouding the entire coater set-up with
a porous, foam-like material.
1lc.6.S VARIABLE CGATING WIDTH

A special feature of curtain coating is its ability
to accommodate changes in coating width easily.
A classic method employs a curtain that is wider
than the substrate, unused liquid being caught
in a pan (Greiller 1972). The coating width can
even be made narrower than the web to leave
uncoated parts on both sides. This may be
important, for instance, when the substrate is a
thin porous material such as paper that is
3
weakened and prone to breakage when wet.
Narrower coating widths can be achieved, for
example, by folding the edge of the running web
downward along both sides of a coating roll so
Figure lle.35 Startup method of curtain coating that the liquid is attracted to the edge guide
(Kawahara and Yamagishi. 1979). rather than the web (for a particular implemen-
tation, see Kawahara et al. (1983)). Further
llc.6.4 AIR SHIELDS improvements in variable coating-width systems
come from sliding catch pans, which are installed
Shielding the curtain sheet from ambient air flow at both edges and slide in the cross-web direction
and protecting the dynamic contact line from to adjust the coating width (Kaneko and Naruse
the impact of the air dragged towards the curtain 1986). Another approach yet is to slant the die
by the substrate are important for achieving in the horizontal plane to reduce the effective
uniform coating. An air shield such as the one coating width as described by Naruse (1985).
in Hughes' patent (Fig. llc.3) is a basic way to
prevent disturbances. Further refinements have
lIe.7 CONCLUDING REMARKS
been made since. Ridley, Denewtown and Stokes
(1980) employed a curved rigid member placed This chapter reviews theories, experiments and
in close proximity to the curtain and combined patents pertinent to curtain coating in order to
with a carefully controlled laminar gas flow to give an overview of current curtain-coating
guide the trajectory of the falling liquid sheet technology, and to indicate directions for future
over its entire height and thereby achieve a developments.
uniform wetting line. Morikawa and Kobayashi Almost a century of continuous improvements
(1987) used a suction box as an air shield. Sone and refinements have transformed curtain coating
and Hanawa (1991) pressed a roller against the into a high-speed production process for multi-
substrate just upstream of the impingement layered materials and many other precision-coated
region. Ghys, Mues and Geerts (1992) used a products. The outstanding feature of curtain
concentric cover plate around a portion of the coating is its operability at high speeds, mainly
coating roll. All these designs are aimed at brought about by the impinging inertia ofa freely
decreasing the impact of air flow on the falling falling liquid sheet. However, at the same time,
References 493
the unsupported liquid sheet is a source of many Do, D. V. and Christodoulou, K. N. 1992. Multilayer
difficulties in practice. The intent of most of the curtain coating flows: stability and sensitivity to
patents reviewed in this chapter is to deal with small disturbances. Paper read at AIChE Spring
National Meeting, New Orleans, LA.
these difficulties, and thereby facilitate operation Fahrni, F. 1976. Coating device. US Patent 3,973,062.
of the process and improve coating quality. Fahrni, F. and Zimmermann, A. 1978. Coating device.
However, several patents typically exist for the US Patent 4,109,611.
same purpose, indicating a lack of a clearly Finnicum, D. S., Weinstein, S. J. and Ruschak, K. J.
optimal method as well as a lack of thorough 1993. The effect of applied pressure on the shape
understanding of the physics. of a two-dimensional liquid curtain falling under
Complete understandings of air entrainment, the influence of gravity. J. Fluid Mech. 255:
647-665.
air entrainment with heel formation, and start- Ghys, J. 1., Mues, W. and Geerts, H. J. 1992. Curtain
up procedures are issues that will have to be coater. European Patent EP 0 489 978A1.
addressed further in the future. The first two Greiller,1. F. 1972. Method of making photographic
phenomena will require a two-phase flow analysis elements. US Patent 3,632,374.
with molecular interaction between liquid and Greiller, 1. F. 1975. Apparatus for production of
solid, breakdown of the entrained air film, and photographic elements. US Patent 3,867,901.
dissolution of the air. Analyses of the instabilities Hartman, R. L. 1989. High speed curtain coating
process and apparatus. World Patent 89/05477.
of the entrained air film and the coated liquid are Hughes, D. 1. 1970. Method for simultaneously applying
keys to the physics of the problems. The third a plurality of coated layers by forming a stable
problem is fairly complicated because design multilayer free falling vertical curtain. US Patent
factors, such as the edge shape of the catch pan, 3,508,947.
influence the phenomenon. Apart from steady-state Kaneko, N. and Naruse, Y. 1986. Coating apparatus.
analysis, stability and transient analyses will be Japanese Patent Publication JP 61-35880A.
important in solving these remaining problems. Kaneko, N. and Scriven, L. E. 1988. Transient analysis
of a curtain coating flow. Paper read at AIChE
With increasing demand for multilayer func- Spring National Meeting, New Orleans, LA.
tional materials, the application ofcurtain coating Katagiri, Y. 1990. Comparative study of unsteady
will become more widespread, provided the behavior in slide and curtain coating. Paper read
difficulties in its operation can be solved with at AIChE Spring National Meeting, Orlando, FL.
further improvements. Katagiri, Y. 1992. Analysis of edge effects in curtain
coating. Paper read at AIChE Spring National
Meeting, New Orleans, LA.
Kawahara, S., Ogura, T., Isayama, S. and Hirose, M.
REFERENCES 1983. Coating method. Japanese Patent Publication
Blake, T. B., Clarke, A. and Ruschak, K. 1. 1994. JP 58-3672A.
Hydrodynamic assist of dynamic wetting. AIChE. Kawahara, S. and Yamagishi, K. 1979. Process for
J. 40: 229- 242. thermosensitive recording sheet. Japanese Patent
Booth, G. L. 1970. Coating Euipment and Processes. Publication JP 54-74761A.
New York: Lockwood Publishing. Kistler, S. F. 1983. The fluid mechanics of curtain
Brown, D. R. 1961. A study of the behavior of a thin coating and related viscous free surface flows with
sheet of moving liquid. J. Fluid Mech. 10: 297-305. contact lines. PhD thesis, University of Minnesota,
Cox, E. R. 1968. Curtain coating of web stock with Minneapolis. Ann Arbor, MI: University Microfilms
hot melts. TAPPI. 51(7): 39A-42A. International.
Degrauwe, D. and Hens, 1. 1986. Limits of curtain Kistler, S. F. and Scriven, L. E. 1994. The teapot effect:
coating: comparison of theory with experimental sheet-forming flows with deflection, wetting and
data. Paper read at AIChE Spring National Meeting, hysteresis. J. Fluid Mech. 263: 19-62.
New Orleans, FL. Kozak, S. 1. 1989. Curtain coating start-up method
Dittman, D. A., Griswold, G. L., Wright, S. E. and and apparatus. World Patent WO 89/07283.
Winkler, D. E. 1976. Curtain coating apparatus. Lin, S. P. 1981. Stability of a viscous liquid curtain.
Research Disclosure. No. 14715. J. Fluid Mech. 104: 111-118.
494 Curtain coating
Lin, S. P., Lian, Z. W. and Creighton, B. J. 1990. Poirier, C. C. 1966. Curtain coating of corrugated
Absolute and convective instability of a liquid sheet. paper board. TAPPI. 49(10): 66A-67A.
J. Fluid Mech. 220: 673-689. Pritchard, W. G. 1986. Instability and chaotic behavior
Miyake, A. 1973. Coating die. Japanese Patent in a free-surface flow. J. Fluid Mech. 165: 1-60.
Publication JP 48-34946A. Reiter, T. C. 1989. Curtain coating method and
Miyamoto, K. 1986. On the nature of entrained air apparatus. World Patent WO 89/10583.
bubbles in coating. Paper read at AIChE Spring Ridley, K. A. 1977. Curtain coating method. US Patent
National Meeting, New Orleans, LA. 4,019,906.
Miyamoto, K. 1991. On the mechanism of air Ridley, K. A. 1979. Curtain coating apparatus. US
entrainment. I ndustrial Coating Research 1: Patent 4,135,477.
71-88. Ridley, K. A., Denewtown, J. E. and Stokes, L. R.
Miyamoto, K. and Scriven, L. E. 1982. Breakdown of 1980. Curtain coating. Great Britain Patent GB
air film entrained by liquid coated on web. Paper 1 559 701.
read at AIChE Annual Meeting, Los Angeles, CA. Sakiadis, B. C. 1961. Boundary-layer behavior on
Morikawa, M. and Kobayashi, K. 1987. Coating continuous solid surfaces. AIChE J. 7: 26-31,
method and apparatus. Japanese Patent Publication 221-224,467-470.
JP 62-289264A. Sone, Y. and Hanawa, T. 1991. Coating method.
Morikawa, M., Kobayashi, K. and Miyamoto, K. Japanese Patent Publication JP 3-94870A.
1989. Coating apparatus. Japanese Patent Publication Sone, Y., Hanawa, T. and Kageyama, T. 1989. Coating
JP 1-304076A. method. Japanese Patent Publication JP 64-51170A.
Mues, W., Hens, 1. and Buiy, L. 1989. Observation of Suga, Y., Kobayashi, K., Sasahara, T. and Miyamoto,
a dynamic wetting process using laser-Doppler M. 1991. Coating method. European Patent EP 426
velocimetry. AIChE J. 35: 1521-1526. 122 A2.
Naruse, Y. 1985. Coating method. Japanese Patent Taylor, G. I. 1959. The dynamics of thin sheets of
Publication JP 60-44077A. fluid. Part III: disintegration of fluid sheets. Proc.
Nishida, S.1990. The formation offlow rate distribution Roy. Soc. London A. 253: 313.
near the edge guide ofliquid curtain. Paper read at Taylor, L. H. 1903. Machine for production of
the SCEJ 55th Meeting, Nagoya, Japan. solidification forming nutriment with whole cover
Nishida, S., Katagiri, Y. and Suga, Y. 1992. Coating as confectionery and pie. German Patent Publication
method. European Patent EP 0517 170A2. DE 145517.
O'Connor, T. R. 1981. Coating apparatus provided Teletzke, G. F., Davis, H. T. and Scriven, L. E. 1988.
with a protective shield. US Patent 4,287,240. Wetting hydrodynamics. Rev. Phys. Appl. 12:
Ogawa, S. and Scriven, L. E. 1990. Analysis of curtain 1473-1484.
coating flow. Paper read at AIChE Spring National Wandtke, L. A. 1964. Curtain coating machine. US
Meeting, Orlando, FL. Patent 3,132,968.
Ogawa, S. and Scriven, L. E. 1994. An analysis of Watanabe, T., Tanaka, K. and Yoshimura, Y. 1980.
curtain coating flow. AIChE 1. (to be published). Coating method and apparatus. Japanese Patent
Ogawa, S., Kashiwabara, Y. and Scriven, L. E. 1992. Publication JP 55-73365A.
Theoretical and experimental study ofcurtain coating Yoshida, T., Ujihara, M., Watanabe, K. and Takagi,
flow. Paper read at AIChE Spring National Meeting, A. 1978. Method to coat onto both sides of web.
New Orleans, LA. Japanese Patent Publication JP 53-115754A.
SURFACTANT EFFECTS IN COATING lId
PROCESSES
P. Randall Schunk* and L. E. Scriven t
ltd.1 INTRODUCTION
Free surfaces and laminar viscous flow are the example, in aqueous coatings of photosensitive
features common to all coating flows which are materials, surfactants are employed as solubilizers,
susceptible to the effects of surfactant additives. sensitizers and dispersants, or to influence specific
Surfactants can change drastically the interfacial characteristics of the final product, such as
physics and chemistry of aqueous solution - and surface roughness and permeability to processing
more rarely organic solutions - because they solutions (Sturge 1977; Knox 1970). But as long
preferentially concentrate, or adsorb, at interfaces. as surfactants have been used for these purposes,
The surfactant concentration there can be orders they have also been used as real or imagined
of magnitude larger than in the bulk liquid, coating aids, ostensibly to make the coating
depressing surface tension far below that of the method more operable, or operable over a wider
solvent. If the liquid flows in such a way that range of operating conditions. It is made plain
fluid elements moving along the surface undergo in Chapter 4 that surfactants playa significant
rapid compression or expansion (area change), role in controlling and sometimes causing coating
then the surface composition changes. If the defects, both through a static (equilibrium) effect
surface deformation is not everywhere the same, and through a so-called dynamic surface tension
neither is the surface concentration: there are effect.
concentration gradients. Concentration gradients The concept ofdynamic surface tension, however,
make surface tension gradients. The consequence is often misunderstood. If a surface-active substance
is in the balance of forces at the interface: the is dissolved in a liquid, the surface tension is lower
depression of capillary forces and the induction than that of the pure liquid. Surface tension is a
of surface tension gradients can stabilize the flow, local equilibrium property and is most reliably
lead to large changes in interfacial shape, and measured when the whole body of solution is at
sometimes induce unsteady flow. equilibrium, i.e. uniform in composition and
Surfactants are often dissolved in liquids to be temperature, and at rest. The dynamic surface
coated to serve sundry purposes, the nature of tension is the local tension measured or estimated
which depends on the type of coating. For somewhere in a solution and surface that are not
at rest, but flowing, and flowing in such a way
* Current address: Sandia National Laboratories, Department that parcels of liquid moving along the surface
of Incompressible Fluid Dynamics, Albuquerque, are deforming. When the surface is expanding or
NM 87185, USA. contracting, the amount of surfactant adsorbed
t Current address: University of Minnesota, Department of
Chemical Engineering and Materials Science, Minneapolis, per unit area changes by diffusion and adsorption,
Minnesota, 55455. or desorption, to re-establish equilibrium. Thus
London. ISBN 0 412 06481 2.
496 Surfactant effects in coating processes
dynamic surface tension is a result of a local of momentum, and models and constitutive
non-equilibrium process of surface area change, equations of surfactant equilibria provide the
adsorption and desorption, diffusion and convective framework for analyzing the mechanisms; these
transport, and even micellization as well. In principles are described in Section Ild.2. Appli-
contrast, (static) surface tension is a well-defined cations of published theories to coating and
thermodynamic property. related flows is the subject of Section Ild.3. The
The primary goal of this chapter is to reconcile highlights there are several analyses of the effects
the concept of dynamic surface tension with the of surfactant at static and dynamic wetting lines.
physical mechanisms by which surfactants exert In the remainder of the chapter (Section Ild.4)
effects and then to identify what surfactant we examine three viscous free surface flows that
properties are relevant so that practical extensions serve as coating-related case studies. The first is
of this research can be made. As such what is of film flow down an inclined plane (Section
presented here complements Chapter 4, which 11d.4.2), the simplest of all viscous free surface
focuses on methods of measuring the relevant flows. The second is of an overflowing pool
surfactant properties (e.g., static surface tension, (Section Ild.4.3), a method which epitomizes the
dynamic surface tension, diffusion coefficients, difficulties of interpreting measurements of
etc.). Together we hope these chapters will provide apparent surface tension of a flowing liquid. The
a guide, as well as valuable insight, towards the third is of slide coating flow (Section Ild.4.4),
correct selection of surfactant systems to aid in the most widely used method of depositing
the production of single and multilayer coatings. multilayered photographic films.
Mechanisms by which surfactants affect free-
surface flows hinge on the complex interplay of
1111.2 THE THEORY OF SURFACTANT AcnON
surfactant convection, diffusion, adsorption and
even micellization (aggregation). The governing The theory describing surfactant action must, of
principles of conservation of mass, conservation course, include the principles that govern small
p
~~
y2) MICELLES GIVE
~ MICELLES TAKE
&~
C?2)
b J
()-J)
p
r----o b
DIFFUSION DELIVERS
&\f
C?2)
d
DIFFUSION REMOVE
d
cf ~ I
0----
b !<Y 0-r'
Figure 1111.1 Surfactant action at and near a free surface.

The theory of surfactant action 497
scale viscous free surface flow. The relevant script referring to the aggregation number.
equations and boundary conditions are the Expressions for qi come from the model of
Navier-Stokes equations, which govern the flow micelle-solution exchange equilibrium, an
in the bulk liquid, the kinematic and traction example of which is described below.
boundary conditions, which account for mass Boundary conditions on (lId.!) are
and momentum conservation at gasjliquid surfaces
and liquidjliquid surfaces, and special conditions n'Vc i =0 for i = 1,2, ... , N at solid surfaces
which control the behavior of gasjliquid/solid (IId.2)
lines of contact (Chapter 2). In addition, the
theory must account for the surfactant inventory,
an inventory generally made up of individual n'Vc l = SCIRe(~:L)
surfactant molecules in solution, which are
surface-active and are free 'monomers' with respect
to aggregation; molecules, or monomers, adsorbed
x [~~ + 2Hn'(u - us)r + vs,(ulIn]
at the surface; and molecules aggregated in
micelles. The mechanisms by which surfactant + SC: (r* )
Sc c.* L v;r at free surfaces (lId.3)
moves into and out of the surface and through
which micelles affect that movement are portrayed r is the surface concentration (moles of adsorbed
in Fig. IId.I. Those mechanisms and the under- material per unit area) in units of the surface
lying theory have three parts: convective diffusion, concentration, P, that is in equilibrium with the
adsorption-desorption, and micelle-solution initial concentration of monomers in the liquid
exchange of surface active monomers. Cr Sc s == J1./(pD s ) is the surface Schmidt number,
Ds being the surface diffusion coefficient. 2H is
the mean curvature of the surface, Un == (0 - nn)' U
Ild.2.l THEORY OF CONVECTIVE DIFFUSION is the velocity vector tangent to the surface, and
The movement of surfactant monomers and Us is the velocity of the surface.
micelles in the bulk liquid is described by the The first condition (IId.2) makes solid surfaces
convective diffusion equations. In dimensionless impenetrable to surfactant. The second condition
form these equations are (IId.3) allows bounding free surfaces to take up
surfactant from the liquid and hold a significant
inventory of surfactant monomers (Gibbs' defi-
nition of so-called 'surface excess', (Gibbs 1906);
see also Edwards et al. (1991)). Condition (IId.3)
for i = 1,2, ... , N (lId.I) says that the diffusive flux ofsurfactant monomers
normal to the surface is controlled by four
where Cj is the bulk concentration of surfactant factors: (1) the rate of accumulation of surfactant
species i in units of ct, the initial concentration per unit area or/at; (2) the rate oflocal creation
of the same. SCi is the Schmidt number, J1./(pD;), of surface owing to motion of a curved surface
with J1. being the viscosity, p the density, and Di normal to itself 2Hn . (u - us)r (a small effect and
the diffusivity (diffusion coefficient) of species i. contributes only for unsteady flows or evaporating
U is the velocity vector and Re is the Reynolds solvents); (3) the rate of local redistribution by
number, pUL/J1., with U being the chosen unit surface convection Vs ' (unn, and (4) movement
of velocity and L the chosen unit of length. qj is within the surface by diffusion relative to mass-
the rate of generation of species i in units of average motion DsV;r. The third term contains
ctu /L. CI is the concentration of surface-active the effect of surface expansion and contraction
monomers and Cj ' i = 2, 3, ... , N are the concen- - that is, free surfaces are mathematically
trations of surface-inactive aggregates, the sub- compressible, unlike the liquid itself, because the
surface velocity field may have nonzero divergence,

i.e., Vs • UII #- O. The fourth contribution can be a n· T 2Hn [da - Es In (I + ;;)]
=
delicate matter when diffusion itself induces
mass-average motion. Because surface diffusivities
are immeasurable this contribution is neglected
-ES(-._l_)VsC
+ a
s (11d.5)
Cs
here except near dynamic contact lines, where
the surface concentration gradients are so steep
==
Here Es (RTr <X»)/p,U is the surface elasticity
number, which measures the relative importance
that some diffusion is required to stabilize the
of viscous force and the decrement of capillary
concentration field at affordable levels of com-
force owing to surface tension lowering by
putation.
surfactant; and Ca is the capillary number (see
==
Chapter 2). a· a/C~ is the dimensionless
Ild.2.2 SURFACE CONSTITUTIVE RELATION
Langmuir constant, or the concentration at
- ADSORPTION MODEL which r = r <X)/2 in units of C~ (Chapter 4). Cs
denotes the dimensionless concentration just
To complete the theory of surfactant transport below the surface, in units of q. What is called
in free surface flows, what is needed is a constitutive Gibbs elasticity ofa surface is defined as (Rusanov
equation that relates surface excess per unit area, and Krotov 1979)
r, and bulk concentration, C l' in the liquid next
to the surface. Here, as an approximation, only
the monomers are allowed to concentrate at the E==(d~:r) (1Id.6)
surface. Generalizations that allow for dimers Gibbs elasticity is a property of an interface that
and small aggregates to do the same are given
acts through interfacial (surface) tension gradients
by Franses (1979). These are not accounted for
that, in turn, develop as a result of differential
in the present analysis.
amounts of expansion or contraction of neigh-
The Langmuir-Syszskowski equation ofsurface boring elements of the interface:
state (equation (4.5)) is a reasonably accurate
description of equilibrium adsorption of many (lid.?)
surfactants. Combining it with equation (4.1) and
Gibbs elasticity E is a true mechanical property
integrating yields the equation for surface tension
only if it is measured in the absence of bulk-
in terms of surfactant concentration:
surface exchange of surfactant. Only in that case
does the shear stress in the surface respond
instantaneously to a surface deformation, without
diffusion-induced stress relaxation. With some
as is the surface tension when the surfactant rearrangement the last term in equation (1Id.5)
concentration is zero, or equivalently the surface can be combined with equation (4.1) to give
( +I)
tension of the solvent; R is the ideal gas constant;
Es V (lnr) (lId.8)
T is the temperature; and r <X) is the surface Es. Vs c1 =.c s s
a c1 a
concentration under saturated conditions. That
(lldA) is a reasonably accurate description of By comparing (1Id.7) with (1Id.8), it is clear that
equilibrium surface tension data of many aqueous Gibbs elasticity implied by the Langmuir-Szys-
surfactant systems is discussed in Chapter 4. zkowski equation is (Es/a*)c s' So the surface
When the constitutive equation (lldA) is elasticity can be determined from input parameters
combined with the balance of traction at the free to the adsorption model, i.e. a* and r A review
<x).
surface, (cf. Chapter 2), a working condition of many attempts to measure surface elasticity
results on momentum transport in which appears directly is given by Lucassen-Reynders (1981);
explicitly the surface concentration: see also Chapter 4.
The theory of surfactant action 499
Ild.2.3 MICELLE-SOLUTION SURFACTANT What the Bury-Hartley model cannot explain
INVENTORY is that the critical concentration of different
Not all surfactants form micelles in water. When properties is often different; that in some cases
they do, the micelles normally show up above a at least there is substantial preaggregation below
certain range of surfactant concentration. This the critical micelle concentration; that micelle
range is often narrow and can resemble a phase sizes, as measured by aggregation numbers, are
transition. For this reason it is termed Critical distributed over a definite range; and that
Micellization Concentration, or simply CMC equilibrium is established in much less time than
(Mukerjee and Mysels 1971). Many ther- the waiting time for simultaneous encounter of
n surfactant molecules.
modynamic properties, surfactant monomer
activity, or chemical potential, solvent activity, Physically a micelle must be built up by a
surface tension, etc., change more or less abruptly succession of binary encounters, as recognized
at the CMC. first by Meyer and Van der Wyk (1937). At
Micelles are equilibrium aggregates ofcolloidal equilibrium the distribution of aggregates is
dimension, i.e., they are made up of 10 to 100 governed by a set of stepwise-addition reactions
monomers, and are stabilized by physical forces. (Hall and Pethica 1967):
They tend strongly to stay in equilibrium with k,
the remaining molecules, or monomers, from R1 + R1 --.+R 2

k,
which they are formed. Various models to account k2
for this equilibrium have been put forward. The R1 + R2 --.+R 3 (1Id.l0)
k,
oldest one by Bury (1927) and Hartley (1936)
regards a micelle as a colloidal entity, made up
of n monomers ofsurfactant, that is in equilibrium kN-l
with the individual molecules in solution: R1 + R N - 1 --.+ R N

kN-l
Here k; is the rate constant for the ith forward

step and k; is the rate constant for the ith reverse
Here Cn is the concentration of micelles of size =
step.K; C;/(C;_lC1)istheequilibriumconstant
n, i.e., made up of n monomers, C1 is the for the ith addition reaction. In reality the i
concentration of monomers in solution and ~Gn different steps could have i distinct equilibrium
is the free energy of micelle formation. Typically constants. But unless some simple functional
in aqueous solutions n, the aggregation number, relationship can be established among all of the
is greater than 20 - often it is around 100. constants this is not practicable.
In this so called Bury- Hartley model a micelle Bidner, Larson and Scriven (1976) examined
is formed supposedly by an n-molecule encounter. a simpler model with just two different equilibrium
This is disconcerting because even a three- constants, and two additional constants which
molecule encounter is improbable and a simul- define the range of aggregation numbers of
taneous encounter of twenty or more molecules micelles in contrast to premicelles, or submicelles,
is so unlikely as to be practically impossible. as depicted in Fig. lld.2. The 'floor' N is the
Nevertheless, the Bury-Hartley model describes lowest degree of aggregation at which a complete
qualitatively the CMC, predicting an abrupt micelle exists. The aggregates that contain less
change in many physical properties ofthe solution than N monomers are referred to as premicelles.
at the CMC (see Kamrath 1984). For this reason The 'ceiling' N + P is the maximum possible
it has found widespread use in the study of degree ofaggregation. The aggregates that contain
surfactant transport near surfaces at which more than N but less than N + p monomers are
monomers concentrate (for example, Miller 1981 referred to as micelles. The N - 1 reaction steps
and Lucassen-Reynders 1981). needed to form a micelle are supposed to be
1, 2, ... ,N - 1. If the equilibrium constants for

the latter are all K == K j , then at equilibrium
(KCl)C l = C 2, (KC l )2C l
c
= C 3 ,···, (KClt-lC l = CN
or, equivalently,
CN = K CN - l
(11d.12)
C eq 1
1
where K eq == K N - l is the effective equilibrium

n constant between the monomers and the micelles
N N+p
of aggregation number N. The dimensional rate
FLOOR CEILING
of formation of each species, Qj == (Ci V /L)q; (see
equation (lld.l», can now be expressed with
Figure Ild.2 Micelle-solution inventory typical of
mass-action principles:
surfactant systems (Bidner, Larson and Scriven 1976).
unfavorable equilibria and are characterized by

a relatively small equilibrium constant K j , at + k~+ 1 C N+ 2 + ... + k~+P-l C N+ p
least small compared with those K == K j which
characterize the p micelle-solution exchange - kNCNC l - ... - kN+p-1CN+p-lCl
steps. Consequently the distribution statistically
favors micelles, and the concentration of
premicelles is insignificant. Bidner, Larson and
Scriven (1976) suggest characteristic values of p
should be 20 or 30, more-or-Iess in accord with
a thermodynamic treatment by Ruckenstein and
- k~CN+l - kN+lClCN+l
Nagarajan (1975) and experimental evidence by
Shinoda et at. (1963). dC N + p
Hereafter the usually small concentration of QN+P = ~ = kN+P-lCN+p-lCl
submicelles is neglected and the entire inventory
is contained between the monomers and the p
different species of micelles: Physically, the difficult reaction steps are the first
N - 1. Hereafter we account for the division of
CT = C l+ NC N + (N + I)C N + 1 + ... surfactant monomers between those free in solution
+ (N + p)C N + P (lld.ll) and those which make up a micelle with the
equilibrium constraint (l1d.12). The kinetic
CT is the total concentration of individual expressions (lld.l3) describe the monomer-
monomers that would be in solution if no micelle exchange process.
micelles were formed. With the model of Bidner, Larson and Scriven
Because the equilibrium constants of the ith (1976) six additional physical parameters must
step, for i > N, are all the same we take kN = be specified in order to account for micellization
k N+ l = ... = kN+pandlikewisek~ = k~+l' = ... in any mathematical model of a flow field: one
= k~ +p', i.e. the rate of monomer emission or for micelle diffusion, i.e. D N , three for micelle-
adsorption is the same for each micelle. The same solution equilibria, i.e. N, p, C~/C~, and two for
assumptions are made for the unfavorable steps micelle-monomer exchange kinetics, i.e. k' and
Theoretical treatments of coating and related flows 501
k. To predict physical realizable values of each lId.3 THEORETICAL TREATMENTS OF
requires measurement of kinetic exchange rates COATING AND RELATED FLOWS
of monomers among the micelles and the
distribution of aggregation sizes that accounts Although experimental studies on the effects of
for the entire inventory of surfactant monomers. surfactant in viscous free surface flows abound
Although numerous theoretical studies have been in the open literature, there has been relatively
made to elucidate the equilibrium aspects of little in the way of theoretical analysis. The
micellization by models that predict the behavior majority of the theoretical treatments have been
of physical properties, e.g. surface tension, kinetics limited to insoluble surfactant systems, the effects
ofmicellization have received much less attention of surfactant being relegated to a free boundary
(Kreshek et al. 1966; Muller 1972; Hall 1987); equation akin to equation (11d.3) and the balance
however, insight has been gained through measure- of normal and tangential stress at the free surface
ments of micellar equilibration rates (Mijnlieff (Chapter 2). When surfactants are allowed to be
and Ditmarsch 1965; Leung and Shah 1986). appreciably soluble, theoretical analysis can
As equilibrium between micelles and monomers become much more difficult: the main problem
is established very rapidly, fast-reaction-techniques is the relatively low diffusivity of surfactants in
are required to study them. Among those which liquid compared with the kinematic viscosity of
have been applied successfully to micellar solutions the liquid, i.e., high Schmidt numbers, which
are temperature jump, pressure jump, and leads to exceedingly thin concentration boundary
ultrasound adsorption. These methods lead to layers. The challenge, analytically, has been to
the relaxation time for a micellar system: the time handle the momentum transport in the flow
needed for micelles, and monomers that make simultaneously with surface tension gradients
them up, to re-establish equilibrium after equi- and surfactant convection and diffusion.
librium has been upset. The relative (relaxation) In this section we make no attempt to cover
step in the present model is the relatively slow the voluminous literature available on dynamic
dissociation of monomers from the micelles. The surface tension, but instead we review the most
characteristic relaxation time of these steps is significant analyses ofsurfactant effects in coating
(Kreshek et al. 1966) and related flows, in both the insoluble and more
importantly the soluble regimes. Most research
or == [k(C 1 + CN) + k'r 1 (lld.14)
on the hydrodynamic effects of surfactant has
Diffusivities of surfactants in solution, both focused on the motion of drops and bubbles, the
monomers and micelles, must also be measured, spreading of liquids on solid surfaces, and the
or predicted, to complete the theory. The task is damping of waves on films. Below we highlight
difficult and only a few methods have been the analytical treatments most significant to each
successful (see Chapter 4). This is especially the of these areas. We then discuss the few contri-
case when the surfactant is ionic and its concen- butions available that directly address the effects
tration is above the CMC. Experimental evidence of surfactants in coating flows.
suggests that even when the surfactant is nonionic Two-phase flow in porous media (e.g., Ginly
the micelle diffusivities are as much as 30% and Radke 1988) and the dynamic behavior of
smaller than that ofthe monomers (Weinheimer, foams (e.g., Hirasaki and Lawson 1985; Edwards
Evans and Cussler 1980). When the surfactant et al. 1991) have interested researchers in the
is ionic the micelle diffusivity can rise with effects of surfactants on the motion of bubbles
concentration above the CMC due to electrostatic through channels of various shapes. Analytical
interactions (Weinheimer, Evans and Cussler treatments of the problem are usually approached
1981). In this chapter, diffusion coefficients are by simplifying the bulk momentum equations
left as empirical parameters and independent of with lubrication theory, which results in surface
concentration. evolution equations that can be integrated (e.g.,
Hirasaki and Lawson 1985; Ginley and Radke urements. They also give a historical review of
1988; Ratulowski and Chang 1990). all previous attempts at solving the problem.
Unfortunately, the lubrication approach pre- The effect of surfactants on the ability of
cludes the analysis of complex flow patterns, i.e., liquids to spread on solids has also been a
recirculations, and for the most part has been popular subject of research. Much of the work
applied to insoluble systems. For flows dominated in this area is closely related to the hydrodynamics
by viscous and capillary forces the boundary- of wetting, the essential part of most coating
element method is an efficient route to a solution processes. In Section Ild.4.4 we cite the analytical
and has been applied extensively to drop and treatments of spreading where they are relevant
bubble hydrodynamics (see Chapter 9; Stone and to our analysis. The interested reader should
Leal 1990; Lu and Chang 1988); unfortunately, consult reviews given by Kistler (1993) and Bose
the method does not lend itself to convective (1993).
diffusion analysis and hence its use has been That surfactants have a significant damping
mostly confined to insoluble systems. Wassmuth effect on waves in quiescent pools and in films
et al. (1993) have recently made a significant flowing down inclined surfaces has been known
advance with their numerical treatment of the for some time (cf Lamb 1879; Emmert and
moving bubble problem that is general enough Pigford 1954; Whitaker 1964). Thin film coating
to handle complex flow regimes and convective- is just one area of technology that has kept
diffusion of surfactant. Like the work we present interest in this problem alive. Early theoretical
in this chapter, they solve the full two-dimensional studies of the wave-damping effect were made
Navier-Stokes and convective diffusion problem. with quiescent liquid pools. The first to explain
The motion of drops in a uniform stream of the damping mechanism was Lamb (1879) who
fluid under the influence of surfactant has also deduced the behavior of small-amplitude
been the subject of extensive research (e.g., Oguz sinusoidal waves by what amounts to linear
and Sadhal 1988; Argrawal and Wasan 1979; stability analysis. Although his analysis was
LeVan and Newman 1976). In fact, this problem limited to insoluble and inextensible films of
benefited from one of the first numerical treatments surface active material, his predictions were
of bulk diffusion of surfactant in more than one accurate as was his insight to the underlying
space dimension (LeVan and Newman 1976); in physical principles. Nearly a century later
that work, however, the effect of surfactant on Dorrestein (1959) extended Lamb's analysis to
the velocity field was not considered. Sadhal and include extensible films but then Levich (1945)
Johnson (1986) have since put forward an exact and independently Hansen and Mann (1964)
solution to the mixed boundary-value problem gave the first explanations of the effects of soluble
for the case of insoluble surfactants. The case of surfactants, i.e., diffusion and adsorption effects
soluble surfactants in the continuous phase have included. Finally, Lucassen (1975) added micell-
been treated by Holbrook and LeVan (1983a,b) ization to the picture (cf Lucassen 1975, Rillaerts
who used the collocation method to solve the and Joos 1982, Lucassen-Reynders 1987), but the
convection-diffusion problem for realistic surfac- more complex interplay of the underlying prin-
tant properties. Their results showed an appreci- ciples which control stability was not laid bare.
able amount of interfacial velocity retardation. Beyond quiescent pools there are falling liquid
Oguz and Sadhal (1988) used the method of films, and waves running down them in many
matched asymptotic expansions to generalize the circumstances grow, as is well established; that
theory for insoluble as well as slightly soluble is, the flow can be unstable (Benjamin 1957,
surfactants. The results oftheir treatment graphi- Krantz and Goren 1971). That adding soluble
cally exhibit the physical phenomena involved surfactant to the liquid can dampen waves, i.e.
in the mass transport within the drop and lead stabilize the flow, is also well established (Emmert
to quantitative agreement with experimental meas- and Pigford 1954; Whitaker and Jones 1966).
Theoretical treatments of coating and related flows 503
Linear stability analysis still pertains to the surfactant monolyer to the substrate. One
evolution of disturbances up to visible wave mechanism involves dissolution of the monolayer
inception, but is complicated by the main flow into the bulk phase followed by re-adsorption
of the liquid film, the 'base flow', which ramifies to the substrate. He concluded that two hy-
Lamb's approach into a combination of the drodynamic parameters are needed to make the
Orr-Sommerfeld equation oflinear flow stability problem well posed: the dynamic contact angle
theory and the linearized convective diffusion and a surfactant transfer ratio. Cox (1986) also
equation. This combination has been solved analyzed the effect of insoluble surfactant on
without major restrictions on the range of validity moving contact lines; however, he did not allow
of predictions only by Whitaker (1964) and Sterzi for transfer of surfactant to the substrate. More
(1982); however, the still more complex interplay recently, Bose (1993) considered soluble systems.
of underlying phenomena remains unexamined His modeling results show that Rame's surfactant
in the literature, so far as the authors are aware. transfer ratio can be used to get both buildup
Among the approximate analyses that apply to and depletion of surfactant at an advancing
limiting cases, those of Anshus and Acrivos dynamic contact line.
(1966) and Lin (1970) for waves that are much The work of Bose and his colleagues is significant
longer than the film is thick are noteworthy, because most practical coating applications, with
because they verify that Whitaker's (1964) results the exception of Langmuir- Blodgett films (see
are accurate in the long-wave limit. But neither Chapter 13), involve surfactants that are appreci-
have they shed light on how stability is controlled ably soluble. Ruschak (1987), in one of the few
by the interplay of underlying phenomena. We other available treatments of soluble systems,
return to this problem in Section Ild.4.2, where studied the effects of several surfactant properties
we show that the damping effect can be explained on the undesirable flow induced by air currents
by the phasing of the surface tension gradient- after the liquid is coated. Wu and Whitaker
induced shear stress relative to the wave amplitude. (1986) examined the effects of surfactant at and
Direct analysis of the effects of surfactant in near static separation lines. They deduced that
coating flows includes research in the area of secondary flows (recirculations) near such lines
Langmuir- Blodgett film deposition (see Chapter provide a convective mechanism by which
13 for review; Rame 1988; Park 1991). Park surfactant molecules are brought to the surface
(1991) used the method of matched asymptotic much more rapidly than those delivered by
expansions to solve the governing equations and diffusion, and, consequently, steep surface con-
found that the surface speed profile is greatly centration gradients there are unlikely. Goet-
affected by the presence ofan adsorbed monolayer. mackers and Scriven (1992), following the work
He built on the analysis of the effects of surfactant we present below, probed the effects of surfactant
monolayers on the movement of a bubble front on the steady operating states and the dynamics
confined in a channel, which takes on a similar of curtain coating. Most recently Giavedoni and
appearance to dip coating (see Ratulowski and Saita (1992) examined the effects of surfactant-
Chang 1990; and Chapter 13). Rame (1988) also induced interfacial viscosity on location of the
examined Langmuir-Blodgett film deposition. dynamic contact line terminating the upstream
With similar analytical techniques he examined meniscus in slot coating.
the effects of the monolayer on the moving Like Giavedoni and Saita (1992) many re-
contact line which is present in the plunging searchers are now reverting back to a science
mode (advancing contact angle; see Chapter 13), called 'interfacial rheology' as a means ofmodeling
as opposed to the withdrawal mode (receding the effects of surfactants in capillary hy-
contact angle). Like Park (1991) he focused on drodynamics (see Edwards et al. (1991)). The idea
insoluble surfactant systems but proposed several is to develop constitutive equations that describe
mechanisms which allow for transfer of the how the state of surface stress depends on surface
deformation and the rate of surface deformation: examine the stability around the base flow, which
effectively, the surfactant transport mechanisms is accessible through a fourth-order differential
are lumped into the mechanical properties of the equation, i.e., the celebrated Orr-Sommerfeld
surface. Before some pioneering work by Scriven equation (Benjamin 1957). In the other two cases,
(1960), the term 'interfacial rheology' had been an overflowing liquid pool and a slide coating
employed mostly in reference to experimental process, the flow is too complex to fit any
studies devoted to qualitatively elucidating the standard coordinate system. The culprits are the
rigidity and viscoelasticity ofadsorbed monolayers free surfaces that bound the flow but relocate
at fluid interfaces. Gibbs elasticity, as we defined and reshape in response to changes in properties
it in Section Ild.2.2, is used to describe the elastic and operating parameters, e.g., flow rate, surface
response, and surface dilatational viscosity and tension and so on. The changes usually require
shear viscosity are used to capture the effects of robust parameterization techniques to follow
rate of surface strain. All of these properties can them successfully. These aspects make the
be built into surface constitutive equations, as Galerkin/finite-element technique, suitably aug-
was demonstrated originally by Scriven (1960). mented to handle free boundaries (Chapter 9,
This approach has been used in the study of Kistler 1984; Christodoulou and Scriven 1992)
drops and bubbles (cf Argrawal and Wasan 1979 and the intense surfactant concentration gradients
and LeVan 1981), the motion waves on a normal to those boundaries (Schunk 1989;
monolayer-covered surface (e.g., Lucassen and Chapter 13), the method of choice for an efficient
Hansen 1967), the stability of liquid threads route to a suitable solution.
(e.g., Rusanov and Krotov 1979), the behavior The results of the first case study, which come
of foams (e.g., Edwards et al. 1991) and even from a numerical solution ofa differential equation
recently in coating flows (Giavedoni and Saita in a single space dimension, were found to be
1992). converged with respect to mesh refinement, i.e.
Other than the work we present in this chapter the results proved to be virtually insensitive to
there have been only a few other numerical the further addition of elements to the mesh (see
treatments of the effects of surfactant on viscous Chapter 9). Unfortunately the same cannot be
free surface flow, at least without significant rigorously claimed for the second and third case
approximation. Three of them are mentioned studies (overflow cell and slide coating flow),
above (Bose 1993; Holbrook and LeVan 1983a,b; although we did find in most cases that the
Wassmuth et al. 1993). Another one that we cover solutions agreed qualitatively, and even sometimes
in Section Ild.4.3 is the finite difference work by quantitatively, to published experimental data
Bergink-Martens et al. (1990), who examined the (see Section Ild.4.3). We believe that sufficient
effects ofsurfactant in an overflowing liquid pool. resolution is lacking only near static and dynamic
contact lines (see Section Ild.4.4) and in regions
where the solution is unimportant (flow over the
lld.4 CASE STUDIES OF SURFACTANT
rim in the overflow cell). We are confident that
EFFECT
we use sufficient resolution of the steep concen-
tration gradients that occur normal to deforming
Ild.4.1 METHODS OF ANALYSIS
free surfaces. Moreover, most of the conclusions
In the coming subsections, the Navier-Stokes we draw agree with published experience, as we
system (Chapter 2) and the convective-diffusion bring out as often as possible in the following
system (equations (1Id.l)-(lld.3)) are solved discussion.
together to investigate the physics of surfactant The reader should be cautioned, however,
effects in three prototypical flows. Flow down when interpreting the results regarding the effects
an inclined plane is the first and possesses a of the ad hoc models employed at the dynamic
simple, analytically tractable base flow. We contact line, models which have never been
Case studies of surfactant effect 505
substantiated by careful comparisons with ex- surface concentration can also be used to explain
perimental data. Because the traditional hy- how surfactants temper wave growth.
drodynamic assumptions break down at the Given below is a complete assessment of the
dynamic contact line, the results obtained with mechanisms of instability (and stability) within
any approach are subject to scrutiny. The reader the bounds of the linear theory for small
is urged to take the results as no more than the disturbances on flowing liquid films which hold
demonstration that, whatever the correct model, a significant inventory of surface-active material
the solutions are extremely sensitive to the in the surface layer. Surfactants that are soluble
physics near the dynamic contact line. The reader are also allowed to diffuse and aggregate (form
is also urged to consult several authoritative micelles).
references on the subject (Kistler 1993; Bose 1993).
Ild.4.2.1 Linear stability analysis
The tools we us are normal mode analyis of the
lld.4.2 FLOW DOWN AN INCLINED SLIDE
linearized vorticity equation, i.e. the Orr-Som-
It is well known that the wave-damping capability merfeld equation, and the principles which govern
of surface active substances, or surfactants, must surfactant inventory, i.e. convective-diffusion
hinge on surface tension gradients. Incipient equations for both the monomer and the micelles.
disturbances, which can be viewed as propagating Disturbances that are small enough to justify the
waves, tend to compress and extend periodically linearization of the vorticity equation can be
the adsorbed layer of surfactant at the free represented mathematically in terms of combi-
surface, thereby inducing a surface tension gradient nations of sinusoids, or Fourier modes. Because
that aids or opposes the disturbance. The gradient most dangerous modes are two-dimensional
is reduced and shifted by diffusion, the more (Squire 1933) it suffices to restrict disturbances to
soluble the surfactant, and can be further altered the type shown in Fig. lId.3.
by uptake and discharge of soluble surfactant When the temporal form of the Fourier modes
from micelles, if they are present. This section is substituted into the Navier-Stokes equations,
uses the theory put forth in Section Ild.2 to the stability formulation results in the Orr-Som-
predict how nonequilibrium processes ofdiffusion merfeld equation (see for example Whitaker
and micellar exchange of surfactant affect the
stability of liquid films running down inclined
planes - 'falling liquid films'.
The key to understanding the mechanism
behind the damping effect of surfactant is the
magnitude and phase angle of the surface
SPATIAL GROWTH RATE e ...... •
concentration response relative to wave height.
These of course determine the surface tension
response and the phasing of surface tension
gradients relative to the disturbance. The latter
induces a shear stress on the underlying liquid
and through it couples the hydrodynamic stability
problem with surfactant transport. Smith (1990)
reconciled these phasing concepts in the surfactant-
free case, arguing that shear stress, properly
phased relative to wave amplitude, heightened
disturbance growth rate. However, few have Figure ttd.3 Flow down an inclined surface: definition
recognized that the magnitude and phase of the sketch.
1964, Yih 1963). The Orr-Sommerfeld equation these equations can be found elsewhere (Whitaker
is a fourth-order ordinary differential equation 1964; Schunk 1989).
and hence requires four boundary conditions: The relevant parameters in the Navier-Stokes
two at the solid wall and two at the free surface. system are the Reynolds number, Re, the capillary
At the wall the two conditions are those of number, Ca, and the angle of inclination p. The
impenetrability and no-slip (Chapter 2). At the first two of these are described in Chapter 2. The
free surface the balance of normal and tangential relevant parameters in the convective diffusion
forces (equation (lld.5)) is employed. system are the Schmidt number of each species
In steady flows of films down inclined surfaces =
(monomer and micelle), SCi J1./(pDJ, the re-
there are no surface tension gradients; the system cruitment depth (or strength of adsorption),
is in complete chemical equilibrium and bulk de =(dr/dC1)/h, and the elasticity number,
concentrations of each species are everywhere =
Es (l/J1.U m )do/dlnr. Here we specialized to a
the same. When this base flow is perturbed linearized version of the Langmuir adsorption
sinusoidally, surface compression and expansion model (equation (4.5)): r = (dr/dC1)C 1. The
by each normal mode cause small departures elasticity and Schmidt numbers are described in
from equilibrium. The convective diffusion Section 11d.2. The recruitment depth is based
equations, i.e. equation (lId.1), can be linearized on surface-subsurface equilibrium and is the
accordingly, and the solution can be represented depth of liquid which contains the same amount
with the same generic Fourier modes as used for of surfactant as the surface, measured in units of
the velocity and pressure, necessarily of the same the film thickness h. The remaining parameters
frequency because the source ofthe concentration appear in the source and sink terms of the
disturbance is wave expansion and contraction. convective diffusion equations and are used to
A second-order ordinary differential equation account for the surfactant monomer-micelle
system results for each species. The boundary exchange process. These are also described in
conditions as described earlier enforce impen- Section 11d.2. The physical and material properties
etrability at the substrate surface and a species needed to determine these dimensionless groups
balance between the excess surfactant at the are given in Table lId.I.
surface and that just underneath (cf. equations The Orr-Sommerfeld equation and its bound-
(11d.2)-(11d.3)). The stability formulation of ary conditions together with the linearized form
Table lld.l Surfactant liquid properties corresponding to sodium-dodecyl sulfate in water at 20·C
Parameter Value Source
Monomer diffusivity D1 0.4 x 1O-9 m2/s Weinheimer et al. (1981)

Dimensional recruitment
depth deL 1O-2 cm Mysels and Florence (1970)
Langmuir constant a 34.3mol/m 3 Prins et al. (1967)
Saturation adsorption roo 18 x 1O- 6 mol/m 2 Prins et al. (1967)
Aggregation number of
smallest micelle N 98 Ruckenstein and Nagarajan (1975)
Corkill and Walker (1972)
Micelle size distribution p 3 Corkill and Walker (1972)
Micelle diffusivity DN 0.5 x 1O- 9 m2/s Weinheimer et al. (1981)
Micelle/monomer
concentration ratio cZ/C~ 1 Kamrath (1984)
Rate constant k 1...... 1 x 108 S-1 Leung and Shah (1986)
Rate constant k' 1...... 1 x 108 molar- 1s- 1 Leung and Shah (1986)
of the convective-diffusion equations and their

boundary conditions are homogeneous. As such
they constitute a complex eigenproblem, the
'-
~
~
Gl
C!
0
eigenvalues of which depend on the disturbance ci

0
forms of interest, i.e., either temporal or spatial t-
U
or even some combination of the two. In the ...

« lD
0
0 EsiRe = 0
:r
temporal formulation, complex values of the ~
0
wave speed w (the eigenvalue) are sought at a a:
specified real wave number IX so as to avoid a «
...
Cl
a:
trivial solution. The temporal growth rate in this ...
0
case is eaWj', where w; is the imaginary part of w. ::E
w
t-
In the spatial formulation, complex values of IX 0.2
::E N
are sought given the real frequency, W == IXrWr • :::>
::E C!
x 0
0.4
The spatial growth rate is eaix , IX; being the «
imaginary part of IX. The corresponding eigen-
::E
0.7~
0.9
functions physically represent the variation of 0
the stream function disturbances and the con- 10' 10 2

(a) REYNOLDS NUMBER Uh/v
centration disturbances· across the base flow
direction as well as the unknown disturbance
amplitude at the free surface.
With the use ofthe finite element method (e.g.,
Schunk 1989) or the finite difference method (e.g.,
Whitaker 1964) the system ofcomplex differential ~
-0
.,;
equations can be transformed into an algebraic a:
eigenproblem. Of interest are those solutions to ~
U II>
<C
the eigenproblem which correspond to marginally ... C!
0
:r 0
stable and most dangerous states with respect to
~
wave number or frequency. The most useful o
a:
Cl
result, however, is the curve of most dangerous oJ 0 I"""'"------~:__------'o,__--_.f
perturbations, that is the wave number or S

:
t-
frequency of the most rapidly amplified pertur- til II>

o
bation at each value of a specified physical C!
<;>
parameter.
o
9 '-----,---...-----r---r-----,r----!
lld.4.2.2 Results o 6
234 5
FREQUENCY HZ
(b)
Effects of surface elasticity
Figure IldAa shows predictions of the effect of Figure lld.4 (a) Temporal growth rate (!Xc;) of the
surface elasticity on the growth rate of the most most dangerous disturbance. de-oo (insoluble limit).
dangerous temporal disturbance in the limit of Ca = 1, f3 = n/2. (b) Spatial growth factor -!Xi vs.
disturbance frequency OJ with varying Schmidt number
negligible bulk-surface interchange of surfactant
SCI (ditTusivity). de = 10- , Ca = 1, f3 = n/2.
2
(this ideal situation is approached as either
de -+ 00 or D 1 -+0 (Sc 1 -+ (0)). We divided the
elasticity number in Fig. IldAa by the Reynolds gradient forces with inertial forces. The curves
number in order to compare our results with show that tangential surface forces owing to
those of Whitaker, who scaled the surface tension surface tension gradients tend to delay at a given
wave number the onset of instability to higher how the Schmidt number Sc 1 affects the rate at
Reynolds number and, moreover, decrease the which disturbances of a given frequency amplify
growth rate of the most dangerous disturbances, or decay in the downstream direction. Actually
provided that Es > 0; for Es < 0 these forces are these curves describe the frequency response,
destabilizing. Surface elasticity supports the appearing similar to those of the output of a
stabilizing effect of capillary pressure, and both band-selective amplifier. As diffusion coefficients
have a greater stabilizing effect for shorter waves, fall, unstable modes are confined to lower
i.e. larger wave numbers. frequencies and the growth rates in the unstable
In the limit approached as the wavelength band decrease.
becomes large, i.e., as the wave number approaches The time scale 't d == (Sc 1 h 2 /v)d; characteristic
zero, the disturbances at the surface penetrate of diffusional interchange is the time needed for
deeper and the influence of the base flow profile diffusion to extend over a distance of order of
curvature is greater. Schunk (1989) showed that the recruitment depth de == (dr/dC l)jh (see Cussler
temporal growth rates of long-waves (small ex) 1984). Plainly 't D depends on the concentration
go as e,,(6Re/5 -E.-cot P-,,2/ 3 /Cal. This stability result of surfactant when the Szyszkowski-Langmuir
applies to insoluble, spread monolayers and is equation for r (equation (4.5)) is employed. For
accurate to first order in ex. It shows that surface a dilute surfactant-water system and a base flow
elasticity has an effect at leading order in ex; film thickness h of 0.1 cm, Sc 1 is of the order of
capillary pressure, on the other hand, enters as 1000 and de is of the order of 10- 2 (Mysels and
a third-order effect. Thus for ex = 10 - 2 surface Florence 1970). The time scale T D is then of the
tension would have to increase 100-fold more order of 0.1 s. Hence, for frequencies of surface
than surface elasticity to achieve the same degree deformation considerably less than to Hz diffu-
of stabilization. For film flow down vertical sional transport has a noticeable influence on
planes, or f3 = n/2, the flow becomes completely surface tension gradients.
stable to disturbances of all wave numbers as
the elasticity number exceeds 6Re/5. This result
agrees well with the authors' predictions (to Effects of micellization
within 1%) and to a lesser extent with those of Because an exhaustive study of the effect of each
Whitaker (1964) (to within 20%) who reported micellization parameter in the model of
a critical value of t.03Re. micelle-solution exchange is beyond the scope
ofthis work, the authors chose here a representative
case study based on micelle kinetic studies of
Effects of diffusion
sodium dodecyl sulfate (SDS) in water described
When diffusion is added to the story, the surface in the work of Leung and Shah (1986) and in
tension variations no longer respond instan- the earlier work of Mijnlieff and Ditmarsch
taneously to surface deformation. In this case (1965). The parameters needed to describe the
the mechanism by which surfactant affects stability micelle dynamics are the last six given in Table
is most easily explained in terms of time scales, Ild.I. Of these, the micelle-solution exchange
not only those of surface deformation but also rates k and k' and the micelle/monomer concen-
those characteristic to diffusion. tration ratio are the most sensitive to the presence
The characteristic time scale of surface defor- of other additives (Leung and Shah 1986). For
mation goes as the reciprocal of the wave frequency, instance, electrolytic, organic, and even polymeric
t/ro. By imposing the disturbance frequency ro additives can change characteristic aggregation
as a parameter the time scale becomes a physical rates by several orders of magnitude. Because of
parameter. Figure 11dAb shows the spatial growth this sensitivity, parameter studies were focused
factor exi (growth rate e"jX) as a function of on k, k', and C~/Cr.
disturbance frequency. The various curves show Figure 11d.5 shows how sensitive the maximum
spatial growth factors are to the characteristic
relaxation time 'm of the micelle-monomer
exchange process; ,;;; I depends linearly on the
rate constants k and k' (equation (lld.I4». The ... 0
results indicate that the more rapid the exchange 0
of monomers between the micelles and solution
the more rapid equilibrium between the surface .-
and the bulk is re-established, and hence the (j
more unstable the flow because of the reduction
a:
of stabilizing surface tension gradients. Rapid 0
I-
micelle-solution exchange clearly facilitates U
<I:
transport of the monomer species to and from u.
J:
the surface. 0
The degree to which micelles help re-establish ~

0
equilibrium, and thereby the degree to which a:
C)
10 -5
gradients along the surface are reduced, depends ..J
<I:
on the time scale over which micelles, which are t=
<I:
surface-inactive, re-establish equilibrium in sol- a.
en
ution relative to the time scale over which
movement of monomers, which are surface active,
by diffusion re-establishes equilibrium between ...
o
the surface and the bulk. The degree to which o+-_-L.~_.....L...-r"'-'--I.._-,-----f
. 0.0 1.0 /..0 3.0 '1.0

equilibrium is upset in the first place depends on
the time scale of surface expansion, or 2n divided FREQUENCY HZ.
by the frequency.
Figure lId.5 shows that when k' and k are Figure lld.5 Effect of micelle-monomer relaxation
such that 'm = 10- 2 s (equation (IId.I4): time on the spatial growth factor vs. disturbance
C I = 0.003 molar, k' = 100 molar-Is-I, frequency OJ. SCI = 1000, d. = 10- , SC N = 4000,
2
k = lOOs-l, and CN/C I = 1), the dependence of f3 = 1[/2, Ca = 1, Re = 1.

spatial growth rate on frequency is nearly the
same as that when no micelles are present. maximum spatial growth factors higher. In the
Clearly micelle-solution exchange is inconsequen- high frequency limit, it is expected that all curves
tial during the time scale of surface deformation, in Fig. IId.S will again become independent of
which is longer than 0.1 s in the range offrequencies micelle dynamics, because in that range the
covered. The diffusion time scale (dr/dC I)2/D I system behaves as if it were insoluble.
ofall curves in Fig. lld.5 is of the order of 0.01 s;
it is shorter than that of surface deformation.
Ild.4.2.3 Mechanism of the stabilizing effect
Figure IId.S also indicates that spatial growth
factors corresponding to each relaxation time fall The predictions in the preceding section indicate,
on the same curve at frequencies less than about in the range of parameters examined, that
1 Hz. In that range micelle effects appear to be surfactants tend to stabilize film flow. The
small, or at least small compared with diffusion sensitivity of surface tension to concentration is
effects. In any case, at frequencies greater than responsible, the magnitude of which is set by the
1 Hz, micelle-solution exchange speeds up the surface elasticity number Es. The physical
diffusion process, surface tension gradients are mechanism begins with local surface expansion
smoothed, and stability is diminished: bands of and compression. Expansion dilutes and com-
unstable frequencies become broader and the pression concentrates any surfactant deposited
or adsorbed at the liquid surface. If the expansion gather near the crests and scatter near the
or compression is not everywhere the same, troughs, by virtue of surface compression being
neither is the surface concentration. Through least near the crests and surface expansion being
surface elasticity, concentration gradients make greatest near the troughs; the maximum and
surface tension gradients. minimum rate of strain is of course a quarter-
The damping effect of surfactant rests on the wavelength displaced, which in this case is at the
phase lag of surfactant response behind the wave nodes where Vs . u is either a maximum or
disturbing wave. Conveniently the phasing is a minimum. The phase lag of surface expansion
explicit in each normal mode of the stability rate behind the wave is -1[/2 and maximum
analysis described above. By virtue of the way surface expansion rate coincides with the rear
complex variables are used to represent quarter-wave points; the maximum surface
sinusoidally periodic change, the normal modes compression rate coincides with the forward
of surface expansion rate V.' u == (I - nn)' V' D, quarter-wave points. Because there is no diffusion,
surface concentration response c 11 y = 0' and surface the surface concentration is in phase with wave
tension gradient response V.a can be readily amplitude, and the surface tension gradient is in
evaluated. Because each normal disturbance mode phase with the forward quarter-wave point. The
varies sinusoidally in space and time, the rate of surface tension gradient is directed from the
change of any quantity necessarily, in time, lags crests towards the troughs and in that way tends
90', or 1[/2 radians, behind its amplitude; and the to support effects of capillary pressure, which
gradient of any quantity necessarily leads 90' acts downwards on the crests and upwards on
ahead of its amplitude. the troughs.
The phase lag of surface expansion rate behind As the Reynolds number Re of the base flow
wave amplitude depends on the flow near the increases, the locations of maximum surface
surface. The phase lag of surface concentration compression and expansion shift relative to wave
depends on the flow, and hence the phase lag of amplitude; so does the surface tension gradient.
surface expansion, and the convective diffusion On the right of Fig. 11d.6 the Reynolds number
process in the liquid beneath the surface: were is 10 and the flow was found to be neutrally
there no diffusion, the location of minimum stable, although it would be unstable were it not
surface concentration would coincide with that for the surfactant - vertical falling films are
of maximum surface expansion, which is a unstable to any disturbance when no surfactant
quarter-wavelength out of phase with that of is present (Krantz and Goren 1971). The phase
maximum surface expansion rate. The phase lag shift relative to the low-Reynolds number case
of the surface tension gradient is exactly 1[/2 on the left is roughly 1[/4. What this shift does
behind the surface concentration: wherever the to the stabilizing effect ofsurface tension gradients
location of maximum surface concentration will is discussed below. First diffusion is added to the
also be the location of minimum surface tension; story.
likewise, wherever the location of minimum Diffusion affects the stabilizing action of
surface concentration will be the location of surfactant in two ways: the first is by causing the
maximum surface tension. phase of the surface tension to lag the wave crest;
Figure 11d.6 portrays the surfactant response the second is by reducing the magnitude of the
on a film flowing down a vertical wall when the surface tension gradient. Figure lld.7 illustrates
surfactant is insoluble, i.e., D 1 -+ O. The left of the effect of surfactant diffusivity on surface
Fig. lld.6 shows the disturbance velocity field concentration response. The Reynolds number
over 1.5 wavelengths. Disturbances travel from Re is 0.001 and the recruitment depth de is 0.01.
left to right, as does the direction of flow. The The contours ofconstant monomer concentration
Reynolds number is 0.001 and the disturbance are shown at a diffusivity of 1O- 6 cm 2/s; the
decays with time. Surfactant molecules tend to Schmidt number is 1000. The lower the diffusivity,
6,,= -1tI4
c;~=o f--+--..,--~---+--
~u
'\
....... - .... ' ......
/ ...... - ~
-', "- '-.
_--I \ ..... -
I \
-.,..- I I \ ....... I \
. .... -
. .
. ............. ....
' ...... ~,....
",., ~
; I ....- I
. . _A. .--
;
*
,
* -- ,-
,.
.. -
..... _ .. .- JI
- .... ,. "
. . . --
., ....
--"
.. ...
.. - . . . -
- - #
~ ,
" ~,
60' .. - - · · ...... ~, ..
GRAVITY -7 GRAVITY -7
Re =0.001 Re= 10
Figure lld.6 Phasing of surface expansion rate, surface concentration and surface tension gradient relative
to a passing disturbance. Predictions are for a liquid film running down a vertical wall with insoluble surfactant.
The capillary number Ca is unity; the surface elasticity number Es is 10.
the less the disturbances in concentration penetrate disturbing wave: diffusion replenishes faster the
into the film. The surface concentration is a local surface deficit of molecules in the surface
maximum near the wave crests, the exact location layer after that layer undergoes maximum
of the maximum being shifted slightly forward expansion in the wave troughs. When the
on the wave. The phase lag of concentration diffusivity is raised to 0.001 cm 2/s, the phase lag
relative to the wave crest is about n/8. Were there of surface concentration is nearly n/2, the point
no diffusion, i.e., if D 1 -+ 0, the phase lag would ofmaximum concentration being near the forward
be 0 at this Reynolds number (cf Fig. Ild.6). quarter-wave point.
Because there is diffusion, surfactant moves to The matter of how surface tension gradients
replenish local surface deficits of surfactant near stabilize is settled by considering the magnitude
the troughs and relieve local surface excess of and phase of the shear stress loading induced by
the same near the crests, thereby tending to those gradients. The phasing idea takes on a
restore equilibrium. At larger diffusivities, diffusion similar form in the problem of wind first raising
is much more effective at re-establishing equilib- water waves, as Lamb (1879) discovered. He
rium during the period ofthe disturbance: gradients deduced, from what amounts to a linear stability
normal to the surface broaden as do those along analysis of waves on still pools, that two
the surface. The higher the diffusivity, the more requirements of a purely tangential force must
the concentration response lags behind the be met to generate and maintain waves: first, the
SURFACE TENSION GRADlENTS
..l-
Figure lld.7 Contours of constant concentration of surface-active free monomer. Calculations are for a film
running down a vertical wall (from left to right). d. = 10- 2 , P= rr./2, Ca = 1, Re = 1.
tangential force must be strong enough to offset loading because it reduces surface tension, and
the damping effect of viscous forces; second, the thereby capillary force, which acts downward at
tangential force must be distributed so as to act the wave crests and upwards at the wave troughs.
forward on the wave crests, in the direction of Surfactants can also affect normal viscous stress
wave propagation, and backward on the wave in the liquid beneath, insofar as surface tension
troughs, changing signs at the nodes. He also gradients alter the flow. Although not illustrated
showed that a tangential force having an opposite here, this research revealed that surfactants do
sign, i.e., one that lags the first case by 180·, not alter significantly the magnitude or phase of
would temper growth or accelerate decay. The the normal stress loading relative to the waves,
situation with surfactant is no different; surface at least in the range of flow and surfactant
tension gradients act entirely within, and therefore parameters examined (Re from 0 to 100, Es from
tangential to, the surface. In the case under study o to 110, SCI from 0 to 1 X 106 , and de from 0
here, however, the liquid is flowing and main- to 1 X 106 ).
tenance of waves is by forces generated from The way in which surfactant distributes along
gravitational potential and not by wind blowing a surface deformed by normal modes is contrary
over the surface. Nevertheless, surface tension to what many have speculated. Levich (1945) and
gradients exert a tangential force, part of which later Lin (1970) proposed that crests tend to
acts backward on the wave crests and forward stretch films and troughs tend to compress them.
on the wave troughs - though this is not so much This would actually give rise to a destabilizing
the case the higher the diffusivity (cf Fig. ltd.7). mechanism similar to the one proposed by Lamb
There is also the issue of normal stress loading. (1879) and others for wind over water. The
Lamb and more recently Smith (1990) examined correct picture is given in Fig. 11 d.7: the kinematics
this as well. Surfactant affects normal stress of a passing wave tend to gather surfactant, and
thereby depress surface tension, near its crest and than one millisecond. Because dynamic surface
scatter surfactant near its troughs. tension is so difficult to interpret only a small
number of techniques have proved to be practi-
cable. Some of them are detailed in Chapter 4.
11d.4.3 THE OVERFLOWING LIQUID POOL
It appears that one method seeing at least
Predictions of surfactant effects in flow down an some use in the aqueous coating industry is that
inclined surface demonstrate that a complete in which liquid wells up in a small circular pool
theory provides a means to probe relevant and overflows the edge, called the overflow
mechanisms of surfactant action. But how can technique. Although it is easier to operate than
specific surfactant systems be characterized in any other method, the flow field is complex and
terms of the degree to which they can exert poorly understood. As near as the author can
dynamic effects? One way is simply to measure tell, a study with the method was first published
all the relevant properties, i.e., equilibrium surface by Piccardi and Ferroni (1951). The approach
tension versus surfactant concentration, surfactant appears to have originated with M. Mclean, a
molecular diffusivity, and micellar coefficients if researcher with Kodak Limited Research in the
concentrations above the critical micellization United Kingdom, in 1951 (Orem 1988). One
range are of interest. An alternative, although overflow cell design is shown in Fig. 11d.8. In it
harder to interpret, is to measure a so-called liquid is pumped into the bottom of an upright
dynamic surface tension, as described in detail dish and allowed to overflow the rim. As an area
in Chapter 4. element associated with a parcel of surface liquid
To interpret dynamic surface tension requires moves radially outward to flow over the rim it
a complete theory of the physical rate processes expands. As this happens any adsorbed surfactant
involved, i.e. diffusion, adsorption and micelliz- has to cover more area, and so the surface
ation. The ultimate goal of measuring dynamic concentration falls locally. Correspondingly, the
surface tension is to deduce basic physical
properties needed to complete that theory. To
date few methods have emerged with that
capability because the flow field is usually too
complex for simplified analyses to be accurate,
especially when surface tension gradients are
important. Possible exceptions are those methods
designed to detect falling surface tension over
the course of several minutes (see Defay and
Petre 1971; Chapter 4). On that time scale, well
established static methods are suitable, e.g.,
spinning drop, Du Nouy ring, etc. (see Adamson ~:::::=::===i====:::::;"'-I ~9 mm
1982), and there is no flow field to contend with. L
Available methods of measuring dynamic OUTFLOW
surface tension - or otherwise measuring or

inferring the surface tension as a function of time
or of surface deformation rate - are categorized
most appropriately by their relevant time scales:
that is, time scales on which measurements can
be interpreted. Coating flows, as well as many
other processing flows, pose an outstanding
challenge because surface deformation time scales
can be exceedingly small - often they are less Figure Ild.8 Overflow cell design from Schunk (1989).
surface tension rises. What is often called the relevant Navier-Stokes system and species
dynamic surface tension is the apparent surface conservation equations are summarized in Chapter
tension inferred from the capillary force that the 2 and earlier in the present chapter. Here only
expanding surface exerts on a Du Nouy ring, the specializations are made for this particular
Wilhelmy plate, or the like. flow; most pertain to the parameters that define
Although some related techniques have been the flow configuration and the boundary condi-
described (e.g., Fainerman 1986; Imaishi et al. tions at the inflow and outflow surfaces; those
1982), only several other published accounts of parameters that pertain to surfactant transport
the overlfow cell exist: Defay and Roba-Thilly remain for the most part unchanged. Again we
(1954), Padday (1957), and most recently Bergink- use the Galerkin/finite-element technique to solve
Martens et al. (1990) are examples (more can be the governing equations, with an automatic mesh
found in Chapter 4). Defay and Roba-Thilly used generation scheme based on solving an elliptic
a funnel-shaped dish and measured the apparent partial differential equation system; for details
surface tension with a Du Nouy ring, concentric see Schunk (1989).
with the rim of the cell. Padday used a cylindrical
vessel like that in Fig. 11d.8 and attempted to
Ild.4.3.2 Results and discussion: parameter
equidistribute the flow by placing a glass sinter
studies
a few millimeters below the surface. Although
this reduces the required inventory of liquid and The dimensions and geometricl features of the
confines the flow so the flow is better controlled, prototype cell match those of the laboratory cell
it does not change the basic aspects. Bergink- shown in Fig. 11d.8. Physical properties and
Martens et al. combined a theoretical and parameters are specified for sodium-<todecyl sulfate
experimental treatment of a funnel-like design to (SDS) in water at a temperature of 20°C. This is
examine the effects of surfactant on the surface the same surfactant system studied earlier iIi the
speed profile. flow down an inclined surface. The relevant
Interpreting measurements ofdynamic surface properties of the system are given in Table 11d.1.
tension by the overflow cell or any related The equations of overall mass and momentum
method is difficult, chiefly because ofan unknown transfer were solved first over a range of flow
flow field. Based on existing data and the author's rates bounded by those attainable with the
experience, a few disconcerting aspects of the experimental setup, i.e. 0.1 to 3cm 3 /s. Figure
method need to be analyzed before its usefulness lId.9 shows the flow structure at 3 cm 3 Is; the
as a dynamic tensiometer can be assessed. These basic aspects are the same at lower flow rates.
aspects are (1) the heretofore unknown nature The liquid stream entering the bottom of the cell
(strength) and extent of the surface deformations broadens under the action of viscosity. A
and how these depend on flow rate and cell momentum boundary layer builds up from the
dimensions, (2) the possibility of inferring from stagnation point and grows along the free surface.
the measurements basic physical properties, e.g., Much of the liquid is turned back into a large
diffusivities, micellization kinetics, etc., and (3) toroidal recirculation in the upper, outer part of
the effect of surface tension gradients on the flow the cell. The liquid stream jets upward, welling
field. up the surface around the center line.
The surface speed profile is of particular
interest because it determines the characteristic
lld.4.3.1 Analysis
time scale of surface deformation, and hence the
Insofar as the cell sits upright, the flow is depth to which diffusion can penetrate in the act
rotationally symmetric around the cell axis (z- of re-establishing equilibrium. For example, the
axis); see Fig. lId.8. Accordingly we needed to slope of the speed profile determines the rate of
solve only an axisymmetric flow problem. The surface expansion (or compression) and hence
the time scale over which equilibrium between
the surface and the bulk is upset. Figure 11d.9
also shows the speed profiles at five different flow
rates. The rate of surface expansion is the relative '"
0.---------------,---------,
IMPINGEMENT
rate of change of surface area per unit of fiducial AEOI~ AAOlAL OlJTFlOW RECION RlN REGtoN
Xl I+-~._------- ..- - - . .
mass. It is governed by the differential equation I
o
of area change: d[ln(dA)]/dt = V' Vs dA is any
~!
I
EXPANSION RATE
infinitesimal control area and Vs is the velocity
field by which elements of the surface that defines
the system move. In terms of a coordinate basis
that lies within an axisymmetric surface the cm 3/s I
VU
Q Q: 1.4
surface divergence of the surface velocity, V. vs, is 0.8
[d(svs}/ds]/s, where s is a measure of arc length I

along the surface in the radial direction. I
The speed profiles reveal that the kinematics III
o
of the surface flow is radially uneven; the top of o
Fig. 11d.9 shows that its nature can be described
in two zones. The innermost zone, which surrounds 0
the center stagnation point, is the one with the 0

I
0.2 0.4 0.6 0.8 1.0 1.2
most intense surface expansion as parcels of I r I
I R I
liquid accelerate rapidly away from the center. I
There the surface expansion is intense and localized I
(200s- 1 at a flow rate of 1.4cm 3/s), taking on I
I
the greatest value at the stagnation point which I
is ordinarily at (or near) the center of the surface.
Hereafter, this region is referred to as the
impingement zone. In this zone the surface speed
increases linearly with radial distance, indicating
that the flow is irrotational in nature (Batchelor
1967). The surface accelerates to a peak speed at
a distance that is between one-tenth and one-fifth
of the distance from the center to the rim. Then
the surface expansion abates so that adsorption
simply controlled by diffusion can move surfactant
into the surface to bring it toward equilibrium
with the liquid beneath. In this region the speed
falls with radial distance as l/s and the expansion
factor falls to nearly zero. Hereafter, this region
is called the radial-outflow zone. Both surface
zones are shown schematically in Fig. lld.9.
The speed profiles also reveal that the concept
of 'surface age' has no physical basis in an
Figure lld.9 Predicted flow field without surfactant.

Lower part shows pattern of streamlines at a flow rate
of3 cm 3/s. Upper part shows the surface speed profiles
at five different flow rates.
overflow cell of this shape and dimension. Surface 4.0 r - - - r - - - - - - - - - - - : - 1

age is usually defined as the time elapsed from
the creation of a surface, without any excess of IMPINGEMENT REGION
r:=0
adsorbed material (surfactant), and a time during
~
which there is no surface deformation (Defay and 3.0
Petre 1971). Defay and Roba-Thilly (1954) and
Padday (1957) referred to the transit time from .
the center of the surface to the edge as the 'surface 00)
2.0
I -6
age', but the speed profiles in Fig. Ild.9 reveal C =1.2 x 10 molar
o
that the extent of effective surface expansion =1.5 x 10·5 molar
C = 1.0 x 10-4molar
('renewal') spans a large portion of the surface
1.0 = 1.0 x 10·3 molar
surrounding the center, especially at low flow
rates. In fact, in the funnel-like cell design examined
by Bergink-Martens et al. (1990), the impingement
zone spans more than half the cell radius. 0
0 0.2 0.4 r 0.6 0.8 1.0
Figure Ild.l0 shows a sampling of computed (a)
results at different concentrations and flow rates R cell
for SDS in water at 20°C. Figure 11 d.l Oa shows 70
the difference between the dynamic surface tension MPINbEMENT REGION
and the equilibrium surface tension at four E
69 0.9 crrf3/s
concentrations. The flow rate is 0.6cm 3 /s. De- ~>.
partures from equilibrium surface concentration 'C 0.6 crJ Is
are greatest when the bulk concentration is the 68 0.3 cm 3 /s
Z
smallest. The highest bulk concentration, 10- 3 mo- 0
lar, is just below the CMC; at this concentration enz
w
departures from equilibrium are small if not ~ 67
negligible, even near the stagnation point. Time w
() EQUILIBRIUM
scales over which diffusion works to re-establish Cl:
u.
equilibrium are proportional to (r/ci)2/D 1 and cc 66
=>
(J)
this is ofthe order of 1ms for SDS at concentrations
above 10- 3 molar (Mysels and Florence 1970).
The rate of surface expansion is 100 s - 1 in the 0.1 0.2 0.3 0.4 0.5 0.6
impingement region and so the time scale over
which equilibrium is being upset is about 0.01 s. (b)
Hence diffusion is able to re-establish equilibrium
much faster than surface expansion can destroy
it. For concentrations less than 10- 4 molar, Figure lld.lO Predictions for SDS in water at 24°C.
however, diffusion time scales are greater than (a) Effect of bulk concentration on the surface tension
0.01 s. In this case diffusion between the surface profile. The flow rate is O.6cm 3 js. (b) Effect of flow
and the bulk is too slow to re-establish equilibrium. rate on the same. The solution is 2 x 10- 5 molar.
In all cases, the surface expansion rate is small
in the radial-outflow zone and the transport of Figure Ild.lOb shows the effect of flow rate
surfactant is controlled by diffusion. Correspond- on surface tension profiles. The predictions are
ingly, the surface tension falls towards equilibrium. computed for 2 x 10- 5 molar SDS in water. The
As the surface passes over the rim, however, rapid diffusion time: scale, (r/cif/D 1, at this concen-
surface expansion again dilutes the surface and tration is of the order of 1s. It is clear from the
surface tension rises sharply. speed profiles in Fig. Ild.9 that the surface
expansion rate is the greatest when the flow rate The effects of micelles on local surface adsorp-
is the greatest. At each of the flow rates for which tion, and hence local surface tension, is difficult
predictions are shown in Fig. lid. to the reciprocal to extract from measurements of surface tension
of the expansion rate is less than the characteristic in the overflow cell. At concentrations near that
time scale for diffusion. The result is a subequilib- at which surfactant aggregation is thought to
rium surface concentration across the cell, even occur for SDS (about 3 x to- 3 molar on the
at the lowest flow rate, 0.3 em 3 Is, because diffusion basis of the equilibrium measurements of surface
is too slow to re-establish equilibrium. tension; see Schunk 1989) little sensitivity oflocal
The predictions of local surface tension for a surface tension to flow rate is observed. So the
2 x to - 5 molar SDS solution agree to within 5% predictions to follow cannot be corroborated by
with the measurements of the same with the wire the measurements. Nevertheless they are exceed-
probe (Schunk 1989). Propagation of the uncer- ingly useful for studying the effect of micelles on
tainties in the force measurement, wire diameter, the rate at which rapidly expanding surfaces
and contact angle sets the upper bound of random re-establish equilibrium. That micelles can indeed
error at about 3%. One possible source of the exert effects has been confirmed in controlled
discrepancy is the surface constitutive equation experiments (Lucassen and Giles 1975).
(1Id.4); the theory is reliable inasmuch as that For the same micellar system we examined in
equation is accurate. Other possible sources are the previous section on slide flows we again focus
uncertainty in the physical properties with which our parameter studies on the rate constants k
the predictions were made and computational and k' and the inventory CZ/C!. Figure lId.11
error caused by insufficient mesh refinement. shows the profiles of monomer and micelle
1.1
Z
0 C N+2
~
a:
C·N+2
z~
OZ
~~ 1.0
<Z
a:o Cl
~(.)
Zw
W(.) C·
c.J< 1
Zu.
0a:
c.J::::I
-- ~6~
~~ ---
Will
c.J:i;
<::::I 0.9
11._
k'
a: a:
+ ~
::::lID
Ill:]
k
5
0
W
-1 ·1 -1
k =1s k' =1 mol s
0.8
0 0.2 0.4 0.6 0.8 1.0
r
R
Figure 1111.11 Surface concentration profiles of monomer and micelle species. The flow rate is 0.5 em 3/ S•
Predictions are for (1.003 molar SDS-water solution.
concentration along the free surface at a flow micelle-solution exchange clearly facilitates
rate of 0.5 cm 3 Is. Evidently the distribution of transport of the monomer species towards the
concentrations of surface active monomers and avid surface.
micelles shifts when chemical equilibrium is The characteristic micelle relaxation time is
disrupted by the expanding surface ofthe overflow given by equation (lld.14) and its magnitude for
cell. Near the cell center micelles emit monomers the case in which k = lOOmol- 1 s-1, k' =
in an attempt to replenish the free surface deficit loos- 1 and e~/ei = I, is 2 X 1O- 4 s. The time
of the latter. The result is a concentration scale for diffusion to and from the surface is 0.01 s
distribution that shifts to favor the 'floor' species for a SDS-water solution that is 3 x 10- 3 molar
eN' A distribution shift that favors the ceiling (Mysels and Florence 1970). The time scale of
species N + P would be the case if the surface surface expansion at a flow rate of 0.5 cm 3 Is is
were being compressed. (25 s- 1) - I, or 0.04 s; see Fig. 11d.9. Hence micelles
Figure I1d.12a shows how sensitive the and monomers re-establish equilibrium amongst
monomer surface concentration profile is to themselves much faster than equilibrium is upset
micelle-solution exchange rate constants k and by surface expansion. In so doing they must
k'. The concentration ratio e~/ei is set to unity. speed up the much slower diffusion process in a
The results indicate that the more rapid the way that keeps the surface more uniform in
exchange of monomers between the micelles and concentration and nearer to equilibrium, more
solution the more rapidly the equilibrium between so than would be the case if micelles were absent.
the surface and the bulk is re-established, and Finally, Fig. lld.12b shows how the micelle
hence the more uniform the surface concentration to monomer concentration ratio affects the surface
and surface tension are across the cell. Rapid monomer concentration profile. Increasing the
k =100 S·l k' =100 mol,l s ,l

Z
Z o
o ~
~a:: Z~
a::
Z~ OZ
QaJ
~u
I=~
c(z
c(z a:: 0
a:: 0 ~u
Zw
~u
Zw wu
wu Uc(
Zu.
Uc(
Z u.
o a::
0.95 8~ 0.95
u::l w(/)
w(/) u::
u:: NO MICELLES
c(::l
u._
~2 a:: a::
a:: a:: ::JlXl
(/):::i NO MICELLES
::llXl
(/):::i :5
:5 ow
ow
0.9 0.9 '----'----'---..............................---'---'---'-..........................

0.01 0.1 0.01 0.1
r r
(a) R (b) R
Figure lld.12 Predictions for 0.003 molar SDS in water at 24°C. The flow rate is 0.5 cm 3 /s. (a) Effect of
micelle-solution exchange rate constants on the profiles of monomer concentration along the surface. The
micelle to monomer concentration ratio is unity. (b) Effect of monomer to micelle concentration ratio on the
same. k' = IOmol- 1 S-I and k = lOs-I.
concentration of surfactant when the system is Knox 1970). Though surfactants are routinely
above the CMC usually raises the ratio of micelle used as coating aids, the zones where they act
concentration to monomer concentration (Kam- and the mechanisms by which they work are still
rath 1984). What is most significant is that the matters of controversy. To settle some aspects
greater the ratio C N /C 1 , i.e. the more the total of this controversy is the main objective of this
inventory of surfactant is in micellar form, the case study.
closer the system stays to equilibrium. Again, the Over the years many surfactants have been
mere presence of rapidly exchanging micelles tested and their effects have been correlated with
tends to drive the system towards equilibrium. their chemical natures and, often, capacities for
lowering surface tension. In this way has grown
an art of picking a surfactant to cope with a
lld.4.3.3 Conclusions
specific difficulty, and to be compatible with the
The essential result of this analysis is the finding rest of the coating formulation. An early report
that the surface speed changes rapidly across the came from Deryagin and Levi (1964). They tested
free surface in the overflow cell and this makes surfactant additives for their ability to improve
for highly nonuniform surface expansion (defor- dynamic wetting in a roll-coating process, i.e.
mation). To complicate matters further, the extent their ability to increase the maximum speed at
of the regions of surface expansion change with which the substrate can be successfully coated
flow rate. Because of this, the so called 'surface without excessive air-entrainment. They also
age' - defined by many researchers as a measure predicted from a crude analysis what their
ofelapsed time during which surfactant molecules experiments confirmed: that only certain surfac-
arrive at the surface in the act of re-establishing tants can improve dynamic wetting, and those
equilibrium - is indeterminate without reliable surfactants have a common chemical structure
theory. Moreover, the material parameters for and interaction with gelatin. Their experiments
the mechanisms proposed in the theory cannot were limited to low coating speeds - at least by
be deduced from the measurements in an overflow today's standards - and it is unclear what
cell, unless those measurements are interpreted criterion for 'successful' wetting they used.
with a detailed analysis of the sort we present In a landmark patent, Knox (1970) revealed a
here, i.e. the material parameters are determined particular combination of surfactant (hydroxy
by matching theory to experiment. propylene oxide) and gelatin, and other additives
There is a possibility that the cell design can too, that resulted in improved quality and
be altered in such a way to make the measurement uniformity of the coated layers in the final
easier to interpret. For example, Bergink-Martens product. In other words, the surfactant made the
et al. (1990) proposed one particular design which formulation easier to coat successfully, or at least
results in appreciable surface expansion across that was Knox's claim.
more than half the cell radius. Unfortunately, the Before Knox's patent, it appears that the most
expansion rates they report are all less than 5 s - 1. common coating aid was saponin, a surfactant
prepared from quillara bark (Sturge 1977; cf
Sterzi 1982). Being of natural origin, saponin
Ild.4.4 SLIDE eOATING FLOW
varies in composition according to the source
Nearly any coating flow could serve as a case and method of extraction and is often plagued
study ofthe potential effects ofsurfactant additives. by the presence of impurities. Growing demands
Here the focus is slide coating (Chapter 11 b). for higher coating speeds and increased number
This process is the workhorse of the photographic of coated layers, together with the growing
films industry, and the use of surfactant additives complexity of the coating formulation, rendered
is nowhere deeper entrenched than in multilayer saponin alone insufficient as a coating aid.
photographic coating technology (see for example Consequently, there came a proliferation of
synthetic surfactants developed for the same characteristic time scales ofmost defect formation
purpose, the first of which appears to be that by phenomena are far less than that on which he
Knox (1970). measured dynamic surface tension. Padmanabhan
The art has remained highly proprietary and and Bose (1988) measured the dynamic contact
little attention is given to it in the scientific angle with a simple plunging tape apparatus (see
literature. Since Knox (1970), only a few reports Chapter 3) in the presence ofsurfactants. Although
have been published on the matter, at least with their measurements were at speeds much lower
any direct mention ofcoating. Zvanut and Cohen than those found in a typical coating operation
(1978) investigated the physical properties of ( < 1 cm/s), they found strong evidence of
gelatin-surfactant combinations, including surface nonequilibrium values of surface tension at the
tension and dynamic surface tension, and claimed dynamic contact line, over a certain range of
that these correlated well with the ability of concentration.
certain surfactants to aid the coating process. A computer-aided theoretical analysis of
They published no proof of this claim, however. surfactant effects in steady and unsteady two-
Greener et ai. (1980) found that a certain surfactant dimensional flow of Newtonian coating liquid in
(Atlox 2081) delayed the onset of the ribbing a slide coater is reported below. The details of
imperfection to higher coating speeds in a two- the techniques can be found in Chapter 9 and
roll coating process. Their case study did not Chapter 13. Of course, as we pointed out earlier,
employ gelatin mixtures, but rather a glycer- there have been several significant theoretical
in-water mixture. Hempt et al. (1985), though developments, the most noteworthy coming from
citing no particular coating method, claimed that Park (1991), Cox (1986), Giavedoni and Saita
improvements in 'coating operability' can be (1992), Ruschak (1987) and Bose (1993). The
made when the right surfactants are added to phenomena of interest here are shown in Fig.
gelatin mixtures. Again their claim is not ltd. 13. Clearly, surfactant can act by various
substantiated with careful coating experiments mechanisms in each zone. Surfactant action on
but instead is based on experience and on surface waves on the liquid flowing down the
measurements of a relevant surface mechanical slide of indefinite length is well understood now
property known as surface elasticity. Zeldes from the linear stability analysis earlier; however,
(1985), in a more recent US patent, demonstrated it remains unclear how end effects, where the
clearly the improvement in slide coating operability flow may not be rectilinear, alter this action -
by adding surfactants of the fluorocarbon type. as in the case when the liquid runs off the end
His claim is that such additives promote defect-free of the slide into the bead region. Possibilities of
coating at higher coating speeds. But the nature surfactant action on the rapidly deforming upper
of the defect to which he refers is unclear. and lower menisci between the slide and substrate
Fruhner, Kragel and Kretzschmar (1989, 1991) - that is, around the 'coating bead' - are complex
examined the effects of several surfactants on the and barely identified. Whether surfactant transport
stability of the liquid bridging the gap between to and from the lower meniscus is fast enough
the die face and web and demonstrated a clear to have a significant impact on the static and
improvement in coating quality at higher web dynamic wetting process is also unclear.
speeds. The degree of improvement, however, Other issues arise when more than one layer
was highly dependent on surfactant type and its is coated, the most notable one being edge
interaction with gelatin. Valentini et al. (1991) control after the liquid has been coated. In
attempted to correlate measurements ofdynamic practice, surfactants are added to each layer in
surface tension with the range of applied vacuum varying amounts, so that the bottom layer has
over which defect-free coating can be achieved. the highest apparent (dynamic) tension, and each
Though he claimed that dynamic surface tension layer above that has a successively diminishing
does affect the so-called 'coatability window', the apparent tension (see Chapter 4 for complete
EDGE STABILITY
SURFACE LEVELING
7
DRYING
I SURFACE
UPPER MENISCUS
SURFACE TENSION
OF ADJACENT LAYERS
MICROSCALE FORCES
CONJOININGJDISJOINING PRESSURE
Figure lld.13 Zones of surfactant action in multilayer slide coating.
discussion). In that way the exposed edges of the a surface-tension lowering effect, we believe this
upper layers cannot contract and drag lower study is typical ofslide coating in the photographic
layers with them after being deposited. But when film industry. As in the previous case studies, an
equilibrium conditions are not attained immedi- aqueous sodium-dodecyl sulfonate (SDS) solution
ately this simple argument fails, especially when was chosen as a model system, and its relevant
adjacent layers contain different surfactants and properties are given in Table lId. I.
their equilibrium tensions are nearly the same Figure Ild.14 shows the computed velocity
(Orem 1988). A quantitative understanding field, streamlines and surfactant concentration
requires a three-dimensional, or at least a contours corresponding to the base case. The
specialized two-dimensional, transient analysis liquid is seen to wet the die face. The expansions
and is not pursued here. However, predictions and compressions that material elements undergo
below reveal the extent to which equilibrium is while moving along the surfaces in the slide
achieved after the liquid has been coated, thereby coating flow can be inferred from the streamline
providing a test for the argument above. structure in the adjacent liquid: where the
streamlines are converging in the direction of
flow, the liquid on the surface speeds up along
Ild.4.4.1 Results and discussion: effect of sur-
its path and initially, at least, any adsorbed
factant on steady operating states
surfactant has to cover more area. In other
The many surfactant solution properties and words, any surface excess concentration per unit
coater design specifications in Table Ild.2 define area of the surface is lowered. Conversely where
a case of a slide coating process. Despite lacking the streamlines are diverging in the direction of
the effects ofgelable polymer, which too can have flow the surface slows down along its path,
Table lld.2 Dimensional operating parameters and sponding surface concentration profiles. If soluble
surfactant liquid properties representative of a slide surfactant is to act dynamically in the coating
coating system. Solution is sodium-dodecyl sulfate in bead, there must be places on the menisci of the
water at 20·C
bead where the time scale of surface strain is
Webspeed Uw 2m/s shorter than the characteristic diffusion and
Gapwidth L gap 250j.lm adsorption times. Figure lId.15b shows that
Vacuum Pb -150Pa such places evidently exist on the upper meniscus
Gravity-web in the bead region and on the lower meniscus
angle p O· close to the dynamic wetting line, as the surface
Slide-web concentration deviates from its equilibrium value
angle IX 70·
by as much as 20%. Because surface tension
Dynamic depends strongly on surfactant concentration,
contact concentration gradients make surface tension
angle 0d 155·
gradients, as depicted in Fig. lId.15c. These in
Static contact
angle Os 70· turn exert a shearing stress on the liquid beneath
and can alter the flow.
Layer A Layer B What matters is how this diffusion time scale,
(lower) (upper) (r oo/CiY/D 1 (Mysels and Florence 1970), com-
pares with the time allowed for surfactant to
Final film re-establish equilibrium. If the surface is under-
thickness 45j.lm 25j.lm
going no strain, such as on the slide or far
Initial film
thickness 940.3j.lm 283.38j.lm downstream on the web, that time scale is the
Viscosity 15mPas 5mPas transit time of the surface from the point at which
Density l000kg/m 3 l000kg/m 3 equilibrium is destroyed. If the surface is being
Solvent surface strained, such as between the slide and the
tension (10 70mN/m 6OmN/m substrate - that is, around the coating bead -
Surfactant then that time scale is of the order of the inverse
concentration 20mol/m 3 20mol/m 3 of the strain rate. Evidently, from Fig. lId.15,
Equilibrium sur-
face tension 42mN/m 38mN/m the diffusion time scale is much larger than the
inverse of surface strain rate on the upper
meniscus, as long as the concentration is less
than 0.05 molar; that is, diffusion cannot deliver
surface elements are compressed and initially, at surfactant to or remove surfactant from the
least, any surface excess concentration per unit surface fast enough to maintain equilibrium. On
area is raised. In each event equilibrium between the lower meniscus diffusion is fast enough to
the surface and the bulk is upset, and surfactant establish equilibrium over a small portion of the
tends to move from the solution to the surface, surface near the static contact line, as long as
or vice versa, to restore equilibrium. The con- the concentration is greater than about 0.02 molar.
centration contours show how the straining The surface strain rate is simply the gradient
motion of the surface causes movement of of surface speed. On the upper meniscus, in the
surfactant by diffusion in the adjacent liquid. It bead region, the surface strain rate is of the order
is noteworthy that the concentration boundary of 1500 s- 1 where the surface is expanded and
layers adjacent to the surface are extremely thin. of the order of 2000 s- 1 where the surface is
Figure lId.15a shows the speed profiles along compressed; on the lower meniscus it is of the
the upper and lower menisci at four surfactant order of 1000 s- 1 where the surface is expanded
concentrations, ranging from 0.002 molar to near the static contact line and of the order of
0.05 molar. Figure lId.15b shows the corre- 20,000 s- 1 where the surface is compressed near
COMPRESSION
EXPANSION
(b)
Figure Ild.14 Base case (a) streamlines, (b) velocity field and (c) concentration contours of surfactant in bead
region with arrows indicating surfactant movement in an out of the surface (vacuum reduced to 300 Pa).
the dynamic contact line. Whether the surface bulk are in equilibrium. There the surface and
liquid undergoes compression near the contact the bulk are assumed to have ample time to
line is an unsettled matter, as we discuss below. establish equilibrium after they emerge from the
Figure lId.15 shows that equilibrium is nearly feed slot; Fig. lId.15. Whether equilibrium is
maintained across the entire lower meniscus at attained before the liquid exits the flow domain
concentrations greater than 0.02 molar, and so downstream depends on how the transit time of
the time scales ofdiffusion at those concentrations the liquid from the base of the slide to the outflow
must be less than (10,000 s- 1) - 1, or 0.1 ms. This plane compares to the characteristic equilibration
is in accord with the characteristic recruitment time of the surface and the bulk, i.e.,(r oo/CiY/D 1 •
time, (r oo/CiY/D 1 , which falls below 0.1 ms at The former is about 1.6 ms when the web speed
concentrations greater than 0.025 molar. is 2 m/s; the latter depends on the bulk concen-
Far upstream on the slide the surface and the tration. At concentrations less than 0.01 molar
(a) UPPER MENISCUS LOWER MENISCUS
t 0,002 molor
~ 0,015 molor 0.002 mola' ~
• II'! 0,025 molar 0.015 molar
:> 0
0.05 molar 0.025 molar
0.05 molar
(b) 0,002 molar

0.015 molar 0.002 molar ~
0.025 molar 0,015 molar
0.05 molar 0.025 molar ~
0.05 molar ""1 /'v1
STATIC
'I
DYNAMIC
CONTACT
CONTACT
UNE
UNE
~ ~==========~
(e)
0.002 molar
0.015 molar N
000'.""
0,015 mola, \ .
0,025 mola,
0.025 molar 0.05 molar
0.05 molar
~ I------"'~~C~=::§§:::=
E 0
~
E 0 \\ -(\
?o ?o '--..\
';" L..-........_ ........_ l . . . -........_....&...._l...-........_...J
8 9 10 11 12 13 14 15 o 0,05 0.10 0.15 0.20 0.25 0.30 O.~ 0.40
DISTANCE ALONG UPPER DISTANCE ALONG LOWER MENISCUS
MENISCUS THROUGH BEAD REGION (1 UNIT 1.22 mm) =
(1 UNIT = 1.22 mml
Figure lId.lS Profiles along upper and lower meniscus in the bead region: (a) surface speed, (b) surface
concentration, and (c) difference between local surface tension and equilibrium tension. System is SDS in water
at 20·C.
the equilibration length can be many times the The situation on the lower meniscus depends
film thickness, e.g., with 0.005 molar solutions on the position of the meniscus relative to the
equilibrium is not attained between the liquid sharp corner at the base of the slide and on the
and the surface until the deposited film is about flow kinematics at and near the contact lines.
5em downstream. That the surface and the bulk Reducing the vacuum from 350 Pa to 150 Pa
have not established equilibrium just after the raises the meniscus from the position of the base
film is deposited may change the widely accepted case to a pinned position, as depicted in Fig.
hypothesis that equilibrium surface tension lld.16. When the meniscus is pinned the surface
controls edge stability in multilayer coating and strain rates are more intense. Although the
is apparently one reason why measurement of surface is effectively 'created' at the static contact
so-called dynamic surface tension is important line, it maintains a near equilibrium inventory
to edge control in a multilayer coating (Chapter 4). of surfactant, even when the solution is as dilute
Figure 1Id.16 Pinned vs. free lower meniscus. (a) Surface shape, (b) surface speed measured in units of web
speed, (c) surface concentration, measured in units of equilibrium concentration, of 0.001 molar SDS in water
at 20·C, and (d) difference between dynamic surface tension and equilibrium surface tension.
LOWER MENISCUS
DYNAMIC CONTACT
,,..,,
LINE
~
,, ,,
FREE...../
STATIC CONTACT , ,
, I
LINE
------' ~
(b) 0 L.._.....~::::::=:.:._...l L._.J
r
•
•••
•
I
~----------------- ._--. ...... '."----",'

I
(c) 0 ....- - - - - - - - - - - - - - - - - - - - -
q~ 0.1 0.2
I....--....
. ...--~--.....- -......
O
0.3 0.4
--.-Io--~
0.5 0.6
(d) DISTANCE ALONG LOWER MENISCUS
(1 UNIT =1.22 MM)
as 0.001 molar. Presumably the infinite residence whether surfactant is capable of exerting any
time of the surface elements immediately at the effects on the lower meniscus, as it has been
static contact line allows the surface to recruit thought that the flow is too rapid and the
surfactant from a broad region like that between residence time of liquid between the die and the
the die lip and lower meniscus; see Fig. lld.14c web is too short for the meniscus to recruit a
(cf Whitaker and Wu 1985). Still the concen- significant excess of surfactant from the bulk
tration at the static contact line is slightly liquid. This controversy is settled above where
depressed because of the slight surface expansion it is shown that the lower meniscus is capable
there. of recruiting surfactant molecules fast enough to
Figures lld.14-11d.l6 indicate that surfactant depress its tension significantly.
adsorbed to the lower meniscus is apparently not Another way surfactant can affect the lower
being transferred to the web at a significant rate meniscus is through the apparent static contact
(see Cox 1986; Rame 1988). Instead it holds up angle, O~, the meniscus makes with the die face.
there and diffuses back into the bulk, contradicting The equilibrium value - for the composition of
the fundamental idea ofa coating process (Kistler the solid surface and the composition of the
1984). It is perhaps relevant that recent experi- adjacent coating liquid - is crucial in positioning
mental evidence indicates 'surfactant hold-up' the static contact line (see Chapter 3). If the
might occur at speeds much slower than we contact angle is within a certain range -less than
employ here (Padmanabhan and Bose 1990). The 60· in this study - that line can attach to the
boundary condition on liquid motion used so edge of the die lip, which has been shown to be
far requires that the velocity of the liquid at the desirable for uniform coating. Otherwise the line
dynamic contact line be equal to the velocity of may locate on the face below the edge. The
the dynamic contact line. This condition results positioning is also influenced by the pressure field
in perfect slip and effectively relieves the stress in the coating bead, which responds to the
singularity that otherwise occurs with the con- vacuum being applied to the underside. All else
ventional no-slip boundary condition (Kistler being fixed, the higher the contact angle, the more
1984). But only certain slip models allow transfer likely the coating bead is to wet the die lip.
ofsurfactant from the free surface to the substrate The static contact angle is an outward measure
(Dussan V. 1976). Rame (1988) proposed several of the tendency of the liquid L to form a film on
indirect mechanisms that promote transfer of or adsorb to the solid S in the presence of a gas
surfactant: these include surface diffusion or G. Young's equation relates the angle to the
imposition of at least a limited amount of surface tension a and to hypothetical excess free
solubility, thus providing for dissolution then energies per unit area of the solid exposed to the
reabsorption through the bulk phase. The only gas, a SG and theliquid, a SL : cos O~ = (aSG - aSL)/a.
other way to remedy this problem is by allowing a SG and a SL both diminish when the surfactant
a multivalued velocity at the dynamic contact adsorbs to the solid, as a is likely to do, and they
line, which would allow surfactant to pass freely are sensitive to unintended contamination as well
to the substrate but brings back the nonintegrable (Chapter 3; Bose 1993; Blake 1984). Surface
stress singularity. As it happens, such a condition roughness and compositional heterogeneity of
brings about dramatic changes in the response the solid can be additional complicating factors,
of the lower meniscus to small disturbances, because they can lead to multivalued contact
changes which warrant further investigation. angles, a phenomenon known as 'contact angle
What effect the unresolved stress singularity has hysteresis' (Chapter 3). The contact angle depends
on the validity of any solution computed with a on whether the liquid is in a receding or advancing
multivalued velocity is unknown. We return to situation (see Johnson and Dettre 1969; Bose
this matter in the following subsection. 1993). Moreover, the presence of surfactant can
There has been a long-standing controversy change the range of the hysteresis between
advancing and receding values. To incorporate the angle data are sparse for materials with which
effect of contact angle hysteresis in this analysis coating dies are often fabricated, e.g., stainless
would require some knowledge of how the liquid steel, this effect is accounted for qualitatively here
originally came into contact with the solid. At the through Young's equation and an empirical
very least this issue begs for a parameter study relationship among CT, CTSG ' and CTSL : CTSG - CTSL =
within the advancing/receding contact angle range, 72.5 - CT (Fowkes 1962). Here CT, CTSG ' and CTSL are
although we do not pursue that here. in units of mN/m and the relationship is an
Here we assume that the current surfactant approximation for aqueous solutions of ionic
system is 'inert' to the substrate and is not volatile surfactants on metals such as copper, iron, and
so that changes in CTSG and CTSL are negligible silver (Johnson and Dettre 1969).
(Bose 1993). Surfactant adsorption at the Figure lld.l7a shows the effect of the amount
gas-liquid interface lowers the contact angle and of added surfactant on the position of the lower
therefore promotes wetting. Although contact meniscus, using the empirical relationship given
CD
(a)
o
0.5
Li} '002
0.015
0.025
.05
molar
molar
molar
molar
Figure ltd.17 (a) Surfactant concentration effect on the free surface shape. (b) Surfactant diffusivity effect on
the free surface shape. The system is 0.025 molar SDS in water at 20°C.
by Fowkes (1962). The greater the concentration The profiles of surface concentration of SDS
the more the liquid wets the solid. The change in Fig. IId.15 show that the surface concentration
in location of the lower meniscus can be attributed gradients can be steep on the upper meniscus,
partly to the depression of tension, and hence to especially in the bead zone. The corresponding
the reduction in static contact angle. To assess steep surface tension gradients do not appear in
how much dynamic surface tension is involved, earlier versions of coating theory but must be
the meniscus position is predicted when equilib- reckoned with because they can alter the flow,
rium is maintained and compared with that when as Fig. 11d.17 indicates. In Fig. lld.17a the
equilibrium is not maintained, as in Fig. 11d.17a. surface elasticity Es is changed by varying the
This is done by simply specifying a large diffusivity, concentration. Plainly the surface tension-lowering
of the order of 1 cm 2 /s. effect of surfactants together with the surface
The results of doing so are shown in Fig. tension gradients they induce can reduce the size
11d.17b. Clearly the dynamic effects of surfactant of the standing capillary wave on the upper
are important, but not as important as the meniscus. It is perhaps relevant that a Dupont
equilibrium effect. The movement of the lower Patent (Zeldes 1985) is based on experimental
meniscus is seen to be slightly less than would observations of such an effect. The patent claims
be the case ifequilibrium were always maintained. that reducing the amplitude of this wave expands
The apparent dynamic contact angle in this the bounds of operability with respect to the
study remains an empirical parameter. Unfortu- ribbing instability.
nately, the fate of adsorbed surfactant on menisci The mechanism underlying this amplitude
which terminate at a dynamic contact line is reduction hinges on the local reduction ofcapillary
uncertain, as is the physics of dynamic wetting forces by the surface tension lowering effect. The
itself. Several studies suggest that surfactant can forces that operate beneath the standing wave
affect the macro-scale features of the dynamic are pressure, which rises in the streamwise direction
wetting process. Cox (1986) demonstrates that due primarily to a velocity decrease, and the
the macroscopic contact angle depends on contact capillary pressure, the resultant ofsurface tension
line velocity and surfactant concentration under in the curved meniscus. The streamwise viscous
certain conditions. Unfortunately, he did not stress immediately beneath the peak of the wave
allow transfer of surfactant from the liquid is about 10% of both forces and rises to about
surface to the substrate, and his analysis is limited 40% at the foot of the slide. Reduced capillary
to insoluble surfactants. Hopf and Geidel (1987) pressure by surfactant leads to a reduced
and Hopf and Volke (1987) investigated the wavelength and a less pronounced thinning of
effects of surfactant on advancing and receding the film just upstream of the bead. On the other
contact angles. They found poor agreement with hand, the curvature of the lower meniscus increases
the results of several noteworthy theories (e.g., to compensate for the lower surface tension
Blake 1988), which they attributed to the effect under the applied vacuum.
of surfactant transfer from the surface to the A small part of the amplitude reduction can
substrate. They also claimed that surfactant can be attributed to the surface tension gradient that
change the macro-scale dynamic contact angle, the surfactant causes. This is indicated in Fig.
at least at speeds less than 1 cm/s. So far we have 11d.17b, where the strength of the surface tension
adopted Cox's idea of not allowing surfactant to gradient is varied while the equilibrium tension
pass across the dynamic contact line. When we is kept the same. This is accomplished by varying
do allow surfactant transfer, as suggested by the surfactant diffusivity over the range ofO.5 cm 2 Is
Hopf and co-workers and in a different way than to 5 x 10- 5 cm 2 Is. The locally rising tension of
advocated by Rame (1988), we find a dramatic the meniscus on the backside of the standing
effect of surfactant on unwanted disturbances in wave, just downstream of its crest, is part of a
slide coating. We describe this below. surface tension gradient that is directed towards
the trough ofthe concave meniscus. That gradient and that is the vacuum. Here it is reduced to
steepens with falling diffusivity and exerts a stress150 Pa (from 350 Pal so that the meniscus pins
that accelerates the liquid underneath the wave at the sharp edge at the end of the slide (the
towards the bead, thereby reducing the liquid equilibrium contact angle is taken to be 60°).
deceleration upstream that leads to the standing Figure lId.18 shows the frequency response of
wave. the total film thickness to small disturbances in
two of the same parameters chosen by Van
Abbenyen et al.; namely, (1) vacuum, and (2)
lld.4.4.2 Surfactant effects on the dynamics of
dynamic contact angle. We also tested four other
slide coating
parameters (gap width, flow rate of the upper
We examine the effects of surfactant on the layer, flow rate of the lower layer, and webspeed)
dynamics of slide coating by imposing small but found no major differences, at least when the
disturbances in various parameters describing meniscus is pinned. The first parameter (vacuum)
the flow and measuring the response in the is an operating condition; the second reflects the
take-away zone on the web, because thickness surface characteristics of the web being coated
variations there may become permanent defects and the fluid mechanics of the liquid flow as it
downstream. As far as theoretical predictions are contacts the web at the apparent dynamic wetting
concerned that position is the outflow plane that line. On the left of Fig. lId.18 the response is
terminates the flow actually computed. When shown at a large surfactant diffusivity, 1cm 2 /s,
the disturbance amplitude is small enough, the and at three different surfactant concentrations
resulting variations in the final coated thickness - 0.00036 molar, 0.015 molar, and 0.05 molar,
can be inferred with a frequency response analysis corresponding to equilibrium tensions of69 mN/m,
(see Chapter 9 for details). 55 mN/m and 35 mN/m, respectively; on the right
The frequency response to five disturbance the response is shown at a realstic surfactant
sources in a two-layer slide coating process of diffusivitY,1 x 10- 6 cm 2 /s, and at the same three
constant viscosity, surface tension, and density concentrations. The results in Fig. lId.18 are all
was established in the work of Van Abbenyen et for the customary 'zero' velocity formulation at
al. (1989). The operating conditions they chose the dynamic wetting line (Kistler 1984). Results
are used for the base case in this study. Here the of the new 'discontinuous' velocity formulation
effect of surfactant on the frequency response is (Schunk 1989) when they are perceptibly different
examined, mainly to determine if dynamic surface are given below.
tension can dampen the response of the uniformity The effect of varying equilibrium tension without
in final thickness to ongoing sinusoidal disturb- the effect of dynamic surface tension, or equiv-
ances. In this section, one surfactant-related alently varying concentration at a high surfactant
parameter is varied, the concentration, and one diffusivity, is such that disturbances in vacuum
physical property is varied, the surfactant dif- and dynamic contact angle are less amplified (or
fusivity. A comparison of the response of the more damped) by falling tension (see for example
system with varying concentration and realistic graphs on left side of Fig. lld.18). The opposite
surfactant diffusivity to that with varying con- effect is seen for disturbances that produce
centration and an unrealistically large diffusivity motion on the slide and cause vibration in the
resolves the importance ofdynamic surface tension. upper layer thickness, e.g., those excited by
Of course were the diffusivity large enough, the variations in the flow rate of the upper layer (not
time scale of re-equilibration would be so short shown here), the damping effect of rising tension
as to make it improbable for the surface and the is likely caused by increased capillary pressure
bulk to stray from equilibrium. gradients - that go along with rising tension -
In the results we report here only one operating which tend to fill the wave troughs by driving
parameter has been changed from the base case, liquid from underneath the crests. This is the
LARGE DIFFUSIVITY, 1 CM 2/S REALISTIC DIFFUSIVITY, 1x10-6CM 2/S
"#.
~ 5 VACUUM 1 VACUUM
w
~ 0.00036 molar (69 dyn/em)
U 4 0.015 molar (55 dyn/em) 0.8
0.00036 molar (69 dyn/em)
~ 0.05 molar (35 dyn/em)
0,015 molar (55 dyn/em)
.J 0.05 molar (35 dyn/em)
zct 3 0.6
u::
ou. 2 0.4
Z
o
~ 1 0.2
c::
:; 0 E::....._-J.....................-J.._~~ 0
1 10 100 1000
"#. DYNAMIC CONTACT ANGLE DYNAMIC CONTACT ANGLE

tn 5 5
tn
W
Z
~
U 4 4
i:
to-
.J
ct 3 0.00036 molar (69 dyn/em) 3 0.00036 molar (69 dyn/em)
Z 0.015 molar (55 dyn/em) 0.015 molar (55 dyn/em)
u:: 0.05 molar (35 dyn/em) 0.05 molar (35 dyn/em)
u.
o 2 2
z
o
~ 1
a:
~ o ~=~~~.:.J:=~~
10 100 1000 1 10 100 1000
FREQUENCY (Hz) FREQUENCY (Hz)
Figure lId.IS Frequency response of a two-layer flow. Variation of final layer thickness caused by 1%
perturbation of backpressure (upper half), and dynamic contact angle (lower half).
same mechanism as that which damps capillary contact angle fluctuates sinusoidally, whatever
waves on the surface of the slide flow examined the cause, the effect of surfactant is such that the
earlier. least concentrated solution, 0.00036 molar, allows
The picture can be different when the surfactant the most amplification, the amplification factor
diffusivity takes on a realistic value, as shown being a little less than 2 at a frequency of 70 Hz,
on the right in Fig. lld.18. When the dynamic but for more concentrated solutions, 0.015 molar
Summary 531
and 0.05 molar, there is damping at all frequencies, necessarily because one is more realistic than the
the peak amplification factor being only 0.2 at other, but because each represents an extreme
a frequency of 50 Hz for the most concentrated case, as we made clear above.
solution. The mechanism evidently involves surface Large differences are seen in the sensitivity to
tension gradients, because raising the concentra- ongoing fluctuations in vacuum and dynamic
tion but maintaining equilibrium (graph on the contact angle. These differences are illustrated in
lower left in Fig. 11d.18) lowers the amplification Fig. lld.19. The lower right half of Fig. lld.19
factor far less than when equilibrium is not shows the effect of the new velocity condition on
established (graph on the lower right in Fig. the fate of fluctuations in vacuum. The most
lld.18), especially at the intermediate concen- dangerous frequency remains unchanged at about
tration where the amplification factor at 50 Hz 70 Hz, but the sensitivity of the flow is far greater
is 0.8 on the left and 0.4 on the right. The same when the velocity is nonzero at the contact line.
effects of dynamic surface tension can be seen in At the lowest concentration, 0.00036 molar, the
the response to fluctuations in vacuum (see Fig. amplification factor at the most dangerous
lld.18), for which significantly more damping frequency is 1.1 when the velocity is discontinuous
occurs at the highest and intermediate concen- and about 0.7 when the velocity is zero, so that
trations, 0.05 and O.ot 5 molar, when the diffusivity the disturbances are amplified instead ofdamped.
is lowered to a realistic level. So the right amount Adding more surfactant results in damping at all
of surfactant can have a desired effect but too frequencies, however, reducing the peak ampli-
much or too little produces little or no effect. In fication factor to 0.4 for concentrations of
the limit of zero frequency, the same response is 0.015 molar and 0.05 molar.
seen at both low and high diffusivity, as they The effect of the 'discontinuous velocity
should because in that limit the surfactant has condition' on fluctuations in dynamic contact
ample time to equilibrate over the infinitely long angle is illustrated in the left half of Fig. lld.19.
period of the fluctuation. This condition has the largest effect on the
Results so far presented are for the customary response of the least concentrated solution,
zero-velocity expedient at the dynamic contact 0.00036 molar, sharpening and heightening the
line (see Kistler 1984). Earlier we demonstrated peak around the most dangerous frequency to
that this leads to a large hold-up of adsorbed about 2.5, from 2.0 in the 'zero velocity' case.
surfactant on the surface just upstream, creating The 'nonzero velocity' condition also increases
a locally lower tension there (see Figs Ild.15 and the sensitivity of the response to the bulk
lld.16). Schunk (1989) proposed a different concentration of surfactant, i.e., increasing the
formulation that allows a nonzero (discontinuous, bulk surfactant concentration from 0.00036 molar
actually) velocity at the dynamic contact line and to 0.05 molar reduces the peak amplification
results in a steep acceleration and hence a higher factor from 1.5 to less than 0.1, making the flow
surface tension at the same place. Moreover, it nearly insensitive to dynamic contact angle
effectively allows more rapid transfer ofsurfactant disturbances. So strong an effect warrants further
from the liquid/gas surface to the solid surface, investigation.
as advocated by Hopf and Volke (1987). Unfor-
tunately the condition results in a nonintegrable
stress singularity, the effects of which are only
lld.5 SUMMARY
as severe as the discretization allows, i.e., the
finer the discretization in the contact line region, Experimental evidence in the form of surface
the stronger the singularity. Nonetheless, it is tension measurements in dynamic situations
relevant to compare the dynamic response of the (Chapter 4) and flow visualization (Valentini et
coating flow with the zero-velocity formation al. 1991; Zeldes 1985; Knox 1970) reveal potentially
and the multivalued velocity formulation, not important effects ofsurfactant additives in coating
DYNAMIC CONTACT ANGLE ZERO

VELOCITY VACUUM
_ _ _ _ CONDITION - - - -
en en
en en
w w
z z
::ll:: ::ll::
(.) g
:I: :I:
to- I-
...I
...I
ex: ex:
Z Z
u: 2.5
DYNAMIC CONTACT ANGLE
u: 2
VACUUM
u. u.
o o
z z 1.6
o
2
o
t=
ex: ~ 1.2
a:ex:
1.5
a:ex:
> > 0.8
0.5 o.~
- - - DISCONTlNUOUS-
a ~r;;~~~;tt~~~~~:;;~;;~ VELOCITY
CONDITION a (!I~~~::::::'-~ .........u..u.u.--=:~~~
1 I0 100 1000 1 I0 I 00 1000
FREQUENCY (Hz) FREQUENCY (Hz)
Figure lld.19 Effect of velocity at the dynamic contact line on frequency response. Upper half is response
of zero velocity; lower half is response of nonzero, discontinuous velocity. Left half is the response to ongoing
fluctuations in dynamic contact angle; right half is the response to ongoing fluctuations in vacuum.
flows - and indeed many closely related flows. beginning with the simplest of all free surface
Some of those effects prove to be desirable in flows: film flow down an inclined surface (Section
coating because they increase 'coating latitude', Ild.4.2). An important understanding of the
or expand the operatibility to a wider range of mechanisms by which surfactant dampens waves
operating parameters. But some of those effects has been developed. Although these mechanisms
prove to be detrimental because they lead to have long been known to hinge on surfactant-
unacceptable nonuniformities in the final coated induced surface tension gradients, we found that
film (Chapter 4). A comprehensive understanding diffusion and micellization undermine those
of the underlying mechanisms will enable more gradients, and shift the phase of the surface
efficient choice of surfactant additives to cope tension response relative to the wave crests and
with specific difficulties and to expand the bounds troughs. This analysis extends what went before
of operability. (Whitaker 1964; Sterzi 1982) to include effects of
In this chapter we have demonstrated potential soluble surfactant that aggregate, or form micelles.
effects of surfactant in three prototypical flows, An overflowing liquid pool represents the
References 533
second situation examined above (Section 11 d.4.3). disturbances that arise as a result of ongoing
Selection of this flow as a case study stems from fluctuations in dynamic contact angle and
its use in making dynamic surface tension webspeed. These strong effects warrant further
measurements, which are accomplished with a investigation.
Du Nouy ring or Wilhelmy plate inserted into
the expanding surface. The outstanding challenge, REFERENCES
however, is to interpret the measurements in
Adamson, A. W. 1982. Physical Chemistry of Surfaces.
terms of the variable processes, and ultimately
4th edn. New York: Wiley.
to deduce from those processes surface-bulk Agrawal, S. K. and Wasan, D. T. 1979. The effect of
re-equilibration rates that are independent of the interfacial viscosities on the motion of drops and
flow field. This would of course lead to relevant bubbles. Chern. Engng J. 18: 215.
physical properties such as surfactant diffusivities Anshus, B. E. and Acrivos, A. 1967. Effect of surface
and micelle relaxation times for specific surfactant active agents on the stability characteristics of
solutions, properties that are necessary to make falling liquid films. Chern. Eng. Sci. 22: 389-393.
Batchelor, G. K. 1967. An Introduction to Fluid
theoretical predictions. Without reliable theory
Dynamics. Cambridge University Press.
to predict the complex flow field and the effect Benjamin, B. T. 1957. Wave formation in laminar flow
of surfactant-induced surface tension gradients down an inclined plane. J. Fluid Mech., 2: 554-574.
on it, however, ambiguous interpretation is Bergink-Martens, D. J. M. et a/. 1990. Surface dilation
inevitable. Only such a theory with reliable and fluid-dynamical behavior of Newtonian liquids
constitutive equations - the latter involving the in an overflowing cylinder. J. Coli. Interf. Sci. 138: 1-9.
effect of surfactants on equilibrium properties Bidner, M. S., Larson, R. G. and Scriven, L. E. 1976.
On the theory of micellization with reference to
which must also be measured - will enable
petroleum recovery. Lat. Am. J. Chern. Eng. App/.
physical properties of surfactant solutions to be Chern. 6: 1-32.
inferred from the surface tension measurements Blake, T. D. 1984. Wetting. In Surfactants. London:
with the overflow device. Part of this program Academic Press. pp. 222-275.
is carried out above. Blake, T D. 1988. Wetting kinetics - how do wetting
Finally we tested the theory in slide coating lines move? 1988 Spring National Meeting of the
flow (Section I1d.4.4). This process is the AIChE. 6-10 March 1989, New Orleans, LA.
Bose, A. 1993. Wetting by solutions. In Wettability,
workhorse of the photographic films industry,
ed. J. Berg. New York: Marcel Dekker.
and the use of surfactant additives is nowhere Bury, L. R. 1927. Adsorption of butyric acid on water
deeper entrenched than in multilayer photographic surfaces. Phil. Mag. 4: 980-997.
coating technology. Under steady operating Christodoulou, K. N. and Scriven, L. E. 1992.
conditions surfactant acts mainly through its Discretization offree surface flows and other moving
ability to lower equilibrium surface tension. If boundary problems. J. Compo Phys. 99: 39-55.
present in high enough concentrations, surfactants Corkill, J. M. and Walker, T. 1972. Light scattering
and diffusion studies in nonionic micellar systems.
can significantly depress the surface tension of
J. Coli. Inter! Sci. 39: 621-631.
the lower and upper menisci, thereby changing Cox, R. G. 1986. The dynamics of the spreading of
the shape of the coating bead as it bridges the liquids on a solid surface. Part 2. Surfactants. J.
narrow gap between the slide die and the moving Fluid Mech. 168: 195-220.
web. Surfactant can also act dynamically on the Cussler, E. L. 1984. Diffusion: Mass Transfer in Fluid
coating bead, inducing flow-altering surface Systems. Cambridge University Press.
tension gradients where the menisci deform rapidly; Defay, R. and Petre, G. 1971. Dynamic surface tension.
In Surface and Colloid Science - Vol. 3, ed. E.
however, these effects in steady operation are
Matijevic. Wiley-Interscience. pp. 27-81.
much smaller than those caused by the depression Defay, R. and Roba-Thilly, 1. 1954. Sur I'existence
of equilibrium tension. It appears that dynamic d'an palier dans la courbe d'evolution aci cours der
effects of surfactant on transient phenomena are temps de la tension superficielle d'une solution. J.
more noticeable, such as in tempering unwanted Coli. Sci. Suppl. 1: 36-52.
Deryagin, 1. and Levi, S. M. 1964. Film Coating Hall, D. G. and Pethica, B. A. 1967. Thermodynamics
Theory. London: Focal Press. of micelle formation. In Nonionic Surfactants, ed.
Dorrestein, R. 1961. General linearized theory of the M.1. Schick. pp. 517- 557. New York: Marcel Dekker.
effect of surface films on water ripples. Ned. Akad. Hansen, R. S. and Mann, 1. A. 1964. Propagation
Wetens. 54: 260-271. characteristics of capillary ripples, I: theory of
Dussan, V., E. B. 1976. The moving contact line: the velocity dispersion and amplitude attenuation of
slip boundary condition. J. Fluid Mech. 67: 665. plane capillary waves. J. Appl. Phys. 35: 152-171.
Edwards, D. A., Brenner, H. and Wasan, D. T. 1991. Hartley, G. S. 1936. Aqueous Solutions of Paraffin
Interfacial Transport Processes and Rheology. MA: Chain Salts. Paris: Hermann et Cie.
Butterworth-Heinemann. Hempt, c., Lunkenheimer, K. and Miller, R. 1985. On
Emmert, B. and Pigford, R. L. 1954. A study of gas the experimental determination of the dilational
adsorption in falling liquid films. Chem. Eng. Prog. surface elasticity and the exchange of matter of
50: 87-93. mixed gelatin surfactant adsorption layers. Z. Phys.
Fainerman, V. B. 1986. Dynamic surface tension and Chemie. Leipzig. 266: 713-720.
the viscosity of an expanding adsorbed layer. Hirasaki, G. 1. and Lawson, 1. B. 1985. Mechanism
Kolloidnyi Zhurnal. 48: 758- 763. of foam flow in porous media: Apparent viscosity
Fowkes, F. M. 1962. Determination of interfacial in smooth capillaries. Soc. Pet. Eng. J. 25: 176.
tensions, contact angles and dispersion forces in Holbrook, 1. A. and LeVan, M. D. 1983a. Retardation
surfaces by assuming additivity of intermolecular of droplet motion by surfactant. Part 1. Theoretical
interactions in surfaces. J. Phys. Chem. 66: 382-399. development and asymptotic solutions. Chem. Engng.
Franses, E. I. 1979. Phase behavior and interfacial Commun.20: 191.
tension studies of surfactant systems. PhD thesis, Holbrook, J. A. and LeVan, M. D. 1983b. Retardation
University of Minnesota. of droplet motion by surfactant. Part 2. Numerical
Fruhner, H., Kragel, 1. and Kretzschmar, G. 1989. solutions for exterior diffusion, surface diffusion,
Investigation on the stability of the liquid bridge of and adsorption kinetics. Chem. Engng. Commun. 20:
the slide hopper type coating technique. Part 2: 273.
Stability criteria for coating bead in slide coating. Hopf, W. and Geidel, Th. 1987. The dynamic contact
J. Inf. Rec. Mater. 17: 273. angle I. Dependence of the receding contact angle
Fruhner, H., Kragel, 1. and Kretzschmar, G. 1991. In- on velocity in the surfactant-containing three-phase
vestigation on the stability of the. liquid bridge of the system. Colloid and Polymer Sci. 265: 1075-1084.
slide hopper type coating technique. Part 3: Mechanism Hopf, W. and Volke, K. 1987. The dynamic contact
of the effect of surfaetants for the stabilization of the angle II. Investigation of the adsorption mechanism
liquid bridge. J. In! Rec. Mater. 17: 273. based on the measurement of the preceding contact
Giavedoni, M. D. and Saita, F. A. 1992. Interfacial angles. Colloid and Polymer Sci. 265: 1085-1089.
viscosity in viscous free surface flows. A sample case. Imaishi, N. et al. 1982. Surface hydrodynamics of
Ind. Eng. Chem. Res. 31: 2222-2231. surfactant solutions. J. Chem. Eng. Japan. 8: 143-149.
Gibbs, J. W. 1906. The Scientific Papers of J. Willard Johnson, R. E. and Dettre, R. 1969. Wettability and
Gibbs. Vol. I Thermodynamics. New York: Longmans contact angles. In Surface and Colloid Science. ed.
Green & Co. 1931; New York: Dover reprint. 1961. E. Matijevic. 2: pp. 85-120.
Ginley, G. M. and Radke, C. 1. 1988. Influence of Kamrath, R. F. 1984. Thermodynamics and phase
soluble surfactants on the flow of long bubbles behavior of single and binary surfactant systems.
through a cylindrical capillary. ACS Symp. Ser. 396: PhD thesis, Purdue University.
480. Kistler, S. F. 1984. The fluid mechanics of curtain
Goetmackers, B. and Scriven, L. E. 1992. Multilayer coating and related viscous free surface flows with
curtain coating: Modeling surfactant and rheological contact lines. PhD thesis, University of Minnesota.
effects. Semiannual meeting of the AIChE, New Kistler, S. F. 1993. Hydrodynamics of wetting. In
Orleans, LA. Wettability, ed. 1. Berg. New York: Marcel Dekker.
Greener, 1., Sullivan, T., Turner, B. and Middleman, Kistler, S. F. and Scriven, L. E. 1983. Coating flows.
S. 1980. Ribbing instability of a two-roll coater: In Computational Analysis of Polymer Processing.
Newtonian fluids. Chem. Eng. Commun. 5: 73-82. Eds 1. A. Pearson and S. M. Richardson. London:
Hall, D. G. 1987. A phenomenological approach to Applied Science Publishers.
micellisation kinetics. J. Chem. Soc. Faraday Trans. Knox, W. J. 1970. Photographic surfactant composi-
83: 967 -983. tions. US Patent 3,514,213.
References 535
Krantz, W. B. and Goren, L. S. 1971. Stability of thin NSRDS-NBS-36, Superintendent of Documents,

liquid films flowing down a plane. Ind. Eng. Chem. US Government printing office, Washington,
Fundam. 10: 91-99. D.C.
Kreshek, G. C. et al. 1966. Determination of the Muller, N. 1972. Kinetics of micelle dissociation by
dissociation rate ofdodecylpyridinium iodide micelles temperature-jump techniques. A reinterpretation. J.
by a temperature-jump technique. J. Amer. Chem. Phys. Chem. 76: 3017-3024.
Soc. 188: 246-253. Mysels, K. J. and Florence, A. T. 1970. Techniques
Lamb, H. S. 1879. Hydrodynamics. 6th edn. New York: and criteria in the purification of aqueous surfaces.
Dover reprint. 1945. In Clean Surfaces: Their preparation and character-
Leaist, D. G. 1986. Binary diffusion of micellar izationfor interfacial studies. Ed. G. Goldfinger. New
electrolytes. J. Coll. Inter! Sci. 111: 230-239. York: Marcel Dekker.
Leung, R. and Shah, D. O. 1986. Dynamic properties Oguz, H. N. and Sadhal, S. S. 1988. Effects of soluble
ofmicellar solutions. I. Effects ofshort-chain alcohols and insoluble surfactants on the motion of drops.
and polymers on micellar stability. J. Coll. Inter! J. Fluid Mech. 194: 563-579.
Sci. 113: 484-498. Orem, M. W. 1988. Private communication.
LeVan, M. D. 1981. Motion of a droplet with a Padday, 1. F. 1957. A direct reading, electrically
Newtonian interface. J. Coll. Inter! Sci. 83: 11. operated balance for static and dynamic surface
LeVan, M. D. and Newman, 1. 1976. The effect of tension measurement. Proc. 2nd Inti. Congr. Surface
surfactants on the terminal and interfacial velocities Activity. London: Butterworth.
of bubbles or drops. AIChE J. 22: 695. Padmanabhan, S. and Bose, A. 1988. The importance
Levich, V. G. 1945. Physicochemical Hydrodynamics. of direct meaurement of dynamic contact angles
Reprinted in 1962. Englewood Cliffs, NJ: Prentice- during the wetting of solids by surfactant solutions.
Hall. J. Coll. Inter! Sci. 126(1): 164-170.
Lin, S. P. 1970. Stabilizing effects of surface active Padmanabhan, S. and Bose, A. 1990. On the anomalous
agents on a film flow. AIChE J. 16: 375-379. behavior of surfactant contaminated drops. J. Coll.
Lu, W. Q. and Chang, H. C. 1988. An extension of Inter! Sci. 139(2): 535-540.
the biharmonic boundary integral method to free Park, C. W. 1991. Effects of insoluble surfactants on
suface flow in channels. J. Comput. Phys. 77: 340. dip coating. J. Coll. Inter! Sci. 146: 382-394.
Lucassen, J. 1975. Adsorption kinetics in micellar Piccardi, G. and Ferroni, E. 1951. Surface tension of
systems. Faraday Disc. Chem. Soc. 59: 76-92. liquids in movement. Annali De Chimica (Rome). 41:
Lucassen, 1. and Giles, D. 1975. Dynamic surface 3-23.
properties of nonionic surfactant solutions. J. Chem. Prins, A., Arcuri, C. and Van Den Temple, M. 1967.
Soc. Faraday I. 71: 217-230. Elasticity of thin liquid films. J. Coll. Inter! Sci. 24:
Lucassen, 1. and Hansen, R. S. 1967. Damping of 84-90.
waves on monolayer-covered surfaces I. Systems Rame, E. 1988. On the spreading of liquids in the
with negligible surface dilatational viscosity. J. Coll. presence of surfactants. PhD thesis, University of
Inter! Sci. 22: 32-44. Pennsylvania.
Lucassen-Reynders, E. H. 1981. Anionic Surfactants. Ratulowski, 1. and Chang, H. C. 1990. Marangoni
New York: Marcel Dekker. effects of trace impurities on the motion of long gas
Lucassen-Reynders, E. H. 1987. Effects of surfactant bubbles in capillaries. J. Fluid Mech. 210: 303.
aggregation in soluble monolayers on dynamic Rillaerts, E. and Joos, P. 1982. Rate of demicellization
surface properties. Colloids and Surfaces. 25: 231-249. from dynamic surface tensions of micellar solutions.
Meyer, K. H. and Van der Wyk, A. 1937. Properties J. Phys. Chem. 86: 3471-3478.
of polymers in solution, VII. Helv. Chim. Acta. 20: Ruckenstein, E. and Nagarajan, R. 1975. Critical
1321-1329. micelle concentration. A transition point for micellar
Mijnlieff, P. F. and Ditmarsch, R.1965. Rate of micelle size distribution. J. Phys. Chem. 79: 2622-2626.
formation of sodium alkyl sulphates in water. Rusanov, A. I. and Krotov, V. V. 1979. Gibbs elasticity
Nature. 5613: 889-891. of liquid films, threads, and foams. In Progress in
Miller, R. 1981. Adsorption kinetics of surfactants Surface and Membrane Science, 13. Eds D. A.
from micellar solutions. Coll. Polym. Sci. 259: Cadenhead and J. F. Danielli. Academic Press.
1124-1128. Ruschak, K. J. 1987. Flow of a thin liquid layer due
Mukerjee, P. and Mysels, K. J. 1971. Critical micelle to ambient disturbances. AIChE J. 33(5): 801.
concentrations of aqueous surfactant systems. Sadhal, S. S. and Johnson, R. E. 1986. On the
deformation of drops and bubbles with varying L. E. 1989. Frequency response of a coating flow:
interfacial tension. Chem. Engng. Commun. 46: 97. predictions for slide coating. Submitted to
Schunk, P. R. 1989. Surfactant and polymer additives AIChE J.
in coating and related flows. PhD thesis, University Wassmuth, F., Laidlaw, W. G. and Coombe, D. A.
of Minnesota. Available from University Microfilms 1993. Calculation of interfacial flows and surfactant
Internation, Ann Arbor, MI. redistribution as a gas/liquid interface moves between
Scriven, L. E. 1960. Dynamics of a fluid interface. two parallel plates. Phys. Fluids A. 5(7): 1533-
Chem. Eng. Sci. 12: 98-105. 1548.
Shinoda, K. et al. 1963. Colloidal Surfactants. New Weinheimer, L. E., Evans, D. F. and Cussler, E. L.
York: Academic Press. 1981. Diffusion in surfactant solutions. J. Coli.
Smith, M. K. 1990. The mechanism for the long-wave Inter! Sci. 80: 524-531.
instability in thin liquid films. J. Fluid Mech. 217: Whitaker, S. 1964. Effect of surface active agents on
469-485. the stability of falling liquid films. Ind. Eng. Chem.
Squire, H. B. 1933. On the stability for three-dimensional Fundam.3: 131-142.
disturbances of viscous fluid flow between parallel Whitaker, S. and Jones, L. O. 1966. Stability offalling
walls. Proc. Roy. Soc. London. A142: 621-628. liquid films: effects of interface and interfacial mass
Sterzi, G. S. 1982. Surfactant effect in film flow. MS transport. AIChE J. 12: 524-531.
thesis, University of Minnesota. Wu, A. and Whitaker, S. 1986. The recirculation zone
Stone, H. A. and Leal, L. G. 1990. The effects of at the entrance of a falling liquid film: consequences
surfactants on drop deformation and breakup. J. for the surfactant adsorption problem. J. Coli.
Fluid Mech. 220: 161-186. Inter! Sci. 110: 389-397.
Sturge,1. M. 1977. N eblette's Handbook of Photography Yih, C. S. 1963. Stability of liquid flow down an
and Reprography. Materials, Processes, and Systems. inclined plane. Phys. Fluids. 6: 321-334.
7th edn. New York: Van Nostrand Reinhold. Zeldes, M. D. 1985. Coating process employs surfactants.
Valentini, J. E. et al. 1991. Role of dynamic surface US Patent 4,508,764.
tension in slide coating. Ind. Eng. Chem. Res. 30: Zvanut, C. W. and Cohen, E. D. 1978. High shear-rate
453-461. viscosity and dynamic surface tension ofphotographic
Van Abbenyen, W., Christodoulou, K. N. and Scriven, emulsions. Int. Congo Photo. Sci. August 1978.
KNIFE AND ROLL COATING 12a
Dennis J. Coyle
This chapter describes the processes of knife and flow in a small gap between a pair of rotating
roll coating from a fluid-mechanical point of rolls is the primary factor controlling the thickness
view. It combines a critical review of what is and uniformity of the coated film. The thickness
known, along with some speculation aimed at of the coating depends primarily on the gap
stimulating new lines of research. Physical between adjacent rolls and their relative speeds.
mechanisms underlying the flow phenomena are Two basic types of roll coaters are distinguished
emphasized, along with their practical implications. by the relative direction of roll surface motion
Section 12a.1 describes a variety of coaters and in the gap: in forward roll coating the roll surfaces
presents a logical scheme for their classification. move in the same direction and in reverse roll
Section 12a.2 introduces the fluid-mechanical coating they move in opposite directions. The
mechanisms important in determining the oper- simplest versions of each are depicted in Fig.
ability of these coaters. The remaining sections 12a.1b,c. In terms of the flow fields, knife coating
deal specifically with knife coating, forward roll is a subset of forward roll coating where one
coating, and reverse roll coating. surface is stationary.
12a.l.l ILLUSTRATIVE EXAMPLES

12a.l CLASSIFICATION OF KNIFE AND ROLL
There are several types of knife coaters, and a
COATING PROCESSES
multitude of machines considered to be roll
Knife coating is a process by which a thin liquid coaters (Higgins 1965; Booth 1970; Satas 1984;
coating is formed on a continuous web by the Scharenberg 1985; Zink 1979). Figure 12a.1
application of an excess of coating liquid which shows selected examples. These machines tend
is subsequently metered by a rigid knife held in to be simple, inexpensive, and extremely versatile;
close proximity to a rigidly supported web. able to coat over a wide range of thicknesses,
Figure 12a.1a depicts a typical roller-fed knife speeds and viscosities. In general, these devices
coater. The thickness of the coating depends are not as precise nor do they coat films of as
primarily on the clearance, or gap, between the high uniformity as premetered devices such as
knife and the web, and upon the geometry of the extrusion, slide, or curtain coaters (Chapter 11).
gap (bevel angle, length, etc.). A notable exception to these generalizations are
Roll coating is a process by which a thin liquid so-called 'precision' roll coaters, which are capable
film is formed on a continuous web by the use of uniformity comparable to the premetered
of two or more rotating rolls, such that the fluid devices, and which can be complex and expensive.
540 Knife and roll coating
Knife coaters, one example of which is shown a liquid supply upstream of the knife, it is referred
in Fig. 12a.la, are simple, low-cost, compact to as pond-fed or hopper-fed knife coating.
coaters used for applying thick coatings of high The various rolls in a roll coater are given
viscosity liquids. Typical operating parameters common names associated with their function,
are 3-120 m/min speed, 1000-10000 mPa s vis- and these names are included in Fig. 12a.1. The
cosity, and 50-2500 ~m wet coating thickness applicator roll applies the coating to the web.
(Zink 1979). The main disadvantages of this The back-up roll supports and often drives the
method are its propensity for streaking and the web at the point of coating application. If a gap
fact that the coating thickness is sensitive to web is maintained between applicator roll and web,
thickness variations. the back-up roll is usually rigid. If the web is
Forward roll coating, sometimes referred to forced into near-contact with the applicator roll
as 'meniscus roll coating', is most commonly used (forming a nip), the back-up roll is rubber-covered
to apply thin optical-quality coatings of low- and sometimes referred to as the impression roll.
viscosity liquids (Satas 1984; Zink 1979). Typical The metering roll is a reverse-moving roll that
operating parameters are 3-60m/min speed, removes ('meters') excess coating from the surface
1-50 mPa s viscosity, and 25-60 ~m wet thickness. of the applicator roll. Its surface must be scraped
The main disadvantages of this method are the (doctored) clean by a blade before contacting the
limitations of low speed and viscosity, and that coating on the applicator roll. A fountain roll
the coating thickness is sensitive to speed and picks up coating solution from a pan and transfers
viscosity as well as gap. Reverse roll coating is it to the applicator roll, and may be run in a
probably the most versatile coating method in forward or reverse direction.
use today (Booth 1990; Satas 1984). A reverse The simple two-roll forward roll coater in Fig.
roll coater can apply uniform wet coatings of 12a.lb is fed by dip-coating from a pan, after
10-1000 ~m thickness at speeds of 3-600 m/min, which the coating is metered by its passage
and can handle coatings of 1-500000 mPa s through the narrow gap between applicator and
viscosity. The primary disadvantages are com- back-up rolls and then splitting downstream.
plexity, cost, and solvent evaporation (Zink 1979). Addition of a third roll to pick up the coating
This chapter deals only with roll and knife and transfer it to the applicator roll gives more
coaters with rigid rolls and knives. Flexible blade flexibility in coating thickness and speed (Zink 1979).
coaters, kiss coaters, squeeze-roll (deformable- Figures 12a.lc-f show four versions ofreverse
roll) coaters, and gravure coaters are considered roll coaters, but there are many other variants
separately with elastohydrodynamic systems (Booth 1970, 1990; Satas 1984). All types use a
(Chapter 12c). reverse-acting metering roll to control the coating
thickness, and the region between these rolls is
called the metering gap. There are two major
classes of reverse-roll coaters. The first, shown
12a.1.2 CLASSIFICATION - PRIMARY AND
in Fig. 12a.lc, uses the transferred film as the
AUXILIARY FLOWS
final product. The film on the applicator roll is
There are several variations of knife coating. In first formed by dip coating or fountain roll
the simplest case the knife, in addition to its transfer. The second, shown in Figs 12a.ld-f,
metering function, also serves the function of uses the metered film as the final product. The
cross-web distribution of coating by forming a implications ofthis significant difference, in terms
rolling puddle upstream. A roller-fed knife coater of thickness and uniformity, are discussed later.
applies excess coating to the web by either Reverse roll coaters can be pan-fed (Fig. 12a.ld),
dipping, roll transfer, or a puddle-feed, and the fountain-roIl-fed (Fig. 12a.le), die-fed, or nip-fed
knife causes excess coating to run back toward (Fig. 12a.lf), depending upon the means of
the applicator. If there is a structure for containing coating supply to the applicator roll.
Classification of knife and roll coating processes 541
Coated Coated
Web Web
Coaled Web
(a)
Coaled
Web
Coated
Web
(f)
Figure 128.1 Common types of knife and roll coaters. The coating thicknesses and gaps between rolls are
magnified for clarity. (a) Roller-fed knife (knife-over-roll) coating; (b) two-roll pan-fed forward roll coater;
(c) two-roll pan-fed reverse roll coater; (d) three-roll pan-fed reverse roll coater; (e) four-roll pan-fed reverse roll
coater; and (f) three-roll nip-fed reverse roll coater.
While there are many knife and roll coating
configurations, the situation can be simplified by
realizing that all such coaters can be described
as a series of separate flows, each between a single
pair of rolls (or a knife and a roll). This leaves
only four types of flow fields, herein called
T
primary flows, that need to be analyzed. These
are: (1) knife coating gap flow, Fig. 12a.2a; (2)
forward roll film-splitting flow, Fig. 12a.2b; (3)
reverse-roll metering gap flow, Fig. 12a.2c; and
(4) reverse-roll nip-transfer flow, Fig. 12a.2d. All
(a)
knife and roll-coating configurations are simply
a combination of these 'building-blocks' with
other non-roll-coating flows, herein called auxiliary
flows, such as dip coating and extrusion coating.
While the auxiliary flows are usually ignored in
analyzing knife and roll coaters, they can be a
source of coating nonuniformity and operating
limits.
The primary flows depicted in Fig. 12a.2 are
often referred to as the coating bead. Preferably,
these flows are laminar, steady, and two-dimen-
sional, but under certain conditions they become
three-dimensional and/or unsteady and the coating
is nonuniform. The following section describes
the physical mechanisms of the primary flows,
the appropriate scalings and dimensionless groups,
Film-Transfer and the major bead failure modes.
Vortex
128.2 FLOW MECHANISMS IN KNIFE AND

(c)
ROLL COATING
12a.2.1 STEADY BASE FLOWS
12a.2.1.1 Lubrication flows

The primary mechanism of flow between rigid
rolls and knives is that of hydrodynamic lubri-
cation - the combined drag and pressure-driven
flow in a slowly converging/diverging geometry
(Cameron (1976); see also Chapter 8 in this
(d) volume). The nearly one-dimensional flow in the
gap is therefore governed by the integrated
Figure 118.2 Definition sketches of primary flows in Reynolds equation:
knife and roll coating. (a) Knife coating gap flow;
(b) forward roll film-splitting flow; (c) reverse roll
metering flow; and (d) reverse roll nip transfer flow.
Flow mechanisms in knife and roll coating 543
where P is the pressure, Jl. is the viscosity, VI and problem is the two-dimensional free-surface flow
V 2 are the roll speeds, Q is the volumetric flow near the film-splitting meniscus governed by the
rate per unit width and H is the separation Navier-Stokes equations, though the simpler
between roll surfaces as a function of main-flow film-profile equations of Kheshgi, Kistler and
coordinate X. It is standard practice in lubrication Scriven (1992) could be used to describe this
theory and contact mechanics to define an region. Matching the pressure gradient of the
equivalent radius of a roll against a flat plate inner and outer solutions determines the flow
which gives an identical geometric description rate and meniscus position. The same technique
ofany two-roll or roll-knife configuration (Johnson could be used to solve the other primary flows
1985). Using this approach, the equivalent radius in roll coating.
Re and the roll surface separation or gap profile Finite-element analysis allows the complete
Hare calculation of two-dimensional free-surface flow
fields for many coating processes (Kistler and
1 1 1
-=-+- (12a.2a) Scriven 1983, 1984). The predictions from these
Re RI R2 computations are often very accurate but can be
X 2 tedious to do, so that the asymptotic approach
H = H o + 2R e (12a.2b) is more efficient at capturing the essence of the
underlying mechanisms.
where H 0 is the minimum clearance (gap) between In the meniscus region of the flow there are
two surfaces of radii R I and R 2 • This leads to several possible mechanisms controlling the flow
the introduction of the main controlling dimen- and its stability. There are three choices of
sionless geometric parameter HO/R e , which is a pressure scaling: viscous, surface tension, and
measure of the converging/diverging nature of inertial
the gap profile. In the gap the pressure scales
with lubrication viscous force P =Jl.U (12a.4a)
Ho
P 1ub
_ Jl.U
- H
(RH )1/2
e
(12a.3)
v
o 0
(J
(12a.4b)
PS1="R
where U is the average roll velocity. The solution c
of the Reynolds equation for roll coating flows Pi = pV 2 (12a.4c)

has been the subject of numerous papers with
various approximate boundary conditions applied where (J is the surface tension, Rc is the meniscus'
at the upstream and downstream menisci. radius of curvature, and p is the density. The
relevant dimensionless groups are the capillary
number, Reynolds number, and the Weber
12a.2.1.2 2D free surface flows
number:
G. I. Taylor (1963) pointed out that roll coating
(and journal bearing) flows naturally separate Jl.V
Ca=-
into two distinct regions: lubrication-type flow (J
in the vicinity of the gap center and two-

pVL
dimensional free-surface flow in the vicinity of Re=-
the coating menisci. Ruschak (1982) formalized Jl.
Taylor's basic idea in terms of matched asymptotic pV L 2
expansions for the case of forward roll coating We=--=ReCa

(J
with a fully flooded upstream region (see Chapter
8). The outer problem is the lubrication flow where V and L are appropriate characteristic
described by the Reynolds equation. The inner velocity and length scales.
12a.2.2 THE RIBBING INSTABILITY (Dowson and Taylor 1979). The same type of
Roll coating flows are especially susceptible to coating defect is also found on films on rotating
rolls, such as those in a paper-making machine
a bead failure mode called ribbing. The thickness
(Yih 1960).
of the coated film varies in a sinusoidal fashion
across the web, such that 'stripes' or 'ribs' appear
to run along the machine direction. This defect 12a.2.2.1 Interfacial force balance
is also refered to as 'corduroy', 'rake-lines', or
'phonographing', and can also occur in extrusion, Ribbing is a manifestation of the instability of
slide, and knife coating (see Chapters l1a and the two-dimensional base flow to a perturbation
11b). Figure 12a.3 is a photograph of ribbing in of the coating meniscus. The mechanism of
forward roll coating film-splitting ofa viscoelastic ribbing can be understood from a simple force
liquid. The defect occurs even with Newtonian balance on the air-liquid interface that forms
liquids. This instability was first observed in the coating meniscus (Pitts and Grei1ler 1961;
flows created by rolls and spreaders by Chalmers Savage 1977b). Consider the meniscus perturbation
and Hoare (1941), Pearson (1960), Sone and of infinitesimal amplitude 13 depicted in Fig. 12a.4.
Fukushima (1960), and Pitts and Greiller (1961). The pressure under the 'bump' or 'rib' (point 2)
Ribbing is similar to the phenomena of viscous is different from the pressure in the unperturbed
fingering, which is observed when a low viscosity region (point 1) for two reasons: there is a
liquid displaces a high viscosity one from a pressure gradient in the flow, and the radius of
confined passageway (Saffman and Taylor 1958). curvature of the meniscus has been changed. If
It is also seen in the peeling of an adhesive film the pressure under the perturbation is the greater
(McEwan and Taylor 1966) and in the splitting of the two, it will generate lateral flow out of the
of an oil film in a ventilated journal bearing perturbation, causing it to level. If the pressure
under the perturbation is the lesser of the two,
it will generate lateral flow into the perturbation,
causing it to grow. This latter situation corresponds
to a flow which is unstable to a ribbing disturbance.
Evaluating the pressures at points 1-3 (Savage
1977b) results in the criterion that the flow is
Figure 12a.3 Photograph of the ribbing instability

in forward roll coating with rigid rolls and flooded Figure 12a.4 Schematic of a sinusoial perturbation
inlet. The roll surfaces move upwards at the center of to a coating mensicus, from which an approximate
the photograph. The fluid is a viscoelastic polyac- stability criterion can be derived. The magnitude (e) of
rylamide/water solution (from Coyle 1984). the perturbation is greatly magnified.
Flow mechanisms in knife and roll coating 545
unstable to ribbing if the case, surface tension is the main stabilizing

force and stable operation is typically limited to
dP _ (J(-b dR c
+ N 2 ) > 0 (12a.5) low speed. The ribbing instability in the specific
dX R dX
c geometries are reviewed in Sections 12a.3, 12aA,
and 12a.5.
where N is the wavenumber of the ribs and X
increases as the meniscus is approached from the
liquid side. Surface tension stabilizes the meniscus, 12a.2.2.3 Inertial mechanism for ribbing
while a positive pressure gradient is destabilizing. Inertia can also be the driving force for the
The sign and relative magnitude of these terms, ribbing instability, as it too can create a pressure
and thus the stability of the flow to ribbing, gradient. The simplest case is the stability of a
depends on the specific geometry and operating liquid film on a rotating roll (Yih 1960). For a
conditions. In addition, the pressure gradient can film of thickness D on a roll of radius R rotating
arise from either viscous forces (lubrication flow) with a surface speed of U, the inertial pressure
or inertial forces, depending on the viscosity of gradient at the free surface is pU 2(1 + D/R)/R,
the liquid under consideration. the radius of curvature of the meniscus is
Equation (12a.5) can be used to evaluate the -(R + D), and dRc/dX = -1. Inserting these
stability of all knife and roll coating flows to expressions into equation (12a.5) gives the criterion
ribbing. To obtain information on the shape of that the flow is unstable if
the unstable disturbances and their initial growth
rate, a more rigorous linear stability analysis is (J [
1 + D/R - pU2R N - (1
2 1 ]
+ D/R)2 > 0
required. Neither approach predicts the shape of
the meniscus at finite amplitude or whether the
(12a.6)
new three-dimensional flow state is steady or
time-dependent. which is equivalent to the result derived by Yih
(1960, equation (46». Equation (12a.6) can be
rearranged in terms of the critical Weber number
12a.2.2.2 Viscous mechanism for ribbing above which the flow is unstable
Viscosity can be the driving force for the ribbing n2 I
instability as the converging/diverging geometry We*---- (1 + D/R)3 (12a.7)
1 +D/R
creates pressure gradients. This is the 'classical'
ribbing mechanism studied by Pearson (1960) As the base flow meniscus radius of curvature is
and many others since. Using a viscous pressure negative, surface tension destabilizes the flow for
scaling such as equation (12aAa), the interfacial long enough waves (small n), and so a rotating
force balance can be nondimensionalized and so film is always unstable. Yih performed a linear
the capillary number appears. The pressure stability analysis to predict growth rates. At low
gradient depends on the solution to the Reynolds Reynolds number (high viscosity) they are small,
equation (equation (12a.l», so the parameter so that the ribs might never be seen in a coater
Ho/R. arises. Thus there is a critical capillary where the film spends a very short time on the
number, which is a function of the gap/radius rotating roll. At high Reynolds number the
ratio, above which the flow is unstable to ribbing. growth rate is proportional to .JU, while the
In reverse roll coating the critical capillary residence time is proportional to I/U, so that
number will also depend on roll speed ratio. If the total growth scales with 1/.JU. This poses
the pressure gradient is negative, which can be the interesting possibility that for low-viscosity
the case in knife and reverse roll coating, the liquids, inertial ribs could appear at low speed
flow is stable at any value of capillary number. but then seemingly disappear at higher speeds.
If the pressure gradient is positive, which is often The contention that this mechanism is only
important for very low viscosity liquids is high enough coating speeds, though there may
supported by the experimental observations that be hydrodynamic assist of wetting in these flows.
for liquids of roughly 10 mPa s or higher viscosity, Even though this limit must exist, there are no
the apparent onset of ribbing in roll coating published studies.
correlates with the capillary number (see Section There is also possible air entrainment in the
12a.4), such that increasing viscosity lowers the auxiliary flows, most notably where the wet
critical speed, which is opposite of what the surface of an applicator or fountain roll plunges
inertial mechanism would predict. In addition, back into the coating pan. If the roll has a
if one computes the Reynolds number based on relatively deep submersion, the critical capillary
a lubrication length scale (equation (12a.3) divided number for air entrainment is (Bolton and
by equation (12a.4c)), the viscosity would need Middleman 1980):
to be on the order of 5 mPa s or lower to make
Ca* = 3.3Po- O. 3 ; Po < 3 (12a.8a)
the Reynolds number greater than one. But for
special cases where the viscosity is low and the Ca* = 5.0Po- O•7 ; Po> 3 (12a.8b)
gap is large, this inertial mechanism could become
significant. where Po =
P )1/3 is the property number.
Another mode of inertially-driven instability
(T
(- -
J1.Ug
could result from the instability of rotating If the roll has a very shallow immersion depth,
vortices in the coating bead, similar to those critial speeds of air entrainment are much lower.
visualized in a blade-coating pond (Trianta- This occurs if the immersion depth is less than
fillopoulos and Aidun 1990; Aidun, Triantafillo- the characteristic interfacial displacement depth,
poulos and Benson 1991) or other rotating flows which is on the order of 1- 5 mm (Ghannam and
discussed by Aidun (1991a,b). The manifestation Esmail 1990).
in a free surface flow could be similar in appearance Another bead failure mode is called starvation.
to the classical viscous-driven ribs, though the If the auxiliary flow, such as dip coating, does
wavelength should be longer. The basic mechanism not deliver enough flow to the gap of the primary
is an unstable stratification ofangular momentum flow, the coating bead will break. This mode of
in a rotating flow (Chandrasekhar 1961). The failure is most common in processes such as
free surface bounding the flow could render it meniscus roll coating (Chapter 12b) where both
much more unstable than the corresponding speed and viscosity are low.
confined flow, such as is the case when one
compares the rotating film with the problem of
Taylor vortices. 12a.2.4 SOURCES OF COATING
NONUNIFORMITY
Yih (1960) also presented experimental data
showing the existence of inertially-driven ribbing. There are other sources ofcoating nonuniformity
Since then there has been no study of this aspect which are not failure modes of the coating bead,
of ribbing in coating flows. This is surprising but are subtleties of the coating process and its
considering that the ranges of coating conditions sensitivity to disturbances which can lead to
claimed for some roll coaters would fall into this defects such as streaks and cross-bars.
regime.
12a.2.4.1 Auxiliary flows

12a.2.3 AIR ENTRAINMENT AND STARVATION
The first category of defects are those that arise
Air entrainment is another mode of coating bead due to the auxiliary flows. Any nonuniformity
failure (see Chapter 3). The three- phase dynamic in a coated film formed by an auxiliary flow
contact lines present in the primary roll coating corresponds to a disturbance of the inlet of the
flow fields are susceptible to air entrainment at primary flow. These disturbances are damped if
Knife coating 547
they must traverse a long lubrication-type flow cross-web variation due to mechanical misalign-
in a narrow gap between rigid surfaces before ment or variation in roll diameter along its axis.
impacting the coated film of interest (Pearson It can also be time-dependent in nature, caused
1985). Such is the case for a knife coater, a by roll run-out, vibrations, and speed fluctuations.
forward roll coater, and a reverse roll coater The sensitivity of the bead to perturbations can
where the metered film becomes the product. In be evaluated by computing the frequency response
contrast, in a reverse roll coater using the (as described in Chapter 9 and illustrated for
transferred film for the product (Fig. 12a.1c) the slide coating in Chapter 11 b). No such studies
disturbance will be transferred directly to the have been done for roll coating.
product with little or no damping.
There are several types ofdisturbances. Ribbing
may occur on the applicator roll, either inertially 12a.2.4.3 Bubbles and contamination
driven or viscous-driven from a previous film-split The third category of defect mechanisms arises
between fountain and applicator roll. Cross-bar from subtleties of the steady-state primary flow
type waves could form due to the instability of field. Roll coating flows often have regions of
a thick rising film flow (Yih 1963). The pond in closed recirculation, or vortices. These can trap
a nip-fed reverse roll coater becomes unstable at bubbles and particulate contaminants and cause
high Reynolds number causing the rolling vortex streaks. The liquid in the vortices has a very long
to become three-dimensional (Aidun 1991b). None residence time in the process which could lead
of these effects have received much study. to degradation of the coating, again producing
For reverse roll coaters (Fig. 12a.1d-f), the streaks or point defects.
doctoring of the metering roll is another auxiliary
flow. The metered film uniformity is dependent
on the complete removal of coating from the 12a.3 KNIFE COATING
metering roll. Incomplete or nonuniform doctoring
will generate streaks in the metered film. Doctored 12a.3.1 STEADY FLOW
blades can be thin (-0.1 mm) flexible steel or
Flow in a knife coater is governed predominantly
thick (> 1mm) plastic material. Trailing con-
by hydrodynamic lubrication, sometimes with
figurations (acute angle between blade tip and
inertia being important as well. Available models
roll surface tangent) are the same as flexible blade
are either derived for a special case (Greener and
coating covered in Chapter 12c. Extremely thin
Middleman 1974), or define redundant dimen-
films ('clean doctoring') require stiff blades at the
sionless geometric groups and are thus confusing
proper application angle and loading. Reverse-
(Sullivan and Middleman 1986; Hsu et at. 1985).
angle doctoring is akin to knife coating (covered
An analytic model for the creeping flow of a
in Section 12a.3) with a short knife length and
Newtonian liquid in a knife coater has been
a nearly zero gap which is controlled by the
derived for this book.
applied load. Neither process has been studied
As pointed out by Secor (1991), the geometry
systematically as related to reverse roll coating.
of a knife coater (Fig. 12a.2a) can be described
in terms of two dimensionless parameters, the
12a.2.4.2 Process perturbations inclination and the curvature:
The second category of defects are those that L 1L L

arise due to perturbation of the process. As the 1= -tan(w)· K =--- (12a.9a,b)
Ho ' 2R e H O
flow rate in a self-metered coating process is
primarily determined by the gap spacing, any where liRe = llR r + llR k is the equivalent
variation in gap directly translates into coating curvature in terms of the roll and knife radii, L
thickness variability. Such nonuniformity can be is the length of the knife, H 0 is the gap, and w
is the inclination angle. In terms of the dimen- If, in addition, the surfaces are parallel (I = 0)
sionless flow direction coordinate ( = X/L, the then t = 1/2, while if the surfaces are steeply
dimensionless gap profile h = H/H o, and the inclined (w = (0) or the blade level very long
dimensionless coating thickness (flow rate) (L/H o = (0) then 1 = 00 and t = 1.
t = T/H o, the gap profile and the Reynolds The case of 1 = 0 gives the limiting flow rate of
equation (equation (12a.1» can be written as
(12a.10) 1 3]
+-2
-1
[
t=
(I( + 1)2 -
dp 6 12t 1(+1+ .jl( tan- 1(.jI()
d( = h2 - hI" (12a.ll)
(I = 0) (12a.14)
were
h
· . I
t h e d ImenSlOn .
ess pressure IS p= -PH~-.
jJ.UL If L/H o = 00 then t = 2/3, which is the limit of
The Reynolds equation can be integrated, a roll rotating in close proximity to a stationary
boundary conditions of ambient pressure at wall under flooded conditions.
both inlet and outlet imposed, as the result re- The limit of t = 0 is obtained when the two
arranged to solve for the dimensionless flow rate surfaces touch, which occurs when 1 < - 1 - I(
if I( < 1 or when 1 < - 2.j;' if I( > 1.
The flow rate as a function of inclination and
curvature is shown in Fig. 12a.5, alongside results
of finite-element calculations of the full two-
dimensional free-surface flow field. The calcula-
tions were done in the same manner as those of
2 Sullivan, Middleman and Keunings (1987),
where D = 1 - 41( and T1 is the integral of l/h
incorporating the downstream free surface and
which is
applying no-traction boundary conditions far
I)
upstream of the knife. Lubrication theory tends
- -+
-2- [ tan _1(21( =
to underestimate the flow rate for short blades
~ .j-D (L/H o < 10-100), mainly due to an entrance
effect shown in the pressure profiles in
- tan - 1(.j ~ D)]; D<O Fig. 12a.6a. This entrance effect was also calculated
and experimentally verified by Sullivan, Middle-
man and Keunings (1987). Lubrication theory
2 2
D =0 accurately predicts both flow rate and pressure
- - 21( + I'
profiles for long blades (Fig. 12a.6b), even when
the blade inclination angle is steep. When inertial
1 I [I +.jD 21( + 1- .jDJ. D>0 forces become significant the entrance is more
.jD n 1 -.jD 21( + 1 -.jD ' important due to an inertial impact pressure
(12a.12b) (Fig. 12a.6b), which is common under con-
ditions of high speed flexible blade coaters
There are several limiting cases of interest. As (Chapter 12c).
curvature effects become negligible (I( = 0), the The effects ofshear-thinning and viscoelasticity
dimensionless flow rate becomes on the flow rate were studied by Sullivan,
Middleman and Keunings (1987). A 1.25%
1+1 CMCjwater solution, which shear-thins with a
t = - - (I( = 0) (12a.l3)
1+ 2 power-law index n = 0.5, gave lower pressures
Knife coating 549
1.0 r-~---'-------r---"'------'----'------'-----,
1=00; t=1
0.9
0.8
0.7
- -------------
1(=00; t=2/3
eli 0.5
0; K=O, 1=0, t=1I2
a:
;:
0 0.4
u:::
0.3
Finite Element Calculations
0.2 o K=O, UG=10 (short)
o 1(=0, UG=100 (long)
0.1
0.0
-4 -2 o 2 4 6 8 10
Inclination, 1=(UHJtan(w)
Figure 128.5 Knife coater flow rate versus inclination and curvature as predicted by lubrication theory, with
comparison to numerical model. Experimental data (not shown) agree with numerical results.
and the flow rate was increased by 10-15%. knife coater where the downstream geometry
Their numerical model was consistent with these diverges and the downstream static contact line
observations. A constant-viscosity viscoelastic is not pinned at a corner, have reported the
Boger fluid gave experimental flow rates reduced critical capillary number for ribbing (Savage
by about 40%. 1977a; Sullivan and Middleman 1979; Dowson
et al. 1980; Bauman et al. 1982; Adachi et al.
1988). At low capillary number the flow field is
sensitive to the contact line where the downstream
12a.3.2 RIBBING INSTABILITY
meniscus separates from the stationary plate at
The primary fluid-mechanical instability in knife some static contact angle (Coyle et al. 1986). In
coating is ribbing. The operating diagram has fact, Adachi et al. (1988) reported that at much
been studied theoretically and experimentally by higher capillary number, the contact line passes
Strenger, Secor and Sramek (1992). For Newtonian through the nip center and the ribs disappear.
and shear-thinning liquids, ribbing appears above These results seem to indicate that increasing
some low value of capillary number only for knife gap divergence destabilizes the flow to
diverging-tip geometries (I < 0). As the liquid is ribbing. In a related geometry, Bixler (1982)
made more viscoelastic, ribbing also occurs at reported that increasing the divergence of the
highly converging geometries (I » 1) due to the flow channel stabilized the flow to ribbing, and
development of elastic normal stresses. similar results exist for forward roll coating (see
A number of studies of a roll rotating next to Section 12a.4); which would seem to contradict
a stationary plate, which can be thought of as a knife coating results.
0.35 .--~-...,.-----r-----,----r-----...------.--- ~
UH.=10
0.30
0.25
./
<0=5° "./
...J 0.20 (1=0.875) /"
:::> /
::1.
"'.....0 /
/
I 0.15 /
a. /
&. /
/
i::J 0.10 /
/
'"'"~ /
-- -
/
0.05
a.
.: / /
0.00 _.··__···_····_····_·-r-·~·c"'.=_:.~_ . __.. _.__..._.
-0.05
-1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2

Flow Direction Coordinate, ~=XIL
(a)
0.16
UH.= 100
0.14
0.12
...J 0.10
:::>
~~
I 0.08
a.
II
Co
i:J 0.06
'"'"
~ 0.04
a.
0.02
0.00 ---
..............................
-0.02
-1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2
(b) Flow Direction Coordinate, ~=X1L
Figure 12a.6 Knife coater pressure profiles at small and large inclinations, comparing lubrication theory
(dashed lines) and finite-element calculations (solid lines), for: (a) short blades; (b) long blades, and (c) diverging
tip geometries.
Knife coating 551
0.25
0.20
0.15
.J
::::l
N~
~ 0.10
:J:
CL
II
Co
0.05
r€
:l
'"~'"
CL 0.00
-0.05
-0.10
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4
(c) Flow Direction Coordinate. x=(X·X...)/Lo
Figure 12a.6 Continued.
The stability of these flows is determined by critical capillary number. Figure 12a.6c shows
the interfacial force balance (equation (12a.5)), pressure profiles for constant minimum gap
not just by the degree of flow channel divergence. clearance H min for three different diverging tip
To make comparisons based on minimum (not lengths (i.e. w = - 50 corresponds to the knife
tip) gap, equation (12a.11) can be re-arranged to tip extending 5° past the point of minimum
give the pressure gradient at the meniscus in clearance around the roll). These profiles clearly
terms of the minimum gap show that increasing the diverging tip region
causes the pressure gradient at the downstream
dP = 6~U (1 _ ~)2 (1 _ 2t) (12a.15) meniscus to go from negative (stabilizing) to
dX H min 411: positive (destabilizing).
Adachi's results can be explained as follows.
This equation indicates that the pressure gradient At the lowest speeds the meniscus locates far
at the meniscus is negative as long as t > 1/2, so downstream from the gap center so that the
that the flow is always stable to ribbing if t > 1/2. pressure gradient is destabilizing but low enough
If t < 1/2, ribbing will appear at some critical in magnitude that surface tension dominates. As
capillary number. Figure 12a.5 shows that the speed is increased, the meniscus recedes into the
flow rate t can be greater than 1/2 for small gap and ribbing occurs. At the highest speed the
negative inclinations (diverging tips) depending wetting line moves far enough into the gap that
on the curvature. Equation (12a.12) can be used t '" 1/2 and the flow becomes stable again. Bixler's
to find the inclination at which t = 1/2(1 1 / 2 ), Any results can be explained as follows. His flow was
diverging-tip geometry with 1 > 11/ 2 will be stable premetered with t < 1/2 held constant. He
to ribbing. If the inclination is less, the tip is increased the divergence by increasing II: which
more diverging and ribbing will occur at some according to equation (12a.15) decreases the
pressure gradient and so stabilizes the flow. A Newtonian or shear-thinning. If the fluid is
similar effect takes place in forward roll coating, viscoelastic, the window shrinks considerably
and is discussed in Section 12a.5. because viscous normal stresses at high inclination
generate another region of ribbing. Experiments
indicated there was no window with viscoelastic
12a.3.3 OPERABILITY DIAGRAMS
polyacrylamide solutions, rather, some degree of
Figure 12a.7, adapted from the work ofStrenger, ribbing was always observed.
Secor and Sramek (1992), shows operability
diagrams for a roller-fed knife coater. The main
12a.3.4 PRACTICAL IMPLICATIONS
failure modes are starvation, where the auxiliary
feed flow does not supply enough coating to the There are several practical implications of the
knife; flooding, where too much flow is supplied; above analysis and results. The coating thickness
and ribbing. The window is large when the in a knife coater is primarily determined by the
inclination is positive (converging) and the fluid clearance between knife and web, and is typically
1/2 to 1 times the gap which necessitates small
gaps for thin coatings. This makes the coating
thickness sensitive to mechanical tolerances, web
thickness variations, knife deflections and physical
damage to the knife tip. Contamination can be
trapped in this gap leading to streaks.
Knife
As the coating thickness is comparable in
Inclination magnitude to the gap, it might be difficult to pin
the static contact line to the downstream corner
of the knife as drawn in Fig. 12a.2a. The liquid
will have a natural tendency to wet up the
downstream surface of the knife, possibly causing
streaky coatings. The wedge spreader (Pearson
Feed Film 1960) is an extreme example of this phenomenon,
Thickness where ribbing results.
(a)
Finally, ribbing is a problem if the geometry
is too diverging or the fluid viscoelastic. The
former simply requires proper knife installation,
Starvation
while the latter might require reformulation,
~
dilution, or a change in coating method.
Knife
Inclination 12a.4 FORWARD ROLL COATING
The simplest forward roll coater is shown in Fig.
12a.lb. The coating thickness applied to the web
is governed by the flow in the nip between the
bottom applicator roll and the top back-up roll
Feed Film around which the web is wrapped. The coating
Thickness liquid is initially deposited on the applicator roll
(b)
by dip coating from the coating pan, although
Figure 12a.7 Schematic operability diagram for a
a fountain roll or die could also be used. The
roller-fed knife coater, showing major failure modes
for: (a) Newtonian and (b) viscoelastic liquids (adapted most-commonly analyzed case is where an excess
from Strenger, Secor and Sramek (1992». of coating reaches the gap so that it is flooded,
and so the details of the inlet region can be rolls were completely submerged, the pressure
neglected. The case where the gap is 'starve-fed' is profile on the downstream side of the gap would
only briefly covered in this section and is dealt be an inverted image of the upstream profile in
with in detail in Chapter 12b. Fig. 12a.8. Downstream of the gap center the
liquid film splits into two films, each coating one
of the rolls. In the vicinity of the film-splitting
12aA.1 FILM-SPUrrING MENISCUS REGION
liquid/air interface (the coating meniscus) the
- STEADY FLOW
flow is two-dimensional.
The lubrication of journal and slider bearings A number of analyses of forward roll coating
(Dowson and Taylor 1979) and gear teeth (Martin flow are based on lubrication theory with various
1916; Gatcombe 1945), along with the knife ad hoc boundary conditions which either ignore
coater ofthe previous section, are all special cases or crudely approximate the influence of the flow
of the geometry of Fig. 12a.2b. Flow in the region near the film-split meniscus. The primary focus
of the gap center is therefore governed by the of these models is the prediction of the dimen-
Reynolds equation (equation (12a.l)). Fluid is sionless flow rate, defined as.A. = Q/(Ho(J) where
dragged into the gap by the moving roll surfaces Q is the total flow rate (per unit width) through
through the action of viscosity. As the geometry the gap and H 0 is the minimum total gap spacing.
is converging, the 'squeezing' action generates Martin (1916) used the Reynolds condition that
high pressures on the inlet side of the gap. the film splits at a point downstream where both
Conversely, the drag flow in a diverging geometry the pressure and its gradient are zero, predicting
results in low pressures (below atmospheric) on .A. = 1.226. Gatcombe (1945) assumed the rolls
the downstream side of the gap. In fact, if the were completely submerged, predicting .A. = 4/3,
2.5
2.0
1.5
Lubrication
1.0
S
oc"
=:)
0.5
to
J: 0.0
Submerged
a-
Il
c.
e
:J
(J)
-0.5
(J)
~
-1.0
a-
Ca=0.1
-1.5
-2.0 Finite Element
-2.5
-3.0 -2.0 -1.0 0.0 1.0 2.0 3.0
Flow Direction Coordinate. C.=XI(2H.R.l'/2
Figure 12a.8 Computed pressure along the main flow direction in forward roll coating as a function of
capillary number. The dashed lines are lubrication theory results for different downstream pressure boundary
conditions.
while Hopkins (1957) proposed the separation Several authors have attempted to develop
hypothesis that the film splits at the first stagnation simpler analytical models that crudely incorporate
point downstream of the nip center, thus predicting the film-split meniscus into a lubrication model
A. = 1.3015. Pitts and Greiller (1961) attempted of the flow (Greener and Middleman 1975, 1979;
an approximate solution to the difficult problem Taylor 1974a,b,c; Benkreira et al. 1981a). Un-
of the two-dimensional flow near the meniscus fortunately, these models give totally inaccurate
with some success, as did Coyne and Elrod predictions of flow rate and meniscus location at
(1970a,b), who essentially solved approximate small capillary number. The exception are those
film profile equations. Williamson (1972) obtained that patch together lubrication pressure profiles
an approximate numerical solution to the with Coyne and Elrod's (1970a,b) approximate
downstream free-surface flow. Savage (1977a) solution, which give more accurate behavior
and Dowson and Taylor (1979) have reviewed (Greener and Middleman 1979).
some of these earlier boundary conditions. Finite-element solutions of the Navier-Stokes
Ruschak (1982) solved the problem accurately equations show details of the flow field (Coyle
(for small Ho/R.) using matched asymptotic et al.1986). Streamlines as a function of roll speed
expansions, while Coyle et al. (1986) used the ratio and capillary number are plotted in Fig.
finite element method to solve the full 12a.9. As capillary number is decreased, the
Navier-Stokes equations for the flow field. meniscus locates farther from the gap center and
...
Ca= 0.05 Ca=5.00

'\
Figure I2&.9 Computed stream lines for forward roll coating flow as a function of capillary number and
speed ratio (Ho/R. = 0.(01).
a pair of recirculations grow. This is due to
surface tension forces at low capillary number
10
limiting the curvature of the meniscus and so
8
forcing it to locate at a point oflarger roll surface 0.6
separation. The positive pressure gradient (Fig. E Data
12a.8) combines with this larger roll surface 0.4 n =0.67
separation to drive the recirculations. .12
ca 4
The computed (Coyle et al. 1986) dimensionless a: 0.2 L..-_L-----,,'C----L:-'
0.4 0.6 0,8 1.0
flow rate A. varies from 1.33 to 1.29 as capillary '"'"
Q)
C
Power Law Inde•• n
.>t<
number is increased from 0.01 to infinity for U
:r:
t-
equal roll speeds, in agreement with experiments E
2
(Pitts and Greiller 1961; Schneider 1962; Hinter- u:

maier and White 1965; Benkreira et al. 1981b).
The computed value of A. is also relatively
insensitive to roll speed ratio, having a value of
A. = 1.23 for U2 /U 1 = 0 and Ca = 00, which is 2 4 6 8 10
also in agreement with experiments (Sullivan and Speed Ratio. UilU,
Middleman 1979; Adachi et al. 1988). The film-split
meniscus location has been measured by several
authors for both symmetric (Pitts and Greiller Figure 12a.l0 Effect ofshear-thinning on film thickness
1961; Coyle et at. 1986), and zero speed ratio ratio in forward roll coating at infinite capillary
roll/plane configurations (Dowson et at. 1980; number, compared with experiments. Decreasing n
causes the film to split more symmetrically (reproduced
Adachi et al. 1988), and agreement with finite- from Coyle et al. (1987), by permission).
element solutions is good (Coyle et al. 1986).
Theory and experiment for shear-thinning
viscous fluids give values of the dimensionless
flow rate A. that increase to around 1.4 as the (1981 b). If the liquid is shear-thinning, decreasing
power law index (n) decreases (Coyle et al. 1987; the power law index at a given speed ratio causes
Benkreira et at. (1981b). Doremus and Piau the film to split much more symmetrically, so
(1981,1983) measured greatly reduced flow rates that the faster roll gets more nearly the same
with viscoelastic polyethylene oxide solutions. coating thickness as the slower one (Coyle et at.
This effect is a result of the high elongational 1987; Benkreira et at. 1981b). This is primarily
viscosity of these fluids interacting with the rapid due to the action of the shear-thinning viscosity
accelerating/decelerating flow kinematics when which partially nullifies the asymmetry in the
the gaps are large and Newtonian viscous forces velocity boundary conditions.
weak. These effects are analogous to those reported The pressure field is one-dimensional except
for knife coating (Section 12a.3). in the immediate vicinity ofthe film-split meniscus.
When the rolls are moving at unequal speeds At high capillary number, the positive portion
the faster roll carries away the thicker coating ofthe pressure curve is much larger in magnitude
(Fig. 12a.1O). Combining lubrication theory with than the negative portion, producing a roll-
the hypothesis that the flow splits at the first separating force. As the capillary number is
stagnation point simply predicts that the ratio decreased, the negative portion of the pressure
of coating thicknesses on the rolls is equal to the curve expands and becomes comparable in
square root of the speed ratio (Savage 1982). magnitude to the positive part, as shown in the
Computations (Coyle et at. 1986) indicate that computed pressure profiles of Fig. 12a.8. Similar
the exponent should be roughly 0.65, which is profiles were measured by Dowson et at.
in agreement with the data of Benkreira et at. (1980).
12a.4.2 FILM-SPUTIING MENISCUS REGION These neutral stability curves are similar to those
- RIBBING shown by Ruschak (1985, Fig. 9). The slope
The most-studied aspect of roll coating instability between roll surfaces is
is ribbing due to the viscous mechanism described 2T 1.34Ca 2/3
in Section 12a.2.2. The pressure gradient at the q---
- HI - cos(a/2)
film-splitting meniscus is always positive (Fig.
12a.8) and thus always destabilizing according (12a.18)
to equation (12a.5). Therefore the avoidance of
Equations (12a.16) and (12a.18) represent the
ribbing in roll coating requires low capillary
neutral stability curve. Combining equations
numbers and large gaps such that the surface
(12a.17) and (12a.18) for Ca~O gives
tension forces generated by the curved interface
can dominate the viscous pressure forces generated
by the lubrication flow. Ca = 0.338(~:r/8 (12a.19)
Many authors have proposed simple models
to predict the onset of ribbing (Pitts and Greiller which is included in Fig. 12a.11. This expression
1961; Fall 1978, 1982, 1985; Savage 1977a,b, 1984, has the correct limiting behavior for Ca ~ 0, but
1991; Gokhale 1981, 1983a,b; Benkreira et al. underestimates the critical capillary number over
1982a). The following illustrates the basic approach. the range of HO/R e of interest. The accuracy can
The critical capillary number for the onset of be improved by changing assumption (4) regarding
ribbing can be simply predicted by equation the meniscus shape (Savage 1991).
(12a.5) by considering the flow between linearly An alternative approach is that of Ruschak
diverging surfaces (Ruschak, 1985). The terms in (1985), who combined his asymptotic analysis of
equation (12a.5) are evaluated by assuming that: the steady base flow with Pearson's (1960) linear
stability analysis to achieve an accurate prediction
(1) A '" 4/3, so that the coating thickness on each of the onset of ribbing. A more rigorous approach
roll is given by T= 2Ho/3; was taken by Coyle et al. (1990), who used 20
(2) the roll separation at the film-split location base flow calculations and 3D normal-mode
is given by the geometric relationship linear stability analysis to obtain more accurate
HI = 2Rc cos (a/2) where a is the diverging predictions for large gap/radius ratios.
angle at the split point; Figure l2a.ll shows measurements of the
(3) dRc/dX = sin(a/2); critical capillary number for ribbing in roll
(4) the meniscus radius of curvature and film coating systems, summarizing results ofa number
thickness are related by the Landau- Levich of workers (Pitts and Greiller 1961; Mill and
approximation T/R e = 1.34Ca2/3; South 1967; Greener et al. 1980; Benkreira et al.
(5) the pressure gradient given by equation 1982a; Carter and Savage 1987; Couder et al.
(12a.l) is dp/dx = 12(I/hi - A/h~) where 1989; Rabaud et al. 1989, 1990; Coyle et al. 1990a;
p = PHo/jJ.U, x = X/H o, and hI = HdHo. Rabaud 1990), and the data is consistent within
Combining these expressions gives a neutral the range of experimental uncertainty. The
stability curve determined by theoretical curves, both finite-element and asym-
ptotic, agree well with the experiments over more
1.34 5/3 1 2 .
than two orders of magnitude ofgap/radius ratio.
Ca - cos(a/2) Ca - 3cos (a/2)slO(a/2) = 0
Both data and theory indicate that speed ratio
(12a.16) itself has little effect on the critical capillary
number when defined with the average roll speed
which, for small a and Ca, has the approximate
(Carter and Savage 1987). In addition, Hasegawa
solution
and Sorimachi (1993) showed that ribbing could
Ca = a/6 (12a.17) be eliminated by placing a string acrosss the gap
10'
OPins & Greiller (1961)
a Mill & South (1967)
o Greener et. a/. (1980)
li Benkreira et. a/. (1982) Theory
+ Coyle et. at. (1990)
x Rabaud (1990) Coyle et. al. (1990)
Hakim et. a/. 1988
Ribbing
Uniform Coating
Equation 12a.19
10-3
Gap/Radius, HjR.
Figure 12a.ll Critical capillary number for the onset of ribbing in forward roll coating as a function of
gap/radius ratio, showing comparison of experiments and theory.
touching the meniscus. This causes the meniscus in non-Newtonian liquids such as butter and
to pin to the string at a larger distance from the printing inks (Fukushima 1976). For non-New-
gap center which reduces the destabilizing viscous tonian paints, ribbing, filamentation, and 'spatter'
pressure gradient. is observed (Glass 1978a-d). Bauman et al.
Linear stability analysis does not predict what (1982) report that adding 10 ppm ofhigh molecular
the new steady or time-dependent flow pattern weight polyacrylamide reduced the critical
will be beyond the critical capillary number, it capillary number by a factor of 2- 5, even though
merely predicts the fastest-growing mode at there was no measurable change in steady shear
criticality. Experiments indicate that the steady viscosity nor could normal stresses in steady
ribbed flows are stable for some range of higher shear flow be measured. Adding 1()() ppm reduced
capillary number, though no detailed data exists the critical capillary number by a factor of 10-30.
(Mill and South 1967; Coyle 1984). Recently, Similar results were obtained by Hasegawa and
Gurfinkel and Patera (1996) computed the Sorimachi (1993). To quantify the effect of rheology
evolution from the unstable two-dimensional on the onset of ribbing, Soules et al. (1988) and
steady states to the stable three-dimensional Fernando and Glass (1988) studied a series of
steady ribbed states by numerical solution of the aqueous blends of hydrolyzed polyacrylamide
unsteady three-dimensional free-surface fluid flow (flexible backbone, high extensional viscosity)
problem using periodic boundary conditions in and hydroxyethyl cellulose (more rigid backbone,
the cross-web direction. The results in Fig. 12a.12 low extensional viscosity). All blends had the
show the evolution of a nonsinusoidal steady- same steady shear viscosity, but differed greatly
state ribbing pattern similar to that of Fig. 12a.3. in extensional viscosity. Dramatic reduction of
Unusual ribbing patterns have been observed the critical capillary number for the onset of
past criticality the ribs get shorter in wavelength,
become time-dependent, and take on filament-
like structure. The ultimate chaotic filamentation
3.0
film-splitting mode is simply the natural state of
the system which is very far beyond criticality,
not the result of vaporous cavitation.
2.5
12a.4.3 ROLLING BANK INLET REGION

2.0
The flow field in the upstream rolling bank region
t
TIME
was qualitatively described by Pearson (1985)
and experimentally determined by Agassant and
1.5 Espy (1985) for calendering of high-viscosity
polymer melts. The flow field may consist of up
to two vortices under the free surface, shown in
1.0 Fig. 12a.13. These have been computed for a
limited number of cases at low capillary number
x (Kistler and Scriven 1984; Chen and Higgins
1988). Benjamin and Scriven (1992) computa-
z tionally mapped out the flow rate versus capillary
number parameter space. For a given capillary
number, reducing the flow rate below the
Figure 12a.12 Free surface profiles across symmetry flooded-gap limit gives two-vortex flows, followed
plane for film-splitting flow in forward roll coating, by one-vortex flows, followed by vortex-free
showing computed evolution to a stable three-dimen- flows, below which the bead breaks at the
sional non-sinusoidal ribbed flow state. (Ca = 0.5,
'starvation' limit. These computations were all
UzIU 1 = 1, HolRe = 0.02, wavelength = 15Ho , and
time is in units of p/pU 2 (reproduced from Gurfinkel for conditions of small enough gap that a
Castillo and Patera (1996), by permission). one-dimensional lubrication pressure profile
developed near the gap center. Chapter 12b
examines cases outside this regime, where the
ribbing was correlated with the magnitude of the structure of the flow is much more complex and
extensional viscosity. more vortices can be present (Gaskell, Savage
Finally, a number of authors have postulated and Summers 1994; Benjamin and Scriven 1994).
that under extreme conditions (high viscosity, The stability of this flow has not been studied.
high speed, small gap), the sub-ambient pressure It is suspected that these vortices will become
downstream of the gap center in forward roll unstable to three-dimensional disturbances at
coating causes vaporous cavitation within the high enough Reynolds number as discussed in
liquid, and therefore represents a fundamentally Chapter 12a.2.2. Air entrainment at the wetting
different mechanism of film-splitting (Banks and line might occur at high enough speeds, but this
Mill 1954; Floberg 1964; Miller and Myers 1958; too has not been studied. Air entrainment might
Myers et al. 1959; Myers and Hoffman 1961; also occur due to formation of a cusp where the
Hoffman and Myers 1962; Dowson and Taylor incoming free surface meets the rolling bank (see
1979). A different point of view is as follows Section 12a.2.3). Finally, excessive feed could
(Kistler 1994). For high viscosity liquids and lead to excessive runback and film instabilities
small gaps, the critical speed for the onset of which could perturb the flow and give coating
ribbing is extremely low. As speed is increased nonuniformities.
Reverse roll coating 559
instability could be high-speed limits. Inertial

ribbing such as seen by Yih (1960) might also be
present. Ofall these limits, only the high-viscosity
ribbing limit has been mapped out thoroughly.
12a.4.5 PRACTICAL IMPLICATIONS

There are several practical implications of the
above analysis and results. As in knife coating,
the coating thickness in a forward roll coater is
primarily determined by the minimum gap, which
necessitates small gaps for thin coatings and
makes the coating thickness sensitive to mechanical
tolerances, web thickness variations, and roll
deflection. In addition, viscosity and surface
tension (Ca) affect the flow field and its stability.
Forward roll coating is especially susceptible to
ribbing, requiring either large gaps, low speed,
low viscosity, rapid leveling after coating, or
less-stringent product uniformity requirements.
12a.5 REVERSE ROLL COATING

Figure 12a.1c-f shows four types of reverse roll
Figure 12a.13 Streamlines in the rolling bank of a coaters. The flow in the metering gap, between
forward roll coater, showing transition between: (a) the applicator roll and the reverse-acting metering
two-vortex regime; (b) one-vortex regime, and (c) roll, is the primary flow field all reverse roll
no-vortex regime as the inlet film thickness is decreased. coaters share. This flow is the main focus of this
(Ca = 0.51, U2/U 1 = 1, Ho/Re = 0.004 (reproduced
from Benjamin (1992), by permission). section. The reverse-roll nip transfer flow of Fig.
12a.2d is also discussed here because it can also
strongly affect the coating uniformity achieved
with a reverse roll coater.
12aAA OPERABILITY DIAGRAMS
Figure 12a.14 attempts to capture the operability 12a.5.1 METERING GAP FLOW - STEADY
diagram of the simple forward roll coater of Fig. FLOW
12a.1b. Two hypothetical diagrams are shown, Simple lubrication-theory models of flow in the
since the high-viscosity behavior will be quite metering gap, ignoring the metered film meniscus
different from the low-viscosity behavior. At high and the dynamic wetting line, predict a simple
viscosity, ribbing imposes a severe limit on speed,
linear relationship between the metered film
while starvation should not be a problem. Air thickness and the speed ratio and gap thickness:
entrainment, either on the web or in a cusp, could
also be high-speed limits. At low viscosity
starvation can be a limit because dip coating will t =
m
Tm
= ~(1
Ho 2
_v,.
Vm ) (12a.20)
supply a thin film to the gap and so low capillary
number (low speed) will be needed to stabilize where the applicator and metering roll surface
the bead. Air entrainment and/or inertial vortex speeds, v,. and Vm , are taken as positive numbers,
Air Entrainment in Pan

or Runback
...
B
\'lI
£
c..
Q.
<[
Starvation
Web Speed Uw
(a)
'0
<II
<II
Q.
UJ
a
IX:
o
'lii
.!:!
c..
Co
<[
Web Speed Uw
(b)
Figure 12a.14 Operability diagrams for the two-roll forward roll coater of Fig. 12a.l b: (a) viscous regime;
(b) inertial regime.
and A., as in forward roll coating, is 1.3 (Ho and the flow as a function of capillary number and
Holland 1978; Greener and Middleman 1981; speed ratio (Coyle et al. 1990b). Similar to
Benkreira et al. 1982b; Coyle et al. 199Oc). forward roll coating, decreasing the capillary
Experimental data of Ho and Holland (1978) and number causes the menisci to be less-tightly
Benkreira et al. (1981b), which are restricted to curved and in doing so move to where the roll
low speed ratio, agree with this expression, with surfaces are farther apart. Therefore the bead
the extensive data of the latter authors yielding gets larger and a recirculation appears in the
a value orA. = 1.26 for equation (12a.20). Equation upstream film-transfer flow. Increasing the speed
(12a.20) predicts negative metered film thicknesses ratio causes the wetting line to become located
at speed ratios above unity, which is not possible. on the upstream side of the gap (the right-hand
Figure 12a.15 shows computed streamlines for side of the figure), rather than on the downstream
Figure 12a.15 Computed streamlines for metering gap flow as a function of: (a) capillary number and (b)
speed ratio. Increasing either causes the bead to shrink and the wetting line to invade the gap (HoIR. = 0.004,
a - Ca = 0.1, b - VmlV. = 0.4). As drawn, the metering roll is directly above the applicator roll (reproduced
from Coyle et ai. (1990b), by permission).
Rev erse roll coating 561
(a)
g g jf :
=-=~::gej[~ :
(b)
side as one might intuitively expect. The recir- experimental and computed coating thickness
culation on the upstream side of the gap shrinks, pass through a minimum beyond which they
and the coating bead becomes small and tightly increase sharply. The minimum film thickness
curved. In all knife and roll coating flows, corresponds to the point where the wetting line
lubrication theory predicts that the net flow is is at the gap center. At this point the gap has
in the direction of the faster surface, and this sets effectively lost its metering ability since the
up a pressure gradient that augments the flow region of minimum roll surface clearance is no
rate by roughly 30% (i.e. A. = 1.3). The reverse-roll longer filled with liquid. Instead, the metering is
metering gap must always have a net flow in the being done further upstream where the clearance
direction of the applicator roll, so that as speed is larger and thus the metered film is thicker.
ratio approaches and passes unity there must be Further increasing speed ratio worsens this
significant deviations from the predictions of situation, and the steep upturn in metered film
lubrication theory due to the curved menisci and thickness corresponds to the wetting line moving
surface tension. further into the upstream side ofthe gap. Lowering
Figure 12a.16 shows the. dependence of the the capillary number has little effect at low speed
metered film thickness on speed ratio and capillary ratios, but it shifts the minimum metered film
number. As speed ratio is increased both the thickness to higher speed ratio values.
Theory Theory
0.8 Ca=5 • Ca=1
~~/
Ca=7.6
/Ca=2.3
;' Ca=1.5
~ 0.7 /! :/.
:.
I- Ca=3.8
~ ;' ....
..........
o
:.. l
~ 0.6
~
u
:E
t-
..........
.....•
.....~ .
...-
.......
//
§ I •••• ...............
u::: 0.5 ...................................
"0
e
~
~
0.4
"" .....,
""'
Lubrication , ,,
0.3 '--_........._--'-_ _"'----_--'-_ _' - - _ . . . . L . . _ - ' "_ _- ' - - _........_--'_-",'--"::::....--.J
0.0 0.1 0.4 0.5 0.6
Figure 118.16 Reverse roll metered film thickness as a function of speed ratio and capillary number, comparing
experiments (dashed lines) and theory. The point of minimum film thickness corresponds to the wetting line
located at the gap center (H 0 = 500 Jlm, R 1 = R 2 = 0.1 m).
Kang and Liu (1991) report the speed ratio at The mechanism of ribbing in reverse roll
which the film thickness attains its minimum as coating is the same as that in forward roll
coating. Surface tension stabilizes the flow to
Vm/v.. = 0.29 Ca- o.ss ; 0.12 ~ Ca ~ 12
ribbing. The hydrodynamic pressure gradient at
(12a.21) the metered film meniscus destabilizes the flow
Unfortunately they do not report a correlation if it is positive. Whereas in forward roll coating
for minimum coating thickness, which would be this pressure gradient is always destabilizing, in
of interest since significant positive deviations reverse roll coating the pressure gradient can be
from equation (12a.20) are present at the minimum, reversed by increasing the speed ratio (Fig.
especially for low capillary number. 12b.18) and in this way the pressure gradient
Fukazawa (1993) matched lubrication pressure becomes stabilizing. Thus ribbing can be elim-
profiles to the Landau and Levich condition for inated in a reverse roll coater even at high
radius of curvature of the metered film meniscus capillary number when surface tension is no
to create a simple model which captures the longer a significant stabilizing influence.
essential physics of this flow, thus predicting the The mechanism of cascade stems from the
presence of a minimum in the film-thickness intrusion through the gap of the dynamic wetting
versus speed ratio curve. line, as shown in Fig. 12a.15. When the wetting
Theory for shear-thinning viscous fluids predicts line is located on the upstream side of the gap,
that there is little change, as compared with the region of minimum roll surface clearance
Newtonian theory, in metered film thickness as does not affect the metering of the coating, and
a function of speed ratio and capillary number, increasing speed ratio increases the coating
providing the latter is defined using the viscosity thickness. At high enough speed ratios, the
at the characteristic process shear rate y = metered film becomes as thick as the minimum
2v../Ho (Coyle et al. 1990d). Experiments with roll surface clearance through which it tries to
shear-thinning polymer solutions confirm this pass. It periodically re-attaches to the metering
(Benkreira et al. 1981b). The biggest difference roll at the gap center, entraps a large bubble of
is a predicted 5-7% increase in film thickness air on the metering roll, and is suddenly much
at low speed ratios. The data confirm that the thinner because the gap at the re-attachment
dimensionless flow rate A. increases to 1.34 (a 6% point (which does the metering) is much smaller.
increase from its Newtonian value of 1.26). The wetting line then moves back into the
upstream region and the film becomes thicker
and the cycle repeats itself. Flow visualization
12a.5.2 METERING GAP FLOW - FAILURE
experiments have confirmed this mechanism
MODES
(DiCarlo 1986).
The stability of the metering gap transferred film Two typical operability diagrams are shown
has not been examined closely. The metered film in Fig. 12a.19. At very low capillary number the
in a reverse roll coater is subject to two instabilities. flow is uniform and stable at any value of speed
The first is the classical viscous ribbing, while ratio. What makes a reverse roll coater so useful
the second is a three-dimensional time-periodic is that at high capillary number there exists a
disturbance called 'cascade', 'herringbone', or stable 'window' of operation: if the speed ratio
'seashore' (Babchin et al. 1981; Coyle et al. Wm/v..) is within a certain range, the metered
1990b). This latter disturbance, shown in the film is uniform and stable. If the speed ratio is
photograph of Fig. 12a.17, consists of irregular below this range, the metered film exhibits ribbing
saw-tooth cross-web bars on the film coinciding of the sort seen in a forward roll coater. At speed
with periodic bead oscillation and air entrapment ratio values above the stable range, the cascade
on the metering roll. The latter gives a nonuniform instability sets in. Transition to ribbing as the
transferred film containing large air bubbles. speed ratio is lowered out of the stable range is
Figure 118.17 Photograph of the cascade instability on the metered film (top view, roll surface moves from
bottom right to top left). The irregular cross-web waves are regularly spaced in the downweb direction
(reproduced from Coyle et al. (1990b), by permission).
gradual, but the transition to cascade is sharp, coater in order to eliminate ribbing. In addition,
occurring just beyond the speed ratio which as Reynolds number increases, the stable operating
produces the minimum metered film thickness. window widens and shifts to higher speed ratios
The onset of cascade shifts to slightly higher (Coyle et al. I990b).
values of speed ratio as the gap is decreased, but Experiments and computations for non-New-
this onset can be approximated for gap widths tonian liquids have been carried out by Coyle et
of 25-250 Ilm as (Coyle 1984) al. (1990d). Experiments with shear-thinning
polymer solutions show that the onset of cascade
Vm/Va = 0.43 Ca- O.46 ; 0.2 ~ Ca ~ 4 (l2a.22)
is unchanged from Newtonian results if capillary
As one would expect, this curve is above and number is based on the process shear rate as
roughly parallel to the correlation of maximum above (y = 2v../H o), which is in agreement with
speed ratio for minimum coating thickness theory. Ribbing was aggravated, and the stable
(equation (12a.21)) defined by Kangand Liu(1991). window disappeared at high polymer concentra-
It is interesting to note that for small gaps tions. Experiments with highly elastic polymer
there is an intermediate range ofcapillary numbers solutions gave no stable operating window due
over which there is no stable window ofoperation. to ribbing. At high concentrations, there was no
This leads to the counter-intuitive fact that in longer a clear transition to cascade, but rather
some cases one must increase the speed of the the metered film gradually became more 'mottled'
2.0
1.5
1.0
~
a:
.
::::>
0.5
~~
J:
a.. 0.0
II
Q.
~ -0.5
::::l
I/)
0.4
...
I/)
Q)
a.. -1.0 0.2
0.0
-1.5
-2.0
-1.0 0.0 1.0 2.0 3.0
Flow Direction Coordinate. ~=X1(2HoR.)1/2
Figure 12a.18 Computed pressure along the main flow direction (X) as a function of speed ratio for reverse
roll coating. The metered film forms on the left side of the figure.
in appearance as speed ratio was increased. ribbing is approximately 1.2 for Newtonian
Deformation rates computed from Newtonian liquids and 1.6 for viscoelastic polymer solutions.
kinematics illustrate the severity ofthe deformation Broughton et al. (1950) found a limiting wipe
rates in a reverse roll coater, and how rapidly a ratio of roughly unity. This general behavior can
fluid element experiences these deformations. be explained according to the viscous mechanism
For example, shear rates of 10 5 s - 1 and extension for ribbing (Section 12a.2.2). The Reynolds
rates of 104 s- 1 are not uncommon, and occur equation predicts a positive pressure gradient at
over a time scale of milliseconds. Considering the film transfer meniscus if the wipe ratio
the extreme nature of the deformation one might exceeds 1.0, so that this meniscus becomes
expect that rheology will be important in potentially unstable to ribbing according to
determining the operability ofa reverse roll coater. equation (12a.5), and this result is consistent with
the viscous mechanism of ribbing. A curious
aspect of Kang, Lee and Liu's results is the
12a.5.3 REVERSE ROLL NIP TRANSFER
insensitivity of the limiting wipe ratio to capillary
The operability of reverse roll nip transfer, shown number and inlet film thickness. The stabilizing
in Fig. 12a.2d, has been studied by Kang, Lee effect of viscoelasticity can perhaps be explained
and Liu (1990). If the web is moved too fast with by considering a polymer solution's extensional
respect to the applicator roll, ribbing is observed. response. Increasing the wipe ratio increases the
The limiting wipe ratio (Vw/v,,) for the onset of relative stretch rate of the coated film, so that a
film. Rising film flow instability could also be a

2.2
Gap=750 11m Gap=250 11m limit, as high coverages from dip coating could
result in run-back, which would introduce a
2.0
defect directly onto the coated product. Air
1.8 entrainment in the pond will occur at high
enough viscosity and applicator roll speed
1.6 Stable (equation (12a.8)). Air entrainment on the web
m might occur, but has not been studied. At low
1.4
52. enough viscosity, metered film ribbing and cascade
Q;
.0
E 1.2 are absent and starvation becomes the main
limit. Inertially-driven ribs and vortex instabilities
:0
Z
~
2! 1.0
might also come into play, but have not been
:a
11l
<.)
0.8
studied.
0.6
0.4 12a.5.5 OPERABILITY DIAGRAMS-

THREE-ROLL REVERSE ROLL COATER
0.2 :-, }"
Stable
St~bie------------
The operability limits of a three-roll reverse roll
0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 coater would be those of the two-roll version
Speed Ratio (V",IV a)
above, with the extra variable of the wipe ratio
as a third dimension to the diagram. In this case,
the main extra failure mode that limits operability
Figure 12a.19 Experimental operability diagrams for is reverse nip transfer ribbing, avoidance of which
metering gap flow with a Newtonian liquid, showing requires a wipe ratio of 1.2 or less. Once again,
operating regions of defect-free coating, ribbing, and
inertially driven ribs and vortex instabilities may
cascade. The triangles on the y-axis are data for a roll
against a stationary plate (reproduced from Coyle also come into play, but have not been studied.
(1992), by permission).
viscoelastic solution might develop locally lower 12a.5.6 PRACTICAL IMPLICATIONS

pressures near the meniscus which could partially There are several practical implications of the
counteract the positive pressure gradient due to above analysis and results. As in knife and
viscous forces. forward roll coating, the coating thickness in a
reverse roll coater is primarily determined by the
minimum gap, which necessitates small gaps for
12a.5.4 OPERABILITY DIAGRAMS -
thin coatings and makes the coating thickness
TWO-ROLL REVERSE ROLL COATER
sensitive to mechanical tolerances and roll
Figure 12a.20 attempts to map the operating deflection. As in forward roll coating, viscosity
diagram for the simple two-roll reverse roll and surface tension affect the flow field and its
coater of Fig. 12a.lc, for both high and low stability through the capillary number. Reverse
viscosity liquids. At high viscosity the main roll coating adds an extra degree of freedom, the
failure modes are metered film ribbing and speed ratio, which can be adjusted to control the
cascade, which are just mapped from Fig. 12a.19. coating thickness, flow field, and its stability. At
It should be noted that there is no evidence, pro low capillary number coating thicknesses much
or con, indicating that ribbing on the metered smaller than the minimum gap can be obtained,
film is associated with a defect on the transferred while stable flows can be achieved with proper
References 567
III III
> >
Air Entrainment in Pan ~
Ql
D-
en
'0
lr
...
SIII
~
0.
D-
c(
Starvation
Web Speed Vw Web Speed Vw

(a) (b)
Figure 128.20 Operability diagrams for the two-roll reverse roll coater of Fig. 12a.lc: (a) viscous regime, (b)
inertial regime.
speed ratio adjustment even at high capillary REFERENCES

number. Both of these effects illustrate the
advantages of reverse roll coating over knife Adachi, K., Murata, Y., Morikawa, T. and Nakamura,
coating and forward roll coating. R. 1990. Thin film coating with a knife or blade.
AIChE Spring National Meeting, New Orleans,
March.
Adachi, K., Tamura, T. and Nakamura, R. 1988.
12&.6 SUMMARY AND CONCLUDING Coating flows in a nip region and various critical
phenomena. AIChE J. 34(3): 456-464.
REMARKS Aidun, C. K. 1991a. Principles of hydrodynamic
The above summarizes most of what is known instability: application in coating systems Part 1:
about the fluid mechanics of knife and roll Background. Tappi J. 74(2): 213-219.
Aidun, C. K. 1991b. Principles of hydrodynamic
coating. There are many gaps in our understand-
instability: application in coating systems Part 2:
ing. The phenomena ofinertially driven instability Examples of flow instability. Tappi J. 74(3): 213-220.
in low-viscosity systems has not been studied, Aidun, C, K., Triantafillopoulos, N. G. and Benson,
nor has the sensitivity of these coating devices J. D. 1991. Global stability of a lid-driven cavity
to perturbations. The subject of dynamic wetting with throughflow: Flow visualization studies. Phys.
and air entrainment in these coaters is not Fluids A. 3(9): 2081-2091.
documented. Doctoring in reverse roll coating Babchin, A. J., Clish, R. J. and Wahren, D. 1981.
Stability and disturbance of coating films. Adv. Coli.
needs further understanding. The dynamics and
Int. Sci. 14: 251-280.
stability of the rolling bank in forward roll Banks, W. H. and Mill, C. C. 1954. Some observations
coating is not understood. Perhaps the least on the behavior of liquids between rotating rollers.
understood and most practically significant Proc. Royal Soc. A. 223: 414-419.
problem in all of these coaters is the effect of Bauman, T. M., Sullivan, T. M. and Middleman, S.
non-Newtonian fluid rheology on their steady 1982. Ribbing instability in coating flows: effect of
operation and stability. These and other specific polymer additives. Chern. Eng. Cornrnun. 14: 35-46.
Benjamin, D. F. 1994. Roll coating flows and multiple
aspects need further research, but we also need
roll systems. PhD thesis, University of Minnesota,
detailed systematic studies of the operating limits Minneapolis.
of specific coating configurations in order to Benjamin, D. F. and Scriven, L. E. 1992. Forward roll
understand how all the building-block flows fit coating: effect of feed condition. AIChE Spring
together and interact. National Meeting, New Orleans, April.
Benjamin, D. F. and Scriven, L. E. 1994. Stability of steady flows, stability, and rheology. PhD thesis,
forward roll coating flows. AIChE Spring National University of Minnesota, Minneapolis.
Meeting, Atlanta, April. Coyle, D. 1. 1992. Roll Coating. In Modern Coating
Benkreira, H., Edwards, M. F. and Wilkinson, W. L. and Drying Technology, eds E. Cohen and E. Gutoff.
1981a. A semi-empirical model of the forward roll New York: VCH Publishers. pp. 63-116.
coating flow of Newtonian fluids. Chem. Eng. Sci. Coyle, D. 1., Macosko, C. W. and Scriven, L. E. 1986.
36: 423-427. Film-splitting flows in forward roll coating. J. Fluid
Benkreira, H., Edwards, M. F. and Wilkinson, W. L. Mech. 171: 183-207.
1981 b. Roll coating of purely viscous liquids. Chem. Coyle, D. J., Macosko, C. W. and Scriven, L. E. 1987.
Eng. Sci. 36: 429-434. Film-splitting flows of shear-thinning liquids in
Benkreira, H., Edwards, M. F. and Wilkinson, W. L. forward roll coating. AIChE J. 33(5): 741-746.
1982a. Ribbing instability in the roll coating ofNewton- Coyle, D. 1., Macosko, C. W. and Scriven, L. E. I990a.
ian fluids. Plastics Rubber Proc. Appl. 2: 137-144. Stability ofsymmetric film-splitting between counter-
Benkreira, H., Edwards, M. F. and Wilkinson, W. L. rotating cylinders. J. Fluid Mech. 216: 437-458.
1982b. Mathematical modelling of the reverse and Coyle, D. 1., Macosko, C. W. and Scriven, L. E. 1990b.
metering roll coating flow of Newtonian fluids. The fluid dynamics of reverse roll coating. AIChE
Chem. Eng. Sci. 37(2): 277-282. J. 36(2): 161-174.
Bixler, N. E. 1982. Stability of a coating flow. PhD Coyle, D. J., Macosko, C. W. and Scriven, L. E. 1990c.
thesis, University of Minnesota, Minneapolis. A simple model of reverse roll coating. Ind. Eng.
Bolton, B. and Middleman, S. 1980. Air entrainment Chem. Res. 29: 1416-1419.
in a roll coating system. Chem. Eng. Sci. 35(3): Coyle, D. 1., Macosko, C. W. and Scriven, L. E. 1990d.
597-601. Reverse roll coating of non-Newtonian liquids. J.
Booth, G. L. 1970. Coating Equipment and Processes. Rheol. 34(5): 615-636.
New York: Lockwood Publishing Company. Coyne, J. C. and Elrod, H. G. I970a. Conditions for
Booth, G. L. 1990. Coating converting processes: the rupture of a lubricating film. Part 1: Theoretical
Traditional methods and new uses. 21st Annual model. J. Lubric. Technol. 92: 451-456.
IGC Coating Systems Conference, Boston, August, Coyne, J. C. and Elrod, H. G. 1970b. Conditions for
Bis-Cap International, Newtonville, MA. the rupture of a lubricating film. Part 2: New
Broughton, G., Egan, L. W. and Sturken, C. 1950. The boundary conditions for the Reynolds equation. J.
reverse roll principle of coating. Tappi J. 33(7): Lubric. Technol. 93: 156-167.
314-317. DiCarlo, M. J. 1986. Reverse roll coating - experiments
Cameron, A. 1976. Basic Lubrication Theory. New and theory. MS thesis, University of Minnesota,
York: Wiley. Minneapolis.
Carter, G. C. and Savage, M. D. 1987. Ribbing in a Doremus, P. and Piau, J.-M. 1981. Viscoelastic
variable speed two-roll coater. Math. Eng. Ind. 1(1): elongational lubrication. J. Non-Newtonian Fluid
83-95. Mech. 9: 389-400.
Chalmers, B. and Hoare, W. E. 1941. J. Iron St. Inst. Doremus, P. and Piau, J.-M. 1983. Experimental study
144: 127. of viscoelastic effects in a cylinder-plane lubricated
Chandrasekhar, S. 1961. Hydrodynamic and contact. J. Non-Newtonian Fluid Mech. 13: 79-91.
Hydromagnetic Stability. Oxford: Oxford University Dowson, D. and Higginson, G. R. 1966. Elasto-
Press. Hydrodynamic Lubrication. Oxford: Pergamon Press.
Chen, K. and Higgins, B. G. 1988. Study of the flow Dowson, D., Smith, E. H. and Taylor, C. M. 1980. An
in the upstream bank of liquid in a forward roll experimental study of hydrodynamic film rupture
coater by the finite element method. Chem. Eng. Sci. in steadily-loaded, non-conformal contact. J. Mech.
43(10): 2867-2875. Eng. Sci. 22(2): 71-78.
Couder, Y, Michalland, S., Rabaud, M. and Thome, Dowson, D. and Taylor, C. M. 1979. Cavitation in
H. 1989. The printer's instability: The dynamical bearings. Ann. Rev. Fluid Mech. II: 35-66.
regimes ofdirectional viscous fingering. In Nonlinear Fall, C. 1978. Surface ribbing of a thin viscous fluid
Evolution ojSpatio-temporal Structures in Dissipative film emerging from a spreader or roller. J. Lubric.
Continuous Systems, eds F. H. Busse and L. Kramer. Technol. 100: 462-466.
Proceedings of the NATO Advanced Research Fall, C. 1982. A theoretical model of striated film-
Workshop, Streitberg, September. rupture. J. Lubric. Technol. 104: 164-167.
Coyle, D. J. 1984. The fluid mechanics of roll coating: Fall, C. 1985. A theoretical model of striated film-
References 569
rupture applied to the cylinder-plane. J. Tribol. 107: Greener, J. and Middleman, S. 1979. Theoretical and
419-422. experimental studies of the fluid dynamics of a
Fernando, R. H. and Glass, J. E. 1988. Dynamic two-roll coater. Ind. Eng. Chem. Fundam. 18: 35-41.
uniaxial extensional viscosity (DEUV) effects in roll Greener, J. and Middleman, S. 1981. Reverse roll
application. Part 2. Polymer blend studies. J. Rheol. coating of viscous and viscoelastic liquids. Ind. Eng.
32(2): 199-213. Chem. Fundam. 20: 63-66.
Floberg, L. 1964. Cavitation in lubricating oil films. Greener, J., Sullivan, T., Turner, B. and Middleman,
In Cavitation in Real Liquids, ed. R. Davies. S. 1980. Ribbing instability of a two-roll coater:
Amsterdam: Elsevier Publishing Company. pp. Newtonian fluids. Chem. Eng. Commun. 5: 73-83.
138-146. Gurfinkle Castillo, M. E. and Patera, A. T. 1996.
Fukazawa, K. 1993. Reverse roll coater model. MS Three-dimensional ribbing in forward-roll film
thesis, University of Minnesota, Minneapolis. coating processes. J. Fluid Mech., in press.
Fukushima, M. 1976. Study of an anomalous flow Hakim, V., Rabaud, M., Thome, H. and Couder, Y.
during rolling of viscoelastic fluid. Part I. Experi- 1988. Directional growth in viscous fingering. In
mental. Jap. J. Appl. Phys. 15(3): 525-530. New Trends in Nonlinear Dynamics and Pattern
Gaskell, P. H., Savage, M. D. and Summers, J. L. 1994. Forming Phenomena: The Geometry ofN onequilibrium.
Flow transformation from classical to meniscus roll Proceedings of the NATO Advanced Research
coating. AIChE Spring National Meeting, Atlanta, Workshop, Cargese, August.
April. Hasegawa, T. and Sorimachi, K. 1993. Wavelength
Gatcombe, E. K. 1945. Lubrication characteristics of and depth ofribbing in roll coating and its elimination.
involute spur gears - a theoretical investigation. AIChE J. 39(6): 935-945.
Trans. ASME. 67: 177-188. Higgins, D. G. 1965. Coating methods. In Encyclopedia
Ghannam, M. T. and Esmail, M. N. 1990. Air ofPolym. Sci. Tech. John Wiley, Vol. 3, pp. 765-807.
entrainment and dynamic contact angles in hy- Hintermaier, J. C and White, R. E. 1965. The splitting
drodynamics of liquid coating. Canadian J. Chem. of a water film between rotating rolls. Tappi J.
Eng. 68(2): 197-20l 48(11): 617-625.
Glass, J. E. 1978a. Dynamics of roll spatter and Ho, W. S. and Holland, F. A. 1978. Between-rolls
tracking. Part I: Commercial latex trade paints. J. metering coating technique, a theoretical and
Coatings Technol. 50(640): 53-60. experimental study. Tappi J. 61(2): 53-56.
Glass, 1. E. 1978b. Dynamics of roll spatter and Hoffman, R. D. and Myers, R. R. 1962. The splitting
tracking. Part 2: Formulation effects in experimental of thin liquid films. II. Cavitation dynamics. Trans.
paints. J. Coatings Technol. 50(640): 61-68. Soc. Rheol. 6: 197-207.
Glass, J. E. 1978c. Dynamics of roll spatter and Hopkins, M. R. 1957. Viscous flow between rotating
tracking. Part 3: Importance ofextensional viscosities. cylinders and a sheet moving between them. Brit.
J. Coatings Technol. 50(641): 56-71. J. Appl. Phys. 8: 442-444.
Glass, J. E. 1978d. Dynamics of roll spatter and Hsu, T. C, Malone, M., Laurence, R. L. and Middleman,
tracking. Part 4: Importance of G* recovery and S. 1985. Separating forces in blade coating of viscous
NI in tracking. J. Coatings Technol. 50(641): 72-78. and viscoelastic liquids. J. Non-Newtonian Fluid
Gokhale, V. V. 1981. Bounds for region of ribbing Mech. 18: 273-294.
instability in some free surface coating flows. J. Johnson, K. L. 1985. Contact Mechanics. Cambridge
Rheol. 25(4): 421-432. University Press.
Gokhale, V. V. 1983a. Improved stability criterion for Kang, Y.-T., Lee, K. Y. and Liu, T.-J. 1991. The effect
lubrication flow between counterrotating rollers. of polymer additives on the performance of a
AIChE J. 29(5): 865-866. two-roll coater. J. Appl. Polymer Sci. 43: 1187-1195.
Gokhale, V. V. 1983b. Exact solution to the ribbing Kang, Y.-T. and Liu, T.-J. 1991. Minimum film
instability problem in lubrication flow by invariant thickness for reverse roll coating. Chem. Eng. Sci.
imbedding. Chem. Eng. Commun. 21: 81-87. 46(11): 2958-2960.
Greener, J. and Middleman, S. 1974. Blade coating Kheshgi, H. S., Kistler, S. F. and Scriven, L. E. 1992.
of a viscoelastic fluid. Polymer Eng. Sci. 14: Rising and falling film flows: viewed from a first-
791-796. order approximation. Chem. Eng. Sci. 47(3):
Greener, 1. and Middleman, S. 1975. A theory of roll 683-694.
coating of viscous and viscoelastic fluids. Polymer Kistler, S. F. 1994. Personal communication.
Eng. Sci. 15(1): 1-10. Kistler, S. F. and Scriven, L. E. 1983. Coating flows.
In Computational Analysis ofPolymer Processing, eds Saffman, P. G. and Taylor, G. I. 1958. The penetration
J. R. A. Pearson and S. M. Richardson. London of a fluid into a porous medium or Hele-Shaw cell
and New York: Applied Science Publishers, pp. containing a more viscous liquid. Proc. R. Soc. Lond.
243-299. A. 245: 312-329.
Kistler, S. F. and Scriven, L. E. 1984. Coating flow Satas, D. (ed.) 1984. Web Processing and Converting
theory by finite element and asymptotic analysis of Technology and Equipment. New York: Van Nostrand
the Navier-Stokes system. Int. J. Num. Meth. Reinhold.
Fluids. 4: 207. Savage, M. D. 1977a. Cavitation in lubrication. Part
Martin, H. M. 1916. The lubrication of gear teeth. 1. On boundary conditions and cavity-fluid interfaces.
Engineering. 102: 119-121. J. Fluid Mech. 80: 743-755.
McEwan, A. D. and Taylor, G. I. 1966. The peeling Savage, M. D. 1977b. Cavitation in lubrication. Part
of a flexible strip attached by a viscous adhesive. J. 2. Analysis of wavy interfaces. J. Fluid Mech. 80:
Fluid Mech. 26: 1-15. 757-767.
Mill, C. C. and South, G. R. 1967. Formation of ribs Savage, M. D. 1982. Mathematical models for coating
on rotating rollers. J. Fluid Mech. 28: 523-529. processes. J. Fluid Mech. 117: 443-455.
Miller, 1. C. and Myers, R. R. 1958. A photographic Savage, M. D. 1984. Mathematical model for the onset
study of liquid flow in a roll nip. Trans. Soc. Rheol. of ribbing. AIChE J. 30(6): 999-1002.
2: 77-93, Savage, M. D. 1991. Meniscus instability and ribbing.
Myers, R. R. and Hoffman, R. D. 1961. The distribution Industrial Coating Research. 2: 47-58.
of pressures in the roll application of Newtonian Savage, M. D. and Gaskell, P. 1990. Meniscus coating.
fluids. Trans. Soc. Rheo/. 5: 317-328. AIChE Spring National Meeting, Orlando, March.
Myers, R. R., Miller, 1. C. and Zettlemoyer, A. C. 1959. Scharenberg, R. T. 1985. Coating methods. In Ency-
The splitting of thin liquid films. I. Kinematics. J. clopedia of Polym. Sci. Eng. 2nd edn. John Wiley,
Colloid Sci. 14: 287-299. Vol. 3, pp. 552-567.
Pearson, 1. R. A. 1960. The instability of uniform Schneider, G. B. 1962. Analysis of forces causing flow
viscous flow under rollers and spreaders. J. Fluid in roll coaters. Trans. Soc. Rheol. 6: 209-221.
Mech. 7: 481-500. Secor, R. B. 1991. Process modeling ofcontact extrusion
Pearson, 1. R. A. 1985. Mechanics ofPolymer Processing. and coextrusion. Tappi J. 74(11): 168-174.
London: Elsevier Applied Science Publishers. Sone, T. and Fukushima, M. 1960. On the stripe
Pitts, E. and Greiller, 1. 1961. The flow of thin pattern during rolling of viscous materials. J. Phys.
liquid films between rollers. J. Fluid Mech. 11: Soc. Japan. 15: 1708.
33-50. Soules, D. A., Fernando, R. H. and Glass, J. E. 1988.
Rabaud, M. 1990. Personal communication. Dynamic uniaxial extensional viscosity (DUEV)
Rabaud, M. and Hakim, V. 1989. Shape of stationary effects in roll application. Part 1. Rib and web
and travelling cells in the printer's instability. In growth in commercial coating. J. Rheo/. 32(2):
Proceedings of the Third International Workshop on 181-198.
Instabilities and Nonequi/ibrium Structures, eds E. Strenger, M. R., Secor, R. B. and Sramek, R. J. 1992.
Tirapegui and W. Zeller. Kluwer Academic Pub- Knife coating of a viscoelastic material. AIChE
lishers. Spring National Meeting, New Orleans, March.
Rabaud, M., Michalland, S. and Couder, Y. 1990. Sullivan, T. M. and Middleman, S. 1979. Roll coating
Dynamical regimes of directional viscous fingering: in the presence of a fixed constraining boundary.
spatiotemporal chaos and wave propagation. Phys. Chem. Eng. Commun. 3(6): 469-482.
Rev. Let. 64(2): 184-187. Sullivan, T. M. and Middleman, S. 1986. Film thickness
Reynolds, O. 1886. On the theory of lubrication and in blade coating of viscous and viscoelastic liquids.
its application to Mr Beauchamp Tower's experi- J. Non-Newtonian Fluid Mech. 21: 13-38.
ments, including an experimental determination of Sullivan, T. M., Middleman, S. and Keunings, R. 1987.
the viscosity of olive oil. Phi/os. Trans. R. Soc. Lond. Use of a finite-element method to interpret
A. 177: 157-234. rheological effects in blade coating. AIChE J. 33(12):
Ruschak, K. 1. 1982. Boundary conditions at a liquid/ 2047-2056.
air interface in lubrication flows. J. Fluid Mech.119: Taylor, C. M. 1974a. Separation cavitation in lightly
107-120. loaded fluid film bearings with both surfaces in
Ruschak, K. 1. 1985. Coating flows. Ann. Rev. Fluid motion. Part 1. Theoretical considerations. J. Mech.
Mech. 17: 65-89. Eng. Sci. 16(4): 147-150.
References 571
Taylor, C. M. 1974b. Separation cavitation in lightly Weiss, H. L. 1977. Coating and Laminating Machines.
loaded fluid film bearings with both surfaces in Converting Technology Co., Milwaukee.
motion. Part 2. Analysis of the cylinder plane and Williamson, A. S. 1972. The tearing of an adhesive
journal bearing configurations. J. Mech. Eng. Sci. layer between flexible tapes pulled apart. J. Fluid
16(4): 150-155. Mech. 52: 639-656.
Taylor, C. M. 1974c. Film rupture for a lubricated Yih, C. S. 1960. Instability of a rotating liquid film
cylinder lightly loaded against a plane. J. Mech. with a free surface. Proc. Roy. Soc. A. 258: 63-89.
Eng. Sci. 16(4): 225-231. Yih, C. S. 1963. Stability of liquid flow down an
Taylor, G. I. 1963. Cavitation of a viscous fluid in inclined plane. Phys. Fluids. 6(3): 321-334.
narrow passages. J. Fluid Mech. 16: 595-619. Zink, S. C. 1979. Coating processes. In Encyclopedia of
Triantafillopoulos, N. G. and Aidun, C. K. 1990. Chern. Tech. 3rd edn. Vol. 6, John Wiley, pp. 386-426.
Relationship between flow instability in short-dwell
ponds and cross directional coat weight nonunifor-
mities. Tappi J. 73(6): 127-136.
MENISCUS ROLL COATING 12b
Philip H. Gaskell and Michael D. Savage
12b.l INTRODUCTION
effective means for unraveling the mysteries of
The reader may well be wondering 'What is forward mode, meniscus roll-coating.
Meniscus Roll-Coating'? This is not altogether The two regimes of forward roll coating are
surprising because, as far as we are aware, there introduced and contrasted in Section 12b.2. An
are no references to this form of coating in the experimental and theoretical investigation of the
open literature and yet it has been around in flow structure in meniscus roll coating is the
various industries for several decades. Meniscus subject of Section 12b.3 which concludes with
roll-coating is used to produce high quality, the identification of three critical flow rates.
ultra-thin films at line speeds of the order of tens These correspond to three key events which arise
of meters per minute and is characterized by two as the feed condition is varied. Section 12bA
key features: identifies the part of parameter space corre-
sponding to the meniscus roll coating regime
• coating is in the low capillary number range
while Section 12b.5 discusses various instabilities
where capillarity plays a dominant role in
to the coating bead.
determining the pressure field;
• inlets are starved; the incoming film is much
thinner than the roll separation. 12b.2 FORWARD ROLL COATING
One reason for the lack of references in the In forward roll coating two regimes of practical
literature is perhaps a tacit assumption by interest can be identified according to their inlet
practitioners and theoreticians alike that this feed condition; these are the classical (inlet-
coating process was essentially the same as the flooded) and the meniscus (inlet-starved) coating
more familiar inlet-flooded, coating process. In regimes. They are shown in Fig. 12b.l for forward
fact this assumption falls quite wide of the mark; roll coating where the rolls are contra-rotating
the two are, in reality if not appearance, quite and therefore move in the same direction through
different with their own characteristic pressure the nip. The lower roll is partially submerged in
and velocity fields. a bath and fluid is transferred to the nip by the
This chapter aims to review the results of a action of viscous lifting, prior to being deposited
five year investigation into forward mode, meniscus on a moving substrate. The meniscus coating
roll-coating. A tripartite scientific approach has bead is formed in one of two ways:
been used combining experimental methods,
mathematical modeling and computational simu- 1. beginning with a small minimum gap and a
lations (Gaskell et al. 1995b,1996a). This combi- flooded inlet, a starved inlet condition is
nation has indeed proved a very powerful and produced by increasing the roll separation;
574 Meniscus roll coating
and V I ' V 2 are the speeds of the upper and lower
rolls. A thin fluid stream can be seen snaking
(a) between and around a pair of eddies occupying
most of the flow region - the relative size of the
eddies being determined by the speed ratio. This
flow pattern contrasts sharply with that typical
of the inlet-flooded regime, where the flow is
essentially unidirectional ahead ofthe downstream
recirculation region, see Fig. 12b.3.
Two nondimensional parameters which help
to distinguish these coating regimes are capillary
number, Ca, and nondimensional flow rate, q,
defined as follows:
Ca=J1.Vrn , q=_Q- (12b.1)

(b) ( - - - - - - - - - - - - - - - - - : : : ; : : o i l
inlet-starved
(J 2HOVrn
where J1. and (J represent fluid viscosity and

surface tension; Q is the actual flow rate per unit
axial length, 2H 0 the minimum gap separation
r--<::=----------- and Vrn = (VI + V 2 )/2 is the average roll speed.
~
---=-- inlet-flooded ~ When inlets are fully flooded q has its maximum
_======:::: value, qrnax' which depends on speed ratio and
- ~ capillary number. In particular, qrnax varies from
1.33 to 1.29 as capillary number is increased from
10- 2 to 00 for equal roll speeds (Coyle et al.
Figure 12b.l (a) A two-roll coater operating in 1986). Classical roll-coating with flooded inlets
forward mode; (b) exploded view of the nip region
is a self-metered coating process in which the
showing a flooded and starved inlet.
geometry and roll speed determine the flow rate,
Q - which is accordingly nondimensionalized by
2. beginning with a suitable gap and a flooded means of the average roll speed, Vrn' and the gap
inlet the incoming flux is decreased by reducing thickness, 2Ho. In meniscus roll coating neither
the speed of the lower roll. the gap nor the upper roll speed play any part
in determining the flow rate Q as all the fluid
A key feature of inlet-flooded roll coating is approaching the inlet is transferred through the
the location of the upstream meniscus - far from nip. Nevertheless it may still be regarded as a
the nip - where it has negligible influence on self-metered process as the coating thickness
both the flow and the pressure field. In addition, applied to the web is determined by the roll
fluid pressures in the nip region are much greater speeds. The flow rate Q is given by:
than those due to capillary action at the menisci.
In the starved inlet regime the upstream meniscus
(12b.2)
is seen to be much closer to the nip where
capillary action plays the crucial role in generating
a sub-ambient pressure field. Characteristic flow where Hi is the thickness of the inlet film attached
patterns are shown in Fig. 12b.2a and 2b for to the lower roll. For meniscus roll coating
speedratiosS p = 1andS p = 2, whereSp = VdV 2 therefore, it is useful to introduce a new dimen-
.
"l
VI
]V
(a)
- V2 •
Figure 12b.2 Streamline flows in the bead of a mensicus roll coater with R,jH 0 = 125 and rolls moving left
to right: (a) Ca = 10- 2 , Sp = 1; (b) Ca = 1.5 x 10- 2 , Sp = 2.
Figure 12b.3 Acharacteristic flow for inlet flooded roll coating with rolls moving left to right and Rm/HO = 110,
Ca = 0.2, Sp = 2.
sionless flow rate, qm, defined by: experimental. Its aim was to identify the char-
acteristic velocity and pressure fields and also to
Hi
(12b.3) measure various quantities such as film thicknesses
2Ho and free surface locations for comparison with
where qm provides a measure of the degree of theoretical predictions. The work involved a
inlet starvation. Typically qm - 10 -1 and is combination of state-of-the-art high speed video
related to q via the equation: photography, image processing and computerized
C: sJ
particle tracking software allied to both dye
q= qm (12b.4) injection and particle seeding of the flow .- two
of the oldest methods for determining the motion
Table 12b.l presents a summary of the two ofa fluid known to experimental fluid dynamicists.
coating regimes with their associated inlet This marriage of old and new was not easily
conditions and flow rates. arranged and a number of major hurdles had to
be overcome - not least the sheer practical
difficulties of making observations and obtaining
12b.3 FLOW STRUCTURE
physical measurements of such a small scale flow
confined within a restricted geometry comprised
12b.3.l EXPERIMENTAL INVESTIGATION
of two stationary free surfaces and two moving
The first investigation of the fluid dynamics of solid boundaries.
meniscus roll coating (Malone 1992) was purely An important feature of the flow visualization
Flow structure 577
Table 12b.l Coating regimes in a forward roll coater
Coating regime Feed/Inlet condition Flow rate
Classical Inlet-flooded Maximum flow rate,

q = qmax
Meniscus Inlet-starved/Ultra-starved Substantially reduced flow rate,
qm = (1 + Sp)q/2 « 1
techniques employed is their ability to provide look along the axis at an illuminated cross-section
information about the complete flow field without of the flow.
physical intrusion. In addition to giving a The generation of meaningful, rigorous experi-
qualitative picture ofthe associated fluid dynamics mental results required the design and development
involved, quantitative information may be derived of specialized equipment and procedures. Figure
by analyzing recorded images of the flow. In 12b.4a shows the cantilevered roll system adopted
order to view the fluid motion in the very small for the study. Its major advantage is ease of
region between the two rolls it was necessary to accessibility to the nip region while retaining the
MICROMETER PLATFORM
MICROMETER (RIGIDLY LOCATED INTO
LOWER HOUSING)
LOWER ROLL SHAFT HOUSING (FIXED)
UPPER ROLL SHAFT HOUSING (VERTICAL

MOVEMENT BETWEEN SPRINGS AND
MICROMETER SHAFT)
(a> (b)
Figure 12b.4 (a) Photograph of the experimental setup showing a cantilevered roll system; (b) schematic of
roll shaft housing showing micrometer assembly.
essential features of an industrial process in the reduction gearbox, pulley and toothed rubber
meniscus coating regime. belt. The latter requires very little tension during
For the purposes of flow visualization operation and therefore loading on each roll
several lower/upper roll configurations could be shaft is negligible. Vibration from the drive
achieved: steel/steel; steel/acrylic; acrylic/hollow system was also minimal.
acrylic. The steel rolls were manufactured from The lower roll bearing/shaft housing is bolted
stainless steel, ground and polished to achieve a to an aluminum base plate while four steel
high quality surface finish. The use of acrylic rolls columns are press-fitted into the lower housing
allowed the nip region to be illuminated more onto which the upper housing is located with a
effectively. Their surfaces were polished (including precision slide fit (see Fig. 12b.4b). The control
the inner surface in the case of the hollow roll) of the proximity of the upper and lower housings
to obtain a glass-like finish. is achieved by a micrometer assembly - the
Steel support shafts of 25 mm diameter were precision slide fit on the columns ensures that
ground on the roll location diameter and the the movement of the upper housing remains
rolls bored to obtain a precision H6 fit (BS 4500) parallel. The bearing end-caps are used to adjust
in order to eliminate effectively roll runout. independently the axial position of the shafts to
Similarly, all roll diameters were finished on their facilitate alignment. In addition they are used to
respective support shafts to ensure concentricity retain and pre-load the bearings via the shaft
- the maximum indicated roll run-out was found collars which reduces any internal clearance and
to be less than 5 ~m. The rolls can be detached takes up any play in the shaft, either axially or
from the steel support shafts allowing ease of roll radially. Also, as the end caps are secured using
interchange - each shaft is supported in a housing four diagonally positioned screws it is possible
by two high precision roller bearings, separated to make precise independent adjustments to
by a suitable distance (100 mm) so as to minimize ensure that the roll ends are aligned.
bending and vibration of the shafts. Although the separation of the rolls can be
The rolls are of sufficient diameter (50 mm) to adjusted accurately (to within 1 ~m) using the
give the radius of curvature required to produce micrometer arrangement outlined above, the
a reasonably high aspect ratio of bead width to actual gap setting relies on accurate calibration
minimum gap thickness over a wide range of of this device. This was done using specialist
operating conditions and enable clear images of plastic wrapping film which is available in
the flow within the nip to be obtained. Three thicknesses as low as 6 ~m. The film was pulled
factors were considered in choosing the length between the rolls as the gap size was decreased
of the rolls: until resistance was encountered, in much the
same way as is experienced with a feeler gauge.
1. the optimum viewing depth for flow visualiz-
However, unlike the latter, film has the advantage
ation;
that there is less of a problem in ensuring
2. accurate representation ofthe industrial process;
horizontal alignment. Error in setting the gap
3. moderate roll weight to minimize deflection
size was estimated to be in the region of 10 ~m
and vibration at high speeds.
(e.g. 3.5%, for a typical gap setting of 350 ~m).
The chosen length of 35 mm gave a more than The viscosity of fluids used in industrial
adequate bead length to gap ratio (e.g. 100, for applications of meniscus roll coating is relatively
a typical gap setting of 350 ~m) as well as a very low, i.e. in the region of 10- 3 Pa s when compared
compact design. with fluids used in classical roll coating where
The rolls are driven and controlled indepen- viscosity ranges from 5 x 10- 2 to 5OOPas. This
dently for maximum versatility. Each motor has posed a problem in the sense that a compromise
a control unit for continuous speed variation and has to be reached between choosing a fluid with
is connected to the roll shaft via a 27: 1 worm a viscosity representative of the industrial process
Flow structure 579
and one which will enable the use of a minimum region of interest. A useful variant to continuous
roll separation large enough for good quality dye injection was to 'pulse' the injection sequence.
images of the flow within the bead. For particle tracking, a number of different
Silicone oils meet the above requirements; materials were examined but only one, Sephasord,
viscosity ranges from 0.65 x 10- 3 to a material used for gel filtration was found to
5.0 x 10- 3 Pa s upwards. These oils are Newtonian meet the necessary criteria of sphericity, opacity,
at shear rates less than 104 s - 1 and optically neutral buoyancy, mono-dispersiveness and size
clear. A less expensive, alternative used for this (5 11m), and produce consistent results when
study, was Shell Tellus R5 oil (a light organic suitably illuminated.
oil). The fluid properties of these oils are given In order to achieve high quality flow visual-
in Table 12b.2. ization the light sensitivity of the system is of
The action of the web in removing fluid from prime importance because of the difficulties
the nip is mimicked by the use of a scraper blade encountered in illuminating a two-dimensional
mechanism. The scraper material has to be rigid plane in the constricted area of a two-roll coater.
enough not to deform under its own weight but The dye injection studies were performed using
flexible enough to be clamped up against the roll a combination of acrylic rolls together with
for the efficient removal of fluid. Both plastic powerful white light sources (two 250 W studio
(polyester) and metal (brass shim, thickness 0.3 mm) lamps), one positioned each side of the nip.
scrapers were assessed and although they both However for the particle tracking work the white
performed well the latter was found to remove light sources had to be replaced by low power
fluid more effectively. With a metal scraper it is (10mW) lasers. Also, to obtain consecutive images
easier to achieve a uniform knife-edge finish in which the movement of particles from one
along its length whereas plastic tends to score frame to the next is sufficiently small so as to be
or become fibrous. The efficiency of scraping was less than half the mean distance between particles
determined using filter paper to absorb any distributed in the flow field required a high
excess fluid left on the roll. The paper was recording frame rate, typically 250 to 1000 full
weighed after being in contact with the roll for frames per second. A Kodak Ektapro 1000 high
measured periods of time and fluid removal for speed video system and imaged intensified camera
the metal scraper was found to be between 95 proved more than adequate in this respect.
and 98% efficient. Obtaining qualitative discrete measurements
The use of an oil soluble, powder dye namely of velocities within complex flows presents a
grasol blue (solvent blue 27) dispensed with the challenging and difficult problem to the experi-
need to match the properties of the liquid dye mentalist. The imaging technique employed was
to those of the working fluid. The dissolved dye developed in close collaboration with R. J. Perkins
was injected via a hypodermic needle into the (1991), specifically for use with high speed video
incoming fluid film, far upstream of the nip so equipment. The software behaves in effect as two
as not to influence the flow in the cross-sectional separate programs - the first analyzes a digitized
Table 12b.2 Fluid properties at 20°C
Viscosity Surface tension Density

Liquid (Pas) (Nm- 1) (kgm- 3 )
R5 oil 8.4 x 10- 3 30.4 x 10- 3 839

Silicone oil 200/0.65 0.72 x 10- 3 16.1 X 10- 3 761
Silicone oil 200/1 1.12xlO- 3 17.5 x 10- 3 818
Silicone oil 200/5 5.45 x 10- 3 19.7 X 10- 3 920
image and identifies and locates all the particles, where P is the pressure and the velocity field
the second uses this information to match particles U = (U, W) is assumed to be in the two-dimen-
frame-by-frame. sional (X,Z) plane. The penalty for neglecting
Figures 12b.5a-5d illustrate the path taken the curvature of the menisci is that the normal
by the injected dye as it progresses through the stress condition on each meniscus has to be
flow field (the rolls are moving left to right with relaxed. Mathematically, the model reduces to
equal speed). The dye enters on the lower roll the solution of the biharmonic equation for the
(Fig. 5a) and moves with the roll surface until it streamfunction '1':
splits in the bottom right-hand corner ofthe bead
(Fig. 5b); some of the dye continues along the (12b.6)
lower roll to form the outlet film thickness, the
subject to appropriate boundary conditions, as
remainder makes a 'V-turn' to travel in the
shown in Fig. 12b.7 where the bead, of length
reverse direction through the flow field (Fig. 5c).
2A and depth 2H, is assumed to be rectangular
Once the dye reaches the upstream free surface
(-A ~ X ~ A, -H ~ Z ~ H). This boundary
(by now it has traveled two bead lengths) it
value problem is similar to the lid-driven cavity
experiences a second 'V-turn' and then moves
flows studied by Joseph (1977) and Canedo and
with the upper roll, eventually leaving the bead
Denson (1989). A solution for the streamfunction
in the fluid layer attached to the upper roll.
is derived analytically in the form of a truncated
A typical associated velocity field, generated
series ofeigenfunctions, and also computationally
by particle tracking, is illustrated in Fig. 12b.6.
using a streamfunction-vorticity, finite-element
It shows quite clearly the presence of the reverse
technique (Gaskell et al. 1991).
jet along the centre of the bead. The velocity of
Results show (Thompson 1992) that the
this jet can be measured and in the case of
streamline patterns obtained, analytically and
symmetric rolling is found to be approximately
computationally, are in agreement with each
half that of the roll speeds. This is in good
other and in qualitative agreement with the
agreement with the theoretical and computational
experimentally observed streamline flow (Fig.
results of a meniscus roll coating model outlined
12b.5) for various speed ratios. An important
in the next section.
result from this zero flux model concerns the
horizontal pressure gradient dP/dX which has
12b.3.2 MATHEMATICAL MODELS-
negligible variation with respect to the cross film
ANALYTICAL AND NUMERICAL SOLUTIONS
coordinate, Z - indicating that fluid pressure
Three mathematical models have been developed depends only upon the X coordinate i.e. P = P(X).
for analyzing and understanding meniscus (inlet- Also, in the central core of the bead, sufficiently
starved) roll coating; two are solved analytically far away from either meniscus, the pressure
and one numerically. In the first, the 'zero-flux' gradient, dP/dX, is effectively constant:
model, an idealized situation is considered in
which flux into and out of the flow domain, or dP
dX = constant (12b.7)
bead, is neglected, i.e. q = O. The cross-section
of the bead is assumed to be rectangular with
- a result which suggests a linear profile for
the curvature of the rolls neglected and the
menisci modeled as planes on which shear stress pressure, quite different from the distribution in
is zero. The governing equations are Stokes' the inlet-flooded regime with its characteristic
equations for an incompressible fluid: maximum and minimum, see Coyle (1996).
A more refined mathematical model is the
0=VP-ttV2U 'small flux' model which takes into account the
(12b.5) small non-zero flux q transported through the
O=V.U bead. It makes use of results from the 'zero flux'
Flow structure 581
~.:-
<.' '
, "
''':~~.
,... ;. '.
(a)
(b)
(e)
(d)
Figure 12b.5(a)-(d) Dye entering into the bead of a meniscus roll coater identifies the 'snake' - the path
followed by the inlet fluid which is transferred to the upper roll (rolls moving left to right with Rm/H O = 125,
Ca = 10- 2 , Sp = 1).
~~ --. _ _
....... .......
~~-
~...
~~ ~~ ... +~.....
~1--"-"
~+- ~-+-~
-++
~~~.-~ ~-- ~""'+...- .... ---
Figure 12b.6 The velocity field in the bead of a meniscus roll coater generated by computerized particle
tracking methods (rolls moving left to right with Rm/H O = 125, Ca = 10- 2 , Sp = 1).
where Rm is an average roll radius defined by:

'V=o
2 1 1
-=-+- (12b.l0)
'V= 0 'V=O Rm Rl Rz
V''V = a
'If" = a 'If" =a and R l , R z represent the radii of the upper and
lower rolls, respectively. The velocity distribution
is quadratic in Z and U(X,Z) = 0 has two roots
in - H(X) ~ Z ~ H(X) for all speed ratios.
Hence there are two positions across the gap at
which the flow is stationary. It can therefore
Figure 12b.7 Boundary conditions for Stokes flow describe the unidirectional flow in the central
in an idealized rectangular bead with zero flux, core of the bead consisting of a fluid transfer jet
bounded by two free surfaces and two moving plates. snaking around a pair of eddies. This jet is the
mechanism for transferring fluid to the upper
roll in order to coat the web and, as such, is
model that the fluid flow in the central core is similar to the transfer jet operating in the upstream
essentially one-dimensional with fluid pressure recirculation region of an inlet flooded roll
constant across the bead. It is precisely these coater, Pearson (1985).
results which indicate that lubrication theory will At this point it is convenient to introduce the
model the flow in the central core with the following nondimensional variables:
velocity distribution U(X,Z) expressed as a
combination of Poiseuille and Couette flow:
1
U(X,Z) = -2 dP (ZZ _ HZ)
J.ldX
(U l - Uz ) Z (U z + U l )
+ 2 H(X) + 2 Figure 12b.8 shows a schematic for flow in a
bead extending from x = - b to x = d. Fluid
(12b.8) approaches the bead in a uniform layer of
H(X), the semi-gap width at station X, is given by: thickness hi and exits by way of two uniform
layers h l and h z attached to the web and lower
XZ roll, respectively. Continuity of flux requires that
H(X) = H o + -2- (12b.9)
Rm U2 H j = U2 H 2 + UlH p whichinnondimensional
Flow structure 583
z There are a number of observations to be made.
1. Equation (12b.15) identifies (CaJRmIH o) as

the important nondimensional parameter in
1...- -+ x meniscus roll coating. In order to obtain
solutions of (12b.15) this parameter must be
of order 1, in order to ensure that the left-hand
I
i
side balances the dominant term on the right-
i hand side, which is of order 1. Therefore:
i
i
i
i
i (12b.16)
x =-b x=d
Figure 12b.8 Schematic of the bead with small flux
2. Once J H 01R m , Ca and q are specified, equation
q and bounded by two menisci and two moving rolls.
(12b.15) gives a relation between the angles
'l b' 'ld which locate the positions of the free
form becomes: surfaces. If 'ld is known then 'l b can be
hi = h2 + Sph 1 (12b.ll) determined. In fact 'ld can be obtained via
As flow rate is constant, equation (12b.8) yields results from Landau and Levich (1942) which
Reynolds' equation: enable the thickness of the uniform film on
each roll to be related to the radius ofcurvature
dp 3 [ q ] of the downstream meniscus:
dx = (1 + X2)2 1 - (1 + x2) (12b.12)
which is solved for pressure by introducing the h = 1.34 Rd (J1.U 1 )2/3 (12b.17)
change of variable x = tan'l: 1 Ho a
p('l) = 3 f(cos 2 'l - qcos 4 'l)d'l (12b.13)

h2 = 1.34 Rd (J1.U2)2/3 (12b.18)
If R b and R d are the radii of curvature of the Ho a
upstream and downstream menisci then the
corresponding fluid pressures are - a1R b and Taking for example UI = V 2 = U, Ca = J1.Ula
-a/R d , respectively, where a is surface tension. andq = ~,equations(12b.14)and(12b.18)give:
It is clear from Fig. 12b.8 that suitable
approximations for Rd and R b are the semi-gap
width and the full gap width, respectively:
R d = H o(1 + d2 ); R b = 2Ho(l + b2 ) (12b.t4)
Clearly, 'ld is known once qm and Ca are
By writing b = tan'l b, d = tan'l d and imposing
specified.
(12b.14), equation (12b.13) yields:
The case VI #- V 2 is a natural extension of
../2(
3 Ca )-1 (COS 2 'lb - 2COS 2 'ld)
the above.
3. Figure 12b.9 shows a typical pressure dis-
JHolR m
tribution for HolR m = 1O- 2 ,Ca = 1.3 x 10- 3 ,
= (sin2'l b + 2'l b + sin2'ld + 2'ld) UI = U2 , b = tan 'l b = 1.55 and d = tan 'ld =
sin 4'lb 3'lb . 2 ) 3.5. The pressure field is entirely subambient
-q ( - 8 - + T + sm 'l b as a result of capillary action at the menisci.
The sharp reduction in pressure over the
sin 4'ld 3'ld . 2 ) upstream meniscus is followed by a linear rise
- q ( -8- + T + sm 'ld (12b.15)
through the central core (Le. a constant pressure
starved inlets and small flow rates, the validity
0.0
of expression (12b.20) needs to be established
experimentally and also computationally using
·2.5 a more refined mathematical model.
cl';;;J"- -5.0 Such a refined model incorporating the effects

::t: ::1. of curved menisci, wetting lines and nonlinear
l::CI~ -7.5 boundary conditions is shown in Fig. 12b.10.
0... Numerical solutions were found to this nonlinear
II
~ -10.0
boundary value problem using a Galerkin,
weighted residual, finite-element, formulation of
the governing equations together with a spinal
-12.5
representation for the free-surface, as defined by
Kistler and Scriven (1983). The region of interest
-15.0-M---r-T---r-TTTTTTTT"T"TTrrrrrrr T lrn--,,-,-,
0.5 1.0 1.5
was tessellated into 1615 triangular V6/P3 elements
-1.5 ·1.0 -0.5 0.0
a (see Fig. 12b.ll) which are taken to satisfy the
LBB stability condition (Babuska and Aziz 1972),
Figure 12b.9 A nondimensional pressure profile for
discussed more fully in Chapter 11. With the
Rm/H O = 100, Sp = 1, Ca = 1.3 x 10- 3 , q = 0.Q75, pressure interpolation one order lower than that
b = 1.55 and d = 3.5 as predicted via the small flux for velocity, the rate of convergence is 'optimal'
model. and no 'locking' (Hughes 1987) occurs. The
reader is referred elsewhere (Gaskell et al. 1995a),
gradient as suggested by the zero flux model) for a fuller description ofthe computational details.
and then a more gradual rise to the downstream Finite-element solutions were obtained for a
meniscus. two roll coater with Ho/R m = 10- 2, flow rates
This subambient pressure field was verified in the range 0.075 < qm < 0.333 and an apparent
by Malone (1992) who took pressure measure- contact angle of 90· at the dynamic wetting line.
ments in a plate-roll configuration. The plate- There are two points to be noted regarding the
roll set-up may be regarded as a two-roll apparent contact angle employed. The value of
system in which the upper roll is stationary 90· was suggested from experiment whilst com-
and of infinite radius. putationally finite-element solutions were ob-
4. Of particular interest to the coating specialist
are the film thicknesses hi and h 2 attached to
n.Q; = 0
the upper and lower rolls. Equations (12b.17)
and (12b.18) suggests a two-thirds power law slip
for film thickness ratio: length~
imposed
hi = S2/3 (12b.20)
h p
2
This is the relation derived by Ruschak (1985)

Flux, A., imposed.
for flooded inlets. Similarly Coyle et al. (1986)
1 dt
obtained a 0.65 power law relationship over n.~ = Ca ds ; u.n = 0
a range of speed ratio, 0 < Sp < 10 using a
finite-element analysis, which was found to be Figure 12b.l0 Nonlinear boundary conditions for
in accord with the findings of Benkreira et al. Stokes flow in a bead incorporating the effects of a
(1981). However, for meniscus roll coating, i.e. dynamic wetting line and curved menisci.
Flow structure 585
Figure 12b.ll Tessellation of the flow domain shown in Fig. 12b.l 0 containing 1615 six node triangular elements.
tainable for a contact angle in the range 80· to structure. Three critical flow rates will be identified
100· with negligible change to the overall solution. corresponding to three key developments in the
Figure 12b.12a-12c show streamline plots for flow.
qm = 0.075 and speed ratios Sp = 1/2, 1 and 2
with corresponding capillary numbers Ca =
1.69 x 10- 3 ,1.8 X 10- 3 and 1.05 x 1O-3,respect- 12b.3.3.1 Pressure field transformation (q = 1)
ively. (In Section 12bA it will become clear that, First it is shown that q < 1 is a necessary
for a given flow rate of qm and geometry ratio condition for obtaining pressure profiles, typical
Ho/R m , speed ratio and capillary number are not of meniscus roll coating, Fig. 12b.13, for which
independent parameters). capillarity at the upstream meniscus plays a
Figure 12b.12a-c reveal a 'double eddy' dominant role.
structure with eddy thickness dependent on Sp If the feed condition is such that the inlet is
- in qualitative agreement with experiment and flooded or moderately starved then the pressure
the predictions of the zero flux model. The flux profile will exhibit stationary points where pressure
of fluid in the transfer jet, subsequently referred has either a local maximum or minimum value.
to as the 'snake', is seen to increase with Sp. At a stationary point dp/dx = 0 and therefore
Corresponding subambient pressure fields are this condition together with Reynolds equation,
displayed in Fig. 12b.13 which are in accord with equation (12b.12), predicts stationary points where:
the predictions from the small flux lubrication
model. x2 = q - 1 (12b.21)
Flow rate, qm, is seen to have no effect on film
thickness ratio as illustrated in Fig. 12b.14 for and the nondimensional flux, q, is given by
various speed ratios from Sp = 1/5 to 5/2. Figure equation (12b.1).
12b.15 shows the variation of h 1 /h 2 with Sp for Provided q > 1 stationary points will be
a constant flow rate including a comparison with symmetrically located about the nip. When q < 1,
hdh 2 = S;/3 obtained from the analysis of Landau equation (12b.21) has no real solutions and
and Levich, equation (12b.18). Agreement is therefore no stationary points occur. Hence
remarkably close over the whole speed ratio range. q = I is the critical flow rate at which there is a
smooth transformation in the shape of the pressure
profile to one which exhibits no stationary
12b.3.3 CRITICAL FLOW RATES
points, is everywhere subambient and is dominated
In this section the aim is to vary the feed by capillary action at the menisci. Furthermore
condition and examine its effect on the flow equation (12b.21) implies that the critical flow
0.5
z 0.0
-0.5
-2 -1 a 2 3
(a) x
0.5
z 0.0
-0.5
(b) ·1 a 2
x
0.5
z 0.0
-0.5
(c)
-1 a 2 3
x
Figure 12b.12(a)-(c) Flow structures in the bead for RmlHo = 100 and qm = 0.075; (a)Sp = 1/2, Ca = 1.69 x 10- 3 ;
(b) Sp = 1, Ca = 1.80 x 10- 3 ; (c) Sp = 2, Ca = 1.05 x 10- 3 •
rate, q = 1, is independent of capillary number, and a two roll coater with H 0/R m = 10 - 2 and
speed ratio and geometry ratio, HO/R m • Sp = 1.
The above results, derived from lubrication
theory, have been confirmed computationally
and also by experiment with a plate-roll geometry
12b.3.3.2 Effect of feed condition on velocity field
(Malone 1992). Figure 12b.16 shows the evolution
of the pressure profile (with pressures measured So one may conclude that q < 1 is a necessary
along the centre line) for q = 0.8, 0.9, 1.0, 1.1, 1.2 condition for meniscus roll coating. It is not
and 1.3, obtained using the finite-element method sufficient however because, in practice, inlets are
Flow structure 587
, ,
,, ,, Sp = 0.5 (Ca =0.00169)
~
," ,,
/t
·1000 ,
, ,
-1500 Sp = 1.0 (Ca = 0.00180)

~ ·2000
~I ~'25oo
R, ,
II .3000
c:..
·3500
Sp = 2.0 (Ca =0.00105)
-4000
~,!
-4500 +-r-,-,--r;rr--rr-rr",-"""",..,.-,-,--r;rrrr-rr,.,-.."..,
-2 ·1 o 1 2 3 4
X
Figure 12b.13 Pressure distributions, taken along the centre line, Z = 0, for Rm/Ho = 100, qm = 0.075 and
corresponding to each of the flows in Fig. 12b.12a-c.
2.5
2.0
1.0 -+- --+- - -+-- -+- _3 __ 3 "" * t. 'I:.. __",-_

Sp =1.0
_+ L __L _ J' _ _ .t_ _ J' __ J- __ J- __ -," _ _ <t __ • __ i: __ .. __ to __ io- __ io- __ ..... _
Sp = 0.5
Sp =0.3333
Sp =0.2
0.0 -+-r-r-r-r-r"TT"TT"TT'T'T''''-'''-TT'TT'TT'TT'T'T''rr-r-r-,...,
0.075 0.1 0.125 0.15 0.175 0.2 0.225
qm
Figure 12b.14 The effect of flow rate, qm, on film thickness ratio, hdh 2 , for Rm/H o = 100 and various speed
ratios: + finite element solutions; - - - - - Landau Levich, equation (12b.18).
starved with q « 1. In order to specify that part velocity field are needed as the flow rate is varied.
of parameter space corresponding to the meniscus In fact such results reveal flow transitions between
coating regime, computational solutions for the the different velocity fields and the critical
1()()
4
so
o _~_1:~(~~ . .... =_
3 -so q.l.2(Ca.Q.050) ~~~
I~I';::,"
~::l.-I()()
q=1.I(Ca=O.029) ~
Q.., q=1.0 (Ca=O.024)
II -ISO
C>. q.0.9 (Ca.Q.021)
-200
q.Q.8 (Ca=0.018)
-2SO
-300 -I-rTTTTTTTTl"TTnTTTTTTl"TTTTTTTTTTTTTTTTTTl
-4 -3 -2 -1 0 3 4
o-M-'-r-T""T""T'"T""T'"'T"T"T"T'-'''''''r-T""r-T""T""T'"'T"T"T"T,-,,....,n X
0.0 0.5 1.0 2.0 2.5 3.0
Figure 12b.16 Pressure profiles for Rm/H O = 100,

Sp = 1 and q = 0.8,0.9, 1.0, 1.1, 1.2 and 1.3 illustrating
Figure 12b.15 The variation of hdh 2 with speed ratio the transformation in shape at q = 1.
for Rm/HO = 100 and constant flow rate, qm = 0.075;
+ finite element solutions; - - - - - Landau Levich,
equation (12b.18).
Figure 12b.18b shows this streamline as the
separatrix and the former separatrix/bounding
streamline is now detached from the saddle point
parameters associated with them including two P3 giving rise to a 'mini-snake' or 'back-jet' in
additional critical flow rates. the lower left of the picture. Fluid entering the
Figures 12b.12a-12c illustrate meniscus coating bead is now able to reach the upper roll via either
with an ultra-starved inlet, q = 0.075, and a the large snake or the mini-snake and in the
characteristic snake carrying fluid to the upper latter case it circumnavigates the now isolated,
roll. Each solution exhibits two large eddies, both lower left, eddy and flows between its bounding
of which consist of a bounding streamline pinned streamline and the separatrix.
to a free surface, an outer circulation, a separatrix The above situation marks the emergence of
and saddle point, and two sub-eddies as clearly a back-jet, the intensity of which increases with
illustrated in Fig. 12b.17 for the case Sp = 1/2. flow rate. In fact, as q increases, the flux in the
As q is increased gradually the additional flux back-jet increases while that in the snake
will cause a pinching of the lower eddy such that diminishes, and the lower left eddy is observed
streamlines in its outer circulation including the to contract and eventually disappear, Fig. 12b.19a
bounding streamline are successively drawn into and 19b. Flow in the snake will decrease to zero
the saddle point. Figure 12b.18a shows a flow at a critical flow rate when the dividing streamline
pattern at a critical flow rate, q = 1/9, Sp = I, at separating inlet fluid which reaches the upper
which the bounding streamline is also a separatrix roll from that which remains attached to the
passing through three stagnation points PI' P 2 lower roll - rises and is attracted to the saddle
and P 3 . Also shown is a streamline (part of the point, Fig. 12b.20a and 20b. This causes flow in
snake) located immediately below the separatrix the snake to switch off and so fluid transfer to
which will be the next streamline to be drawn the upper roll is entirely by means of the fully
into the saddle point as q is increased further. developed back-jet. Mathematically it is possible
Flow structure 589
0.5
z 0.0
-0.5
-2 -1 o 2 3
x
Figure 12b.17 An illustration of the important streamlines and features which make up the 'lower eddy' for
Rm/HO = 100, Ca = 2.25 x 10- 3 , qm = 0.025 and Sp = 0.5.
(a)
(b)
Figure 12b.18 The onset of the back-jet for Rm/HO = 100, Ca = 2 x 10- 3 and Sp = 1.0: (a) critical flow rate
ij = 1/9; (b) q = 0.13 the back-jet is established.
z 0
-1
-2 -1 o 2 3 4
(a) x
z0
-1
-2 -1 o 2 3 4
(b) x
Figure 12b.19 The effect of increasing the flux for Rm/H O = 100, Sp = 1.0 showing that flux in the back-jet
increases while that in the snake diminishes: (a) qm = 0.18, Ca = 2 x 10- 3 ; (b) qm = 0.3, Ca = 3.25 x 10- 3 •
to use lubrication theory to predict these two that is, the nondimensional coordinates z 1 and
critical flow rates. Z2 are solutions of a quadratic equation and
functions of Sp and q.
The second step is to calculate the flux, Q, of
12b.3.3.3 Switching on the back-jet (q = q)
fluid flowing between the lower roll and the lower
The first step is to locate the z-coordinates of of the two saddle points (z = z 1):
the two saddle points, Fig. 12b.18a. As they are
f
stagnation points, equations (12b.8) and (12b.12) z'
Q= UdZ (12b.23)
with H(X) ~ H o give: -H
o
(1 - q)(z 2 - 1) + -2(Sp
---1) z + -2 = 0 When conditions are critical and the back-jet is
3 Sp +1 3
switched on the flux Q is identically the total
(12b.22) inlet flux, Q = U2 H i and so equations (12b.8),
Flow structure 591
z 0
-1
-2 -1 o 2 3 4
(a) x
zO
-1
-2 -1 o 2 3 4
(b) x
Figure 12b.20 Switching off the snake when Rm/H o = 100 and Sp = 1: (a) critical flow rate q* = 113,
Ca = 3.25 x 10- 3 ; (b) qm = 0.35, Ca = 3.6 x 10- 3 , fluid transfer to the upper roll is entirely via the back-jet
and the saddle points have now moved apart.
(12b.12) and (12b.23) yield: great deal of torturous algebra the following
solution is obtained:
3
-(l-q)
4
[z~
--z --
3 1
2J + -+-1)(zi--1)
3
-
(Sp -
Sp 1 4
Zl = [3(1 - q)(3q(1 + Sp) - (1 - Sp))
- (1 - Sp)J(1 + Sp)/[3(1 - q)(3q - 1)(1 + Sp)2
(z 1 + 1)
+ 2
=q - (1 - Sp)2J (12b.25)
(12b.24) q
= - =
q
~[3
9
+ 2Sp - 2.jS; + 3Sp
l+S
p
J (12b 26)
.
Equation (12b.24) is a cubic in Zl which has to This critical flow rate, q, is speed ratio dependent,
be solved simultaneously with equation (12b.22) yet independent ofcapillary number and geometry
to produce a solution for both Z 1 and q. Mter a ratio.
Solving equations (12b.27) and (12b.22) will,

0.4 in principle, determine z 1 and the critical flow
rate q*. The problem, however, is that we do not
0.35
have a general expression for h 2 except for the
0.3
particular case Sp = 1 when h2 /[(1 + S)] =
hJ[2(1 + S)] = q/2.
0.25 The case of equal roll speeds is the only one
for which q* can be predicted. The solutions for
E
CT 0.2 (12b.27) and (12b.22) are:
0.15
1
\ Zl = 0 and q* = - (12b.28)
0.1 3
"-
0.05
Computations reveal this critical flow rate to be
0.0
q* = 0.3327, Fig. 12b.20a, showing yet again the
0 3 4 ability of simple lubrication theory to predict
important features accurately.
For unequal roll speeds q* has to be determined
Figure 12b.21 The onset of the back-jet for
computationally and thus far an overall pattern
Rm/HO = 100. Theoretical curve for critical flow rate has not emerged.
it as a function of Sp: + indicate data points obtained
from finite-element solutions; the curve is that predicted
via lubrication theory, equation (12b.26). 12b.4 THE MENISCUS ROLL COATING
REGIME
Figure 12b.21 shows a plot of q against Sp as In summary, meniscus roll coating is an inlet
given by equation (12b.26) together with a number starved process, q « 1, for which it is useful to
of data points for the onset of the back-jet distinguish between two degrees of starvation
obtained computationally. according to the fluid transfer mechanism. If the
inlet is ultra-starved (q < q), fluid is transferred
only via the snake, whereas if the inlet is starved
12b.3.3.4 Switching off the snake (q = q*)
(q < q « 1) fluid is transferred by both the snake
Once again the starting point is to consider the and the back-jet. This distinction is summarized
flux, Q, flowing between the lower roll and the in Table 12b.3.
lower of the two saddle points (z = z d, as given The remaining problem is to determine that
by equation (12b.23). When conditions become part of parameter space, i.e. the range of capillary
critical, the snake is switched off and the flux Q numbers and geometry ratios, in which meniscus
is carried away on the lower roll. Therefore, roll coating can arise.
Q = U2 H 2 , and equation (12b.23) along with
(12b.8) and (12b.12) give:
Table 12b.3 The meniscus coating regime
~(1 - q*) (Z3~ - z 1 - D_~: ~ ~ (zi ; 1) Fluid transfer
1)
Inlet condition mechanism Flow rate
Zl + h2
+ ( - 2 - = (l + Sp) Ultra-starved Snake
Starved Snake and back-jet it < q « 1
(12b.27)
The meniscus roll coating regime 593
Because a two roll coater operates in a three- meniscus moves inwards. Experiments by
dimensional parameter space defined by capillary Malone (1992), suggest that once the meniscus
number, Ca, geometry ratio, Ho/R m and flow reaches the minimum gap position it loses
rate, q, it is highly desirable to have a relationship stability giving rise to the phenomena of the
involving these parameters - a relationship which bead break described in the next section.
defines the meniscus roll coating regime. With Similarly b increases when Ca is reduced and
so many parameters involved, it would be difficult the upstream meniscus moves away from the
to obtain such a relationship from computational nip. In practice, however, the meniscus only
data alone. However, the small flux model provides moves out so far. There is a point, beyond
such a relationship in the form of equation which, solutions predicted by equation (12b.29),
(12b.15), as shown in equation (12b.29), above. are not realized nor are they obtainable
This gives valuable insight into the delicate computationally.
balance between the governing parameters. The The remaining problem now is to explain
following observations can be made. why the meniscus only moves out so far when
Ca is reduced and why a certain class of
1. For meniscus coating with q« 1 it is clear solutions is not realized in practice. In fact
that Ca and H 0/R m cannot be chosen randomly. such solutions correspond to flows in which
As mentioned previously with regard to capillary pressure at the upstream meniscus
equation (12b.15) Ca(R m /H o)1/2 must be of is comparable to that at the downstream one
order unity for the equation to have a solution. and so the pressure field has 'essentially
If a solution cannot be found for a given set collapsed', i.e. the assumption on which the
of parameter values it means that the coating small flux model is based - that pressure and
bead does not exist for these values. viscous forces are in balance, with inertia
2. It is observed both experimentally and com- negligible - is no longer valid and therefore
putationally that once an equilibrium flow is such equilibrium flows do not exist. Both
established with Ca, HO/R m and q all fixed experiments and computations suggest that
then if anyone of these parameters is varied, the position of the upstream meniscus,
the position ofthe upstream meniscus changes b = tan CX b , usually lies within the range
immediately whereas the downstream meniscus O<b< 1.75.
changes its position more slowly. Furthermore 3. Equation (12b.29) can be written in the form:
it is found that the upstream meniscus
is restricted in the sense that it cannot move
too far in towards the nip nor too far out. Ca ( H:
R )1/2
= !(q, a b, ad) (12b.30)
This behavior can be predicted via equation
(12b.29). where! is a function of the variables q, a b and
As an illustration consider those flows in ad; then one way of displaying results is to fix
which the downstream meniscus is far from q and plot solutions as data points on a graph
the nip so that ad::::: 1[/2. For each flow rate of Ca(R m / H 0)1/2 against b, (b = tan a b ). Figure
q, equation (12b.29) gives capillary number as 12b.22b identifies a corner of parameter space,
a function of b, (b = tan a b ), the position of bounded by the axes and the curve AB, in
the upstream meniscus. Figure 12b.22a shows which solutions are to be found. The data
a curve of Ca(R m /H o)1/2 against b for flow points correspond to finite-element solutions
rates q = 0.075 and q = 0.2 and ad = 1[/2. As for which qm = 0.2, 0.1 < Sp < 2.5 and
Ca increases b decreases and the upstream 2.2 < d < 3.9 where d = tan ad. The curves
0.3
0.25
0.2
~
:EI~
-0.15
(3
0.1
0.05
0.0
0.0 0.25 0.5 0.75 1.0 1.25 1.5 1.75 2.0
(a)
C
0.3
0.25
~ 0.2
-
:EI~
~
c..>
0.15
0.1
0.05
B
M D
0.0 0.25 0.5 0.75 1.0 1.25 1.5 1.75 2.0
(b)
Figure 12b.22 (a) Theoretical curves of Ca(R m/H o )1 /2 against b for IXd = 7[/2 and flow rates qm = 0.075 and
0.2; (b) + data points obtained from finite-element solutions for the latter flow rate, AD and CD are theoretical
curves derived from equation (12b.29) when IXd = 7[/2 and 57[/12, respectively.
Instabilities 595
AB and CD represent solution curves obtained
analytically via equation (12b.29) when ()(d = n/2
and 5n/12, respectively.
12b.5 INSTABILITIES
In meniscus roll coating Malone (1992) identified
a number of characteristic instabilities and their
associated coating defects, including bead-break,
bead-collapse and bubble generation which are (a)
described below. Little is known about the
mechanisms causing these instabilities and what
is reported here are the results of that preliminary
investigation.
12b.5.1 BEAD-BREAK
This instability manifests when speed ratio is
increased from a stable operating condition
where the bead has uniform free surfaces upstream
and downstream of the nip. As the speed of the
upper roll increases, the upstream free surface
moves towards the nip and subsequently exhibits (b)
a local indentation or 'necking', see Fig. 12b.23a.
However, there is evidence to suggest that the
free surface can pass through the nip before
losing stability. The necking may then lead to
either a local form of bead-break, Fig. 12b.23b,
or several bead-breaks producing a cellular array
of beads, Fig. 12b.23c. As VI is further increased,
additional neckings arise until a quasi-stable
situation is reached with equispaced beads
distributed along the axis of the rolls in the nip
region. The effect is to produce alternate wet and
dry stripes on the upper roll - which in practice (c)
would be the web. When developed fully this
form of instability bears some similarity to Figure 12b.23 (a) Local indentation of the upstream
'ribbing' by virtue of its regular, periodic pattern free surface; (b) bead-break; (c) multi-bead-break (rolls
along the axis. Unlike ribbing however, it appears moving right to left).
to arise as a result ofa two-dimensional disturbance
to the upstream meniscus, Gaskell et al. (1996b).
In pre-metered bead coating operations such as characteristic shape. It is conceivable that these
slide or slot coating this instability is very may be due to the onset of bead-break when the
common and is referred to as 'rivulets'. Other bead may well be alternately breaking and
defects, observed on the web, have been referred reforming, i.e. alternating between Fig. 12b.23a
to as coating misses or 'cigar misses' due to their and 12b.23b.
12b.5.2 BEAD-COLLAPSE which is equivalent to - 3000 Pa) and therefore

Excessive inlet starvation would appear to be degassing is the most likely cause, although
the cause of this particular instability. If flow further work needs to be done.
rate qrn is reduced beyond a critical value or if
Sp is increased by reducing U2 and simultaneously ACKNOWLEDGMENTS
reducing flow rate then the bead will break
somewhere along its length followed by the The authors wish to record their gratitude to Dr
disintegration of the remainder of the bead - J. L. Summers (of the University of Leeds) who
leaving a layer of fluid on the lower roll only. collaborated in the preparation of this manuscript
and kindly generated many of the finite-element
results. Thanks are also due to leI for their
12b.5.3 BUBBLE GENERATION generous sponsorship of this work.
A common defect in meniscus roll coating is the

appearance of bubble lines on the coated web REFERENCES
arising from the presence of bubbles in the fluid
Babuska, I. and Aziz, A. K. 1977. Lectures on the
bead. When they first appear they tend to grow Mathematical Foundations of the Finite Element
in size and wander axially along the nip. A bubble Method, ed. A. K. Aziz. New York: Academic Press.
may lodge up against the free surface at the p. 1.
downstream side of the nip, Fig. 12b.24, giving Benkreira, H., Edwards, M. F. and Wilkinson, W. L.
the appearance of a local indentation of the 1981. A semi-empirical model of the forward roll
surface and giving rise to a bubble line which is coating flow of Newtonian fluids. Chem. Eng. Sci.
36: 423.
a down web line of localized, low coat weight. Canedo, E. L. and Denson, C. D. 1989. Flows in driven
As the fluid bead is a region of subambient cavities with a free surface. AIChE J. 35(1).
pressure, as a result of capillary action at the Coyle, D. 1996. Liquid Film Coating, (eds S. F. Kistler
menisci, it is not surprising that it should provide and P. M. Schweizer). Chapman and Hall, Ch. 12a.
a source of attraction and a resting place for any Coyle, D., Makasko, C. W. and Scriven, L. E. 1986.
bubble that might emerge. Two possible sources Film-splitting flows in forward roll coating. J. Fluid
of bubble generation have been suggested; Mech. 171: 183.
Gaskell, P. H., Savage, M. D. and Thompson, H. M.
vaporous cavitation and degassing as dissolved
1991. Creeping flow: novel analytic and finite element
gas is expelled from the solution. However the solutions. Proc. 7th Int. Conf. Num. Meth. Lam. and
measured pressures (Malone 1992) appear to be Turb. Flow, eds Taylor, Chin, Homsy. pp. 1743.
an order ofmagnitude smaller than those required Gaskell, P. H., Innes, G. E. and Savage, M. D. 1996a.
for vapourization (for example, the vapour pressure An experimental investigation of meniscus roll
ofa typical industrial fluid, toluene, is - 23 mm Hg coating. J. Fluid Mech., in press.
Gaskell, P. H., Innes, G. E. and Savage, M. D. 1996b.
Bead break instability in meniscus roll coating.
effective 'local' free surface Submitted to J. Fluid Mech.
Gaskell, P. H., Savage, M. D., Summers, 1. L. and
Thompson, H. M. 1995a. Flow topology and
transformation in fixed gap symmetric forward roll
coating systems. Proc. 9th Int. Conf. Num. Meth.
-
Lam. and Turb. Flow, eds Taylor, Durbetaki. pp. 984.
Gaskell, P. H., Savage, M. D., Summers, J. L. and
Thompson, H. M. 1995b. Modeling and analysis of
meniscus roll coating. J. Fluid Mech. 298: 113.
Hughes, T. J. R. 1987. The Finite Element Method:
Figure 12b.24 A bubble in the bead giving rise to a Linear Static and Dynamic Finite Element Analysis.
bubble line defect. Joseph, D. D. 1977. The convergence of biorthogonal
Notation 597
series for biharmonic and Stokes flow edge problems: nondimensional outlet film thicknesses
Part 1. SIAM J. Appl. Math. 33: 337. unit normal vector
Kistler, S. F. and Scriven, L. E. 1983. Coating flows. pressure
Computational Analysis of Polymers. pp. 343.
Landau, L. D. and Levich, V. G. 1942. Dragging of
nondimensional pressure
liquid by a moving plate. Acta Physchim, URSS. flow rate per unit axial length
XVII(I-2): 42. nondimensional flow rate
Malone, B. 1992. An experimental investigation of roll qmax maximum value of q
coating phenomena. PhD thesis, University of Leeds. qm modified nondimensional flow rate
Pearson, J. R. A. 1985. Mechanics ofPolymer Processing. ii,q* critical, nondimensional flow rates
pp. 378-380. London and New York: Elsevier R 1, R 2 upper and lower roll radii
Applied Science Publications.
Perkins, R. J. 1991. Department ofApplied Mathematics
R b , Rd radius of curvature of upstream and
and Theoretical Physics, Cambridge. Private com- downstream menisci
munication. Rm average roll radius
Ruschak, K. 1. 1985. Coating Flows. Ann. Rev. Fluid s nondimensional surface coordinate
Mech. 17: 65. Sp speed ratio (nondimensional)
Thompson, H. M. 1992. A theoretical investigation of t unit tangent vector
roll coating phenomena. PhD thesis, University of upper and lower roll speeds
Leeds.
VI' V 2
Vm average roll speed
V,W velocity components
NOTATION X,Z coordinate directions
x, z nondimensional coordinate directions
A bead half length
b a nondimensional length, defined by
b = tana b Greek letters
Ca capillary number IX nondimensional angular coordinate,
a nondimensional length, defined by defined by x = tan a
d
d = tan ad V (0 a
del operator ox' OY' OZ
0)
H bead half depth
Ho half minimum gap separation Jl. fluid viscosity
Hi inlet film thickness on lower roll 7t pi
HI' H 2 outlet film thicknesses on upper and 'I' streamfunction
lower roll (J surface tension
hi nondimensional inlet film thickness !! shear stress tensor
ELASTOHYDRODYNAMIC COATING 12c
SYSTEMS
Ferdinand R. Pranckh and Dennis J. Coyle
12c.l INTRODUCTION
reduced coating thicknesses with less sensitivity
Elastohydrodynamics is the interaction between to mechanical tolerances. Reduced sensitivity to
hydrodynamic traction forces exerted by a flowing operating parameters and process perturbations
fluid and elastic restoring forces imparted by a is an additional bonus. As will become apparent
deformable solid that forms part of the confinement throughout this chapter, however, this is not
of the flow. Elastohydrodynamic effects are always achieved. For that reason, there are
negligible in most coating flows. However, some distinct trade-offs between rigid and elastohydro-
coating methods rely on elastohydrodynamic dynamic coating applicators.
interactions for coated film metering, formation, The various elastohydrodynamic coating sys-
and transfer. Figure 12c.l depicts six of the most tems illustrated in Fig. 12c.l are distinguished
common elastohydrodynamic coating methods. by the type ofelastic confinement they incorporate.
In all but the tensioned-web slot coater and the Flexible-blade and membrane coaters employ
gravure coater, the thickness of the coated film static plates, or sheets, of metal or plastic as
depends primarily on the balance between the metering elements. In flexible-blade coating (Fig.
hydrodynamic forces exerted by the liquid and 12c.1 a), the upstream edge of a rather thin strip
the internal elastic restoring forces and external of metal or plastic is clamped in a rigid holder
loads resisting the deflection of the deformable and loaded against a liquid-laden, moving
flow confinement. In general, therefore, elasto- substrate backed up by a roll so that the blade
hydrodynamic coating configurations are self- and substrate form a converging wedge. The
metered (see Chapter 1 for a systematic classifi- liquid supplied to that wedge typically is dip-coated
cation of coating processes). In the two noted upstream or applied with a more sophisticated
exceptions, the coating flow is nearly pre- device, such as a confined-flow feed box that
metered, but the elastohydrodynamic effects avoid becomes an integral part of the blade installation
the need to build coating devices with extremely (see Chapter 12d). In membrane coating (Fig.
narrow, pre-set gaps between a stationary 12c.lb), the applicator is a compliant trailing
applicator and the moving substrate and thereby wiper film that is wrapped around and extends
allow deposition of much thinner liquid films past a cylindrical, inflatable 'air bag' or membrane.
than are achievable by their rigid counterparts of The membrane-backed wiper together with the
die and roll coating. This points to the major moving substrate form a converging, wedge-shaped
reason for using elastohydrodynamic coating region very similar to that in blade coating. That
systems, namely their ability to apply greatly region is typically fed from a puddle that sits
600 Elastohydrodynamic coating systems
Coated Web
Back-Up Roll
Coated
Web
Trailing
Film
(3) (b)
Coated
Web
Applicator Die
(d)
(c)
Coated
Web
Coated
Web
(e) (f)
Figure 12c.l Common types of e1astohydrodynamic coating methods: (a) flexible-blade coater; (b) membrane
coater; (c) tensioned-web roll coater (forward mode); (d) tensioned-web slot coater; (e) squeeze-roll coater; (f)
direct gravure coater. The coating thickness and gaps between elastic flow confinements are magnified for clarity.
Introduction 601
right above the wedge and is supplied by a pipe parts of the coating devices. The normal load
or a more elaborate flow distributor. that arises from the resultant of the tension in
In both coating configurations, the wedge- the curved substrate presses the substrate against
shaped region between the metering element and the roll or die. In kiss roll coating, the normal
the substrate is key to setting the coating thickness. load together with speed, viscosity and possibly
The hydrodynamic lubrication forces generated substrate stiffness meter the amount of liquid
by the drag flow into the converging wedge tend that passes through the nip between the roll and
to bend and lift the metering element. This action substrate in a manner very similar to that of the
is opposed by the external load which, in pneumatic air load controlling coating thickness
flexible-blade coating, comes from rotating or in membrane coating. Running the roll and
translating the blade holder or from forcing a substrate at differential or opposite speeds provides
pressurized bladder or some mechanical device additional options to influence the final coating
against the blade; in membrane coating, the load thickness. In tensioned-web slot coating, in
comes from the pneumatic pressure within the contrast, the final coating thickness is premetered
membrane, and is transmitted through the with the pump that supplies liquid to the die.
compliant wiper. As a result of the similarities Changes in normal load, resulting from adjust-
between flexible-blade and membrane coating, ments in tension, modifications in die-lip design,
the elastohydrodynamic balance between hydro- or alterations in speed or viscosity are now used
dynamic forces and external load produces an to ensure that the elastohydrodynamic floating
operating response that is quite comparable in height of the substrate above the die lips can
both configurations: the metered coating thickness accommodate the desired, pre-metered amount
depends primarily on the external load, the of liquid. The basic elastohydrodynamic action,
bending stiffness of the flexible plates or sheets nevertheless, is quite similar in both tensioned-web
involved, the liquid viscosity and possibly other coating systems.
rheological properties, and the coating speed. At Squeeze roll and gravure coaters, shown
sufficiently high speeds, inertia effects can also schematically in Figs 12c.1e and 12c.1f, incorporate
come into play (see Chapter 12d). compliant elastic layers of rubber or rubber-like
Tensioned-web roll and slot coaters, illustrated elastomers bonded to steel rolls as metering
in Figs 12c.lc and 12c.ld, rely solely on the elements. Some substrates, like paper, are com-
deformable substrate without any external pressible and hence also interact elastically with
mechanical support to provide the elastic flow the coating process. Squeeze-roll coaters, some-
confinement and imposed load in the coating times referred to as nip-roll coaters, are in essence
application step. In tensioned-web roll coating, configured like forward roll coaters, with either
often also called kiss coating, liquid is transferred one or both rolls being rubber covered. Nip
onto a tensioned web as the web passes over a control is achieved by either direct pressure
rotating applicator roll. The liquid film on the loading of the rolls, or by setting a gap or a
applicator roll is typically supplied by dip coating, positive engagement with mechanical stops. To
sometimes reduced in thickness with additional achieve thin coatings, several nips can be arranged
metering roll steps of the sort described in in sequence in multi-roll configurations. The
Chapter 12a. In tensioned-web slot coating, a liquid supply system to the nip can be rather
liquid film is applied directly out of a slot of a crude, as illustrated by the pipe-fed puddle or
coating die onto a web that travels in free span dip coating systems shown in Fig. 12c.le. Gravure
between two idler rolls. coaters are quite similar to nip roll coaters,
In both tensioned-web coating configurations, except that the rigid roll has cells, grooves or
wrapping the substrate over a certain, albeit other patterns of a specific volume engraved into
sometimes small, angle around the roll or slot- its surface. The liquid supply system is also more
coating die loads the substrate against those rigid sophisticated. An excess of coating solution is
applied to the gravure roll, usually by dipping blade (Section 12c.3), membrane (Section 12cA),
the roll into a pan or by using a more precise tensioned-web roll (Section 12c.5), tensioned-
orifice- or fountain-type applicator. This excess web slot (Section 12c.6), squeeze roll (Section
is then doctored by a flexible blade which leaves 12c.7), and gravure coating (Section 12c.8).
the cells only partially full with liquid. This Throughout, the main emphasis is on identifying
doctoring step, ofcourse, constitutes an additional the physical mechanisms of viscous flow and
elastohydrodynamic coating system very similar elastic deformation, and explaining how their
to the flexible-blade coater introduced above. interaction results in the observed operating
As for the basic elastohydrodynamic mechan- behavior. Where sufficient information is available,
isms at work, the nip interface between the rolls the chapter also comments on instabilities and
is quite similar in squeeze-roll and gravure how these affect uniformity and operability limits.
coating. The mechanical deformation resulting Many of the hydrodynamic instabilities that
from the positive engagement, or at least close are seen in elastohydrodynamic coating systems
proximity, between the rolls builds a pressure are the same as those encountered in other
hump akin to that in Hertzian contact between coating configurations and discussed throughout
two elastic rolls in dry contact. This pressure the entire volume. Most prominent are perhaps
hump must be balanced by the hydrodynamic air entrainment (e.g., Chapter 3) and ribbing (e.g.,
lubrication flow pressures in the liquid, and the Chapter 12a). Ribbing, for instance, tends to
thickness of the liquid film between the rolls is occur in flexible-blade coating under highly
a direct consequence ofthat balance. The purpose loaded conditions that cause the blade tip to lift
of the elastohydrodynamic interaction is quite off the substrate wrapped around the back-up
different, however, in the two coating configur- roll. A related defect in squeeze-roll and tensioned-
ations. In squeeze roll coating, the intent is to web roll coating is the film-split pattern shown
allow the elastohydrodynamics to meter the in the photograph of Fig. 12c.2. Its amplitude
coating thickness. The thickness depends primarily and wavelength change with coating thickness
on the liquid viscosity, coating speed, load, and but, for all practical purposes, the defect cannot
elastic modulus as well as thickness of the rubber be avoided in squeeze-roll and tensioned-web
layer. In gravure coating, in contrast, the wet roll coaters when operated in the forward mode.
coating thickness is premetered in the sense that In this case - and in quite a few other elastohyd-
it is dominated by the engraved volume and is rodynamic coating applications - acceptable
relatively insensitive to the adjustable process coating uniformity depends on leveling before
and fluid variables. The main task of the solidification or a lax criterion of what is
elastohydrodynamic interaction is to allow the 'acceptable'.
substrate wrapped around the rubber-covered Other flow instabilties are more peculiar to
nip roll- or the nip roll directly in offset-gravure elastohydrodynamic, or at least self-metered,
configurations in which that roll serves as transfer coating systems. Many originate in the puddle
roll - to conform to the textured surface of the or rolling bank of liquid that builds upstream of
gravure cylinder and thereby ensure proper the metering element. This puddle, similar to that
pick-out of part of the liquid from the cells or discussed for forward roll coating with rigid rolls
grooves. in Chapter 12a but possibly considerably larger,
The main purpose of this chapter is to provide contains one or more vortices. At high Reynolds
a critical review of the published literature on numbers, these might become unstable and form
elastohydrodynamic coating operations. Section a three-dimensional vortex structure or even
12c.2 recounts the mechanisms of elasto- undergo transition to turbulent flow. Even at
hydrodynamics and outlines the governing low and moderate Reynolds numbers, rolling
equations. The remaining sections discuss each banks are susceptible to air entrainment and also
of the coating methods in turn, covering flexible- to entrapment of dirt, particle agglomerates,
Elastohydrodynamic mechanisms 603
Figure 12c.2 Photograph of film-split pattern in squeeze-roll coating of a Newtonian liquid. The coating
direction is left-to-right; the upper region has a thicker wet coverage than the lower.
bubbles, and other sources ofdefects. The puddles that the coating liquid exerts on elastically
can also have a particularly strong influence on deformable flow confinement arise from both the
the coating near the edges of the web. If too little isotropic pressure and viscous stresses:
fluid is supplied to an upstream puddle, the
1;,n == 00: T =- P + 00: D (12c.la)
coater becomes 'starved' and the coating width
shrinks while maintaining a near constant Tnt == to: T = to: D (12c.1 b)
thickness at the center of the web. If too much
Here, Tnn and 1;,t are the normal and the tangential
is supplied, the puddle can overflow, resulting in
traction components made non-dimensional with
excessive edge beads, run-back, or other problems.
j1U/ L, where j1 is the characteristic viscosity, U
In addition to hydrodynamic instabilities,
is the characteristic speed, usually the substrate
mechanical instabilities can also degrade the
or coating speed, and L is a characteristic length;
coating quality resulting from elastohydrodynamic
o and t are the local normal and tangent unit
coating processes. Examples include vibrations
vector to the boundary; p is the pressure and T
ofdoctor blades, or tension fluctuations, fluttering
is the total stress tensor, both measured in units
and buckling in tensioned webs.
of j1U/L; and D == Vv + VvT is the rate of strain
tensor, where v is the dimensionless velocity field.
12c.2 ELASTOHYDRODYNAMIC If the narrow clearance ofthe flow confinement
MECHANISMS that results from the elastohydrodynamic inter-
action varies slowly in the stream-wise direction,
12c.2.1 LIQUID FLOW the lubrication approximation applies and the
Navier-Stokes system (2.7) and (2.21) reduces to
Steady, two-dimensional coating flows ofinelastic the much simpler Reynolds equation which
liquids, to which this section is restricted, are describes the pressure distribution along the flow
governed by the continuity equation (2.7) (Chapter channel (see Chapter 8 for details). With the
2), the momentum conservation equation (2.19), deformable boundary as the reference surface,
and an appropriate constitutive equation to the non-dimensional form of this equation is:
dP)= ~(h
describe non-Newtonian viscous behavior. For
Newtonian liquids, the last two reduce to the ~(h3 VI + V2 ) (12c.2)
Navier-Stokes equation (2.21), and the traction ds 12 ds ds 2
Here, s is the arc length along the reference surface 12c.2.2 ELASTIC LAYERS
measured in units of L; h is the channel height The simplest approach to modeling solid layers
measured in units of L; and Vi and V 2 are the of rubber, rubber-like elastomers or compressible
projections of the boundary velocities onto the materials such as paper is to assume their
reference surface, both measured in units of U. In deformation is purely elastic (Section 12c.7 briefly
the lubrication approximation, the normal and addresses the complications that arise from
tangential traction components on the deformable viscoelastic behavior of solid layers). In most
flow confinement depend only on the pressure coating configurations, the deformation ofelastic
profile along the narrow flow channel and reduce to: layers has no cross-web component nor any
Tnn = -p cross-web gradients. It is therefore governed by
(12c.3) the equations of plain-strain elasticity. Deforma-
T. = _ (~ dp _~) tion is measured by the Finger strain tensor F
nt 2 ds h whose components are
From the pressure distribution, it is nonetheless
straightforward to calculate the stream-wise F .. = oX i (12c.6)
velocity u(s, z) or the shear rate y(s, z) at any point
I] oXj
(s, z) if desired: where Xi = Xi + ui are the deformed coordinates
~ z)
written as the sum of the reference coordinates
u(s,z) = _~ :~ (zh _ Z2) + (h Xi and the displacements U i • The equation of
continuity requires that the determinant of the
Y(s, z) == du = _ ~ dp (h - 2z) _ ~ Finger tensor (or its inverse F- i ) minus one
dz 2 ds h equal the volume dilation
Here, z is the coordinate perpendicular to the OU i + oU 2 + oU i oU 2 + oU i oU 2 = _ 3(1 - 2v) P
reference surface measured in units of L, U is
oX i oX 2 oX i oX 2 oX 2 oX i E
measured in units of U, and y in units of U fL.
For the purpose of comparing and assessing (l2c.7)
different coating flows, it is sometimes useful to
calculate the total shear suffered by the fastest Here, E is Young's modulus, v is Poisson's ratio,
moving 95% ofcoated liquid as it passes through and P is the isotropic pressure. The equation of
the coating device, r 9S' and the total pressure mechanical equilibrium (momentum balance) is
load that the liquid exerts on the substrate or simply
the deformable coating device, P s (for complete V'(J=O (12c.8)
details, see Pranckh and Scriven (1990». These
quantities are defined as where (J is the Eulerian stress tensor. For a solid
that follows Hooke's law, the stress is proportional
y(s, z) ]
r 9S == JL - -1- [h"(S)
J --dz ds (12cA) to strain so that
o h9S (s) 0 u(s,z)
L (Jij = -Pbij + 2Geij (12c.9)
Ps==J p(s)ds (12c.5)
o Here, eij are the components
where h9S (s) is the dimensionless height from the
moving substrate to the streamline that demarcates
95% of the total liquid flow Q, i.e.,
f
h.,(S)
of e, the Almansi (Eulerian) strain tensor, and G
0.95Q= 0 u(s,z)dz.r 9s ismeasuredinunits
is the shear modulus, which is related to Young's
of U, and P s in units of pD. modulus via E = 2(1 + v)G.
Elastohydrodynamic mechanisms 605
Equations (12c.7)-(12c.1O) are readily solved Here, hand ho are measured in units of L; 1;,n
by the finite-element method as part of a coating- is the normal traction component, given by
flow simulation algorithm (cf Chapter 9). The equation (12c.l) or (12c.3); and El = J1.Uj(LK) is
hydrodynamic boundary loads can be determined the elasticity number that compares the relative
from the full Navier-Stokes system or the importance of characteristic viscous stresses with
Reynolds equation. For sufficiently small defor- that of characteristic elastic stresses.
mations, the gradients can be evaluated in the
undeformed reference state and the quadratic
terms become negligible. The mathematical form
of the resulting linear plain-strain equations is 12c.2.3 THIN PLATES, SHELLS AND
identical to that of the Navier-Stokes system in MEMBRANES
the limit of zero Reynolds number (see equation Flexible metering devices in blade, roll or gravure
(2.27) in Chapter 2). If the deformable layer can coaters and the webs in tensioned-web roll or
be regarded as a deep elastic slab, i.e. a semi-infinite tensioned-web die coaters typically start out flat
Hookean solid, its local indentation can be in their undeformed state and thus are, in
expressed as an integral ofthe pressure distribution, mechanical terms, plates or membranes. Because
thereby further simplifying the analysis as discussed they are most often very thin compared with
in Section 12c.7 (see equation (12c.l8) and also their overall length, the local deformations of
Carvalho and Scriven (1993)). infinitesimally small segments remain small, even
If the normal tractions that act on the elastic in highly bent states during coater operation.
layer have no steep gradients, an additional, Hence, the resistance to bending and stretching
dramatic simplification arises from the so called can be described by linear elastic models, as in
Winkler elastic foundation approximation (Joh- standard thin-plate theory (Fluegge 1960). On a
nson 1985): the elastic layer is taken to act like macroscopic scale, however, the deformations
an array of infinitesimally narrow springs, with can be large, i.e. the changes in position of the
the displacement of a particular spring having deformable members can be ofthe same magnitude
no effect on its neighbors. With non-local effects as their characteristic length or larger, giving rise
ignored, the displacement at any point along the to large curvatures and associated geometric
deformable layer is simply proportional to the nonlinearities. These nonlinearities can be ac-
loading at that point: counted for by what the solid mechanics literature
H - Ho = KP n (12c.ll) refers to as cylindrical shell theory (Fluegge
1960). In its simplest form typically used in
Here, Hand Hoare the loaded and unloaded coating flow analysis, that theory is limited to
elevations, Pn is the local normal load, and K is thin, inextensible shells and thereby retains the
the bulk modulus of the material. Johnson (1985) linear bending model of thin-plate theory. In the
shows how K relates to the Young's E modulus context of viscous free surface flows, shells can
for several limiting cases. The Winkler approxi- be thought of as interfaces that not only have
mation in the form (12c.ll) has been successfully surface tension but also have surface bending
applied to squeeze-roll coating (Coyle 1988b; see stiffness.
also Section 12c.7) and high speed blade coating If there is no misalignment or buckling in the
of a compliant paper substrate (Pranckh and cross-web direction in the first place, and variations
Scriven 1988). of the viscous tractions in the cross-machine
When the normal load arises from the coating direction are negligible so that the base flow is
flow alone, as described by the Navier-Stokes two-dimensional or has, compared with the
equations or the Reynolds equation, equation elastic forces, only small cross-machine disturb-
(12c.ll) becomes ances like those resulting from a ribbing instability,
h - ho = EI1;,n the blade or web shapes are two-dimensional. In
this idealization, the governing equations for the
shells become one-dimensional:
dQ
ds + KN + Po = 0
dM _ Q = 0
ds
M= -K
Free Edge
Here, PI and Po are the tangential and normal
load distributions along the shell; N is the local FI.X;bl~l~
resultant of the internal normal stresses acting
in the direction of the tangent t to the shell which,
if positive, is in effect the tension in the blade or Clamped Edge
web; M is the bending moment resulting from (b)
those normal stresses; Q is the resultant of the
shear stresses in the direction of the normal n to
the thin shell; K is the local curvature; and s is
the arc length along the shell. Some of the
variables are displayed in Fig. 12c.3a. Nand Q
are measured in units of D/L 2 and M in units of Membrane
D/L, where L is a characteristic length of the
shell and D its bending stiffness
D == E<5 3 /[12(1 - y2)J; <5, in turn, is the shell
thickness, and E and yare Young's modulus and
Poisson's ratio respectively of the shell material.
Free Edge
K is made nondimensional with I/L; PI and Po Pinned Edge
Trailing
with D/L 3 ; and s with L. Film
The equations listed above represent the (c)
tangential force balance, the normal force balance,
the moment balance, and the constitutive equation Figure 12c.3 Definition sketches for theory of
that relates the bending moment to curvature. deformation of thin plates, sheets, and membranes into
The contributions KQ and KN are a direct cylindrical shells: (a) definition of tangent, t, normal,
consequence of the geometric nonlinearities that n, radius of curvature I/K, tangential force, N, normal
result from the large deflections. The constitutive force, Q, and bending moment, M; (b) boundary
equation describes the linear response of a conditions for analysis of flexible-blade coater;
(c) boundary conditions for analysis ofmembrane coater.
Hookean solid under the Kirchhoff-Love as-
sumptions for thin plates and thin shells (for an
in-depth discussion, see Fluegge (1960)). For
computational purposes, it is convenient to
eliminate Q which, given the load distributions (12c.12)
along the shell, yields two coupled nonlinear
ordinary differential equations K and N:
Flexible-blade coating 607
From the curvature, the x- and y-coordinates of membrane coating, the force and momentum
the shell, measured in units of L, are readily balances that govern each are coupled via a set
calculated from: ofconditions that describe the contact mechanics.
The simplest approach is to assume that the two
d!3 dx !3 dy '!3 members can freely slide over one another without
ds = K, ds = cos, ds = sm
any friction, as discussed by Pranckh and Scriven
Here, the auxiliary variable !3 is the local angle (1989b). Special attention must be paid to the
of inclination of the blade or web. Eliminating end points of contact zones where two elastically
this variable reduces the three first-order equations deformable members join or part. At such points,
to two second-order equations: a concentrated force of reaction gives rise to a
jump, or discontinuity, of dK/ds or, equivalently,
d 2x dy Q for both members. This must be accounted for
ds2 + K ds = 0 (12c.13)
with a special set of end point conditions (for
2 details, see Pranckh and Scriven (1989b».
d y_ K dx = 0 The tangential and normal load distributions
ds 2 ds PI and P n in equations (12c.12) can have
The appropriate boundary conditions for the contributions from both hydrodynamic forces
force balances (12c.12) and the geometric relations and from externally imposed forces such as the
(12c.13) depend on the physics of the situation, loading that a pressurized air hose exerts on a
illustrated for a blade coater and a membrane coating blade, viz.,
coater in Figs 12c.3b and 12c.3c respectively.
When a flexible blade is rigidly clamped, its
position given by the coordinates Xc and y c at P n = El T..n + Pn,exl
the clamp and the slope !3c at the clamp must
be prescribed: Here, El is the elasticity number, El == J.lU L 21D,
and Pl,exl and Pn,exl are additional distributed
dx dy. external loads, measured in units of DIL 3 ; T..l
ds = cos!3c, ds = sm!3c, X = Xc, Y = Yc and Tnn are the tangential and normal traction
components as defined at equations (12c.l).
When the edge of an elastically deformable Equations (12c.12) and (12c.13) or equivalent
member is pinned along a line but free to rotate forms were used by Saita and Scriven (1985) and
along that line, as is approximately the case for Pranckh and Scriven (1990) with the Reynolds
the membrane wrapped around the holder in equation to describe blade coating; by Shibata
membrane coating, the boundary conditions are: and Scriven (1986) with the full Navier-Stokes
K = 0, X = xp, y = yp equations to describe tensioned web slot coating;
by Pranckh and Scriven (1988) with the Navier-
The first condition expresses that, due to uncon- Stokes equations to describe high-speed blade
strained rotation, there is no clamping moment. coating; and by Pranckh and Scriven (1989a, b)
At a free edge of a trailing blade, there is no with the Reynolds and the Navier-Stokes
bending moment nor any tension or normal load equations to describe membrane coating.
and, hence, the boundary conditions are
dK =0 K = 0, N= 0 12c.3 FLEXIBLE-BLADE COATING

ds '
In flexible-blade coating, introduced already in
When two elastically deformable members are Section 12c.1 above, a liquid film of controlled
in contact, as is the case for the trailing film thickness is metered onto a moving, roll-supported
wrapped around the inflated membrane in substrate by means of a blade that is loaded
against an over-supply ofliquid applied upstream loaded with an air bladder or hose that is
(see Fig. 12c.1a). Flexible-blade coaters are most mounted in a rigid holder and positioned against
commonly used to apply mineral pigmented, the blade away from the clamping point. Con-
latex, or functional coatings onto paper at very trolling the bladder or hose pressure sets the
high speeds. Chapter 12d focuses in depth on external loading P n,ext. The air hose can also be
this particular industrial application, whereas replaced with bar or fulcrum that is loaded
this section is limited to describing the elastohyd- mechanically or pneumatically to deliver a
rodynamic mechanisms of flexible-blade coating controllable normal force F n,ext per unit length
in detail. Other examples of the use of flexible in the cross-web direction at a fixed distance SF
blades in liquid film coating come from the from the clamping axis. Figure 12c.4 illustrates
doctoring elements in gravure and roll coating; the latter two loading configuration schematically.
these are discussed further in Section 12c.8 and Yet another loading method is to press the blade
Chapter 12a respectively. against a fixed stop, which acts as a secondary
The central component in all flexible-blade pivot axis as the entire blade holder is translated
configurations is the coating or metering blade, and forces the blade tip against the substrate
which can range from easily bent plastic strips (Sommer et al. 1987; Nakazawana et al. 1987;
to fairly stiff spring-steel plates. Typical steel Akesson and Olsson 1988). The feed system in
blades are 0.25 to 0.5 mm thick and 20 to 100 mm a blade coater can be a dip-coated applicator
long. External loading can be applied in several roll positioned upstream of the blade coating
ways. Simplest is to pivot the blade about the station, a confined feed box (often referred to as
clamp to a working angle <Pw which is larger than 'fountain applicator' - see Chapter 12d), or
the initial installation angle <Po at which the tip simply a puddle feed between the blade and the
of the undeflected blade first touches the substrate: moving substrate.
see Fig. 12c.4. Alternatively, the blade can be The key strength ofblade coating is its capability
FREE EDGE - - 7
STAGNATION LINE
STREAMLINE AT h 95
STAGNATION LINE
LOADING DISTANCE SF
'I<----lINE LOAD Fn,c,t
UNIT NORMAL 8 3
UNIT TANGENT 8,---;<-'"
' r - - - - WORKING ANGLE ~w
r---+- INSTALLATION ANGLE $0
RIGID CLAMP
.. c----..l..-~
CLAMPED EDGE
Figure 12c.4 Schematic of flexible-blade coater, including geometric parameters and loading variables needed
for modeling blade deformation and flow (see text for further explanation of the variables).
to lay down fairly thin layers at very high speeds. loaded, and the strength of the hydrodynamic
In paper coating, for instance, speeds can be in forces.
excess of 10 or even 20 mis, with viscosities of Motivated by observations of Munter (1981)
0.05 to 20 Pa s and wet coating thicknesses of 15 and Eklund (1984), Saita and Scriven (1985) were
to 150 Jlm being typical (cf Chapter 12d). The the first to elucidate, from first principles, that
main weakness of blade coating is related to the there is a continuous progression from stiff-blade
tendency of particles to get trapped under the to flexible-blade coating. That progression gives
blade, thereby producing streaks or scratches. rise to a characteristic non-monotonic operating
The resulting coating flaws can limit the usefulness curve for coating thickness as a function of blade
of blade coating for demanding precision coating loading, as shown in Fig. 12c.5 and to be
applications. Nevertheless, special configurations explained in more detail below. Saita and Scriven's
of flexible blades are used with great success to (1985) insights are based on a finite-element
smooth the overall uniform but locally patterned analysis of elastic bending and viscous flow in
coatings of magnetic suspensions or viscous flexible-blade coaters operating at moderate speeds
adhesives metered onto a substrate by means of and low blade angles. In this regime, the
a gravure roll. hydrodynamic action in blade coating is often
dominated by viscous lubrication effects, as
described by the Reynolds equation (12c.2). If
12c.3.1 BLADE COATING ACTION AND
the coating speed is high and the angle of attack
OPERATING RESPONSE
of the blade is large, inertia effects can give rise
Blade coaters can operate in two distinct regimes. to significant impact pressures that can be
In the so called stiff-blade, beveled blade, understood in terms of inviscid flow theory.
straight-blade or trailing-blade regime - which Chapter 12d discusses this regime at great length
is the most common mode of operating a high- (see also Kahila and Eklund (1978) and Pranckh
speed coater for particulate suspensions in the and Scriven (1988».
paper industry - the metering blade in its mounting The elastohydrodynamic action in a blade
is deformable but its angle of attack remains coater is as follows (Saita and Scriven 1985). The
rather high. In this regime, the metered film moving substrate, supported by a back-up roll,
thickness is dominated by the dynamics in the carries liquid into the wedge-shaped region
immediate vicinity of the blade tip. The ultimate between the blade and the substrate, causing the
limit where the metering blade is practically rigid, lubrication pressure to rise. The resulting adverse
often encountered in household and laboratory pressure gradient induces a flow component that
spreading applications, is of course the same as opposes the flow driven by viscous drag from
the knife coating discussed in Chapter 12a. On the moving substrate, and much if not most of
the other hand, in the so called flexible-blade, the liquid supplied to the blade is rejected. Only
low-angle-blade, or bent-blade regime, the blade a small fraction passes through the narrow gap
is quite highly flexed and more or less conforms between the blade and the substrate to become
to the substrate over an extended region, which the coated layer. By necessity, rejected and
controls the elastohydrodynamic response of the coated liquid are separated by a stream surface
blade coating system. that terminates at a stagnation line on the blade
Much of the older literature on blade coating (for a schematic representation, see Fig. I2cA).
postulates a priori that the blade operates in one Both hydrodynamic pressures and viscous shear
or the other of these two regimes. However, stresses act as loads on the blade and deform it
whether a blade should be classified as 'stiff or elastically. The configuration of the deformed
'flexible' depends not only on its flexural rigidity blade changes the shape of the flow passage,
as given by its elastic properties and thickness, which in turn alters the flow field and the
but also on the way the blade is mounted and hydrodynamic pressure and the viscous stresses
vi
f:j'"
Z
QN
~ I-'~----+--
vi
(f) I-
W W
zo
~~
~ 0 L----.J=:=.J._--'-_~
02468
u 5 LOADING NUMBER
I
f-
f-
Wlf)
~
10 20 30 40
LOADING NUMBER
(a)
o
10
EI = 2.5-10-'
EI = 1.75.10-'
EI = 1.10->
vi°
(/)1")
w
Z
~ 2 4 6 8
~o LOADING NUMBER
IN
l-
I-
w .........._ - - EI = 10.10->
3 52 _ _- - - E I = 5.10->
o 5 10 15 20
---!--
25
EI = 2.5. 10->
=---==:::::::""--EI = 1.75.10->
o ~_--=:l::...... _ _..L-_ _....I...._ _....L.._ _ EI = 1.10->
LOADING NUMBER
(b)
Figure 12c.S Operating response of a flexible-blade coater: (a) characteristic curve of coating thickness versus
blade loading (£1 == j.lUL 2/D = 10- 3 ); (b) effect of elasticity number on characteristic response curve (for a
complete listing of all the design and operating parameters, see Pranckh and Scriven (1990)).
in it - resulting in a quintessential elastohyd- with any externally applied loads, balance the
rodynamic interaction. In an equilibrium operating hydrodynamic pressure and the viscous shear
state, the blade-internal restoring forces that stress on the blade. This balance controls the
result from the elastic deformations, together amount of coated liquid that passes under the
blade. Thus, for a given speed and viscosity, the to the same coating thickness. In Fig. t2c.5a, for
final coating thickness is readily adjusted by example, a wet coating thickness of hw =
changes in blade stiffness, blade mounting and 1.5 X 10- 4 (measured in units of blade length L)
positioning, and external blade loading. can be produced with loading numbers of 1.02,
When the external load on a deformable blade 5.85, or 33.4. These are marked as points 1-3 in
is first increased past the point of initial contact the figure. With a 40 mm long, 0.381 mm thick
with the liquid-laden substrate, the coating steel blade at a line speed of 13.33 mjs and a
thickness falls abruptly. Figure 12c.5a illustrates 50 mPa s liquid, the wet thickness would be 61!m
this behavior for a blade with a fixed clamp and the three line loads F N would be 680, 3900
installed at 15° and a normal load F n •• xl applied and 22300 Njm, applied here at 14mm from the
at a distance of 0.65 L from the clamp. Here, L tip of the blade.
is the blade length, and the dimensionless loading The extrema in the operating curve mark the
parameter used in the figure is F n •• xI L 2 jD, where boundaries between the operating modes of the
D= E<Pj[12(1 - v2 )] is the bending stiffness as blade. Point 4 in Fig. 12c.5a divides the stiff-blade
defined in Section t2c.2 above. The initial decay mode, or beveled-blade mode, to the left from
of the wet coating thickness is so rapid because, the flexible-blade mode, or bent-blade mode, to
at low loads, the blade deflects little and the the right. The region from the minimum at point
hydrodynamic pressure forces concentrate very 4 to the inflection in the vicinity of point 2 is
close to the blade tip where the flow channel sometimes called the intermediate region. The
between the blade and the substrate needs to term bent-blade mode is often reserved for the
tighten to build the pressures necessary to region to the right ofthe inflection, and higWy bent-
compensate for a higher external load. As the blade for the region to the right of the maximum
load increases further, however, the blade deforms at point 5.
more and more, expanding the range of the For practical applications, it is important to
narrow flow channel that contributes to the know the mode in which the blade operates,
buildup of hydrodynamic lubrication pressure. because of the different response to changes in
Eventually, the net load from hydrodynamic blade loading. For a blade operated in the
pressure increases more rapidly than the external beveled mode, small errors in the blade-to-roll
load, giving rise to a minimum in wet coating alignment or small variations in cross-web load
thickness and a subsequent, seemingly counter- distribution can have pronounced effects on the
intuitive increase ofcoating thickness with higher cross-web coat weight distribution because the
blade loads. In this regime, the slope of the flow coating thickness is so sensitive to blade coating.
channel near the blade tip continues to decrease, In the bent-blade mode, in contrast, deficiencies
and the pressure peak broadens and shifts in blade alignment or load distribution have less
upstream, i.e. away from the blade tip. Eventually, pronounced effects on the cross-web coating
the blade deforms so much that the portion close uniformity because the increase and subsequent
to the tip is nearly parallel to the substrate, decrease in coating thickness require more
reducing the pressure forces around the tip. This substantial increases in blade loading.
gives rise to the maximum in the coating-thickness In the bent mode, the details of the blade-tip
versus load curve (point 5 in Fig. 12c.5a). Past shape can also have a significant effect, for the
the maximum, the blade tip forms a diverging blade tip itself becomes part of the narrow flow
flow channel with the moving substrate. In this channel between the blade and the substrate.
flow geometry, hydrodynamic pressure oscillates This effect was not accounted for by Saita and
close to the blade tip and can become sub ambient, Scriven (1985), but was included in a follow-up
possibly pulling the blade toward the substrate. analysis by Pranckh and Scriven (1988) (see also
Because the operating curve is non-monotonic, Chapter 12d). Another coating configuration in
different blade loadings can, of course, give rise which the tip geometry of a deformable blade is
all important is that of a reverse-angle doctor clamp installed at 15° and a normal load F n
blade, i.e. a blade with an angle of attack larger applied at a distance of 0.65 blade lengths from
than 90°. This setup is quite common in situations the clamp (Pranckh and Scriven 1990). At elasticity
where complete wiping of a roller surface is numbers above a certain value (10- 2 for the
desired, such as in reverse-roll coating, gravure example in Fig. 12c.6), the coating thickness is
coating, or rotogravure printing (e.g., Weiss rather high and its operating response to increasing
(1985», but no rigorous elastohydrodynamic load becomes monotonically decreasing. On the
analyses have appeared in the published literature. other hand, toward lower elasticity numbers, i.e.
In either the beveled mode or the reverse- smaller viscosities, lower speeds or stiffer and
angle mode, blade-tip wear can be significant shorter blades, the coated films become thinner
and can alter blade coater performance over time. and the extrema in the operating curve become
Changing from one blade operating regime to more pronounced. In particular, near the mini-
another can also have a substantial effect on the mum, the wet film thicknesses can become
amount of liquid driven into a porous substrate, extremely thin, amounting to a nearly complete
such as paper. For the example given in Fig. wiping by the stiff blade (see inset in Fig. 12c.5b).
12c.5a, the total pressure load P s that acts on Reductions in coating thickness can also come
the substrate rises from 570 dimensionless units from shifting the external blade load toward the
in the beveled mode to 3180 in the bent mode, blade tip, or increasing the initial blade installation
and climbs to 27100 in the highly bent mode. angle (Pranckh and Scriven 1990). Both changes
This illustration suggests that the pressure driving reduce the distance between the normal force
force for penetration into a porous substrate load and the blade tip, where viscous forces
would be nearly nine times greater in the highly concentrate, and hence have the same effect as
bent mode than in the bent mode. shortening and stiffening the blade. Decreasing
The distribution of average shear along the the radius of the substrate-wrapped back-up roll
blade reveals further differences between the flow also can decrease the coating thickness, but the
regimes (Pranckh and Scriven 1990). In the effect is significant only for radii less than two
beveled mode of the example given in Fig. 12c.5a, blade lengths (Pranckh and Scriven 1990). The
nearly all of the shear is concentrated close to reduced roller radius steepens the wedge between
the blade tip. The total shear strain suffered is the blade and the substrate close to the blade
107 dimensionless units. In the bent mode the tip, which has an effect similar to that ofincreasing
shear distributes more evenly, the maximum in the installation angle.
average shear shifts upstream, and the total shear Alternative blade mounting and loading mech-
suffered mounts to 236. In the highly bent mode anisms can also influence the final coating
the last quarter of the blade is nearly parallel to thickness for a given blade speed, viscosity and
the substrate, the average shear rises to a maximum blade length and stiffness, but do not change the
at 70% of the total blade length and decreases characteristic operating response (Saita and
little from there to the blade tip, so that the total Scriven 1985; Pranckh and Scriven 1990). Pivoting
shear climbs to 2110. the blade about its clamp induces an external
torque at the blade clamp. Elastic restoring
forces, which are internal reaction forces induced
by the blade deformation, transmit the torque
12c.3.2 SENSITIVITY TO BLADE DESIGN,
along the blade to the regions of high hy-
INSTALLATION AND OPERATION
drodynamic pressure and high wall shear.
The characteristic operating curve of a blade Increasing the working angle leads to an operating
coater is quite sensitive to the value of the curve that is similar to that of increasing the
elasticitynumber,El == Ji,UL 2 /D. This is illustrated normal force loading of a blade with a normal
in Fig. 12c.5b, again for a blade with a fixed force loading mechanism (Pranckh and Scriven
Membrane coating 613
1990). In the beveled mode in particular, the including cavitation (Eklund 1984). These can
differences in operating curve are small because even eject liquid from the blade tip (cf Chapter 12d).
the blade remains nearly straight and thus the Flexible-blade coating is also quite prone to
moment of the normal force about the clamp is cause coating defects. Its propensity for streak
equivalent to the external torque applied at the formation in particular is perhaps the primary
clamp. In the bent and highly bent modes, on reason why simple blade-coating configurations
the other hand, blades loaded by an external are not used more often for demanding precision
torque at the clamp deform more in the upstream coating application - in spite of their flexibility
half as they transmit the torque from the clamp and low cost. Streaks possibly originate from
to the regions of high viscous forces downstream. particles or bubbles trapped in recirculating
Thus the pivoting mechanism, while effective at flows on the upstream side of the blade. Saita
low loadings, is not as effective as normal force and Scriven (1985) linked streaking and scratching
loading closer to the tip in the bent and the highly defects in coating of shear-thinning suspensions
bent modes (Pranckh and Scriven 1990). to the region around the stagnation line (cf Fig.
12cA) where shear rates are small and shear-
thinning viscosities are high. They contend that
12c.3.3 COATING OPERABILITY AND
close to the stagnation line particles can get
UNIFORMITY
trapped, which leads to streaks, and suspensions
Once the characteristic operating response is can form particle aggregates, which leads to
understood, a flexible-blade coater is fairly simple coating defects if aggregates get dislodged and
to set up, operate and control. Also, unless the the shear they suffer downstream does not suffice
blade is very stiffand operated near the minimum to break them up. In the bent mode, the stagnation
in coating thickness, it is unlikely to fail completely line is away from the blade tip and the coated
in putting down some layer of liquid on the liquid is subsequently subjected to prolonged
substrate. Still, flexible-blade coating can suffer shear which may break-up dislodged particle
from several instabilities that can seriously degrade aggregates. In the beveled mode, in contrast, the
coating uniformity. For instance, if a puddle of stagnation line is very close to the blade tip and
liquid occupies part of the wedge region instead downstream of the stagnation line the liquid is
of a turnaround flow that cleanly rejects the scarcely sheared at all. This is presumably why
excess liquid supplied to the blade, the resulting the beveled mode is more susceptible to streaks.
rotating flow is susceptible to instabilities, in Another reason may be that a blade in the
particular break-up into cells in the cross-web beveled mode is more susceptible to irregular
direction that can cause down-web striations. and erratic wetting past the tip which, in either
This sort of coating flaw has been investigated mode, could also induce striations and streaks.
most thoroughly in the context of high-speed
paper coating (Chapter 12d). Ifimproperly shaped,
12c.4 MEMBRANE COATING
installed, or loaded, flexible coating blades are
also prone to cause ribbing. In the stiff or beveled In membrane coating, also known as hydro-
blade-coating mode, ribbing can occur when the pneumatic coating (Pipkin 1986), a thin, uniform
beveled edge at the blade tip is at the wrong liquid film is metered by a highly flexible sheet,
angle and forms a diverging flow channel with or membrane, that is bent round to form a sector
the substrate. In the highly bent mode, ribbing of a cylinder and is pressurized pneumatically to
arises when the blade is bent so much that its force a compliant trailing wiper against the liquid
extremity near the tip forms a diverging section to be coated, as described in Section 12c.l above.
with the substrate wrapped around the curved This simple, yet surprisingly effective coating
back-up roll. In either case, at sufficiently high configuration seems to have evolved in parallel
speeds, more chaotic instability modes take over, in two different industries. For magnetic media
coating, Pipkin (1982a,b) started from an appli- that are shear-thinning, for orienting nonspherical
cator that consisted of a wiping film equipped particles, and for breaking up weak flocs in a
with a loading mechanism (Eichler and Brueck suspension. On the other hand, if the coating
1972) to develop a practical membrane coater liquid degrades under prolonged shear, short
that can even be adapted for simultaneous two- trailing films may be required.
sided coating (Pipkin 1982b, 1984). For coating
suspensions of clay and other materials on paper,
l2e.4.l MEMBRANE COATING ACTION
meanwhile, Taylor and Fridhandler (1985) modi-
fied a conventional blade coater loaded by a The basic action of a membrane coater is the
pressure hose by substituting a thin plastic film elastohydrodynamic metering that takes place
for the stiff blade. when the membrane is pressurized and forces a
The key components of a membrane coater, portion of the wiper film against the substrate
shown in Fig. 12c.lb, are an inflatable membrane which, in turn, is pressed against the rotating
with provision for pneumatic pressurization; a back-up roll. Pranckh and Scriven (1989a,b) (see
wiping film that wraps around the membrane also Pranckh 1989) have analyzed the elastohy-
and trails past it; a positionable mounting fixture drodynamics of membrane coating in some detail,
that holds both the membrane and the trailing and this entire section is largely based on their
film; and a rigid back-up roll that carries and work.
supports the coating substrate. On the upstream Where the wiper film and the substrate converge,
side of the membrane, the film may separate from a wedge-shaped region forms that dominates the
the membrane and is then attached separately. thickness of the liquid film coated downstream
The upper part of the trailing film together with very much in the same manner as in blade coating
the substrate from a wedge-shaped hopper into (see Section 12c.3 above). As a result, the rate at
which the coating liquid is supplied, either with which the liquid actually passes into the narrow
a simple feed pipe or a more sophisticated feed passage between the membrane-backed film and
device (e.g., Pipkin (1983); Berdinner and Throck- rigid-roll-backed substrate depends primarily on
morton (1984)). Not shown in Fig. 12c.lb are the the viscosity and possibly other rheological
end dams or other provisions that prevent the properties of the liquid, the speed ofthe substrate,
resulting liquid puddle from leaking out sideways and the pneumatic membrane loading which,
and thereby maintain a certain coating width. together with the hydrodynamic loads, determines
Also not shown are the edges of the rigid holder the shape of the converging wedge.
that seal the pressurized membrane. What is different from flexible-blade coating
In comparison with other coating devices that - and what causes a different operating response
must rely on precision machined parts, the to increasing loading - is that the metering action
strengths and weaknesses of membrane coating in the converging wedge zone can be made
are quite similar to those of flexible blade coating independent of the flow features downstream.
(cf Section 12c.3). A particularly attractive feature Depending on membrane pressurization and the
of the membrane configuration is that the liquid membrane position, the membrane-backed film
supply and distribution system is directly incor- can conform to the rigid-roll-backed substrate
porated in the coater head assembly (see Fig. over a sector of considerable length, as described
12c.1b). Also, the membrane and trailing film are in detail by Pranckh and Scriven (1989a). In this
inexpensive and are fairly easily and quickly so-called contact zone, the channel height between
replaced. Moreover, the shearing strain to which the trailing wiper film and the substrate is nearly
the coating liquid is subjected can be increased constant. If the contact zone is sufficiently long,
or lessened merely by changing the length of the a Couette flow establishes itself at a relatively
trailing film. This can be used to great advantage constant, elevated hydrodynamic pressure that
for achieving uniform coatings of suspensions balances the membrane pressure. The contact
Membrane coating 615
zone ends near the place where membrane curves

back to its downstream attachment on the rigid 60
holder. The trailing film conforms to the membrane
over a short separation region over which the
pressure falls to ambient. Downstream, the 50
membrane and film split and the remainder of
the trailing film adheres to the coated liquid E
~
layer. In this trailing film zone, the film is kept '"'"
Ql
40
under tension by viscous drag transmitted through ~

c:
o
the liquid layer from the moving substrate. :.E
I- 30
0)
.£
iii
12c.4.2 OPERATING BEHAVIOR o
()
20
The operating behavior of the membrane coating Cii
3:
process is quite simple, and makes for easy
process control. Figure 12c.6 shows predicted 10
and measured coating thicknesses as a function
of elasticity number El at different membrane
pressures. Elevated viscosities, higher coating
speeds or more pliable membranes - which all 0.4 0.8 1.2 1.6 2.0
increase the elasticity number - produce thicker Elasticity ~~umber

coatings. So do more pliable wiper films. Toward
higher elasticity numbers, the sensitivity ofcoating
Figure 12c.6 Characteristic operating behavior of
thickness to El diminishes. Increasing the pneu- membrane coating: wet coating thickness versus
matic pressure load shifts the curves of coating elasticity number at different membrane pressures.
thickness versus elasticity number to thinner The curves are theoretical predictions, the vertical bars
coatings, but does not alter the monotonic connect the highest and lowest of three measured
character of the operating curves. Thus, in coating thicknesses, and the symbols indicate
applications where a constant film thickness is their average. The elasticity numbers Es =
jiUL 2 12(1 - v2 )/E{)3 is defined here with the unsup-
desired, changes in speed or viscosity are quickly
ported arc length L of the membrane between its
compensated with changes in air pressure, making clamping points, the modulus E of the membrane, and
for a robust and readily automated coating process. its thickness {) (for a complete listing of all the design
The operating behavior of a membrane coater and operating parameters, see Pranckh and Scriven
and associated control strategy can easily be (1989a)).
understood by analogy with a lubricated slider
bearing: in the case of constant inclination, the Changing the position of the membrane holder
load carrying capacity of such a bearing rises as with respect to the back-up roll alters the amount
the lubricant thickness falls. In a membrane of engagement (i.e., the hypothetical interference
coater, this effect is advantageously adopted to of the undeformed, pressured membrane and the
lower the coating thickness by raising the air back-up roll), but has little effect on the coating
pressure load in the membrane (see Fig. 12c.6). thickness, unless the engagement is so small that
The higher that pressure, the more the liquid the liquid forces are too weak to deform the
pressure must increase in the converging wedge membrane substantially and make it conform to
leading to the contact zone and, hence, the the curved shape of the back-up roll. In that case,
steeper the positive pressure gradient becomes. the membrane acts more like a rigid metering
Therefore, a larger fraction of the liquid gets element and coating thickness does become
rejected and the coated layer becomes thinner. dependent on engagement (Pranckh and Scriven
1989a). Another exception arises when the down-web ridges of wavelengths substantially
substrate being coated is porous. In that case, larger than the wet film thickness, a film of nearly
the elevated pressure in the contact zone over uniform thickness can emerge, with the amplitude
which the membrane and the trailing film conform in thickness variations depending on elasticity
to the roll tends to impregnate the substrate with number.
coating liquid by pushing the liquid into the pore Even though the basic operating behavior of
spaces. The amount of impregnation depends, of membrane coating is simple and the resilience
course, on contact length. This length, in turn, to cross-web nonuniformities seems robust, the
is readily controlled by changes in membrane- window of successful operation can be limited
to-roll engagement. by several factors. For one, the dynamic wetting
Changing the length of the trailing wiper film line at which the liquid in the pool must push
past the membrane also has little effect on the away the air in contact with the moving substrate
final coating thickness (Pranckh and Scriven is quite unprotected and hence susceptible to
1989a). This finding is consistent with the notion massive air entrainment past a critical speed (the
that the elastohydrodynamic interaction in the data given in Chapter 3 for the plunge-tank
converging wedge zone sets the coating thickness, configuration probably provide a good first
as explained above. However, extending the estimate for that speed). As in many other coating
trailing film lengthens the trailing film zone in configurations, air entrainment may preferentially
which the liquid undergoes simple shearing flow start at the edges where some sort of mechanical
and thus raises the total shear strain imparted barriers delimit the pool in the cross-web direction
on the liquid. Hence, ifthe final properties of the and thereby put an additional restriction on
coated film depend on the total shear, they can practicable coating speeds. Depending on pool
be controlled independently of coating thickness. size and viscosity, the maximum coating speed
may be further limited by inertia-driven pool-flow
instabilities similar to those seen in so-called
12c.4.3 eOATING UNIFORMITY AND
short-dwell applicators used for high-speed paper
OPERABILITY
coating (Aidun 1990) (see also Chapter 12d). Such
A major attraction of membrane coating is its instabilities cause a segmentation of the recir-
potential to generate uniform coatings (Pipkin culating flow in the pool into three-dimensional
1982a,b; Pipkin 1984). Basicelastohydrodynamic cells. These, in turn, are prone to cause down-web
concepts suggest that, where the flow channel striations of coating nonuniformity (Triantafil-
converges more because of faulty applicator lopoulos and Aidun 1990). The recirculating flow
fabrication, substrate imperfections, or blemishes in the pool may also allow suspensions to
or debris on the back-up roll, the liquid pressure flocculate or agglomerate, and volatile solvents
rises, causing the highly deformable applicator to evaporate from quiescent parts of the free
to widen the passage locally and to increase the surface. The resulting, unwanted concentration
flow rate there. The resulting more or less ofinhomogeneities may intermittently be entrained
self-equilibrating action epitomizes the desired by the liquid being coated and cause unacceptable
ability of elastohydrodynamic coating devices to coating defects. A membrane-coated liquid film
compensate for less stringent mechanical tolerances may even be susceptible to the ubiquitous ribbing
than are typically built into rigid coating instability because, toward the trailing edge of
applicators. Pranckh (1989) (see also Pranckh the wiper film, the pressure gradient can be
and Scriven 1989b) attempted to quantify this slightly positive (Pranckh and Scriven 1989b)
action with a highly simplified model in which (for a detailed discussion of the ribbing instability
the membrane/film structure is approximated by in general, see Chapter 12a). A further limit on
linear two-dimensional plate theory. His prelimi- the practical applicability of membrane coating
nary results confirm that, for substrates with may be related to difficulties in coating very thin
Tensioned-web roll coating 617
liquid layers, associated for instance with inor- curiously little attention in the published literature.
dinately high membrane pressures needed to For that reason, only a sketchy overview of the
overcome the hydrodynamic lubrication pres- operating principles can be given in what follows.
sures, or with abrasive wear and scratching from The elastohydrodynamic interaction between
particulate-filled coatings. the tensioned web and the viscous liquid film
determines the height at which the web floats
above the coating roll. An estimate for that
12c.5 TENSIONED-WEB ROLL COATING height can be found in the literature for so called
As previewed in Section 12c.1, the essence of foil bearings in which a thin flexible film replaces
tensioned-web roll coating, often also called kiss one of the (nearly) rigid components of a
coating, is the elastohydrodynamic transfer of conventional bearing, forming an elastohydro-
part or all of a liquid film from a rotating dynamic system very similar to a kiss coater.
applicator roll onto a moving web that 'kisses' Foil bearings have been studied extensively, as
the roll without back-up support. Figure 12c.lc summarized by Gross (1980). Practical applications
shows the roll running in a forward mode, but include, for instance, magnetic recording devices
a kiss coater can also be operated in a reverse in which the magnetic tape is the moving foil
transfer mode. The positioning of upstream and and the recording/playback head acts as the shaft
downstream idler rolls (not included in the figure) or roll. Foil bearings are also a common occurrence
controls the angle over which the substrate wraps - and possibly a nuisance - on web lines where,
the applicator roll, usually between 5° and 15° at sufficiently high speeds, the web can become
(Grant and Satas 1984). Common applications 'air-borne' on smooth idler rolls. The characteristic
include adhesives, hot melts, and silicone coatings regions in foil bearings and in tensioned-web roll
on paper and plastic film (Booth 1970; Zink 1979; coaters are in essence the same: a wedge shaped
Grant and Satas 1984). Compared with other entrance zone that constitutes a high pressure-
roll coating configurations - which typically gradient metering zone, as explained for a blade
require expensive mechanisms accurately to coater in Section 12c.3 above; a constant channel
maintain a close gap between two high-precision zone of Couette flow at an elevated pressure; and
rolls and, at a given gap, are limited to a rather an exit zone. In kiss coating, of course, the flow
narrow range of coating thickness (cf Chapter in the entrance and exit zones is complicated by
12a) - a primary attraction of kiss coating is its the presence of a meniscus which could possibly
simplicity. In lieu of the gap, web tension and have some influence. Eshel and Elrod (1965)
wrap angle provide readily adjusted control calculated the constant channel height H for a
variables in kiss coating. They allow successful perfectly flexible foil bearing as
operation over wide parameter ranges, especially
for coating thickness. Standard handbooks suggest
speeds up to 5 m/s with viscosities between 50
H = 0.64 R(6Jl:) 2/3 (12c.14)
and 1000 mPa s and wet coating thicknesses as
R U
low as 25 Jlm (e.g., Zink (1979); Grant and Satas where is the roll radius, the foil speed, the Jl
(1984)). A potentially serious drawback of fluid viscosity, and T the foil tension. In 12c.14,
tensioned-web roll coating, on the other hand, the influence of bending stiffness ofthe foil is taken
is that the unsupported web becomes part of the to be negligible. With rising stiffness, the internal
metering device. Hence, variations in web char- stresses in the direction normal to the foil start
acteristics, such as its thickness or stiffness, or to carry some of the load (cf equation (12c.12)),
web instabilities, such as flutter or buckling, tend lowering the constant floating height H. However,
to degrade coating uniformity. the effect typically is small (Gross 1980).
In spite of its simplicity and practical import- Given the lack of more refined models that
ance, tensioned-web roll coating has received account for the details of the free-surface flows
at the inlet and outlet, the simple relation (12c.14) ceptible to additional instabilities related to
furnishes at least a leading-order estimate for the flexibility and deformability of the substrate.
lubricating film thickness between web and roll Pearson (1985) investigated the stability of the
in a tensioned-web roll coater, provided the angle liquid flow in forward kiss coating. He approxi-
of wrap is sufficiently large to allow for a mated the web as perfectly flexible, accounting
constant-channel zone between the inlet and only for the tangential stress resultants in equation
outlet regions. As is to be expected for an (12c.12). He described the flow with the lubrication
elastohydrodynamic coating system with given approximation, imposing ambient pressure and
geometry, the film thickness is set by the an ad hoc approximation to the film split as
competition between viscous pressure forces, J.lU, boundary conditions. For stream-wise disturb-
and mechanical restoring forces, in this case ances in the inlet zone, he calculated an ampli-
primarily the normal resultant from the externally fication factor of
imposed web tension T (in 12c.14, variations in
the tangential shear resultant N due to viscous
shear tractions are taken to be negligible). In the
case of kiss coating, both the foil and the roll are where 4>R is the length of the constant gap zone,
moving, and the speed that enters (12c.14) is and k and v are the real wavenumber and the
U = UWeb + URon ' i.e., the sum of the web and wave speed of the disturbance respectively.
roll speed. In either the foward or the reverse According to Pearson's result, disturbances in
mode, the thickness of the coated film that is the machine direction introduced in the inlet
metered through the contact region between the zone are always dampened. For cross-web
web and the roll is closely linked to the floating disturbances, in contrast, Pearson (1985) found
height H via the linear Couette velocity profile an appreciable positive amplification factor, i.e.,
between the two moving surfaces. A noteworthy cross-web disturbances grow with distance
special case arises for reverse kiss coating at a downstream.
speed ratio near I: I for which, according to
equation (l2c.l4) the floating height and hence 12c.6 TENSIONED-WEB SLOT COATING
the metered film thickness tend toward zero. This
result provides a simple explanation for why a In tensioned-web slot coating, a liquid film is
reverse kiss coater can be used to transfer a liquid metered out of a slot of a coating die onto a web
film completely from a smooth roll to a web in free span between two idler rolls (see Fig.
while wiping the roll almost perfectly clean, as 12c.ld). The web is wrapped around at least the
mentioned by Grant and Satas (1984). downstream lip of the die, forming an elastohyd-
It is to be expected that kiss coaters have rodynamic system with a viscous coating layer
several operating limits. For one, there are between the die lip(s) and the web.
practical limits to the web tension T that can be Tensioned-web slot coating shares many
applied, which limits the minimum coating similarities with tensioned-web roll coating
thickness for a given liquid and a given line speed, discussed in Section 12c.5 above, for it merely
in particular in the forward mode. Also, tensioned- replaces the applicator roll with a slot-coating
web roll coating most likely suffers from some die. The main advantage of the die over the roll
of the same instabilities as conventional rigid roll is that the liquid distribution system is fully
coating (cf Chapter 12a), but little has been enclosed and the final coating thickness is set by
documented in the open literature. Ribbing, for the liquid supply pump to the die, as in other
instance, is probably equally prominent, especially pre-metered coating systems.
in forward kiss coating; and reverse kiss coating Tensioned-web slot coating is also very closely
must exhibit some sort of 'cascade' instability. related to conventional slot coating, discussed at
Tensioned-web roll coating is furthermore sus- length in Chapter lla. A major advantage of
Tensioned-web slot coating 619
tensioned-web slot coating over standard slot studies, nevertheless, captures some ofthe essential
coating is the simplicity of the overall setup: there design and operating considerations. For instance,
is no precision back-up roll, nor is there a need the analysis shows how changes in web tension,
for a vacuum system to stabilize the coating bead. which alter the normal resultant forces generated
Moreover, the positioning of the die with respect by the curved web, must be offset by changes in
to the web path can be considerably less precise. the liquid tractions. With increasing tension, the
Instead ofthe narrow gap between a slot-coating lubrication-flow pressures generated by the die
die and the back-up roll- which must be pre-set lip downstream of the slot exit rise and the
and maintained with tight tolerances - the upstream free surface with its dynamic wetting
clearance between the web and the die lip that line is pushed further upstream. Changes in the
is necessary to coat a certain film thickness is flexural rigidity of the web also considerably
automatically accommodated by the elastohyd- influence the pressure field, and resulting web
rodynamics. Changes in tension, angle of wrap, deformation and upstream meniscus position.
and possibly even lip design may be necessary Likewise, changes in downstream lip design and
to cover wide ranges of film thickness, viscosity, the relative position of the two lips normal to
or speed but, if judiciously managed, allow for the web have a significant effect. Tanaka and
thinner coatings and higher speeds than may be Noda (1984), for instance, claim that for the
feasible with conventional slot coaters. Tanaka particular design shown in Fig. 12c.lc the angles
and Noda (1984), for instance, claim to have between the points ABC and ABD and the
achieved coating speeds of 5 mls for dry coatings distances AB and BC are ofparticular importance
only 5 ~m thick whereas, with a conventional because they all control the pressure profile
slot coater for the same liquid, the minimum under the substrate from D to A.
thickness was 20 ~m and the maximum speed Proper control of the position and shape of
1. 7- 2.5 m/s. The main disadvantage oftensioned- the upstream meniscus seems to be key to
web slot coating is similar to that of tensioned-web keeping the coating bead within the window of
roll coating: the web becomes part of the coating successful operation (Chino, Hiraki, Sato and
device and any nonuniformity, tension variation, Chikamasa 1988a; Shibata 1988). If the bead is
vibration, or instability in the web directly starved, for example due to an insufficient pumping
translates into poor coating uniformity. rate or too Iowa web tension, the meniscus with
Tensioned-web slot coating devices have been the dynamic wetting line is pulled downstream
in use for quite some time for less demanding and triggers premature air entrainment. On the
applications (e.g., Van Guelpen (1946)), but have other hand, the bead swells if the pumping rate
more recently achieved prominence for precision or the tension is too high, pushing the upstream
coating, especially ofmagnetic tapes and diskettes meniscus, including the static contact line, well
(e.g., Pipkin and Schaefer (1979); Tanaka and past the slot exit (e.g., point D in Fig. 12c.ld).
Noda (1984)). Very recently, tensioned-web slot Eventually, part of the liquid that is supplied to
coating technology has been extended to allow the die flows over and dribbles down on the
simultaneous depositon of two or more layers upstream side of the die. A substantial part of
(e.g., Chino, Hiraki and Sato (1988); Shibata et the range between these two operating limits if
al. (1992)). not the entire range can presumably also be
In relation to the large number of patent covered with relatively minor changes in the
publications on tensioned-web slot coating, the angle at which the incoming web arrives. In fact,
patents and other published literature provide careful control of that angle - and elimination
disappointingly limited insights into the elasto- of unacceptable web vibrations and instabilities
hydrodynamic mechanisms at work. Shibata and - may be the crux to making a tensioned-web
Scriven's (1986) (see also Shibata et al. (1992)) slot coater work: if the angle is too large, the
finite-element analysis of a few selected case bead starves and breaks up; if the angle is too
small, the bead flows over or, ifthere is insufficient action generates high pressures. These pressures
flow supplied to the die, the upstream lip contacts cause elastic deformation of the roll surfaces,
and scars the incoming dry web. which alters the converging geometry and thus
The sensitivity of the tensioned-web slot-coating affects the hydrodynamic pressure. It is the
process to web tension suggests that cross- relative balance of viscous pressure force and
web tension variations can produce substantial elastic restoring force that controls the flow in
coating thickness variations and even local coating the nip. Compared with these forces, surface
failure. On the other hand, controlled tension tension is insignificant in determining the flow
variations could also be used to compensate for rate and can be ignored for all cases but the rigid
skewed cross-web flow profiles generated by the roll limit at low capillary number.
die manifold, as demonstrated by Chino et al.
(1988b). These authors employed idler rolls that
could be tilted by as much as 25 mm per meter 12c.7.l NIP MECHANICS
width upstream and downstream of the coating
There are two ways of operating a deformable
head. Downstream of the coating head, another
nip - preset position (gap) or preset load (force).
idler was tilted to cancel out the induced variation
The first fixes the distance between the roll axes
in tension across the web.
using mechanical stops, such that the distance
between undeformed roll surfaces (gap) is specified
(the gap can be negative). The load which tries
tk.7 SQUEEZE-ROLL COATING to force the rolls apart is a function of the
operating parameters. The second method of nip
The essence of any squeeze-roll coater is the control is constant applied load (usually expressed
deformable nip between two cylinders, one or as pli, pounds/lineal inch, or N/m). One of the
both of which have an elastomeric covering. One rolls can slide towards or away from the opposite
version is illustrated in Fig. 12c.le. Squeeze-roll fixed roll, and is loaded either hydraulically or
coaters are widely used to apply waxes, hot melts, pneumatically up against the fixed roll. The gap
adhesives, silicone fluids, and the like onto paper between the rolls is now a function of operating
or plastic films. Another common example is the parameters.
size press on a paper making machine, which is In a manner analogous to Section 12a.2.1 for
used to impregnate the paper with water, binder the case of lubrication between two rigid roll
and pigment. Speeds up to 6 mis, viscosities in surfaces, the elastohydrodynamic lubrication of
the range from 10 to 5000 mPa s, and wet coating two cylinders can be approximated by a cylinder
thicknesses as low as 5 J.lm and as high as 100 J.lm loaded against a semi-infinite elastic half-space,
are possible (Booth 1970; Zink 1979; Satas 1984). where again we define an effective radius R e (cf
The main attraction of these coaters is that they equation (12a.2a» and an effective plane strain
are simple to operate and inexpensive. Some modulus E e
disadvantages are that the coating thickness is
sensitive to viscosity and speed, and the coated
1 1 - vi 1 - v~
layers exhibit a rough pattern that results from -=--+--
film-split instabilities (see Fig. 12c.2 above). Ee E1 E2
The mechanisms governing the metering action
of the roll nip are in essence the same as those The competition between forces is quantified by
governing the elastohydrodynamic lubrication two dimensionless parameters: the elasticity
of bearings and gears (Dowson and Higginson number and the load parameter,
1966). Fluid is dragged into the nip by the moving
roll surfaces through the action of viscosity, and El -- EJ-lU ,.va=~ (12c.14)
as the geometry is converging the squeezing R' rYl
e e EeR e
Squeeze-roll coating 621
The elasticity number is the ratio of viscous and Thus the ratio of hydrodynamic to elastic pressure
elastic force, while the load parameter is the ratio is proportional to
of applied load (W) and elastic restoring force.
W Wa
For a constant gap operation, the load parameter g3 = (jlUE e R e )1/2 = El1/2 (12c.17)
is replaced by the geometric factor H 0/R e , where
H 0 is the gap between undeformed roll surfaces The rigid roll limit is g3 = 0, while the dry contact
and takes on negative values for most cases of limit is g3 » 1.
practical interest. The dependent variable of In general, both hydrodynamic and elastic
primary interest in coating operations is the flow contributions must be accounted for. The flow
rate through the nip. It is equal to the local gap in the gap between roll surfaces determines the
between deformed roll surfaces at the point of pressure in the liquid which in turn enters the
maximum pressure, usually expressed as the normal traction boundary condition for the
dimensionless group H m/R e. Two other variables equations governing the elastic deformation of
of interest are the pressure and gap profiles, the solid boundary, summarized in Section 12c.2.
P(X)/E e and H(X)/R e , both as a function of the For infinitesimal deformation of a Hookean
flow direction coordinate X. solid, the gap profile becomes
There are two limiting behaviors which are
helpful to understand. The first is the rigid roll
limit, where hydrodynamic (viscous) forces
H(X) = H o + X 2
2R e
- ~ 7 P(s) In IX s- sjdS
nEe _ 00
dominate and elastic deformation is negligible
(12c.18)
so that the theory of Section 12a applies. The
second is the elastic dry contact limit, where elastic This can be combined with the Reynolds equation
forces dominate and hydrodynamic forces are to get a single integral equation for H(X) which
negligible so that Hertz contact theory, or an can be solved numerically (Herrebrugh 1968).
improvement upon it, applies. The relative The solution depends on the single dimensionless
importance of hydrodynamic and elastic forces parameter g3' and results in the dimensionless
can be measured by a single dimensionless group gap profile
g3' which is the ratio of the pressure generated
hydrodynamically in the liquid film to the pressure
n_ H(X)W _ H(X) Wa (12c.19)
- RejlO - ~EI
to produce elastic deformation (Johnson 1985).
The hydrodynamic pressure scale is derived from An accurate large-g 3 solution can be obtained if
lubrication theory (see equation (12a.3)). As the one assumes a nearly parallel gap profile combined
gap can be expressed in terms of the load as with approximate analysis of the wedge-shaped
H o = 4.9jlOR e/W(Coyle 1992), the hydrodynamic inlet zone to give (Hooke and O'Donoghue 1972)
pressure scales as
nm = 3.12 g~.8 (12c.20)
W 3/2 where nm is the value of n at the point of
Plub = (jlU)1/2 R (12c.15)
e maximum pressure and thus the dimensionless
film thickness. Equation (12c.20) agrees with
The pressure for dry contact can be approxi- Herrebrugh's solution for g3 > 1. Later studies
mated by Hertz's contact theory by Cudworth and Higginson (1976) and Hall and
Savage (1988a, b) used improved numerical
X)2)1/2 WE )1/2 methods but did not note a significant difference
P=Pmax ( 1- ( A ' P max = ( nR: ' in the results. A simple one-dimensional spring
model has also been shown to give similar
A = (4WR e )1/2 (12c.16a, b,c) predictions (Coyle 1988b). Comparison with rigid
nEe roll theory of Chapter 12a (cf Coyle (1988b))
shows that the low-g 3 limit gives lim = 6.0 if the over which the pressure profile is nearly Hertzian.
Reynolds boundary condition is used, lim = 0.0 These are the classic elastohydrodynamic gap
if the submerged boundary condition is used, profiles first measured by Crook (1961a,b). If, at
lim = 3.8 if the high-capillary number numerical constant load, the viscosity or speed (contained
solution to the flow is used. in El) is increased (reducing g3)' the pressure in
Another parameter which can be important is the inlet region increases and, as the area under
the thickness of the deformable cover of the rolls each pressure curve is equal, the pressure in the
(LjR.). If the surface deformations are ofthe same central nip region falls. At the same time the rolls
order of magnitude as the cover thickness, the move apart and the inlet region takes on a less
contact width will be reduced, the load more sharp gap profile.
concentrated, and the coating thinner. The relative The limiting cases for constant gap setting are
importance of the cover thickness can be evaluated different. Figure 12c.8 shows the pressure and
by calculating the Hertzian dry contact half-width gap profiles for two constant gap settings. At
(A in equation (12c.l6c)). For example, consider positive gaps, low values of El reduce to the rigid
a coater where R 1 = R 2 = 10 cm, and only one roll limit, while increasing El increases pressure
roll is rubber-covered. For soft covers, typical and deformation of the roll surface. At negative
conditions would be E = 1 MPa and W = gaps, the low El limit is dry contact, while
1750Njm (lOpli) so A = 0.91 cm. Typical roll increasing the viscosity or speed (increasing El)
cover thicknesses are 1.2 cm, so that the cover again increases the pressures and deformations.
thickness should be important. Hooke (l986a)
produced solutions of the appropriate equations
12c.7.2 COMPARISON WITH DATA
over a wide range of loads and layer thicknesses
by compiling and extending earlier results (Gupta Equation (12c.20) can be rearranged in terms of
1976; Hooke 1977, 1986b). Jaffar(1990) produced the operating variables as
similar results. Finite-element calculations of H m = 3.12 (JiO')0.6 W-O. 2 E;0.4 R~·6 (12c.21)
Coyle (1990) incorporated the effects of nonlinear
finite deformation of the elastic solid layer, and to give a dimensionally consistent theoretical
gave a thickness dependence somewhat weaker prediction of coating thickness.
than Jaffar's. Several authors have performed experiments
As roll covers for coating applications are using constant force operation of the nip and
typically made of rubber, which is a viscoelastic Newtonian oils. The most extensive data are
solid rather than a Hookean elastic solid, theory those of Smith and Maloney (1966), which can
based on the latter may not be adequate in be correlated (in SI units) by
describing a roll coating nip. The dry rolling H m =40 0.640'0.58W-O.34E-o.35Ro.58
contact of viscoelastic solids has been investigated .Ji e e
(Bapat and Batra 1984), but the only corresponding (12c.22a)

analysis for the lubricated contact so far attempted
but can also be correlated by
failed due to lack of numerical convergence at
relaxation times much lower than those of lim = 3.13 g~.71 (12c.22b)
experimental interest (Coyle 1990).
with a correlation coefficient of 0.997. Coyle
Results of finite-element calculations (Coyle
(1988a), presented data which are significantly
1990), which happen to be for the case of an
lower and can be correlated (in SI units) by
elastic roll against a rigid one, can be used to
show the details of the pressure and gap profiles. Hm = 4. 1( Ji 0')0.49 W- O.43 E-e O.41 Ro.e 42
Figure 12c.7 shows such profiles for a constant
(12c.23)
load operation. At high load (low El, high g3)
there is an extended nearly parallel gap region but which can not be correlated in terms of lim
0.0 r----.,,.,-.,---,,--r------...---------,----.,.----.,---,
0.0
rr.-
J:
ci.
ctI
~
I
0.0 I
I
I
I
I
I
I
I
/
0.0 L~_ __'_~_---'-~:::.::::;:===:;::===:I::::==~~/~_J
-0.4 -0.3 0.2
0.12
0.10
.... -- "-
"- ,
\
Dry Contact \
\
\
0.08 \ Hertz
\
w- EJ:1.0XIO~ "-
0: EI-:l.2xIO"
I
I' I
~ 0.06 EI.I.OXIO"
II
:0
<Il \\
<Il EI~.2xIO" \
",
\
~
a. EI:1.0xIO"
\\ \
0.04 EI~2xIO·7 \ \
\ \
I ,,
\
I ,
0.02 I \
I I
I ,,
I
I ,,
I
0.00
-0.4 -0.3 -0.2 -0.1 0.0 0.1
Flow Direction Coordinate, XlR.
Figure tk.7 Computed pressure and gap profiles in a squeeze-roll nip as a function of elasticity number at
constant load. The dry contact limit is approached as El is decreased (W = 0.0268, E 2 /E[ = 00, R 2 /R[ = 1.67,
U 2 /U[ = 1, L/R[ = 0.167).
0.0 I I i ' , , , I I I
0.020 I i \\\' i I
0.Q15 0.0
~. EI.'.0X10"
0.010
~. 0.0
ci ci
os os EloJ.2Jll0·
Cl Cl
X1O
•
0.005 ~ / ~ 0.0
EI.l.o.l0·'
~ k
UXlU
I ",.,. :S ' I
0.000 ' " " , , 0.0 ' ' , I , ' I
-<l.5 -0.4 -<l.3 -<l.2 -<l.1 0.0 0.1 0.2 -<l.5 -<l.4 0.1 0.2
Flow Direction Coordinate, X/R.
0.16 , , I I::;::O=C<:;:: i
0.05 , ' I ' I
0.14
0.04
0.12
, 0.10
l!:!' 0.03 EI.1.0X10"
[l, ~
e EloJ.2lc'O· e 0.08
::l
~ l:l
gJ EI.1.0x10" l!!
Ii: 0.02 [l, 0.06
0.04
0.Q1
0.02
0.00 -=:::=:==:J " \\ \, \ \J

I
0.00 t:::::: ---- ~\. \ I -<l.5
=-==== ---
-<l.4 -0.3 -<l.2 -<l.1 0.0
-<l.5 -0.4 -<l.3 -0.2
=2s:-<l.1 0.0 0.1 0.2
Flow Direction Coordinate, X/R, Flow Direction Coordinate, X/R.
(a) (b)
Figure 12c.8 Computed pressure and gap profiles in a squeeze-roll nip as a function of elasticity number at constant gap (E 2IE! = 00, R2IR! = 00,
U 2IU! = 0, LIR! = 0.1). (a) At positive gap settings, the rigid roll limit is approached as El is decreased (HoIR! = +0.033). (b) At negative gap
settin~s, the dry contact limit is approached as El is decreased (Hn/R, = -0.033).
versus g3 nor as hm versus Wa and El. The amount of elastic deflection, may correspond to
exponents in the correlations are all quite similar, large relative changes in the film thickness.
indicating agreement in terms of the sensitivity Adachi et al. (1988) and Kang, Lee and Liu
ofcoating thickness to changes in the parameters. (1991) present coating thickness data for squeeze-
Varnam and Hooke (1977) and Swales et al. roll nips operated at pre-set constant gaps. Their
(1972) present limited data which also fall results, alongside theoretical predictions (Coyle
significantly lower than equation 12c.22. 1990), are plotted in Fig. 12c.10. For fixed
Figure 12c.9 shows a direct comparison of positive gaps, the rigid roll limit is evident as the
results of finite element calculations (Coyle 1990) plateau of coating thickness at low elasticity
with some of the experimental data. The agreement number (H m = 0.65 H 0)' At high elasticity number
between theory and data of Smith and Maloney the coating thickness begins to increase as viscous
(1966) is excellent. The data of Coyle (1988a) forces overcome the resistance to elastic defor-
show the same sensitivity to viscosity, speed and mation. For fixed negative gaps there is, of
load, but the coating thicknesses are lower. The course, no rigid roll limit and the coating thickness
latter data do not correlate with the three steadily decreases as elasticity number decreases.
dimensionless groups available (Wa, EI, LIRe), Making the gap setting more negative reduces
indicating that there is another parameter of the coating thickness, though the sensitivity to
importance. An obvious candidate is viscoelasticity even further reductions in gap decreases. The
of the rubber roll covers. As the deflection of the agreement of theory and experiment is good,
solid boundaries is large compared to the film except for one set of data.
thickness, small changes in the elastic response For negative gaps and low EI, equation (12c.20)
of the solid, and thus small differences in the can be rearranged, eliminating the load W in
OW = 2,452 Nlm
oW = 5,605 N/m
~ W = 12,262 N/m
2
E 10
::1.
0
E
J: 0 0
rJ) 0
rJ)
Ql 0 0 ~
c: 00 0
-""
<..> 0
:c 0 ~
~
t-
~~
.§ ~
u::
~ 10'
t- ~
"0 0
0
10"
Viscosity"Speed Il"U, Pa·m
Figure 12c.9 Computed (lines) and measured (symbols) flow rate in a squeeze-roll coater as a function of
viscosity, speed, and load; (£1 = 3 MPa, £2/£1 = 00, R 2 /R 1 = 1.67, U 2 /U 1 = 1). The dotted lines represent
data of Smith and Maloney (1966), equation (12c.22b).
Ho=0.35mm 0
H,=-0.5mm
*
10-8 10.7 10-6
Elasticity Number, Es
Figure 12c.l0 Computed (curves) and measured (symbols) flow rate in a squeeze-roll coater as a function of
elasticity number and gap setting. The data of Adachi (1988), represented by open symbols, are for a
rubber-covered roll against a stationary flat plate. The data of Kang, Lee and Liu (1991) (*) are for a
rubber-covered roll and a steel roll at equal speeds (H o = -0.3 mm, upper set for 60 durometer, lower set for
90 durometer).
favor of the gap setting H 0 12c.7.3 OPERABILITY
Operability limits for a squeeze-roll coater are

not as clear-cut as for its rigid-roll counterpart.
Kang, Lee and Liu's data fit the expression As mentioned earlier, there is always a film-split
pattern (similar to ribbing) for loaded systems
producing thin coatings. Starvation of the nip is
nearly impossible for pan-fed systems, although
where the latter gap and modulus exponents are puddle-fed devices will tend to coat a narrower
a result of the one data set that deviates significantly width stripe of constant coating thickness if the
from theoretical predictions. feed rate is too low. Flooding of the nip is
Kang, Lee and Liu (1991) also present data possible, which might result in excessive edge
for non-Newtonian fluids. Shear-thinning lowers beads or overflow of the ends of the roll nip. As
both the coating thickness and the sensitivity to nips always tend to produce a rolling bead or
speed - trends that would be expected. Viscoelas- puddle of excess coating liquid upstream, the
ticity of the liquid increases both the coating possibility of vortex instability and air entrainment
thickness and the sensitivity to speed, but also exists, similar to the rigid roll cases in Chapter
causes large ribbing defects; results analogous to 12a. For example, a horizontal size press on a
those for rigid rollers. paper-making machine can have turbulent
Gravure coating 627
splashing flow in the puddle, leading to coating speed. The primary advantage of gravure coating
defects. is that very thin films are attainable at high
speeds and with consistent overall uniformity.
The main disadvantage is that it is difficult to
12c.7.4 PRACTICAL IMPLICATIONS
get films free of local nonuniformities related to
The most obvious difference between squeeze- the engraved pattern.
roll and rigid-roll coating, besides the much In all configurations, the key component that
thinner coatings, are the difference in exponents distinguishes gravure coating from other roll
in the above coating thickness equations. For coating methods is the gravure cylinder, whose
example, the coating thickness is proportional chrome-plated or ceramic surface is engraved
to the gap for rigid rolls, but only depends on with cells or grooves of a specific, precisely
the 0.47 power of the gap (experimentally) for a maintained volume. Because the coating thickness
deformable-roll system. These exponents indicate applied to the web is strongly tied to the engraved
reduced sensitivity to operating parameters, roll volume, gravure could be classified as a pre-
misalignment, roll run-out and other inaccuracies metered coating technique. It is nonetheless
and perturbations to the system. These improve- included in this chapter here because it includes
ments are analogous to those of blade or two elastohydrodynamic processes: flexible-
membrane coating compared with rigid knife blade doctoring and squeeze-roll nip transfer.
coating. Both can have a dramatic effect on the appearance
of the remnant gravure pattern in the coated
layer, and can also be used to adjust coating
12c.8 GRAVURE COATING
thickness to some extent. In differential speed
As described in Section 12c.l, the essence of and reverse gravure coating, moreover, adjust-
gravure coating is the filling of cells or grooves ments in speed ratio provide the option for
on a gravure cylinder with liquid, and the transfer substantial changes in coating thickness at a
of part of that liquid onto a substrate or fixed cell or groove volume.
intermediate transfer roll. Gravure coating has The pattern of cells or grooves is engraved
its origins in rotogravure printing, but has since onto the surface of the roll either by mechanical
evolved into many configurations (Booth engraving (knurling), by chemical etching, or by
1970,1990; Satas 1984). In direct gravure coating, electromechanical engraving (Satas 1984). Figure
illustrated in Figure 12c.1f, the liquid is transferred 12c.ll shows two versions of mechanical engraving
directly to the substrate backed up by the rubber (trihelical grooves and quadrangular cells) and
nip roll. In offset gravure coating, the gravure one example of electromechanical (or electronic)
cylinder transfers liquid to an intermediate, engraving. Knurling uses a rotating tool which
so-called offset roller which, in turn, applies the is run over the surface of the cylinder under
coating to the substrate at a second nip. Both pressure to mechanically form depressions in the
modes can furthermore be operated with counter- surface that are an image of the tool surface.
rotating rolls at matching nip surface speeds Electronic engraving uses a radially vibrating
(forward gravure) or at differential speeds (dif- diamond stylus to cut cells into the roll surface
ferential gravure). The rolls can even be co-rotating as the roll is rotated on a lathe and the stylus
to generate speeds at the nip in the opposite slowly translated across the roll's axis. The
direction (reverse gravure). Gravure techniques 'pitch', 'cell count', or 'screen' is the number of
are very common for applying silicones, adhesives, grooves or cells per unit length measured
magnetic recording layers and many other coatings perpendicular to the pattern angle, and typically
(Booth 1970,1990; Satas 1984). Typical conditions ranges from 4 to 160 cells/cm. The 'volume factor'
are I to 50 ~m for coating thickness, 15 to is the volume of the cells per unit surface area
1500 mPa s for viscosity, and up to 10 m/s for of the roll, and typically ranges from 4 to 300 ~m
(c)
Figure 12c.ll Common cell patterns used for gravure rolls, showing top views and cross-sections. The land
areas are shaded, and as shown here the roll axes are horizontal: (a) tn-helical, knurled; (b) quadrangular,
knurled; and (c) electro-mechanical engraving.
(cubic microns per square micron of surface). The initial application of excess coating liquid to fill
flat surface between the cells is called the land the cells. The second is the doctoring of the
area, and the angle between the sides of the cells gravure roll by the flexible blade, which removes
is called the stylus or tooth angle. excess coating from the land areas. The third is
No matter what the machine configuration or the nip flow and cell transfer process by which
engraving type, there are three fundamental part of the coating liquid in the cells is applied
operations in gravure coating that must be to the web. These three processes are discussed
understood. The first is the feed, which is the below. The book Gravure Process and Technology
Gravure coating 629
(Gravure Association of America 1991) provides the average coating thickness and the closest
detailed descriptions of machine configurations approach distance between the land and the
and gravure applications. blade. The theory predicts two fundamental
regimes ofoperation which depend on the balance
between roll roughness and blade stiffness. For
12c.8.l FEED
shallow grooves or a highly-flexible blade, there
The purpose of a feed system is simply to fill the is a large clearance between blade and land, a
cells with coating liquid. The simplest method is significant amount of coating is left on top of
a pan of liquid in which the gravure cylinder is the land, and the response of the system is in
partially immersed. Entrapment of air in the cells essence the same as for a smooth roll. This regime
will determine the upper limit of speed and corresponds to either small volume factors (and
viscosity with the feed system. The limit will high pitch) or high elasticity numbers. For deep
depend on the cell geometry and the rheology grooves or a stiff blade, there is a very small
of the coating liquid. Enclosed feed systems and clearance between blade and land, all the
fountain dies have been developed to overcome hydrodynamic force on the blade is concentrated
pan feed problems associated with high speeds, over the land, and the flow is metered almost
such as air entrainment, foam generation, large exclusively by the cell volume. In this regime,
volume liquid holdup and excessive solvent which corresponds to large volume factors (and
evaporation (Gravure Association of America low pitch) or low elasticity numbers, the average
1991). There exists no published fundamental film thickness no longer passes through a
understanding, experimental or theoretical, of pronounced minimum and a maximum as the
the feed step. blade loading is increased. Rather, as the blade
tip approaches the land surface the average
coating thickness on the gravure roll becomes
12c.8.2 DOCTORING
much less sensitive to the load and is more or
The doctoring ofa coating from a smooth surface less equal to the cell volume per unit area.
using a flexible steel blade is covered in detail in Toward higher loads, the cells are left only
Section 12c.3, and is also addressed in Chapter partially filled. The results for tri-helical patterns
12d. On smooth surfaces, the response of coating should be qualitatively valid for other patterns,
thickness to increases in blade loading follows a though the exact amount of liquid remaining in
characteristic, nonmonotonic pattern with a the cells depends on a pressure-driven component
minimum followed by a maximum (Fig. 12c.5a). of flow, and thus will be different for discrete
At any load, increasing the elasticity number (i.e. cells and continuous grooves.
the ratio of viscous to elastic forces) lifts the blade Some of this behavior has been observed
and increases the coating thickness (Fig. 12c.5b). experimentally. Lindblad, Tift and Watson (1988)
In gravure coating, it is often desired that the observed the minimum in coating thickness as
land areas of the cylinder are doctored cleanly. blade force increases for high pitch engravings
The insights gained from Section 12c.3 suggest and flexible (high El) elastomeric doctor blades.
that wiping of the lands would be most complete Their data at high loads showed a much steeper
toward the upper end of the stiff-blade regime increase in thickness than would be expected
where blade-coated films on flat surfaces can from their data of blade penetration into dry
become extremely thin. grooves, illustrating the significant hydrodynamic
The theory offlexible blade doctoring has been lifting of the blade which occurs when operating
extended to include trihelical engraved roll surfaces in the bent-blade mode. They also measured the
(Kistler 1988). The shape of the blade is only poorer doctoring (thicker coated films) achieved
weakly affected by the presence of the grooves, with longer blades and with blades made from
but the grooves are important in determining an elastomer with a slower relaxation time.
The data of Patel and Benkreira (1991) illustrate The amount of coating carried by the gravure
similar effects but also include variations in roll into the transfer nip is thus controlled by
elasticity number and volume factor. A normalized the metering action of the doctor blade. This
excess coating thickness can be defined as the amount of metering and its sensitivity to changes
difference between liquid and engraved vol- in operating conditions can be large or small,
ume/area, normalized by the engraved vol- depending on the regime in which the doctor
ume/area. Figure 12c.12 shows the approximate blade is operating. In this manner the doctoring
excess coating thickness as a function of blade can affect both quantity and quality of the
loading and elasticity number, replotted from coating transferred to the web.
Patel and Benkreira's original data. At low
elasticity numbers, here corresponding to low
12c.8.3 CELL TRANSFER OR PICK-OUT
viscosity and speed, there is no excess coating
thickness and the coating volume is determined The transfer of the coating from the gravure roll
solely by the cell volume. At high elasticity to the web can be thought of in terms of two
numbers there is substantial excess coating on limiting regimes. If an excess ofcoating is supplied
the roll, and the increase in thickness with blade to the nip, a rolling bank of liquid is formed, the
coating indicates that conditions correspond to coating on the web is continuous, and the process
the intermediate region of bent-blade operation. is closely related to squeeze-roll coating. [f the
The high elasticity number curves also show how proper amount of coating is supplied to the nip,
rolls with higher volume factors tend to operate printing of the gravure pattern will occur. This
more towards the regime where the lands are leads to two limiting views of the fluid flow. The
scraped clean. first case is the splitting of a continuous liquid
3.5 , - - - - - - , - - - - - - - - , - - - - - - - - - r - - - - - - - ,
3.0
~2.5
f/)
f/)
Cl>
~ 2.0
~
Cl
.~ 1.5
~
()
f/)
f/)
~ 1.0
LJ.J
0.5
Low Es
0.0 _ ___'__ _ _ _---'---_ _ _ _
'-----_~ ~ ~ _ ____=_'
.l....-_~
1.0 2.0 3.0 4.0 5.0

l/gap (mm". i.e. Increasing Blade Load)
Figure 12c.12 Experimental measurements of excess coating thickness on three gravure rolls as a function
of blade loading and elasticity number (data replotted from Patel and Benkreira (1991».
Gravure coating 631
film that occurs in all roll coating processes. The a given volume factor, an equivalent gap between
second is the uniaxial stretching and rupture of smooth rolls which would give the same coating
discrete liquid filaments. The second view especially thickness as calculated, and from this gap a
emphasizes the importance of viscoelastic and natural frequency of ribbing is found. If the pitch
elongational response of the coating liquid. In is chosen to match that ribbing frequency,
reality, this view is somewhat simplistic because single-line pickout will occur up to a critical
other more subtle effects complicate matters. speed. If the pitch is much higher than the natural
The experimental work of Pulkrabek and ribbing frequency, the competition between the
Munter (1983) and Munter (1990) describes the natural ribbing frequency and roll pattern will
transfer operation and the effects of design and result in unstable multiple-line pick-out. Third,
operating variables. While this work concentrated the ratio of land to volume factor is important.
on tri-helical patterns, there is likely to be some High values encourage printing, and worsen the
parallels with other patterns, such as cells. Most leveling process (small lines of coating that are
of the results are furthermore limited to direct widely spaced).
forward gravure coating. They illustrate the two Cell-transfer is also affected by operational
modes of transfer, referred to as merged and variables, though these are secondary compared
printing. There are three major problems which with the gravure pattern design. Operating with
can arise during transfer leading to flaws on the a rolling bank on the upstream of the nip is the
coated product. The first, called flashing, occurs primary means of adjusting the pick-out, creating
when one or more grooves are only partially merged pick-out patterns which level best. The
picked out, leaving a bare region on the web. bank can be adjusted by supplying excess coating
The second, called combined or multiple-line from the doctor blade or by increasing the nip
pick-out, occurs when liquid from two or more pressure to increase the penetration of the web
adjacent grooves is combined to form a single into the grooves. The pick-out can also be
line of coating on the web, thus doubling (or adjusted by changing the take-off angle of the
tripling) the line spacing on the web. This type web or the viscosity of the coating, both of which
of pick-out is not stable, does not level properly, can affect the natural ribbing frequency (Kistler
and can lead to other flaws such as pinholes and and Munter 1986). Generally, operational variables
craters. The third problem occurring at high are more useful in controlling pick-out stability
speed is misting, where fine droplets of coating than in controlling coating thickness. Further
are ejected from the nip. evidence of insensitivity of coating thickness to
The most important parameter for controlling operational parameters is found in the results of
the type and stability of pick-out is the engraving a study of reverse gravure coating with a tensioned
pattern design (Pulkrabek and Munter 1983; web (Benkreira and Patel 1993). Their data
Kistler and Munter 1986; Munter 1990). Three suggest that the thickness of the coating liquid
design variables are critical. First, the volume applied to the web is about 33% of the engraved
factor is the primary means of controlling the volume, independent ofengraving type (tri-helical,
average coating thickness. The amount ofcoating pyrimidal or quadrangular), ratio of web to
transferred to the web is around 60% of the cell gravure roll speed, or liquid viscosity. The large
volume for forward direct gravure coating (see difference in applied coating thickness from that
also Lindblad, Tift and Watson (1988)). If excess obtained in forward direct gravure coating needs
coating is supplied to the nip, a rolling bank is further investigation. Limited data at low speeds,
formed on the upstream side. In most cases the obtained for forward gravure and reverse gravure
web is actually pressed into the grooves or cells with an impression roll, are consistent in terms
to contact the liquid. Second, the pitch (for a oflow sensitivity ofcoating thickness to operating
given volume factor) is the primary means of variables (Stasyuk and Peglovskii 1981; Stasyuk,
controlling the type and stability of pickout. For Peglovskii and Fedotkin 1983).
Model studies have also been performed using elastohydrodynamic coating methods also sheds
fluids that do not level and thus preserve the light on when and why these methods are
pick-out pattern for subsequent analysis (Munter advantageous over their rigid counterparts.
1990). In terms of amplitude of the remnant In general, there are three main advantages to
engraving pattern, the coating quality is improved using an elastohydrodynamic coating method
by: (1) increasing the pitch; (2) flooding the rather than rigid-type applicators. First, elas-
inlet; (3) not running a highly loaded nip, or tohydrodynamic systems are less sensitive to
(4) running reverse rather than forward gravure mechanical tolerances, both spacial and time-
coating. dependent. The former results in less sensitivity
While many of the above results are limited ofcross-web coating uniformity to variables such
to tri-helical engravings in forward direct gravure as roll alignment or bowing. The latter results
coating, they nevertheless illustrate the limiting in lower sensitivity to roller run-out, vibration,
regimes of operation and the general effects of web thickness variations, etc. This decreased
engraving design and operating parameters. In sensitivity also means that the acceptable range
addition, the strong influence ofcoating rheology of machine settings and variability in setup is
on the operation of gravure coating is well larger than in other coating methods. The second
recognized in the trade literature but is not well advantage is the ability of elastohydrodynamic
understood. coating methods to coat very thin coatings with
simpler, easier to operate, and less expensive
machinery than is needed for other precision
12c.8.4 PRACTICAL IMPLICATIONS
coating techniques. The third advantage of
The main strength of gravure coating is its ability elastohydrodynamic systems is their development
to apply thin coatings at high speed. The coating of a long extended high-pressure nip region,
thickness is nearly proportional to the engraved which is primarily responsible for the lower
volume, so that changing coating thickness more sensitivity cited above. This region also serves
than a small amount requires changing the to damp unsteady planar inlet disturbances
gravure roll. The uniformity of the coating is (Pearson 1985) and thus helps coating uniformity.
strongly influenced by engraving design and Elastohydrodynamic coating methods also have
coating formulation rheology, but there are few some disadvantages. The most severe is coating
published guiding rules to assist the practitioner uniformity, especially for demanding product
in specifying either. This uncertainty is com- applications. Squeeze-roll and tensioned-web roll
pounded by the additional influences ofdoctoring, coating always result in a film-split pattern on
feeding, solvent evaporation and rubber roll the coated surface, and may not damp steady
response, among other factors. ribbing-type inlet disturbances (Pearson 1985).
Flexible-blade coaters and membrane coaters
tend to trap particles that create streaks in the
12c.9 SUMMARY AND CONCLUSIONS
coating. Gravure coaters create a patterned
This chapter describes and analyzes some of the coating depending on the engraving pattern and
most common elastohydrodynamic coating fluid rheology. All elastohydrodynamic coating
methods. Their behavior, in terms of coating methods are sensitive to variations in elastic
thickness and sensitivity to operating conditions, properties. Rubber covers on rolls will age,
is fairly well understood. It results from the embrittle or swell in response to exposure to
balance between viscous hydrodynamic forces solvent, oils or ozone. Tensioned web coaters are
generated in converging drag flows and elastic sensitive to web tension control and the stiffness
restoring forces of the deformable boundaries uniformity and straightness of the web. In addition,
confining the flow field. The understanding gained application of thin coatings requires higher
from the theoretical and experimental study of tensions which can stretch or tear the web.
References 633
Ultimately, the advantages and disadvantages of IGC Coating Systems Conference, Boston, August,
every coating method must be carefully weighed Bis-Cap International, Newtonville, MA.
in terms of the requirements of the specific Carvalho, M. S. and Scriven, L. E. 1993. Effect of
deformable roll cover on roll coating. Proceedings
product to be manufactured.
of 1993 TAPPI Polymers, Lamination, and Coating
Even though the basic mechanisms, advantages, Conference. Atlanta: T APPI Press.
and disadvantages are quite well laid out, several Chino, N., Hiraki, Y. and Sato, T. 1988. Coating
essential gaps remain in our knowledge of apparatus. US Patent 4,854,262.
elastohydrodynamic coating systems. The oper- Chino, N., Hiraki, Y., Sato, T. and Chikamasa, N.
ation of a squeeze-roll coater depends on the 1988a. Method of applying a liquid to a moving
nonlinear, time-dependent elastic deformation of web. US Patent 4,717,603.
Chino, N., Sato, T., Tanaka, F., Fukumura, K. and
an elastomer layer, which remains poorly under-
Hiraki, y. 1988b. Thickness correcting coating
stood. The details ofthe operability of a tensioned method. US Patent 4,776,997.
web coater with respect to geometry has not been Coyle, DJ. 1988a. Experimental studies offlow between
thoroughly explored. Most aspects of gravure deformable rolls. AIChE Spring National Meeting,
coating are poorly understood, including filling, New Orleans, March.
doctoring, and 'pick-out' ofcells of both Newtonian Coyle, D. 1. 1988b. Forward roll coating with deformable
and non-Newtonian liquids. As all elastohyd- rolls: A simple one-dimensional elastohydrodynamic
model. Chem. Eng. Sci. 43(10): 2673-2684.
rodynamic coating methods have a converging Coyle, D. J. 1990. Nonlinear theory of squeeze-roll
wedge inlet region on which the coating thickness coating. AIChE Spring National Meeting, Orlando,
depends, all should be sensitive to the viscoelastic March.
forces developed by liquids in such a rapidly Coyle, D. J. 1992. Roll coating. In Modern Coating
accelerating velocity field. Finally, the details and Drying Technology, eds E. Cohen and E. Gutoff.
controlling the uniformity of the coated films are New York: VCH Publishers.
not well understood. Crook, A. W. 1961a. Elastohydrodynamic lubrication
of rollers. Nature. 190: 1182-1183.
Crook, A. W. 1961 b. The lubrication of rollers. II.
Film thickness with relation to viscosity and speed.
REFERENCES Phil. Trans. R. Soc. Lond. 254: 223-236.
Cudworth, C. J. and Higginson, G. R. 1976. Friction
Adachi, K., Ishuzu, Y. and Nakamura, R. 1988. of lubricated soft surface layers. Wear. 37: 299-312.
Coating flows between a rotating deformable roller Dowson, D. and Higginson, G. R. 1966. Elasto-
and a stationary flat plate. AIChE Spring National Hydrodynamic Lubrication. Oxford: Pergamon Press.
Meeting, New Orleans, March. Eichler, W. and Brueck, R. 1972. Casting apparatus
Aidun, C. K. 1990. Principles ofhydrodynamic instability with flexible wiper film. US Patent 3,688,738.
in coating systems. 1990 TAPPI Coating Conference Eklund, D. E. 1984. Influence of blade geometry and
Proceedings. pp. 275-298. Atlanta: TAPPI Press. blade pressure on the appearance of a coated
Akesson, R. and Olsson, A. 1988. Neue Billiblade- surface. TAPPI J. 67(5): 66-70.
Konstruktion verbesserter Qualitiit und Runnability, Eshel, A. and Elrod, H. G. 1965. The theory of the
Wochenblatt Papierfabrikation. 8: 352. infinitely wide, perfectly flexible foil bearing. Trans.
Bapat, C. N. and Batra, R. C. 1984. Finite plane strain ASME, J. of Basic Eng. 87: 831-836.
deformations of nonlinear viscoelastic rubber- Fluegge, W. 1960. Stresses in Shells. Berlin: Springer.
covered rolls. Int. J. Num. Meth. Eng. 20: 1911-1927. Grant, O. W. and Satas, D. 1984. Other knife and roll
Benkreira, H. and Patel, R. 1993. Direct gravure roll coaters. In Web Processing and Converting Technology
coating. Chem. Eng. Sci. 48(12): 2329-2335. and Equipment, ed. D. Satas. pp. 60-80. New York:
Berdinner, E. J. and Throckmorton, J. P. 1984. Coating Van Nostrand Rheinhold.
apparatus. US Patent 4,475,478. Gravure Association ofAmerica. 1991. Gravure Process
Booth, G. L. 1970. Coating Equipment and Processes. and Technology. Rochester NY: Gravure Association
New York: Lockwood Publishing Company. of America.
Booth, G. L. 1990. Coating converting processes: Gross, W. 1980. Fluid Film Lubrication. New York:
Traditional methods and new uses. 21st Annual Wiley.
Gupta, P. K. 1976. On the heavily loaded elastohyd- Nakazawana, T., Morita, H. and Booth, G. L. 1987.
rodynamic contacts of layered solids. J. Lubric. The vari-dwell-time blade coater.1987 TAPPI Coating
Technol. 98: 367-374. Conference Proceedings. Atlanta: T APPI Press.
Hall, R. W. and Savage, M. D. 1988a. Two-dimensional Patel, R. and Benkreira, H. 1991. Gravure roll coating
elastohydrodynamic lubrication. Part 1: The asso- of Newtonian liquids. Chem. Eng. Sci. 46(3): 751-756.
ciated dry contact problem. Proc. Instn. Mech. Pearson, J. R. A. 1985. Mechanics ofPolymer Processing.
Engrs. 202(C5): 347-353. London and New York: Elsevier.
Hall, R. W. and Savage, M. D. 1988b. Two-dimensional Pipkin, D. J. 1982a. Apparatus for forming a coating
elastohydrodynamic lubrication. Part 2: Solution on a substrate. US Patent 4,345,543.
of the line contact problem. Proc. Instn. Mech. Pipkin, D. J. 1982b. Method and apparatus for
Engrs. 202(C5): 354-360. forming a coating on both sides of a substrate. US
Herrebrugh, K. 1968. Solving the incompressible and Patent 4,327,130.
isothermal problem in elastohydrodynamics through Pipkin, D. J. 1983. Coating apparatus and method.
an integral equation. J. Lubric. Techno!' 90: 266. US Patent 4,387,124.
Hooke, C. J. 1977. The elastohydrodynamic lubrication Pipkin, D. 1. 1984. Method for forming a coating on
of heavily loaded contacts. J. Mech. Eng. Sci. 19(4): a substrate. US Patent 4,442,144.
149-156. Pipkin, D. J. 1986. Membrane coating. AIChE Spring
Hooke, C. J. 1986a The elastohydrodynamic lubrication National Meeting, New Orleans, April 7-10,1986.
ofa cylinder on an elastomeric layer. Wear. Ill: 83-99. Pipkin, D. J. and Schaefer, D. W. 1979. Method for
Hooke, C. J. 1986b. A note an the elastohydrodynamic applying a viscous liquid to a substrate. US Patent
lubrication of soft contacts. Proc. I nstn. M echo Engs. 4,142,010.
200(C3): 189-194. Pranckh, F. R. 1989. Elastohydrodynamics in coating
Hooke, C. J. and O'Donoghue, 1. P. 1972. Elastohyd- flows. PhD thesis, University of Minnesota, Min-
rodynamic lubrication of soft, highly deformed neapolis, MN.
contacts. J. Mech. Eng. Sci. 14(1): 34-48. Pranckh, F. R. and Scriven, L. E. 1988. The physics
Jaffar, M. 1. 1990. Two-dimensional elastohydrodynamic of blade coating of a deformable substrate. TAPPI
lubrication of elastic strips. Wear. 139: 335-350. J. 73(1): 163-173.
Johnson, K. L. 1985. Contact Mechanics. Cambridge Pranckh, F. R. and Scriven, L. E. 1989a. Elastohyd-
University Press. rodynamics of membrane coating. University of
Kahila, S. J. and Eklund, D. E.1978. Factors influencing Minnesota Supercomputer Institute Report No.
the coat weight in blade coating with beveled blades: 89/83.
theory and practice. 1978 TAPPI Coating Conference Pranckh, F. R. and Scriven, L. E. 1989b. Analysis of
Proceedings. pp. 13-29. Atlanta: TAPPI Press. membrane coating. University of Minnesota Super-
Kang, Y.-T., Lee, K. Y. and Liu, T.-J. 1991. The effect computer Institute Report No. 89/85.
of polymer additives on the performance of a Pranckh, F. R. and Scriven, L. E. 1990. Elastohyd-
two-roll coater. J. Appl. Polymer Sci. 43: 1187-1195. rodynamics of blade coating. AIChE J. 36(4):
Kistler, S. F. 1988. Flexible-blade doctoring in gravure 587-597.
coating. AIChE Spring National Meeting, New Pulkrabek, W. W. and Munter, J. D. 1983. Knurl roll
Orleans, March. design for stable rotogravure coating. Chem. Eng.
Kistler, S. F. and Munter, J. D. 1986. Effect of take-up Sci. 38(8): 1309-1314.
angle and substrate deformation in forward roll Saita, F. A. 1984. Elastohydrodynamics and flexible
coating. AIChE Spring National Meeting, New blade coating. PhD thesis, University of Minnesota,
Orleans, April. Minneapolis.
Lindblad, N. R., Tift, R. and Watson, P. K. 1988. The Saita, F. A. and Scriven, L. E. 1983. Analysis of the
metering of ink in the grooves of a gravure roll. J. flexible blade coater. Lat. Am. J. Chem. Eng. Appl.
Imaging Tech. 14: 140-143. Chem. 36: 587-597.
Malvern, L. E. 1969. The Mechanics of a Continuous Saita, F. A. and Scriven, L. E. 1985. Coating flow
Medium. Englewood Cliffs, NJ: Prentice-Hall. analysis and the physics of flexible blade coating.
Munter, 1. D. 1981. Personal communications to Saita 1985 TAPPI Coating Conference Proceedings. pp.
and Scriven (1985). 13-21. Atlanta: TAPPI Press.
Munter, J. D. 1990. Attenuation of cell patterns for Satas, D. (ed.) 1984. Web Processing and Converting
gravure coating. AIChE Spring National Meeting, Technology and Equipment. New York: Van Nostrand
Orlando, March. Reinhold.
References 635
Scharenberg, R. T. 1985. Coating methods. In Ency- dynamic conditions. Proceedings of the 1972 Sym-
clopedia of Polym. Sci. Eng. 2nd edn. Vol. 3, pp. posium on Elastohydrodynamic Lubrication, Inst.
552-567. New York: John Wiley. Mech. Eng., pp. 22-28.
Shibata, N., Chino, N., Sato, A. and Chikamasa, H. Tanaka, Y. and Noda, S. 1984. Coating apparatus.
1992. Simultaneous video tape dual-layer coating US Patent 4,480,583.
technology. Fujifilm Research and Development. 37: Taylor, J. A. and Fridhandler, R. M. 1985. Apparatus
107-111. for and method of metering of coating on a moving
Shibata, N. and Scriven, L. E. 1986. Theory of coating web. US Patent 4,532,158.
unsupported web by a slot die. AIChE Spring Triantafillopoulos, N. G. and Aidun, C. K. 1990.
National Meeting, New Orleans, April. Relationship between flow instability in short-dwell
Smith, J. W. and Maloney, J. D. 1966. Flow of fluids ponds and cross-directional coat weight nonunifor-
between rotating rollers. Tappi. 49(11): 63-66. mities.1990 TAPP] Coating Conference Proceedings,
Sommer, H. 1985. Influence of metering element pp. 309-324. Atlanta: TAPPI Press.
geometry on coat quality. 1985 TAPP] Coating Van Guelpen, L. J. 1946. Apparatus for coating webs.
C01iference Proceedings, pp. 23-34. Atlanta: TAPPI US Patent 2,464,771.
Press. Varnam, C. 1. and Hooke, C. 1. 1977. Non-Hertzian
Stasyuk, I. A. and Peglovskii, V. I. 1981. Thickness of elastohydrodynamic contacts: an experimental
polymer coats laid by raster rollers. Khim. Neft. 9: 9. investigation. J. Mech. Eng. Sci. 19(5): 189-192.
Stasyuk, I. A., Peglovskii, V. I. and Fedotkin, I. M. Weiss, H. L. 1985. Rotogravure and Flexographic
1983. Polymer coating on roll substrates by a Printing Presses. Milwaukee, Wisconsin: Converting
screened spindle. Khim. Neft. 2: 22-23. Technology Corp.
Swales, P. D., Dowson, D. and Latham, 1. L. 1972. Zink, S. C. 1979. Coating processes. In Kirk-Othmer
Theoretical and experimental observations of the Encyclopedia of Chemical Technology. Vol. 6, pp.
behavior of soft elastic materials under elastohydro- 386-426. New York: Wiley.
HIGH-SPEED BLADE COATING 12d
Cyrus K. Aidun and Nick G. TriantaJillopoulos
12d.l INTRODUCTION
The flow is primarily due to the shear stress
Blade coating is a popular method among the induced by the substrate motion which is dragging
many techniques used to apply a uniformly thin coating in the channel formed between the
film of liquid continuously onto moving webs. substrate and the blade, the so-called blade nip.
Although still a self-metered, as opposed to a In contrast to the meniscus-metered class of
pre-metered process, blade coating falls within coating processes, such as dip coating (see Chapter
the class of processes where the coating layer is 13), where the coating thickness is determined
metered prior to the meniscus. In that sense, the by the shape of the meniscus, parameters
blade acts as a precise smoothing device which independent of the meniscus control the coating
removes excess coating - applied onto the web film thickness of blade-metered coating systems.
upstream with some kind of a mechanical feeding Generally, the film thickness, its uniformity and
system or puddle - and lets a thin coating film overall quality, depend on the characteristic
pass through the narrow channel formed between blade nip geometry, the rheology of the coating
the blade and the moving substrate (Fig. 12d.l). and properties of the web. Difficulties encountered
-
Bevelled
Blade Mode
-
Bent
Blade Mode
Application .-/
Inflatable Tube
for Blade Load
Excess Coating
Clamping Blade Holder
Figure 12d.l Schematic of the flow approaching the blade in a coater with an applicator roll.
London. ISBN 0 412 06481 2.
638 High-speed blade coating
in understanding, predicting, and controlling the on the high-speed blade coating process for paper
blade-metering process arise from interactions webs.
of complex phenomena, which will be discussed Similar to many other coating techniques,
in more detail in the following sections. blade coating technology has evolved from simple
Blade coating refers to the broad family of and basic applications. Many day-to-day appli-
processes where a liquid layer emanates from the cations are, in principle, forms of blade coating.
shallow channel formed between a stationary A simple process of spreading butter by knife
blade and a moving substrate supported against over a piece of toast is a simple example. In fact,
a roll. The liquid may be a dispersion or suspension Booth (1970) uses this example to describe the
of solid particles; the blade can be rigid or effect of coat weight in relation to the angle
flexible; and the supporting roll can be hard or formed between the blade tip and the tangent to
soft. The subject of rigid knife coating is covered the substrate, i.e., the so-called blade angle. The
in Chapter 12a. Here we focus on flexible blade original coating apparatus appeared in two forms
coating, where the blade can be loaded with a (Trist 1945). In one form, a coating head or
mechanical member or an inflatable pressured trough was used to contain the coating, hence
hose located at some distance from the edge of the flexible blade formed the boundary at the
the blade - the blade tip (Fig. 12d.l). The blade leading edge of the trough. This is a 'puddle' type
may also be loaded by rotating the blade holding of coater, which was common during the early
assembly, so that the blade bends against the development and use ofthe blade coating process
substrate and the supporting (backing) roll. (Fig. 12d.2). In a second form, excess coating was
Industrial applications of this process include applied onto the moving web with an applicator
addition of adhesives and coating to paper, roll and metered downstream with the blade. The
paperboard and films, as well as oxide coatings excess coating was then recycled back into the
onto magnetic recording tapes. pad feeding the applicator roll. This configuration
Although flexible blades are commonly used was termed the 'inverted' blade coater. In both
to apply liquid films onto smooth and incom- cases the substrate was drawn around a rubber
pressible polymeric substrates (e.g., in magnetic covered supporting roll, also called the backing
suspension technology), they are most popular roll, which drives and supports the substrate.
in surface application of coatings for publication Continual upgrading ofthe blade coater process,
papers and paperboard. The reason for surface however, was needed as machine speeds increased
treatment of paper and paperboard is to develop for productivity and film thicknesses decreased
a uniformly smooth surface for printing and
superior appearance. Blade coating generates a
smooth surface with improved optical properties,
gloss, and printed ink density, while capturing
halftone dot highlights during printing. However,
these substrates are rough, porous and compress-
ible, properties which make the physics of this
blade coating process particularly challenging.
Blade metering is the coating process of choice
for paper and paperboard applications because
it is conductive to relatively high machine speeds
and coating viscosities. Blade coaters for paper
run at speeds above 5 mis, while the steady shear
viscosity ofmineral dispersions for paper coatings
varies from several hundred to over a thousand
mPa s. The discussion in this chapter concentrates Figure 12d.2 Schematic of a puddle coater.
Introduction 639
due to rising mailing costs. Nowadays, there is 12d.3. In cases (b) and (c) the blade runs on its
a whole family of newer blade coaters where trailing (heel) or leading (toe) edge, respectively,
differences exist on: (1) the application of excess something which may lead to coating defects.
coating (or pre-metering) with roll applicators or Therefore, the exact blade geometry is a critical
jets; (2) variability in the dwell time between issue in bevelled blade coating, as will be discussed
application and metering; (3) pre-metering with in more detail in Section 12d.7.
one blade first before final metering with a second In the second configuration the underside of
blade, or (4) utilization of a short (2.5-5.0 cm in the blade is parallel to the web over a finite area
length) pressurized pond for application of the (Fig. 12d.1). This is the so-called flexed, low-angle
coating before metering, hence the flexible blade or bent blade mode. The blade angle in this case
itself acts as a boundary that keeps the coating varies from almost O· to 25·. The main difference
in contact with the substrate and as a metering between these two blade modes is the operational
device. This latter design is the popular type of window of the process, i.e., the range of machine
blade coater known in the paper industry as the speed and film thickness, or coat weight, for
'short-dwell coater'. These and other blade coater which satisfactory coating is feasible. (The
designs are discussed in detail in Section 12d.2. papermakers use coat weight to express coating
Blade coaters for paper are also classified into film thickness in grams per unit area, in square
two categories, depending on their operational meters of paper). The bent mode is preferred for
mode. In the first configuration the tip of the comparatively high film thicknesses and low
blade remains nearly tangent to the web surface coater speeds (e.g., paperboard coating), while
and the backing roll at the area of contact the bevelled blade is selected for applying thin
between the blade and the web. This is the films at relatively high speeds (e.g., light-weight
so-called straight, trailing, or bevelled blade (Fig. coated publication papers).
12d.l). The blade angle, formed between the There are several process parameters in high-
blade and the tangent to the backing roll (Fig. speed blade coating systems which influence the
12d.3), is comparatively large, typically varying surface quantity and quality of coating applied
between 25· and 50·. In practice, blades are onto a substrate. Generally, the coating film
prehoned to nominal bevel angles - typically, thickness, or coat weight, depends on the
45·, 40·, 35·, etc., so that one needs to set the interaction between the various operating par-
blade angle to the bevel angle of the blade in ameters; namely, the coater geometry, coating
order to run the blade on its bevelled mode. rheology, and properties ofthe substrate. Usually
However, metallic blades wear out over time the mechanical or pneumatic load on the blade
during normal operation and the bevel angle is adjusted to control the final coating film
changes. When this occurs, the blade is not thickness of coat weight. The blade thickness is
running anymore on its bevel, resulting in typically between 0.1 to 1.0 mm and the blade
configurations (b) and (c), as illustrated in Fig. length (i.e. the distance between the edge of the
blade, or blade tip, and the location where the
blade load is being applied) is about 2 to 10 cm.
:\~.I!~. ,
Blade
A typical time scale of the metering process at
the blade nip is 0.01 milliseconds for 17 mls
coater speed (Gane, Watters and McGenity
1992). Although most of the commercially utilized
blades are made out of steel, newer blades are
Paper
ABC manufactured with ceramic tips which resist wear
and provide a better control of coating film
Figure 12d.3 Schematic of a bevelled blade running thickness in the long run. The manner by which
on (a) parallel, (b) toe, and (c) heel. these parameters influence, and therefore can be
manipulated to predict, the coating film thickness would cause coating film defects and deposits
is discussed in Section 12d.3. appearing at the blade surfaces. In practice,
Blade metering is a simple and rather primitive consistent and sustained productivity requires
process flow, but presents a complex hy- adequate knowledge of the physics of the
drodynamic system which has been under blade coating flows, as well as the character-
investigation since its original recognition as a istics of both the substrate and the coating
superior coating system, as early as the 1950s. material.
Originally blade coaters for paper operated at The focus in this chapter is on high-speed blade
relatively low speeds, i.e., 5 to 10 mis, but modern coating of paper webs. Section 12d.2 describes
blade coating systems operate at speeds in excess the various designs and their evolution over the
of 20m/s. The coating fluid (or dispersion) is years to the high-speed precision coaters of
forced from one extreme of flow conditions to today. Section 12d.3 discusses how the different
another, while the fluid and dispersed particles process parameters interact to determine the film
experience changes in the shear rate from almost thickness, or coat weight. It also includes a review
zero upstream of the blade to nearly 106 s - 1 of analyses and experimental studies on the
inside the passage under the blade tip in a period physics ofblade-metering. Although the emphasis
oftime less than one millisecond. In the convergent is on blade-metering on rough, porous and
flow through the blade nip, regions with high compressible substrates, limited reference is made
pressure gradients are adjacent to nearly constant to rigid knife coating of smooth, incompressible
pressure areas (or pressure plateau). Furthermore, substrates to illustrate qualitative differences
the coating materials are non-Newtonian, exhi- between the two processes. Sections 12dA through
biting complex rheological behavior including 12d.5 deal with hydrodynamic and other phe-
dilatancy, yield stress, thixotropic recovery and nomena which influence the process and the
viscoelasticity. Paper coatings, for example, uniformity of the film thickness. Finally, Sections
comprise highly concentrated dispersions (i.e., 12d.6 and 12d.7 cover respectively aspects of
30-40% by volume) of anisometric, orientable paper coating rheology and blade coating defects,
particles suspended in an aqueous phase con- their appearance and origins.
taining polymeric additives. Therefore, the ability
to predict and control the coating process many
12d.2 BLADE COATER DESIGNS
times becomes a formidable task.
In addition to the above complications, one The first blade metering on a continuous coating
could hardly design a process which, with respect machine is credited to the system designed by
to modeling, involves more complex boundaries. Trist (1945) which was primarily used for coating
The blade is flexible, the substrate is generally an oil phase emulsion on bread wrapping paper.
deformable and permeable, and there are free- The superior potential of this technique was
surfaces with static and dynamic contact lines. recognized in the 1950s when a growing number
On the other hand, the requirements for a ofapplications started to adopt the blade coating
successful coating process are also extreme. The system. A wide variety of blade coating equipment
flow needs to remain steady and two-dimensional has been invented since then. Richardson (1957)
at high machine speeds for maintaining a uniformly and Booth (1970) review the earlier blade coaters
thin coating film. Uniformity needs to be sustained and their designs.
in both the machine line and the cross machine There are in general three important regions
directions, as film thickness nonuniformities would in an application system: (1) the wetting or the
not be acceptable from the standpoint of both dynamic contact line (see Section 12d.5) where
visual appearance and physical attributes. Fur- the fluid first comes into contact with the substrate;
thermore, anomalies occurring at the bounding (2) the dwell region where the coating liquid
walls, as well as the bulk of the process flow, remains usually undisturbed adjacent to the solid
Blade coater designs 641
surface, and (3) close to the metering region application method before metering. These
where the fluid layer approaches the blade. comprise the puddle, roll, jet (slot), and contact
The flow in an ideal application process should application systems. The common process involved
be steady state and two-dimensional. If the in all of these systems is that they supply the
coated layer approaching the blade has uniform blade with excess coating fluid which is, subse-
thickness and if the blade has no defects, then quently, metered by the blade. Typically, fifteen
the resulting coated film would have a constant to twenty parts of the applied coating are removed
thickness. However, nonuniformities or unsteadi- by the blade for every part applied on the moving
ness upstream of the blade may result in coating web. Each one of these application systems can
defects. The mechanisms whereby flow nonunifor- be combined with any blade mode, i.e., bevelled
mities upstream of the blade could result in or bent. A list of the different commercially
coating defects have been analyzed by Miura and available types of blade coaters are listed in Table
Aidun (1992). 12d.1. Figure 12d.4 presents schematics of the
The maximum coating speed is determined different blade coating equipment for paper.
primarily by limitations set by the application Currently, many of the high speed blade
unit. The fundamental limit on the coating speed coating systems operate with a roll applicator
depends on the maximum speed at which the air where a thick layer of coating fluid is transferred
adjacent to the solid surface can be replaced by to the surface of the substrate. The coating layer,
liquid at the wetting region. This is sometimes which is typically between 100 Jlm to 1 mm thick,
referred to as the maximum wetting speed and is metered downstream by the blade to provide
depends on the flow field in the vicinity of the a final film thickness in the range of 10 to 50 Jlm.
dynamic contact line. Above the maximum wetting This operation is commonly referred to as the
speed, small air bubbles entrain into the liquid inverted blade coater; also known as the flooded-
and result in coating defects, as explained in nipTM blade coater. Although this is a quite
Section 12d.5. Flow instabilities in the application common blade coating system, it is now clear
unit upstream of the blade sometimes impose that other techniques such as jet and pond
more severe limitations on the coating speed than application systems are superior. In particular,
air entrainment. These instabilities, as discussed the so-called short-dwell coaters have recently
in Sections 12d.4 and 12d.5, result in operational found wide acceptance by the paper industry due
difficulties and coating defects. Therefore, air to their compactness, ease of operation, and
entrainment and flow instabilities are the main capability to keep the equipment clean, as well
issues of concern in the development of modern as their ability to apply comparatively thin
high speed coating systems. coating films at high machine speeds (i.e., greater
There are four different categories of blade than 25 m/s). The main advantage of the short-
coating systems that depend on the coating dwell coater is that it can provide the same coat
Table 12d.l Classification of commercial blade coating application systems
Coater type Coating application mode Blade configuration
Inverted blade Roll applicator Bevelled (stiff) blade

Slot orifice Bent blade
Jet fountain Zero-angle blade
Contact applicator Billblade™
Puddle coater
Short-dwell coater
Simultaneous two-sided coater
(a) Roll applicator
(b) Jet application (c) puddle coaler
Figure 12d.4 Schematic of (a) roll, (b) jet, and pond application systems which include (c) puddle coater and
pressurized pond application systems with (d) short-dwell time and (e) long-dwell time modes of operation.
weight onto a paper web as a system with an 12dA and 12d.5) where both sides of the substrate
applicator roll but at substantially lower blade are coated and dried simultaneously. In principle,
pressures, something which significantly improves these coaters cannot operate at high speed because
productivity because it minimizes the possibility of air entrainment at the wetting line which
for web breaks. The different application systems, results in foaming of the puddle and, in extreme
presented schematically in Fig. 12dA, are discussed cases, prevents contact of the fluid with the
in the following sections. substrate.
These coaters are commonly used today in
various applications, such as speciality paper and
12d.2.1 PUDDLE-TYPE COATERS
board coating. The pond level in normal operation
There are a variety ofcoating application systems is between 150 to 170 mm high and the backing
in this class. The two principal categories are roll has a soft rubber cover (-70 P&J). The two-
one-sided and two-sided coating systems (Figs sided puddle coaters (Fig. 12d.5), such as, the
(d) short-dwell coater (e) pond applicator
Figure 12d.4 Continued.
Billingsfors-Langed, better known as the Bill- faster than the web and forming a 'take-off' angle
blade™ system (Holton and Klass 1986), have ofthe web away from the roll completely eliminates
remained in operation despite their limitations film split and the misting problems on the roll side.
in speed and film thickness control. In addition
to air entrainment, misting is also a problem with
12d.2.2 ROLL COATERS
these coaters. At high speed, coating droplets
form a mist of coating fluid, usually at the Roll applicators are perhaps the most popular
diverging region between the substrate and the and traditional form of coating systems. The
roll where a meniscus forms. Misting occurs transfer or applicator roll picks up the coating
because of meniscus instability and breakup (film fluid from a supply pan or reservoir, as shown
split), similar to the spraying mechanism in roll in Fig. 12d.4a, and transfers the fluid to the
applicators. It is reported (Holton and Klass substrate being carried by the backing roll. The
1986) that by running the backing roll 3-5% clearance between the surface of the two rolls
Figure 12d.5 Double-sided puddle or (Billingsfors-Langed) Billblade™ coater.
can vary from 100 J.1m to 1mm depending on the are difficult to operate, clean and maintain.
particular operation. The roll coating modes can Manufacturing light-weight coated paper and
be forward where the two rolls rotate counter to film requires metering most ofthe applied coating
each other, or reverse, where the rolls rotate in from the surface by exerting a relatively large
the same direction. The amount of coating fluid pressure on the blade. This causes frequent web
transferred to the substrate by the applicator roll breaks with an enormous waste in coating fluid
depends on the fluid properties, the gap, and the and substrate material. Because ofthese difficulties
relative speed between the substrate and the roll. with roll applicator systems, a number of more
There are a number of roll applicator systems advanced application systems are being developed.
with multi-roll metering units which have been These fall within the class of jet and pressurized
developed by various companies. For examples application systems.
of these systems, the reader is encouraged to see
the review by Booth (1970). For a more extensive
12d.2.3 JET COATERS
analysis of the mechanics and instabilities in roll
applicators, see Section 12d.4 and Chapter 12a. These systems, as the name implies, consist of a
Roll applicators with blade metering units, in rectangular liquid jet which comes into direct
general, provide a good quality coated surface. contact with the substrate, as shown in Fig.
However, because offlow instabilities and ribbing, 12d.4b. We confine the term 'jet coaters' to those
roll applicators have limited use in high-speed systems in which a liquid jet emanating from the
blade coating. Also, as speed increases, the nozzle directly impacts the substrate. Some in
meniscus splits forming droplets. This spraying the industry refer to the pressurized pond system
action causes difficulties in maintaining a clean of Fig. 12d.4c as jet coaters, as well. However,
and smooth layer of coating fluid prior to the the fluid flow characteristics and the coating
blade. As indicated above, it is important that features between the two systems presented in
the coating layer remains smooth and uniform, Figs 12d.4b and 12d.4c are substantially different.
as a nonuniform layer approaching the blade With a jet coating system, the entire fluid leaving
results in coated film thickness nonuniformities the nozzle is transferred to the substrate and
and other defects. In addition to this problem, supplied to the doctor blade. In contrast, in
from a practical point of view, roll applicators pressurized pond application systems some of
the fluid is always rejected from the pond. Also, in a pressurized reservoir adjacent to a metering
jet coaters are usually used for thicker liquid film blade. The main purpose ofincreasing the pressure
application on the surface. Thinner film appli- inside the coating reservoir is to avoid 'skip
cations can better be achieved with short-dwell coating', that is to force the fluid to remain in
time or other pressurized pond application coaters. contact with the substrate at high speed and
To ensure a uniform coated surface, the jet avoid air entrainment at the wetting line (see
must have uniform constant momentum across Section 12d.5 for more detail).
the coater. This is achieved, in practice, by This system ofcoaters has received considerable
injecting the coating fluid from a large reservoir attention in the industry for several reasons. The
into a long channel before forming a two- coating head is compact and easy to operate and
dimensional free-surface jet. As the volume of maintain. These pressurized pond systems are
the reservoir is relatively large, the pressure frequently used in short-dwell mode (referred to
inside will be essentially static and uniform as Short-Dwell Coaters) where a very thin layer
across the nozzle (see Chapter 10 for a more of coating is applied on the surface of a substrate
detailed discussion of manifold design). A long without causing excessive web breaks.
slightly converging nozzle will allow the flow to With these coaters, the pond is adjacent to the
develop fully and stay attached to the wall. In blade and the coating liquid is transferred to the
other words, a converging nozzle ensures that surface of the substrate immediately prior to the
pressure drops in the flow direction and therefore, metering section. In contrast to the long-dwell
the shear stress retarding the flow near the walls time coaters where the coating layer travels
will not cause flow separation or flow reversal adjacent to the substrate for a relatively long
at the wall. time before being metered off, the coating in short-
The control parameters are the opening gap dwell coaters has little time to penetrate into the
of the jet, d, the jet angle relative to the substrate, substrate. These coaters are particularly used for
(x, the velocity of the substrate, U, and the jet light-weight printing and publication grade of
Reynolds number, Re == pdVIJ1-, the capillary papers as well as many other products including
number Ca == J1-Vlu where p, J1-, and u are the pre-metered size press applications. These coaters
fluid density, viscosity, and surface tension, have several advantages over other coating
respectively. The jet velocity scale, V, is defined application systems. The main advantages of
as the mass flux divided by the jet's cross-sectional these coaters as compared with roll applicators
area. The jet could swell or contract depending are the compactness, operational efficiency, and
on Re and Ca. Here we are assuming that the increased productivity of thinner coated film and,
coating liquid behaves like a Newtonian fluid, therefore, lighter-weight coated layers. It is
although at low shear the viscosity of coating reported (Closset 1986) that with these coaters,
fluids usually depend on the rate of shear. In blade pressure can be reduced by 40% resulting
practice, the jet is usually from 1 to 1.8 mm thick in a considerable reduction in web breaks. This
at the vena contracta and extends about 12 to increases productivity by reducing the loss time
25 mm in length. The coating flow rate per unit due to web breaks by as much as 28% as
width is in the range of 1.3 to 3 m/s. The relative compared with blade metering systems with roll
velocity of the jet to the substrate, VIU, varies applicators. Furthermore, a 40% reduction in
from 0.1 to 0.5 depending on the coating thickness coating liquid loss is reported by the industry.
and other parameters. The lower blade pressure results in a 60% longer
blade life time.
As the coating speed is increased, problems
12d.2.4 PRESSURIZED POND COATERS
appear in all coaters. Although the short-dwell
In this class of coating application systems, the coater can operate at speeds in excess of 20 mis,
coating fluid comes into contact with the substrate the coated surface quality deteriorates quickly
beyond a critical speed which depends on several incorporated only a limited number of the physical
factors including the coating rheology. Usually phenomena involved. Fundamental understanding
these problems appear in the form oflong streaks of the blade metering process for paper has only
which have a coating thickness deficit of about recently been attained with the help of both
20 to 60%. Section 12d.7 describes the various computational models and experiments.
coating defects in blade coating systems. The In the present section we review key relation-
most severe coating defect with short-dwell coaters ships between operational variables and coat
are long streaks about 1-3 cm wide and 10 to weight, with a focus on high-speed blade
100cm long. Extensive experiments (Triantafil- metering of paper webs. We will refer to previous
lopoulos and Aidun 1990) have shown that the analyses and experiments which shed light on
parameters which have the most significant the physics of this process. Both the bevel and
influence in generating streaks are coating speed, bent blade operating modes are discussed. Topics
low shear viscosity of the coating liquid, and covered in detail in other chapters of the book,
consequently, the concentration of the suspension. i.e., elastohydrodynamic coating systems covered
Suspensions with larger particle aspect ratio (thin in Chapter 12c, are only briefly mentioned here.
disk-shaped) and larger shear viscosity tend to Differences in the operational behavior of high-
be more susceptible to coating defects, in general. speed blade metering and low-speed blade and
Additional problems for short-dwell and other knife coatings are also discussed in order to point
pressurized pond application coaters are the out qualitative difference between the two systems.
instabilities that occur inside the pond as well The mechanism of low-speed blade coating
as air entrainment issues at the dynamic contact has been studied by several authors (Middleman
line. These problems are fundamental to any 1977; Hwang 1979; Sullivan 1986). When inertia
pressurized pond application system. Sections effects are neglected, lubrication theory is an
12d.4 and 12d.5 treat these issues in more detail. acceptable first approximation to describe the
It is also possible simultaneously to coat both viscous forces in the gap between the blade and
sides of a moving web passing between two the web (or roller in the case of knife-over-roll
blades located one against each other. A Bill- coating). The lubrication model predicts both the
blade™ is the configuration where the web passes pressure profiles and the coating thickness.
through the nip of a bent blade, on one side, and Generally, the coating film thickness for
a roll on the other side. Two separate puddles Newtonian fluids is approximately equal to one
of coating exist, one on the roll side and the other half the minimum separation between the blade
one of the blade side. These puddles allow for and the moving web. As shown by Sullivan (1986)
application ofcoating onto the moving web from and Strenger, Secor and Sramek (1992), the final
both sides simultaneously. coating thickness also depends on the size,
curvature, inclination and inlet conditions of the
upstream (feeding) region of flow. The key
12d.3 BLADE METERING PROCESS
characteristic of this process is that coat weight
The physics of high-speed flexible blade coating decreases with increasing viscosity and web
has been the subject of interest in many studies speed. In contrast, coat weight increases with
over the years. The motivation was to document viscosity and web speed in high-speed blade
the interaction of the various forces acting on metering of paper coatings with bevelled blades
the blade and to determine how process variables (Kahila and Eklund 1978b). These differences
affect coat weight. Most of the work focused on suggest that, at least for the case of high-speed
development ofsemi-empirical relationships which bevelled blade coating, inertia effects, which are
can be readily used to manipulate the process ignored in pure lubrication flows, may be
and control coat weight. These relationships important.
were based on rather simplistic analyses which Typically, the development of coat weight, of
Blade metering process 647
wet film thickness, as a function of blade pressure

in high-speed flexible blade metering follows the
behavior depicted in Fig. 12d.6. In the bevelled
mode, the blade pressure lowers the coat weight
(region I), up to a point where pressure changes
bring a minimal change in coat weight. When
the flexible blade runs on its bent or low blade
angle (i.e., blade angles below 20°) mode, coat
weight increases with increasing blade pressure
(region III). As the blade pressure of a bent blade
increases, the operating blade angle decreases
and, consequently, coat weight increases. In
well-flexed blades, the coat weight passes through BLADE PRESSURE
a maximum upon continual increase in blade
pressure, as shown in Fig. 12d.6. Beyond this Figure 12d.6 Schematic illustrating change in coat-
point (region IV), coat weight decreases with ing weight as a function of the blade pressure in blade
rising pressure because the excessive bending on coating of paper webs (Eklund 1984).
the blade causes formation of a divergent nip
between the blade tip and the surface of the are the shear and normal viscous stresses arising
moving web. Under these circumstances, the final from the coating dragged into the converging
film thickness may acquire a film-splitting pattern region under the blade and through the blade
due to instabilities developed at the free surface nip. Coating fluid in these regions also develops
of the coating film emerging from the channel a pressure and pressure gradient, similar to those
between the blade tip and the web (Eklund 1984). occurring in purely lubrication flows. In turn,
This is similar to the film-splitting phenomenon the viscous and pressure forces exerted on the
that occurs in roll coating. blade cause it to deflect due to generation of
The relationship between coat weight and elastic forces. In addition to the above, inertia
blade pressure depicted in Fig. 12d.6 was first forces are also present. These forces arise from
reported by Eklund (1984) and it was analyzed the impact pressure of the upstream coating fluid
by Saita (1984) using elastohydrodynamic prin- layer, which contacts the blade and changes
ciples (see Chapter 12c). Although this relationship momentum to be expelled away from the blade
applies to coating any type ofa web, experimental nip. It is worth noting that other forces, such as
data presented by Eklund (1984) were based on gravity, surface tension, colloidal, etc., are present
a paper web. In practice, it takes a wide range but playa minor role in high-speed blade metering.
of blade pressures, irrespectively of how these The mechanism of blade metering is based on
are applied (i.e., with an inflatable hose or a the balance of all the forces described above. As
mechanical device), to generate this graph. A key the moving web drags coating between it and
observation, however, is that a certain coat the blade, hydrodynamic forces - in conjunction
weight can be achieved with the flexible blade with inertial forces - stress the blade, which
operating at two, or even three, different conditions. deforms proportionally to its flexural rigidity.
To understand the behavior of this system one Internal elastic forces and external loading (i.e.,
needs to consider the forces involved in flexible tube pressure) oppose blade deformation, so that
blade metering. finally the blade assumes a shape that balances
Several dynamic forces develop as the coating all forces acting on it. In turn, the flow re-
fluid is dragged by the moving web onto the organizes to balance the hydrodynamic forces in
converging flow under the blade (Kahila and the coating field of flow, while the paper web
Eklund 1978a-c; Saita and Scriven 1985). First (and the backing roll) compresses under the
influence of the pressure pulse under the blade. lift which has been considered either inside the
As a result, at least for the case of paper coating, blade nip or upstream of the nip entrance on the
the non-Newtonian rheology of the coating and underside boundary ofthe blade. This lubrication
the compressibility and absorptive properties of flow is similar to the one occurring in a rotating
the web also influence coat weight development. slider or a journal bearing, which has been a
Both these and the fluid dynamic phenomena benchmark flow in fluid mechanics.
occurring during blade metering of paper coatings For the case of bevelled blade coating, the
are discussed in more detail in the following blade nip geometry has been considered as both
paragraphs. a converging (Bohmer 1969; Bliesner 1971;
Many attempts have been made over the years Modrak 1973; Hayward 1973) and a parallel
to analyze and understand blade coating of (Turai 1971) channel. In the former configuration,
paper. These led to deriving models of the physics the hydrodynamic lubrication lift developed in
of blade coating and to developing semi-empirical the convergent flow is considered responsible for
relationships which connect operational par- blade lifting. In the latter configuration, the blade
ameters, coating and substrate properties to tip is supposed to remain parallel to the paper
coating film thickness or coat weight. These surface at all times, so that the gap between the
analyses and models considered three basic kinds blade tip and the paper surface is equal to the
offundamental phenomena: (1) the hydrodynamic final film thickness. As the drag flow induced by
and inertial phenomena at the vicinity of the the moving web can only account for half the
blade nip which are responsible for development coating film thickness, a pressure gradient is
of the forces described above; (2) the filling of needed to derive the other half. That pressure
pores in the substrate, which depends on the gradient leads to calculating the pressure force
surface roughness and compressibility of the which lifts the blade. Experimental results based
web; and (3) the penetration of the continuous on lubrication (Follette and Fowells 1960)
(dispersing) phase of a coating dispersion into indicated that coat weight is inversely proportional
the substrate. It should be emphasized that most to blade load. Coating weight increases with
paper coating applications are considered high increasing machine speed; solids concentration
speed, i.e., in excess of 5 mis, and the substrates and steady shear viscosity of the coating; the
are viscoleastic networks ofcellulosic fibers which absorbancy and roughness of the paper; and
can rapidly absorb water and swell. blade extension. It decreases with increasing
The simplest models are based on hydrodynamic blade load and the blade thickness, and a backing
lubrication at the blade nip. This phenomenon, roll made out of a harder rubber. (Extension here
more or less influenced by paper absorption, is means the distance from the tip to the location
considered the principle mechanism for coating on a blade beam where the blade load is being
deposition (Follette and Fowells 1960; Bohmer applied). In addition, coat weight decreases with
1969; Bliesner 1971; Turai 1971; Modrak 1973; an increasing blade angle.
Hayward 1973). In these analyses the blade is Gartaganis, Cleland and Wairegi (1977) studied
considered as a cantilever beam that deflects the connection between various operating par-
under load, while the flow through the blade nip ameters and coat weight in both bevelled and
is based on hydrodynamic lubrication. Coating bent blade coating of paper. For the case of a
in the region close to and under the blade tip is bent blade mode, coat weight increases with
assumed to develop a load-bearing capacity that increasing machine speed, blade extension, coating
lifts the blade in the normal direction to the web solids content and steady shear viscosity, the
movement. The hydrodynamic lift exerted by the roughness and absorbancy of the web and the
coating is counteracted by the applied blade load. backing roll radius. Coat weight decreases with
According to this model, the net force acting on increasing blade angle and decreasing blade
the blade is that ofthe hydrodynamic lubrication thickness. These authors also documented the
relationship between coat weight and blade load. does not take into consideration the free surfaces,
Coat weight initially increases with blade load, both upstream and at the terminating edge of
up to a limiting load, thereafter decreasing as the blade. The pressure drop across the blade
the load continues to increase. The two regimes nip has been considered nonexistent as both the
before and after the maximum coat weight upstream and downstream pressures are con-
correspond to regions III and IV in Fig. 12d.6. sidered atmospheric. This assumption, however,
Early studies demonstrated that, in both the may not always be valid, especially for the case
bevelled and bent operational modes, coat weight of pond-type of blade coaters where a puddle is
strongly depends on properties of the substrate. used to feed the blade nip (Fig. 12d.4c-e). Guzy
As some fibers at the paper surface cross over and Higgins (1982) considered a nonzero pressure
other underlying fibers and air voids exist between differential across the blade nip. These authors
fiber crossings, paper webs have rough surfaces studied the effect of pressure differential on the
and are porous. Under the same blade configur- load bearing capacity of the blade and found
ation and load, more porous and compressible that, for the case of a negative differential, there
substrates provide comparatively high coat weights is no bound on film thickness. However, when
(Luciani and Galloni 1978). Among many vari- the differential is positive, there is a limit on the
ables, the porosity of the web depends on the final film thickness.
papermaking furnish, mineral additives, paper Drawing from elastohydrodynamic computa-
formation and surface treatment. In particular, tions, which are discussed in detail in Chapter
externally applied surface treatments onto paper 12c, Saita and Scriven (1985) developed a
webs are a practical way to improve their lubrication model which accounts for both the
coatability. Paper compressibility, on the other shear-driven (Couette) flow through the blade
hand, is primarily dependent on the papermaking nip and the pressure-driven (Poisseuille) flow for
furnish, formation, and consolidation of the convergent gaps. In addition to hydrodynamics,
fibrous network forming the paper web. these authors considered the elastic forces of the
The simple lubrication model cannot fully blade, but excluded inertial effects and effects
explain the physics of high-speed blade coating. arising from free surfaces and the compressibility
The blade boundary may not be parallel to the and permeability of the web. This model covers
web under all possible operating conditions. The the operational behavior of both a low-angle stiff
paper may deform under the pressure applied blade and a highly bent blade at moderate speeds
on it by the blade tip and the blade nip gap may (i.e., 4.5 m/s). The coating thickness in the bevelled
therefore be far from parallel. The uniformity of blade mode was found to be approximately equal
the pressure gradient upstream of the blade nip to the blade nip gap, i.e., the clearance between
is questionable, so is the assumption that the the tip of the blade and the surface of the
blade can be analyzed as a cantilevered beam. substrate. In the bent blade mode, the coating
Lubrication approximation does not take into thickness was about one half the width of the
account physical phenomena arising from the blade nip. In addition to predicting the qualitative
free surfaces upstream of the blade and the physics of flow (Fig. 12d.6), this model provided
coating penetration into the substrate. Finally, quantitative results which were in agreement
there is no consideration of the pressure (inertial) with measurements (Saita 1984) of relatively
force upstream from the blade nip where the moderate-speed coating of Newtonian and
excess coating is deflected away. Because of all shear-thinning fluids onto smooth, nonporous
these complications, more advanced models are and incompressible webs. Based on this work,
required to describe the blade metering process hydrodynamic lubrication dominates coating
of paper. deposition in bent blade metering.
All of the previous lubrication flow analyses The above model was advanced by Pranckh
are based on a one-dimensional model which and Scriven (1988, 1990) to incorporate the two-
dimensional flow in high-speed (i.e., 15 m/s) lift under the tip ofthe blade may exist temporally
bevelled blade coating. It accounted for the fluid and only under unsteady conditions, i.e., when
inertia upstream of the blade nip, the viscous the blade tip is not exactly parallel to the web
forces on the face and side of the blade, and the surface. Eventually, over an extended period of
pressure force in the blade nip. Computations operation, the process attains equilibrium as the
based on this physical model illustrated that steel blade wears out to assume a parallel-
blade load and the working blade angle influence plate geometry under the blade tip. Thus, studies
coating deposition in bevelled blade metering, by Kahila and Eklund (1978a-c) neglect the fluid
where coating fluid inertia becomes important viscosity and use potential theory with a simple
at high speeds (Pranckh 1989). Derived qualitative momentum balance over a control volume at the
relationships are in agreement with experimental blade to estimate the force exerted by the coating
results published by Kahila and Eklund (1978b,c), onto the blade. The inertia generates an 'impulse'
Eklund (1984) and Kuzmak (1986a, b). This model force which is similar to the stagnation pressure
was further expanded to incorporate porous incorporated in the elastohydrodynamic model
substrates, hence laden air is being displaced by (Pranckh and Scriven 1988,1990). This force is
the coating due to pressure- and capillary-driven significant at high-speed blade metering, i.e., the
penetration of the coating into the porous substrate impulse force is by two orders of magnitude
(Chen and Scriven 1989). greater than the hydrodynamic lubrication when
Computations illustrated that the amount of the blade angle is greater than 28° (Kahila and
air trapped within the substrate is critical for Eklund 1978b,c; Eklund 1984). Triantafillopoulos
determining the hydrodynamic pressure profile and Altug (1991) found that the inertial phenom-
at the blade nip. More detail on the elastohyd- enon is dominant for determining coat weight
rodynamic models appear in Chapter 12c. in bevelled blade coating of paper at speeds
However, quantitative results of these models between 3 and 5 m/s. The impulse force accounted
remain to be verified by controlled experiments. for more than 90% of the dynamic forces acting
Key issues in high-speed blade metering of on a blade. The strong influence of the impulse
coatings for paper are the role of inertia, from force was also verified by Lyons (1993) who
the excess coating striking the underside of the analyzed blade metering in a pond-type laboratory
blade upstream ofthe blade nip, and the roughness coater. Computations showed that the impulse
and compressibility of the web. Windle and force was by an order of magnitude greater than
Beazley (1967) and Kahila and Eklund (1978a) the hydrodynamic lubrication force. It is therefore
emphasized these phenomena, instead of the important to realize that inertial effects are
viscous forces and pressure gradient at the blade significant in high-speed bevelled blade coating
nip. According to these authors, high-speed of paper.
metering evolves by filling depressions (or cavities) Besides the blade forces acting in the normal
ofthe web, while the blade rides almost in contact direction to web travel, there is the shear force
with fiber crossings at the web surface. Hy- which is parallel to web movement. Kartovaara
drodynamic lift develops only in the convergent (1991) showed that the lateral shear force can be
flow upstream of the blade nip entrance, so that as high as 30% of the dynamic forces acting on
lubrication flow exists only in this region. the blade for coater speeds between 5 and 16 m/s.
According to these authors, the focal phenomenon The lateral force is proportional to coater speed,
involves the coating layer which comes in contact coating viscosity and length of the blade tip (i.e.,
with the blade and changes direction of flow as blade thickness).
excess coating is removed from the blade. In Practical experience has shown that, for bevelled
doing so, the coating layer transfers momentum blades, coat weight is directly proportional to
to the rigid blade, causing it to deflect. steady shear viscosity of paper coatings and
In the case of bevelled blades, hydrodynamic machine speed. When viscosity increases, i.e., by
raising the volumetric concentration of paper running on the toe in order to maintain a certain
coatings, the blade load to obtain a certain coat coat weight. Under constant loading, an increase
weight also increases. Although coating viscosity of the blade tip length, i.e., by using a thicker
does not appear in the impulse force equation, blade or running at the heel, would increase coat
it influences the mass flow rate approaching the weight because the effective loading force at the
blade which is proportional to the impulse force. blade tip decreases. When the blade tip area
Comparatively high viscosity increases the amount increases, blade loading is distributed over a
of coating picked up by the web at the applicator wider area, the force per unit area decreases and,
roll nip and, consequently, raises the impulse consequently, coat weight decreases. A constant
force. Another observation from practice is that, coat weight in bevelled blade coating is maintained
under a constant load, coat weight increases only when the tip of the blade is running parallel
proportionally with the coater speed (Fig. 12d.7). to the web surface.
This is because the impulse force is directly The sensitivity of the system to the exact
proportional to the momentum of the coating geometry at the blade tip is an important issue
layer dragged by the web into the region under in bevelled blade coating of paper. This issue was
the blade. Both trends for coating viscosity and discussed by Ramp (1983) and Roper and Attal
speed observed at high-speed blade metering of (1993) who pointed out that running the blade
paper are the opposite of those observed in on its toe or heel may lead to web breaks and
low-speed knife coating where lubrication flow coating defects, as discussed in more detail in
is predominant. Section 12d.7. The parallel nip geometry is
The contribution of the excess coating layer usually the physical state of operation of bevelled
carried by the web onto bevelled blade nip was blades in practice. Even though sometimes the
also considered by Kuzmak (1986a, b). In contrast blade tip is not set up correctly, the steel blade
to Kahila and Eklund, Kuzmak considered wears out to attain the parallel nip geometry.
lubrication flow inside the blade nip, i.e., between Under a certain blade load and blade angle, there
the web surface and the blade tip. This work is only one blade position where the blade nip
emphasized the effects of running the blade on is a parallel channel and the process remains in
its leading (toe) or trailing (heel) edge on coat equilibrium. Changes in blade loading (i.e., tube)
weight and blade loading (Fig. 12d.3). In particular, pressure to change coat weight lead to deviations
it is necessary to continually increase blade load from the parallel nip geometry, something which
over an extended period of operation when may increase the possibility for coating defects
running on the heel, or decrease the load when (Roper and AttaI1993). When the blade loading
pressure increases, the blade tip shifts closer to
the moving web at the toe and reduces coat
120
e 04 m/s weight. However, this configuration causes the
c- 100 ... S In/s blade tip to form a converging channel with the
'3, web, as depicted in Fig. 12d.3b. A decrease in
110
,.:
:: the blade loading, on the other hand, allows the
"i;j
60
blade tip to run on its heel and apply more or
~ 40 less coating on the web, depending on the
!-
<: 20 sharpness of the blade tip at its heel (Ramp 1983).
0
0 This is illustrated in Fig. 12d.3c. The parallel
0 100 200 300
channel geometry between the blade tip and the
BLADE LOADI G. kPa
web can be maintained only if the mechanical
Figure 12d.7 Coat weight increases with machine mounting mechanism of the flexible blade is
speed in a bevelled blade coater. Experiments by adjusted to compensate for changes in blade
Triantafillopouslos and Altug (1991). loading. It is therefore important to recognize
that, when one changes blade pressure to influence the web, in particular at the nip of an applicator
coat weight, one also needs to adjust the blade roll. These pressures can be as high as 200 kPa
mounting device accordingly to maintain parallel in roll applicators, and about 100 kPa in fountain
nip geometry at the blade tip. applicators (Eklund 1984). In contrast, the pressure
Although blade wear during the operation of in the case of a pond or short-dwell coater is less
steel blades is an important subject, it has drawn than 7 kPa (Korpela, Palsanen and Pitkanen
little attention. Hassell (1981) illustrated that 1986), which is the reason for the widespread
blade wear is proportional to the mass flow rate preference of this application technique. The low
of the coating transferred under the blade and, pressure, in combination with the comparatively
therefore, depends mainly on coater speed and short coating-paper contact time period in
coat weight. Under a constant load, wear reduces short-dwell coaters, minimizes water absorption
as speed increases, the difference increasing at by the paper (cellulose) fibers prior to metering.
high blade pressures. Wear also decreases at high In contrast, substantial absorption occurs before
coat weights. However, it increases with the the blade in the case of a coater with an
abrasiveness of the mineral pigment used in the applicator roll.
coating formulation and the roughness of the An indirect consequence of water leaving the
web. This data indicated that blade wear is coating and entering the substrate is that cellulose
independent of the coating viscosity. As a result fibers absorb water from the coating and swell.
of wear, which changes the blade nip geometry, Swelling influences both the coat weight pickup
new steel blades require readjustment of blade at the blade, as well as paper properties dependent
pressure and holder positioning after a few on changes occurring at the paper surface.
minutes in operation. Recently, ceramic tip blades Qualitative differences which exist between papers
are used for paper coatings which are free of coated with an applicator roll and a short-dwell
wear. However, these blades require proper initial coater have been attributed to fiber swelling and
setting because they have tips with multiple its occurrence before or after the blade metering
bevels. Under normal operational conditions, process (Eklund 1984). For coating with an
ceramic tip blades have a longer life than steel applicator roll, fiber swelling occurs prior to
blades, although care is needed to avoid chipping metering. When the fibers swell, the web surface
the brittle tip of the blade. becomes rougher, requiring heavier blade loading
In addition to fluid dynamics, properties of to deposit a certain coat weight as the blade fills
the coating fluids and the web influence coat in the volume of the surface voids of the web.
weight development in blade metering of paper The blade acts as a smoothing device by covering
webs. One consideration is the penetration of the the surface contour of paper. In contrast, fiber
dispersing phase, i.e., the water with all soluble puffing and swelling take place after the blade
components and small colloidal particles such metering step in the case of a short-dwell coater.
as latex, into the porous and absorbent paper This leads to a relatively rough final coated
substrate. This takes place prior to metering and surface which adversely affects optical and printing
when the coating contacts the moving web for properties. Consequently, papers coated with a
the first time. This process fills in the voids in short-dwell coater have inferior surface charac-
the paper surface and forms a relatively immo- teristics to those coated in a blade coater with
bilized layer, referred to as the 'filtercake', through an applicator roll.
capillary and pressure penetration. This filtercake In addition to substrate properties, the non-
has a much higher solids content than the bulk Newtonian flow behavior of paper coatings
of the coating which is fed into the process. influences coat weight. Typically, higher steady
Pressure penetration seems to be the more shear viscosity leads to higher coat weight under
important mechanism (Sandas and Salminen a constant blade load, or higher load to obtain
1987) as substantial pressure pulses apply onto a targeted coat weight. A complication is that
Hydrodynamic instability in blade coating 653
the blade metering process involves regions with metering, there has not yet been a comprehensive
a wide range of shear rates (Pranckh and Scriven model which also incorporates phenomena
1988). Stagnation regions appear next to areas occurring during the application of the coating
where the shear rate is substantially high. Typical prior to metering. In addition, the influence of
paper coating dispersions are shear thinning, the paper absorbancy and compressibility to the
while they exhibit significant yield stress and final film thickness need to be further investigated,
thixotropic recovery. They also demonstrate in particular as these properties influence forma-
viscoelastic behavior, at least at low strain rates. tion ofa filtercake prior to metering and correlate
When the volumetric concentration of the with swelling of paper fibers and roughening of
dispersion exceeds a certain level, paper coatings the sheet. The exact mechanism of filtercake
become dilatant. Low shear viscosity - obtained formation has not yet been studied in depth, so
with single-point, rotating spindle type of instru- that the operational behavior of the process is
ments - is used to determine the ability to bring difficult to predict and control. A point worth
paper coatings to the metering zone, while high noting in the case of bevelled blade coating is
shear viscosity - obtained with a concentric the influence of blade nip geometry. The blade
cylinder's viscometer - is used to determine the remains in equilibrium only when the geometry
required blade load to attain a certain coat is a parallel channel. When the channel is
weight. Bohmer (1969) was the first to estimate convergent or divergent, coating defects appear,
that, when considering the coating film thickness even though the rheology of the coating fluid
at the blade nip to be about 10 J.1m, the shear has been optimized. Rheological phenomena
rates in the blade nip of a coater running at occurring during the high-speed blade metering
speeds of practical interest can be as high as process and coating defects are discussed in
106 s -1. Viscometric data at these high shear Sections 12d.6 and 12d.7, respectively.
rates are not readily available for paper coatings,
although measurements with capillary viscometers
have recently appeared in the literature (Roper
12d.4 HYDRODYNAMIC INSTABILITY IN
and Attal 1993). Considerations of the non-
BLADE COATING
Newtonian rheology of paper coatings are
discussed in more detail in Section 12d.6. The ideal coating flow is steady state and two-
In summary, the physics of high-speed blade dimensional, that is, there are no temporal or
metering of paper webs involve complex phe- cross-stream flow variations. Any instability would
nomena and their interactions. Viscous forces, almost always result in a transition to a three-
developing in the blade nip and at the underside dimensional or unsteady state. Consequently, the
of the blade, and pressure forces in the blade nip, film thickness becomes nonuniform. If the
contribute to deflecting the blade, but the inertia transition is to a two-dimensional time dependent
of the coating layer reaching the blade also plays state, there will be coating film thickness
a significant role. Although lubrication flow nonuniformity in the streamwise (machine)
seems to be the dominant mechanism for bent direction. Transition to three-dimensional flow
blades, the impulse of the coating layer striking will result in cross-stream variations referred to
the side of the blade is an important mechanism as streaks.
for deflecting bevelled blades. The interaction of There are a number of regions in high-speed
the blade forces under conditions of practical blade coating where the flow could become
interest, i.e., where non-Newtonian effects of the unstable. The most critical modes of instability
coating fluid and the absorbancy ofthe web come are the transitions from a two-dimensional to a
into the picture, is still an issue for investigation. three-dimensional state (i.e., a symmetry breaking
Although computational fluid dynamics shed bifurcation). This transition could take place at
light on the mechanisms of high-speed blade the blade nip, in the case of a diverging gap, or
at other locations upstream of the blade (Aidun scale. The boundary conditions are no slip walls,
1991a-c). given by
In pond application systems, the most serious
ul oo, = (V,O,O) and ulan-an, = (0,0,0) (12d.2)
instabilities are of a centrifugal type which occur
inside the pond. In the pond, the base state where 00 is the boundary of 0 and oOt represents
consists of recirculating eddies which are almost the top surface. At a sufficiently small value of
two-dimensional over most of the coater width, Re, the solution to this system represents a two-
except near the side walls where the Eckman-type dimensional (2-D) flow given by
layer forms. In actual coaters, the coater width U 2D = [u(x,Y), v(x,Y),O] (12d.3)
is at least two orders of magnitude larger than
the pond depth. Therefore, the effects of the The boundary integral form of this solution is
sidewalls on the flow can be assumed to be a available for creeping flows, and in numerical
perturbation of the two-dimensional flow. form, for a wide range of Re by many investigators
In this section, we focus on the pond of these (for example see Bozeman and Dalton 1973); or
coaters which is very similar to a lid-driven Nallasamy and Krishna Prasad (1977)). In practice,
cavity, a rectangular section where the top surface however, this solution becomes unstable at a
moves with a constant speed. In practice, the top critical Reynolds number, Rec ' and most likely
surface is the substrate that is being coated. As gives rise to a steady cellular flow (symmetry
we are interested in the stability of the flow inside breaking pitchfork bifurcation). The value of Rec
the cavity, we assume that all of the surfaces are and the wavelength of the critical mode, A.c '
smooth and impermeable. The substrate is usually depend on the length, H, to depth, D, aspect
being transported by circular rolls. However, as ratio, H/D. The three-dimensional (3-D) solution
the radius of the roll is much larger than the near onset can be approximated by separating
streamwise length of the cavity, we assume that variables in the linearized disturbance equation
the top surface is flat. In the experiments outlined (see Chapter 8 for more information on linear
below, the roll diameter is 61 cm where the stability analysis) and using a trigonometric
streamwise length ofthe cavity is only 5.08 cm long. representation in the z-direction. The critical
The parameters that govern the flow charac- disturbance mode at onset is then given by
teristics are the cavity aspect ratio and the cavity u' = [u' (x, y) cos Z,V'(x,y) sin Z,w'(x,y) sin Z]
Reynolds number
(12d.4)
Re = pUD where Z = 2n(z - rx)/A., A. is the wavelength of
Jl. the periodic cellular flow, and rx represents an
where U is the lid velocity, D is the cavity depth, arbitrary phase of the disturbance structure. As
Jl. and p are the fluid viscosity and density. Let the flow close to the critical Reynolds number
us begin with the simplest case - a lid-driven is a superposition of the two-dimensional base
cavity with infinite span. The Navier-Stokes and flow and the critical disturbance mode, represented
Continuity equations given by in (12d.4), the new state is a three-dimensional
flow pattern with periodicity in the spanwise
U,t + Re(u· vu) = Vp + V2 u in 0 (12d.la) direction.
Now consider a cavity with finite span, L, and
V·u=O inO (12d.lb) free-slip end walls where the boundary conditions
govern the flow in the cavity domain defined by at the end-wall planes, oOe' are given by
O. The velocity vector, u, pressure, p, and time,
t, are scaled with the lid velocity, V, pressure
scale, pvV/D, and time scale, D2 /v, respectively,
w= ° and
au OV
oz oz
- =- = ° on 00 e
and the cavity depth, D, is used as the length (12d.2')

Hydrodynamic instability in blade coating 655
Here also the steady two-dimensional solution

(12d.3) satisfies the boundary conditions (12d.2').
The bifurcating 3-D solution at onset can be
represented by (12d.4) with the eigenfunction, Z,
modified to nrcz/L, where n is the wavenumber
of the disturbance structure inside the container,
and the phase angle, IX, being no longer arbitrary,
is set to zero. Modes representing an even or
odd number of cellular patterns can become 4 L
critical and destabilize the 2-D base solution as ~(L)
shown in Fig. 12d.8. Here the vertical (horizontal)
plane represents solutions with an odd (even) Figure 12d.9 An example of the stability boundary
number of cellular patterns. showing the self-similarity of various cellular patterns
(Aidun, Triantafillopoulos and Benson 1991).
As the slip condition at the end wall given by
(12d.2') is also a symmetry condition, the similarity
mapping used by Aidun (1987) applies. Therefore, transitions from a two-dimensional flow, ideal
from the stability boundary for a single cell for coating systems, to a three-dimensional flow
pattern, R~(L), the entire stability boundary for which results in coating nonuniformities.
the initial bifurcation point can be mapped by Let us next discuss the real driven cavity
the transformation system where the no-slip condition applies to the
end wall as well as any other solid boundary.
R~(L) = R~(~) where n = 1,2'00' (12d.5)
In a series of experiments, Aidun, Triantafil-
lopoulos and Benson (1991) and Benson and
Aidun (1992) investigated the stability of flow in
The expected form of the stability boundary a driven cavity with span to depth aspect ratio
predicted by equation (12d.5) is shown in Fig. of 3 to 1. Their experiment setup consisted of a
12d.9. This figure shows the effect of the cavity roll driving the fluid inside a rectangular cavity,
span on the critical Reynolds number for as shown in Fig. 12d.l0. This is similar to the
-£ells
.....................
...........;.;; I cells
• _ _ .-... Rc
Figure 12d.8 The state diagram for a lid-driven cavity with free-slip end walls, an ideal representation of a pond
application coater.
setup of Pan and Acrivos (1967), although in A very thin film of fluid adheres to the roll
these experiments, the roll diameter is considerably and escapes the cavity through the downstream
larger than the cavity and, therefore, the penet- lip EC. This layer of fluid is scraped off the roll
ration is about 2% compared with 8.5% (of the by a sharp blade downstream of the cavity. At
cavity width) in Pan and Acrivos' experiments. the bottom of the cavity adjacent to the GK
The cavity, placed on top of a lower compart- corner, a narrow slot (3 mm opening) opens into
ment (Fig. 12d.10a), makes contact with the roll the lower section which is connected to a pump
and has length H = 5.08 cm in the direction of and acts as a reservoir for replenishing the fluid
roll motion, depth D = 5.08 cm, and span that escapes the cavity.
S = 15.24 cm transverse to the direction of roll Fluid is supplied to the lower section of the
motion. Thus, the characteristic geometric par- cavity via a 1.00 cm pipe tap from one side. The
ameters are 1: 1 and 3: 1 for depth- and span- observed flow patterns show that the flow through
to-length aspect ratios, respectively. The lower the slot into the cavity is uniform over the span.
compartment has the same dimensions but a At a low Reynolds number, the flow is steady
depth of only 2.54 cm and serves the purpose of and almost two-dimensional. However, as the
feeding fluid uniformly into the cavity. Reynolds number increases above 825, the steady
flow destabilizes and gives rise to a time periodic
state with an oscillation frequency of about
0.1 UIt. Flow visualization of the time periodic
state shows small-amplitude time periodic waves
appearing on the downstream secondary eddy
starting out at the middle symmetry plane and
~z.w traveling towards the side walls. As Re increases,
tQ t, qc.~m I
X.lI these traveling waves become more clearly visible
(see Fig. 12d.11b) until a second incommensurate
frequency, 0.0056 UIt, appears at Re above 990.
L.-..,r--~- FLUID I LET Benson and Aidun (1992) show that the flow
HYDRAULI remains quasiperiodic up to Re = 1055 where
ARM
(a) many other modes are excited and a broad band
of frequencies appear signaling the existence of
a low dimensional chaotic state. Figure 12d.lld,
shows the flow at Re = 1900. At this state, the
mushroom shaped vortices appear and disappear
randomly inside the cavity. These structures are
similar to Gortler-like vortices that Koseff and
Street (1984a-c) have reported at higher Reynolds
numbers.
The length scale of the vortices presented in
primary vortex Fig. 12d.11d is comparable to a class of streaks
.. s b that are observed at high speed with pond
(b) application systems such as the short-dwell coater.
However, other issues such as air entrainment
become important mechanisms which promote
Figure 12«1.10 Schematic of the experimental setup defects in coaters ofthis type. Recent experiments
showing (a) the coordinate system x, y, z and the
corresponding velocity components u, v, w, and (b)
(Veverka and Aidun, 1991; Li and Burns, 1992)
recirculating primary and secondary eddies of the indicate that air entrainment is another significant
steady state primary mode. cause of coating streaks.
Air entrainment 657
12d.5 AIR ENTRAINMENT to a moving substrate. The speed of the substrate
and, therefore, the rate of coating is limited by the
Application of a thin liquid film on a flat substrate maximum critical speed of displacing the air.
involves displacement of air by liquid. With flexible Beyond the maximum wetting speed, air cannot
substrates, such as paper, photographic films, or be effectively displaced, and some patches of air
magnetic tapes, the coating is continuously applied remain on the surface or entrain into the liquid.
Figure 12d.ll Front-view photographs of flow patterns. From top to bottom: (a) two-dimensional steady flow
with primary and secondary rolls at Re = 500 (notice the straight separation line); (b) wavy separation line for
Re = 700-1000 (time-periodic flow); (c) appearance of mushroom-like structures for Re = 1100; (d) unsteady
state at Re = 1900 (Aidun, Triantafillopoulos and Benson 1991).
With the increasing demand for higher pro- at the dynamic contact line. Miyamoto and
duction rates, air entrainment has emerged as Scriven (1982) analyzed the two-dimensional
one of the leading problems in the coating idealization of this flow and concluded that
industry. By air entrainment we mean a physically Kelvin-Helmholtz-type instabilities are likely to
isolated patch of air or bubble that enters into occur for thicker gas films, while a 'disjoining
the liquid phase. In pond application systems, collapse' instability at the interface will dominate
the air gradually accumulates in low pressure for thinner air layers. Miyamoto (1991) presents
areas. These air pockets interact with the substrate some experimental evidence of massive shear
and generate coating defects mostly in the form instability ofthe air film and formation of'craters'.
of streaks on the surface (Aidun 1989). This section focuses on the macroscopic regime
The critical speed for onset of air entrainment of air entrainment where the entrained bubbles
has been measured by many investigators (Ken- are too large to dissolve in the liquid and,
nedy and Burley 1977; Burley and Kennedy therefore, remain as air bubbles in the liquid.
1976; Bolton and Middleman 1976; Blake and To study the kinetics of wetting, Deryagin and
Ruschak 1979; Burley and Jolly 1984; Sullivan Levi (1964) used a simple slot-like coating
and Middleman 1979), see Chapter 3 for more apparatus to reproduce the process of coating
detail. Some of these measurements, however, an emulsion on a flexible substrate. They used
are subjective and show inconsistent behavior. a wetting hopper with a square cross-section with
For example, some studies (Kennedy and Burley sides equal to 2.4 cm. Using visualization tech-
1977; Burley and Kennedy 1976) indicate that niques, they observed the deformation of the
the substrate entry angle significantly affects the wetting line as they increased the substrate speed.
critical air entrainment speed, while other studies At low base velocity of a wettable liquid/solid
(Bolton and Middleman 1976; Blake and Ruschak system, the contact angle is small, and 'perfect
1979) show no relation between the two. The wetting' is observed. As the coating speed increases,
effect of surface roughness on the critical speed the meniscus deforms, and the contact angle
for air entrainment has also remained unclear; increases to 180° as shown in Fig. 12d.12. They
some studies (Burley and Jolly 1984) show a report that at this critical speed, a shift in the
strong interdependence, while others (Sullivan contact line is observed by the appearance of a
and Middleman 1979) indicate otherwise. It is triangle where contact between the liquid and
probably the different experimental setups and the substrate is broken. As the substrate speed
substrate materials that are responsible for some increases further, the triangular area grows to a
of the inconsistencies. point where the brake extends over the entire
In this section, we focus on the physical length ofthe contact as illustrated in Fig. 12d.l2b.
mechanism of air entrainment in blade coating The pioneering work of Deryagin and Levi
systems. In particular, we investigate the sequence (1964) reveals an important event and limitation
of events which leads to formation of air bubbles in the wetting of solids by liquid. A contact line,
and their entrainment into the liquid. We recognize originally straight, breaks into a sawtooth pattern
at least two regimes of air entrainment (Scriven at a critical speed where the contact angle reaches
1982): (1) a microscopic regime, proposed by 180°. In the ideal situation where the contact line
Miyamoto and Scriven (1982) and further inves- is infinitely long, this behavior can be viewed as
tigated by Miyamoto (1991), and (2) a second a transition from a two-dimensional flow to a
regime at macroscopic scale where air bubbles three-dimensional state.
in the order of few hundred microns and larger Blake and Ruschak (1979) attribute this behavior
form at the wetting line and penetrate into the to a maximum wetting speed above which the
liquid. contact line has to break into a V-shaped pattern
In the microscopic regime, it is hypothesized so that the effective wetting speed falls below the
that a thin air layer, less than a micron, forms maximum value. A rigorous analysis of the
Non-Newtonian rheology in blade coating 659
contact line instability and formation of triangular patterns under ultraviolet light after the coated
air pockets remains to be undertaken. samples were dried. The substrate used was paper
Many experiments, as outlined above, show and the coating speed was varied from 15 mls to
that the prerequisite for air entrainment in the a maximum of 25 m/s. To examine the effect of
wetting process of a smooth surface is the air accumulation in the pond, Li and Burns
formation of these structures. The events that (1992) placed a ventilation tube inside the pond.
follow this transition and lead to air entrainment A series of experiments was completed with the
in low-speed roll coating systems are discussed ventilation tube extracting air from the pond. A
by Veverka and Aidun (1991) and Aidun, Veverka second set of experiments was conducted under
and Scriven (1992). There is some evidence that the same conditions with the ventilation tube
the same sequence of events also leads to air blocked so that no air could be extracted from
entrainment with high-speed pressurized pond the pond. In their experiments Li and Burns
application blade coating systems. The difference (1992) clearly document that in every case,
is that with pressurized pond systems, the fluid extraction of air significantly reduced the coating
is forced out of the cavity at the overflow baffle thickness nonuniformity and streaks on the surface.
near the contact line (see Fig. 12dAe), substantially
delaying air entrainment to higher speeds.
12d.6 NON-NEWTONIAN RHEOLOGY IN
There is now direct evidence of the serious
BLADE COATING
adverse effects of air entrainment in pond
application systems. In a series of experiments Paper coatings comprise highly concentrated
with actual pressurized pond coating systems, Li dispersions of mineral particles, with volumetric
and Burns (1992) showed that the accumulation concentrations in excess of 40% by volume. Flow
of air inside the coating cavity results in streaks characteristics are generally determined by the
at speeds much below the 'skip' coating regime. volume fraction of the disperse phase, the
They used typical paper coating liquids with a magnitude ofinterparticle forces and the structure
small amount of fluorescent dye. The dye was of the flocculates formed. In addition, the
added to detect the nonuniform coating thickness dispersing phase of paper coatings contains
(a) u ..
-9-->180 ~,-
(b)
Figure 12d.12 Schematic of the boundary displacement during wetting (Deryagin and Levi 1964).
polymers which mayor may not be water soluble. the effect of rheology on high-speed blade coating
Although these polymers are added at small for paper is an evolving field of research, there
amounts (typically 0.1% to 5%), they substantially exist some documented correlations between
affect coating rheology. Polymers function by coating rheology and coatability. The following
inducing a network structure which physically paragraphs point these out, while addressing the
bounds polymer molecules and/or dispersed main unresolved issues. For comparison, reference
particles. This structure leads to development of is also made to rheological effects in the relatively
yield stress and viscoelasticity at low shear. At low-speed blade (or knife) coating.
high shear rates, forces holding the network Viscoelasticity in high-speed blade coating of
structure are disrupted leading to shear thinning. paper can be assessed from the Deborah number,
This structural breakdown is also associated De = A/t, where A is the characteristic relaxation
with time-dependent rheological behavior, known time of the coating liquid or dispersion, and t is
as thixotropy. Typical coating dispersions, the characteristic residence time in the process
therefore, have high viscosity at low shear rates flow. Most coating materials have a very broad
and low viscosity at high shear rates. At increased relaxation behavior and, therefore, exhibit a slow
particle concentrations, however, paper dispersions change of storage and loss moduli with frequency
become increasingly non-Newtonian, independent (Goodwin 1989). For example, typical values of
of the flow behavior of the dispersing phase. the relaxation time for paper coatings are in the
At low stresses, coating dispersions show not range of 0.1 to 5.0s (Roper 1990; Salminen and
only plastic behavior but also viscoelasticity. In Fors 1992). These values correspond to the time
addition to polymeric network structures, these period for the shear modulus to decrease half of
rheological phenomena arise from three-dimen- its value at 0.Q1 s in a dynamic rheometer. Values
sional network structures due to intermolecular obtained with this method represent quantitative
interactions between pigment particles. When approximations of the characteristic relaxation
subjected to low shear stresses, these structures spectra. Considering that a typical residency
attain a dynamic equilibrium between structural period at the blade nip is approximately in the
breakdown and re-formation under the applied order of magnitude of 10- 5 s (Kline 1985), the
shear stress and exhibit primarily elastic defor- De is much larger than 1. Such high Deborah
mation and a high viscosity. Consequently, in numbers suggest that paper coating dispersions
addition to flow curves which describe the could behave like elastic solids rather than
relationship between shear rate and shear stress, mobile liquids during their application.
rheological characterization of paper coatings Similar or larger Deborah numbers are common
requires controlled-stress and dynamic viscoelastic in coating of polymeric liquids onto smooth and
measurements. impermeable substrates. The effect ofviscoelasticity
The rheological behavior of paper coatings on coating film thickness has been well
plays an important role during processing and documented for the case of lubrication flow in
application in a high-speed blade coater. It rigid blade coating (Middleman 1981). Prediction
determines not only the coating flow at the blade of the film thickness (or coat weight) requires
nip, but also the flow feeding the application consideration of the normal forces arising from
zone. Rheology influences the hydrodynamics the viscoelasticity of the polymeric liquid in
upstream of the blade through the steady shear addition to viscous forces (Hsu, Malone and
viscosity and plastic yield. Once at the blade nip, Laurence 1985). Theoretical and experimental
high-shear viscosity and, possibly, viscoelasticity studies have shown that, while the non-Newtonian
affect the blade deflection and the final film shear viscosity influences the flow and pressure
thickness (or coat weight). After the blade nip, distributions at the nip, viscoelasticity affects
thixotropic recovery contributes to the post blade loading (Greener and Middleman 1974;
metering behavior of the coating layer. Although Hsu, Malone and Laurence 1985; Hsu 1986). In
Non-Newtonian rheology in blade coating 661
particular, viscoelasticity contributes to the rheology of polymeric liquids, which exhibit both
separating (normal) force which tends to push viscous and elastic behavior, it is difficult
the bounding surfaces ofthe web and blade apart. exclusively to attribute their flow behavior at
The separating force increases with flow velocity, blade nips to elasticity. Many times elastic effects
i.e., the speed of the backing roll (Hsu 1986). For compete with viscous behavior (shear thinning),
relatively high flow velocities, the load capacity making prediction and control of the final film
on a stationary blade over a moving plate can thickness difficult to achieve.
be up to fifty times greater for a non-Newtonian The role of viscoelasticity in high-speed blade
viscoelastic lubricant that for a Newtonian fluid coating of paper is further complicated. The
(Doremus and Piau 1981). In addition, much less highly concentrated coating dispersion contains
fluid is carried through the gap in the case of orientable particles and polymeric additives which
viscoelastic fluids, while both the stagnation interact with the dispersed phase in many different
point ofthe flow and the maximum normal stress ways to affect rheology. In addition, paper webs
abscissa move closer to the gap (formed between are compressible, nonuniform, permeable and
the blade and the plate) center (Doremus and filled with air. Water soluble binders and additives
Piau 1983). Theoretical analysis (Greener and - such as starch, carboxymethylcellulose, hy-
Middleman 1974) of the lubrication flow in a droxyethylceIIulose, polyvinyl alcohol, etc. - added
blade nip indicates that, as compared with a to paper coatings induce viscoelasticity and
Newtonian fluid, the pressure developed under increase viscosity at low shear rates. However,
the blade decreases for a polymeric viscoelastic they also make the coatings shear thinning at
fluid and increases for an elastic Newtonian high shear rates. Therefore, it is difficult to
(Boger) fluid. Sullivan (1986) demonstrated that separately assess the viscous and elastic contri-
film thickness increases with purely viscous butions of rheology to the flow at the blade.
shear-thinning fluids, while viscoelastic fluids, Results by Sullivan (1986) suggest that shear
which have both viscous (shear-thinning) and thinning, characteristic of paper coatings, may
elastic behavior, cause the film thickness to either offset the effect of elasticity in the case of
increase or decrease depending on the relative viscoelastic coatings.
contributions ofshear-thinning and elastic effects. Normal forces due to viscoelasticity could
For the case of fluids with the same viscosity but influence the flow and the force balance in and
different elasticity, film thickness decreases as around the blade nip during high-speed blade
elasticity increases. This trend is also valid for a metering of paper coatings. This, in turn, could
Newtonian fluid having elastic properties, i.e., a affect the final coat weight obtained under a
Boger fluid. In a blade-over-roll geometry, the given set of process conditions. Elasticity is
coating film thickness increases with both the expected to generate substantial normal forces
blade angle and blade extension, although coating perpendicular to the flow direction when the
films of highly elastic Boger fluids are independent coating is flowing out from the restricted channel
of blade nip configuration (Sullivan, Middleman under the blade and the relaxation time is
and Keunings 1987). Furthermore, the operability sufficiently long (Ginn 1974,1977). Then, normal
window of blade coaters shrinks with viscoelastic forces and viscoelasticity need to be taken into
liquids (Strenger, Secor and Sramek 1992). The account when setting up the operating parameters
limiting factor is the appearance of ribbing of the process. Windle and Beazley (1967,1968)
instability on the coating film. This defect is analyzed the flow beneath a bevelled blade based
associated with converging blade geometries on the hydrodynamic lubrication flow in a
where inclusion of elastic normal stresses in the converging wedge and considered paper com-
lubrication model predicts, for high inclinations pressibility and roughness in determining coat
(i.e., large blade angles), a rising pressure at the weight. These authors suggested that the hy-
exit from the channel. Because of the complex drodynamic lift acting on the blade arises, at
least partially, from the normal forces of coating costant shear rate (i.e. 17.4 x 104 s- 1). For the
viscoelasticity which contributes to the total clay/carboxymethylcellulose coatings tested by
thrust of the flow through the nip. However, Grankvist and Sandas (1990), normal forces
laboratory coater experiments with coatings increased exponentially with shear rate above
containing clay, styrene butadiene latex and 10 5 s- 1. At constant shear rate, normal forces
starch, and having the same low-shear viscosity increase with the molecular weight of the
but different low-strain elasticity did not yield carboxymethylcellulose and its addition level.
significantly different coat weights. Windle and Laun and Hirsch (1990) investigated the normal
Beazley (1968) attributed their inability to forces of two coatings (64% solids by weight),
document the effect of viscoelasticity to (1) failure one of the coatings demonstrating absence of any
to separate the viscous from the elastic contri- normal forces, while the second one - which was
butions to the dynamic forces acting on the blade, dilatant at shear rates between 103 and 10 5 S-1
and (2) the possibility that the passage of the - had normal stress differences which were five
coating through the blade nip might be viewed times greater than the corresponding shear stresses.
as fracture of a weak elastic solid rather than It is possible, therefore, for high-solids coatings
flow of a viscous liquid. which are dilatant, and consequently behave like
It is important here to recognize that charac- elastic solids as shear rate increases, to demonstrate
terization of coating viscoelasticity is usually substantial normal forces in shear flows when
based on oscillatory dynamic measurements at the shear rate is sufficiently high, i.e., above
low deformations. These measurements are not 10 5 s - 1. The dependence of normal forces on
sufficient to characterize the behavior of coatings shear rate follows the well-known relationship of
at the very fast deformation process in the blade normal stresses being proportional to the square
nip. Similar to steady-shear viscosity measure- of shear velocity (Grankvist and Sandas 1990).
ments, viscoelastic measurements cannot be Normal forces may also arise from shear
extrapolated from low to high deformations. The induced structures (Brady and Bossis 1988) formed
problem is that it is difficult to measure viscoelastic across shear planes existing between the blade
properties at shear rates of the order of 10 5 to and the web. Computer simulations (Bousfield
106 S - 1 which are experienced under the blade 1990) illustrated that shear in the gap, formed
and inside the blade nip. between the paper and the blade surface, can
Attempts to measure the normal forces exerted cause an ordered arrangement of pigment particles
by paper coatings are based on viscometric such that 'bridging' between the two boundaries
measurements in capillary flows. Kurath (1965, occurs. Formation of bridges develops an addi-
1974) documented that normal stresses are much tional normal force, which can be as much as
larger than shear stresses for clay dispersions in four times the normal force occurring originally.
a starch solution, increasing with the shear rate Bridging frequency increases rapidly with in-
and volumetric concentration. However, the creasing solids content of the coating.
dependence of normal stresses on concentration The recent growth of interest in understanding
was limited to clay/starch coatings and did not the physics ofblade metering and coating rheology
make any difference in a coating containing clay, is partly due to higher machine speeds and lower
calcium carbonate and starch. It is also worth coat weights. Currently, many industrial coaters
noting that the magnitude of normal stresses run in excess of 24 m/s and apply from 4 to 10
depends not only on the type of mineral pigment grams of coating per square meter of paper.
used, but also the size and shape of the pigment. When the machine speed increases, the deforma-
Kurath (1974) found that, when a fine particle tion process under the blade becomes faster and
clay was substituted with a delaminated (platey) the Deborah number increases, while large
clay in a starch formulation, both shear and transient stresses develop due to the relatively
normal stresses increased substantially under high blade loadings. Such process conditions,
Blade coating defects 663
coupled with the trend toward coatings with high 12d.7 BLADE COATING DEFECTS
volumetric concentrations, make consideration
of normal forces in high-speed blade metering of As with any coating operation, a defect free
paper webs more realistic. surface is important in maintaining consistent
In summary, the non-Newtonian behavior of and substainable productivity and product quality
paper coating dispersions arises from interparticle in high-speed blade coating. Although there are
interactions and the network structure of polymeric many different types of defects appearing in the
additives. Most paper coatings are viscoelastic various applications of the blade coating process,
at low strain rates, demonstrate substantial yield here we concentrate only on blade coating defects
stress, and are shear thinning thixotropic over a on paper webs. Furthermore, the focus is on
broad range of shear rates. The magnitude of the defects arising from process flow dynamics and
characteristic rheological parameters depends properties of paper coating dispersions.
primarily on volumetric concentration and the Good runnability usually means coating without
amount and type of polymeric additives and, visual film defects on the coated surface and a
secondarily, on the particle size and shape of the uniform coat weight (or film thickness) in both
mineral pigments used. The viscoelasticity of the machine and cross-machine directions. In the
paper coating dispersions is substantially lower development of a coating formulation strategy,
than that of polymers. Although viscoelasticity good runnability also means being able to obtain
prevails at low strain rates, the elastic network a certain coat weight at comparatively lower
structures breakdown with shear so that, at high blade loads. This leads to lower blade forces
shear rates, paper coatings behave like viscous acting onto the web, thus minimizing the possibility
suspensions. The degree ofshear thinning depends for web breaks. Indeed coater runnability becomes
on the coating formulation, i.e., coatings with more important as the speed of commercial
soluble polymers have relatively high low-shear coaters increases beyond 24 mjs and the coating
viscosity and low high-shear viscosity. Suspensions film thickness continually reduces.
of coarse, highly anisometric pigment particles Most of the common visual defects in blade
demonstrate comparatively high elasticity at low coating appear in the form of scratches, skips or
shear and viscosity at high shear. Although most streaks on the coated surface. Scratches usually
paper coatings do not demonstrate significant appear in the form of fine hair-like indentations
elasticity at high shear rates, coatings with (less than 3 mm in width) in the coating layer
extremely high volumetric concentration become running along the machine line direction. Skips
dilatant and behave like weakly elastic materials. are associated with a patterned surface appearance
The large Deborah number makes processing of because of areas devoid or depleted of coating.
these type of coatings in a high-speed blade The appearance of skips is associated with air
coater problematic. However, there is not yet a inclusion or entrapment into the coating pond
clear connection between rheological properties or in a region upstream the metering blade.
ofcoatings and their processability as high-speed Streaks, on the other hand, appear in various
coaters. This is partly due to interactions between forms. Narrow streaks take the form of relatively
the coating dispersion and the substrate, such as narrow indentations, 5 mm and up to a few
formation of a filtercake on top of the web prior centimeters in width. Wider streaks can be as
to metering and the swelling of cellulosic fiber broad as several centimeters and form bands of
of paper when they come in contact with a heavier coat weight running along the machine
coating dispersion. Furthermore, it is not yet direction. All of these defects are easy to observe
known how the coating layer rheology after when paper samples are viewed with transmitted
metering and before drying influences the ap- ultraviolet light. Streaks appear as variation of
pearance ofcoated surface defects and the end-use opacity across the web, while scratches appear
properties of paper. as thin, less opaque lines. The origin and
mechanisms responsible for these defects are are carboxymethylcellulose, hydroxyethylcellulose

numerous, some of them discussed in more detail (and their associative derivatives), alginates,
in the sections about air entrainment and polyvinyl alcohol, polyurethane associative
hydrodynamic defects (Sections 12d.4 and 12d.5). thickeners, alkali swellable lattices, etc. Soluble
The appearance of scratches depends on binders and their modification, such as starch
properties of all three components in a blade and proteins, are also inducing some water
coating operation; namely, the base paper, the retention in paper coatings.
coating dispersion, and process parameters or Experiments by Gane, Watters, and McGenity
blade nip geometry. Usually this defect is a (1992) illustrated that the coating layer, or
problem with bevelled but not bent blades. filtercake, formed next to the web prior to
Huang (1984) separated scratches into mechanical metering has a higher solids concentration than
and rheological based on their origin. 'Mechanical' the coating at the application station or feeding
scratches are attributed to the entrapment of stream. Coatings with good water retention,
foreign matter in the blade nip. These materials however, do not loose substantial water and
may be loose fibers from the web, contaminants soluble material into the basesheet at the applicator
in the coating stream, nondispersed or aggregated roll and the blade nips. Because of that, the
mineral pigments in the coating, poorly prepared filtercake build onto the web before metering
soluble binders or thickeners, etc. Additionally, maintains a solids content close to that at the
hard spots on the bevel of the blade or chipping application zone.
of a ceramic blade can cause scratching of the Associated with water retention is the term
coated surface. 'immobilization solids', which represents the
'Rheological' scratches are associated with volumetric concentration at which a coating
properties of both the coating dispersion and the dispersion has almost infinite shear viscosity -
basesheet. The coating needs to have good 'water the coating becomes dilatant. For a coating
holding', also known as water retention. Water dispersion with a certain solids content, the
holding defines the capability of a coating characteristic immobilization solids are closely
dispersion to control release of its dispersing connected to water retention. Ideally, coating
phase when it makes contact with the porous immobilization should occur after the blade
web during processing (Eklund 1981). Water and metering step of the application process (Eklund
solubilized materials are released from the coating 1981). Excessive dewatering prior to metering
by capillary and pressure penetration, but the can cause the filtercake formed onto the web to
dominant mechanism for dewatering is pressure become dilatant. In turn, a dilatant filtercake
penetration experienced by the coating dispersion reaching the blade nip can generate scratches
at the nip of the applicator roll and under the and narrow streaks (Huang 1984, 1986).
blade (Sandas and Salminen 1987). Water retention Coating dewatering also depends on properties
is associated with the physical and chemical state ofthe basesheet; namely its surface pore structure,
of network structure present in the dispersion compressibility, smoothness and absorbancy. An
due to interparticle interaction and the presence 'open', absorbant basesheet tends to take up more
of polymeric additives. Consequently, it depends water from the coating and, therefore, increases
on the type of mineral pigment used, and its the solids content of the filtercake layer formed
ability to form an array ofclosely packed particles prior to metering next to the web. Coating
under the influence of an external pressure pulse; immobilization before the blade nip is responsible
the low shear viscosity of the coating and the for the appearance of scratches or narrow streaks
liquid phase; and the nature ofstructure networks due to entrapment of particles or particle agglom-
due to the presence of polymeric additives. Water erates at the blade nip. Furthermore, scratches
holding additives used in paper coatings may be and streaks can be generated from poor water
soluble or insoluble in water. Typical additives retention at the stagnation line, appearing on the
blade boundary, of the converging flow upstream cal and viscoelastic properties of the continuous
from the blade nip (Pranckh and Scriven 1988). phase. Furthermore, it seems that the rapidly
Eventually, new coating material reaching the accelerating flow at the converging geometry
blade will push large agglomerates deposited at near the blade nip would generate elongation (or
the stagnation line through the blade nip. As a stretching) along the streamlines of the flow, thus
result, streaks and scratches due to this mechanism introducing the elongation viscosity of the coating
are temporal in nature, disappearing over a long as an additional variable. In that sense, the flow
time of operation. in the blade nip can be viewed as a time-dependent
The exact mechanism of interaction between extensional deformation. Usually, the elongational
water retention and the substrate is not yet viscosity is at least three times greater than the
completely understood. Water leaving the coating steady shear viscosity. This rule may also apply
layer on top of the traveling web is absorbed by for highly concentrated dispersions of paper
cellulosic fibers of the web, thus causing them to coatings containing soluble polymers (Schurz
swell and rise. Fiber rising, in turn, influences 1984). However, recent measurements at shear
the quality of the coated surface and properties rates of 5 x 10 5 S - 1 demonstrated that paper
like gloss. In systems with an applicator roll, coatings without soluble additives (i.e., starch,
fiber swelling takes place prior to the blade nip, carboxymethyl or hydroxyethyl cellulose, asso-
while in the case of a short-dwell coater, swelling ciative thickness, polyvinyl alcohol, etc.) do not
takes place after the blade since the fibers do not have a measurable elongational viscosity (Carreau
have enough time and the coating does not and Lavoie 1993). These coatings flow as
experience high normal pressures prior to metering. Newtonian fluids at high shear rates. Elastic
Consequently, there are differences in the final structures in suspension, which prevail at small
quality of papers coated with each one of these rates of deformation, are destroyed at high shear
two types of coaters (Eklund 1984). Although the rates, i.e., structures are sensitive to shear and
rise in coating solids concentration between the disappear at high strains or shear rates.
application and metering steps has been In addition to the above defects a common
documented, little is known about the exact problem in many blade coating operations with
mechanism of the dewatering process. Recent paper webs is the appearance of almost dried
computational work suggests that dewatering out coating extruded from under the blade at
occurs through a filtration process (Letzelter and the outgoing nip. This defect is known as feathering,
Eklund 1993). weeping, whiskering, or bleeding, and is more
The origins of visual coating defects and large common in the bevelled than the bent blade
scale (i.e., several centimeters in width) coat mode. Many times, bleeding leads to localized
weight nonuniformities at high coater speeds are wet deposits at the exit from the blade nip which
still under investigation. In addition to other grow over time to eventually form dried out
blade coating variables, such as the basesheet deposits, known as stalagmites, hanging onto the
and blade nip geometry, the coating formulation upper side of the blade. Therefore, often bleeding
plays an important role. The major formulation is qualitatively classified as wet or dry, depending
variables are: the choice ofplatey pigment particles on its appearance. Stalagmites as long as several
(i.e., delaminated clays) and their size distributions, centimeters have been observed in practice. Their
the continuous (dispersion) phase rheology, and appearance is a localized phenomenon which
the solids content of coatings. may not adversely affect the coated paper quality
Experimental evidence suggests that the ability until stalagmites detach from the blade and
of highly anisometric particles to orient parallel deposit on top of the paper, or grow to a size
to the flow direction in the blade nip might be where they start scratching the surface. When
critical (Gane and Coggon 1987). Among other this happens, wet and dry bleeding lead to
variables, particle orientation depends on rheologi- visually observed coating surface defects.
There have been numerous attempts over the away with the web. Scanning electron micro-
years to correlate coating and process variables photography (Vodnick et al. 1993) indicated that
with the occurrence of wet and dry bleeding. all weeps have a bulbous crown atop a comb of
Generally, bleeding is promoted by high coater separate trunks which join, two-by-two, in their
speeds; highly concentrated coating, i.e., the upper reaches. Evidence from these observtions
solids content of coatings; poor water retention; suggests that trunks and crowns grow through
increased coating viscoelasticity; the type of different mechanisms and that, at least partly,
mineral pigment particles used, i.e., clays are trunks grow by blockage oflarge pigment particle
more susceptible to bleeding than calcium booklets or lumps trapped at the blade nip.
carbonates (Hofmann and von Raven 1986; The viscoelastic character of paper coating
Eklund and Fors 1988; Gane, Watters and formulations can also cause the appearance of
McGenity 1992; Triantafillopoulos and Grankvist coating surface defects. These defects in turn
1992). Additionally, bleeding reduces with high influence the aesthetic appearance, properties
blade angles and decreased blade thickness and and quality of the coating layer. The 'memory'
blade extension in the case of bevelled blades induced by viscoelasticity at low deformations
(Engstrom and Rigdahl 1987). Although many can dramatically change the spreading response
of these conditions are sufficient for the occurrence and consolidation of coatings after metering.
of bleeding, they are not always necessary. For Immediately upon cessation of shear stresses and
example, it was frequently thought before that other forces acting during the fast deformation
shear thickening in the form of dilatancy under at the blade nip, the stored energy in the coating
the blade tip could cause formation of aggregates begins to relax. The coating layer would then
and the appearance of bleeding. However, return, at least partially, to its previous state of
experimental work has demonstrated that this is rest. Expansion (or spreading) of the coating
not always a requirement (Engstrom 1984). Many layer after the blade tip, due to removal of
times, shear thinning coatings are associated with external stress, promotes coverage of the substrate.
the buildup of stalagmites at the blade. Gane The recoverable strain may be related to coating
and Coggon (1987) speculated that plug flow or patterns and defects, such as scratches and
slippage at the blade tip of a bevelled blade cause various types of streaks (Engstrom and Rigdahl
bleeding and stalagmites formation. Furthermore, 1987). This phenomenon would be more pro-
the coating film at the exit of the nip fractures, nounced for coatings with a more elastic character
thus causing emission of coating droplets and because, to some extent, levelling-out of blade
aggregates. Gane and Coggon (1987) gave no induced streaks is controlled by viscoelastic
explanation of the mechanism which causes the properties of the coating (Adolfsson, Engstrom
film fracture. and Rigdahl 1989). The relaxation time, which
Recent studies with high-speed visualization characterizes coating viscoelasticity, is often
and microphotography have shed light on the comparable to the time available before the
phenomenon of blade bleeding, or weeping. coating layer has been immobilized due to
Stalagmites are initially formed by a stepwise dewatering into the substrate. This is especially
process which involves formation ofsmall columns true in the case of coaters which run at high
of nucleation, or trunks, which then branch out speeds. When one compares the relaxation times
(Branston et al. 1993). The appearance and of different coatings - determined by stress relax-
disappearance of new growths is a continuous ation measurements - with the residual width of
process. Two ways that weeps grow are through an initially rather broad streak deliberately induced
trunks which continually acquire coating that into the wet coating layer, the longer the relaxation
flows up their web sides, and crowns that acquire time the wider is the remaining streak (Adolfsson,
coating during temporary contact with the moving Engstrom and Rigdahl 1989). Appearance of this
web. Overgrown weeps eventually are propelling type of coating defect at the surface of paper can
be, at least partially, attributed to the inability of to be as close as possible to parallel to the web
the coating film to spread and level out after surface (or tangent to the backing roll at the
metering and before drying. Furthermore, at high blade nip region). Under this configuration the
coater speeds, viscoelastic coating formulations blade is truly running on its bevel, as shown in
have a strong tendency to form stalagmites at the Fig. 12d.3a, and the coat weight is proportional
blade surface upstream of the blade nip exit (Eng- to high-shear viscosity measured with a capillary
strom and Rigdahl 1989). The correlation of the viscometer (Roper and Attal 1993). When the
presence of blade defects to the characteristic relax- blade is off its bevel, the system is not in
ation time of a coating has also been illustrated equilibrium and coating defects are more likely
recently with measurements of stress relaxation of to occur. Running on the toe of the blade (Fig.
paper coatings (Young and Fu 1990, 1991). Paper 12d.3b) can lead to dry bleeding, formation of
coatings containing soluble thickeners exhibit sub- stalagmites at the blade surface and web breaks
stantial visoelasticity. These coatings which have at high blade loading. Running on the heel (Fig.
comparatively slow relaxation would most likely 12d.3c), on the other hand, generates a divergent
exhibit formation ofstreaks in the final coated paper. nip where film splitting at the exit from the blade
A few experimental studies have been made induces visual defects, even though the coating
over the years to investigate the influence of the rheology may be acceptable (Hofmann and von
viscoelasticity on defect-free runnability and Raven 1986; Roper and Attal 1993). It is therefore
bleeding in high-speed blade coaters. Sandas important to recognize that, in addition to all
and Salminen (1991) investigated the effect of the other variables discussed earlier, the blade
starch, carboxymethylcellulose, polyvinyl alco- geometry is a critical variable which needs to be
hol and an alkali-swellable thickener on rheologi- monitored and controlled.
cal properties, including viscoelasticity, and on In summary, the appearance of surface defects
coating performance (the pigment was a platey in high-speed blade coating depends on all three
English china clay). The oxidized starch, which components of the process: the paper web, coating
had the lowest elastic modulus at low deforma- rheology, and the metering geometry of the
tions, showed the best performance at the coater. blade. Although the exact mechanisms leading to
A similar study (Triantafillopoulos and Grankvist coating surface defects have only recently been
1992) with a variety of pigments and three explored, many cause-and-effect relationships
different starches demonstrated that wet bleeding have been established through practical experience.
and scratches appear when the low-strain elasticity For instance, too porous and absorbent webs
is relatively high. Particularly, coatings with are detrimental to defect-free runnability because
cationic starch have high elasticity which sub- of excessive dewatering of the coating dispersion
stantially reduces at high strains and develops prior to metering and an increase in the solids
to excessive shear thinning. Dry bleeding, content of the filtercake reaching the blade.
on the other hand, does not seem to depend on Water entering the fibrous matrix of paper causes
elasticity but rather the geometry of the blade fibers to swell, a phenomenon which influences
nip. both the coat weight and end-use properties of
Sustained and consistent coater runnability paper. Generally, coatings with comparatively
without defects and a uniform coat weight does high elastic moduli at low strain rates and high
not only relate to viscoelasticity. It is also a plastic yield are more difficult to process up to
function of a combination of formulation par- the application zone and are associated with the
ameters including: steady shear viscosity at low appearance of visual defects, appearing both on
and high shear rates, water retention, and the blade and the web. In addition, relatively
immobilization solids. In addition, the blade nip high high-shear viscosity is often associated with
geometry in the case of a bevelled blade plays poor runnability because it requires higher blade
an important role. Preferably, the blade tip has pressures to obtain a certain coat weight,
something which may lead to coating defects coated film thicknesses (10-25 J.1m), while bent
because the stresses applied onto the web also blades are commonly used at lower speeds (below
increase. Beyond the rheological parameters, the 7 m/s) for heavier coat weights. However, bevelled
water retention property of a particular formu- blades are more susceptible to visual coating
lation plays a role in its processability. Coatings defects and web breaks as the blade loading
with the ability to hold water, particularly under pressure - and, therefore, the stress applied onto
the influence of a pressure pulse, can be processed the web - continually increases. It is also important
trouble free at high coater speeds. Furthermore, to realize that the operation of bevelled blades
the parallel channel geometry needs to be always is very sensitive to the geometry under the blade
maintained at the tip of a bevelled blade, as tip. Deviations from a parallel channel geometry
convergent or divergent flows at the blade nip lead to covergent or divergent flows at the blade
will lead to coating defects. If this requirement nip which are responsible for the appearance of
is not met, coating defects may appear even coating surface defects, even when both coating
though the substrate and coating rheology have rheology and web properties hve been optimized.
been optimized for good runnability. Bent blades are generally less susceptible to
coating defects.
Variations in the feeding system to the metering
12d.8 CONCLUSION
blade are also responsible for qualitative differences
High-speed blade metering of paper webs is a in the coating film. Systems with an applicator
qualitatively different process from knife coating. roll generate a substantial nip pressure which
Blade coating of paper is a precision process causes coating dewatering, fiber swelling and web
where the blade is flexible, the substrate is porous roughening prior to metering. The blade acts as
and compressible, and the speeds are three to a metering device which also compresses the web,
four times faster than those prevailing in knife filling in its surface voids with coating and,
coating. This process comprises mass transfer therefore, creating a smooth contour of coating.
phenomena coupled with inertia and viscous In contrast, pond type and short-dwell blade
effects, flow through porous media, nip dynamics coaters induce little dewatering prior to metering,
of resilient rolls, and non-Newtonian fluid dynam- but fiber swelling and web roughening take place
ics of highly concentrated dispersions. The forces after metering. Consequently, the coating contour
at the blade nip and the interaction ofthe coating is more nonuniform than in the previous case,
material with the web are the principal mechanisms an effect which may lead to inferior surface
for coating deposition. Because of the high speed properties. Although pressurized pond coaters
involved, inertia effects at the blade nip may are superior coaters compared with roll appli-
become important, as do viscous and elastic cators, these coaters also have limitations in
effects. Furthermore, the quality and uniformity terms of hydrodynamic instabilities and air
of the coating layer also depends on coating entrainment in the pond. Both ofthese phenomena
dewatering prior to metering and the roughening may generate coat weight nonuniformities across
ofthe web at the application zone due to capillary the machine line direction, manifested as bands
penetration and fiber swelling. Therefore, the or streaks when the web is viewed with transmitted
strong interaction between coating rheology, light. These nonuniformities impose limits in the
paper web and process variables make prediction processability of high-speed blade coaters for
and control of the process difficult to sustain. paper, with respect to both the coater speed and
In addition, there are physical differences shear viscosity.
between the operational modes of the blade (i.e., Non-Newtonian fluid dynamics introduce an
bevelled versus bent) and coating feeding systems additional complexity in the physics of high-
into the metering zone. Bevelled blades are speed metering of paper coatings. Coating
preferred for high speeds and relatively low formulations, comprising highly concentrated
References 669
dispersions of anisometric mineral particles and and surface scratches on the coated web. It is
including polymeric additives, demonstrate a only with thorough, systematic studies based on
whole range of non-Newtonian effects under the mathematical analyses and experimentation that
large strains occurring during the coating process. understanding of the process can yield useful
Interparticle interactions and network structure results to predict and control the appearance of
due to the presence of polymers are responsible these and other coating defects.
for plastic yield and viscoelastic behavior at low The high-speed blade metering of paper is an
strain rates. At high shear, most coatings are area rich with opportunity for advancement.
shear thinning and exhibit thixotropic recovery. Conceptual models are needed to explain the
When the concentration of pigment particles interaction of coatings with the paper web to
increases, coatings become dilatant and impose better describe the dewatering process and the
strong viscoelastic effects at high rates of shear. local nonuniformity in coating coverage. The
However, most coatings do not demonstrate physical phenomena occurring at the blade nip
viscoelasticity at high shear rates. Particle are still not well understood as there are only a
orientation and packing under the high shear few experiments that have been executed under
flow conditions prevailing close and inside the well controlled conditions. Furthermore, there
blade nip determine the rheological behavior of needs to be a more systematic approach in
paper coatings. Orientation of platey pigment establishing relationships between rheological
particles parallel to the machine line direction and dewatering parameters of coatings and their
reduces the effective shear viscosity in the blade processability at high-speed coating operations.
nip. However, slipping phenomena at the blade These relationships should also include the
tip surface may lead to coating defects as film influence of the various coating components.
fracture occurs at the exit from the blade nip. In Finally, the mechanisms for hydrodynamic
addition to rheology, the dewatering properties instabilities and air entrainment need to be
of paper coatings influence their processability. investigated in depth in order to assess their
Poor water retention is detrimental to coater contribution to process limitations and to the
runnability, as water leaving the coating layer appearance ofcoating defects. However, successful
prior to metering increase the effective solids contributions to part or all of these areas would
content of the filtercake reaching the blade. require the combination of theoretical analyses
The physics ofhigh-speed flexible blade metering with carefully designed experiments.
remains an active field for research. Practical
experience has played a significant role in
establishing many of the relationships existing REFERENCES
today between application variables and final
Adolfsson, M., Engstrom, G. and Rigdahl, M., 1989.
film thickness and properties. Semi-empirical The effect of the relaxation time of coating colours
relationships, however, are not sufficient to explain on the elimination ofblade streaks. Coating Conference
the complex physics involved, as these relationships Proceedings, pp. 55-58. Atlanta: TAPPI Press.
often do not find universal application to all Aidun, C. K. 1987. Stability of convection rolls in
different types of coating formulations and paper porous media, bifurcation phenomena in thermal
processes and convection. AMD, 89, eds H. H. Bau,
webs. For instance, many relationships connecting L. A. Bertram and S. A. Korpela.
coating rheology and web properties with the Aidun, C. K. 1989. Fundamentals of Coating Systems.
appearance of visual coating defects and film Technical Report to the Project Advisory Committee,
nonuniformities are not always accurate. A good Institute of Paper Science and Technology, Atlanta,
example of this is dilatant coatings which have GA.
Aidun, C. K. 1991a. Principles of hydrodynamic
often been associated with the buildup of blade instability: applications to coating systems - Part
deposits (stalagmites). However, many shear I: Background. Tappi J. 74(2): 213.
thinning coatings can also lead to stalagmites Aidun, C. K. 1991b. Principles of hydrodynamic
instability: applications to coating systems - Part technology. TAPPI Annual Meeting Proceedings.
II: Examples of coating flow instability. Tappi J. pp. 57-62. Atlanta: TAPPI Press.
74(3): 213. Deryagin, B. M. and Levi, S. M. 1964. Film Coating
Aidun, C. K. 1991c. Principles of hydrodynamic Theory. London: Focal Press.
instability: applications to coating systems - Part Doremus, P. and Piau, 1.-M. 1981. Viscoelastic
III: Generalization of coating flows as dynamical elongational lubrication. J. Non-Newtonian Fluid
systems. Tappi J. 74(4): 209. Mech. 9: 389-400.
Aidun, C. K., Triantafillopoulos, N. G. and Benson, Doremus, P. and Piau, 1.-M. 1983. Experimental study
1. 1991. Global stability of a lid-driven cavity with of viscoelastic effects in a cylinder-plane lubricated
throughflow: flow visualization studies. Phys. Fluid contact. J. Non-Newtonian Fluid Mech. 13: 79-91.
A. 3(9): 2081. Eklund, D. E. 1981. Rheology and water retention of
Aidun, C. K., Veverka, P. 1. and Scriven, L. E. 1992. high-solids coating colors. Coating Conference
Onset of air entrainment: effects of surface charac- Proceedings. Atlanta: TAPPI Press. pp. 49-57.
teristics. Paper presented at the International Eklund, D. E. 1984. Influence of blade geometry and
Symposium on Thin-Film Coating, Spring National blade pressure on the appearance of a coated
AIChE Meeting, March 30-April 1, paper no. 55. surface. Coating Conference Proceedings. pp. 37-43.
Benson, J. and Aidun, C. K. 1992. Transition to Atlanta: TAPPI Press. Tappi J. 67(5): 66-70.
unsteady nonperiodic flow in a lid-driven cavity via Eklund, D. and Fors, S. 1988. Observation on stalagmite
torus bifurcation. Phys. Fluid A. 4(10): 2316. formation and blade bleeding. Wochbl. Papierfabrik.
Blake, T. D. and Ruschak, K. J. 1979. A maximum 116(10): 400-404.
speed of wetting. Nature. 282: 489. Engstrom, G. 1984. Coating colors with high solids
Bliesner, W. C. 1971. Basic mechanisms in blade content: runnability on the blade coater. Wochbl.
coating. Tappi J. 54(10): 1673-1679. Papierfabrik. 112(6): 184-187.
Bohmer, E. 1969. The relationship between coating Engstrom, G. and Rigdahl, M. 1987. The implication
weight and other variables during blade coating. of viscoelasticity on coating rheology and structure.
Norsk Skogindustri. 23: 308-324. Tappi J. 70(5): 91-94.
Bolton, B. and Middleman, S. 1976. Air entrainment Engstrom, G. and Rigdahl, M. 1989. Stalagmite
in a roll coating system. Chem. Eng. Sci. 35: 597. formation during blade coating - effects of process
Booth, G. L. 1970. Coating Equipment and Processes. conditions. Coating Conference Proceedings. pp.
New York: Lockwood Publishing. 59-63. Atlanta: TAPPI Press.
Bousfield, D. W. 1990. The simulation of pigment Follette, W. 1. and Fowells, R. W. 1960. Operating
motion during blade coating. Coating Conference variables of a blade coater. Tappi J. 43(11): 953--957.
Proceedings. pp. 325-334. Atlanta: TAPPI Press. Gane, P. A. C. and Coggon, L. 1987. Coating blade
Bozeman, J. D. and Dalton, C. 1973. Numerical study geometry: its effect on coating color dynamics and
of viscous flow in a cavity. J. Compo Physics. 12(348). coated sheet properties. Coating Conference Pro-
Brady, 1. F. and Bossis, G. 1988. Stokesian dynamics. ceedings, pp. 7-18. Atlanta: TAPPI Press. Tappi J.
Annual Rev. Fluid Mech. 20: 111-157. 70(12): 87-96.
Branston, R. E. et al. 1993. Weeping in blade coating. Gane, P. A. c., Watters, P. and McGenity. 1992.
Part I: High-speed video and still macro photography. Factors influencing the runnability ofcoating colours
Coating Conference Proceedings. pp. 79-86. Atlanta: at high speeds. Coating Conference Proceedings. pp.
TAPPI Press. 117-132. Atlanta: TAPPI Press.
Burley, R. and Jolly, R. P. S. 1984. Entrainment of air Gartaganis, P. A., Cleland, A. 1. and Wairegi, T. 1977.
into liquids by a high speed continuous solid surface. Blade mechanics of extended blade coaters: theory
Chem. Eng. Sci. 39: 1357. and practice. Coating Conference Proceedings. pp.
Burley, R. and Kennedy, B. S. 1976. An experimental 145-155. Atlanta: TAPPI Press. Tappi J. 61 (4): 77-81.
study of air entrainment at a solidjliquid/gas interface. Ginn, R. F. 1974. A Fundamental Rheological Study
Chem. Eng. Sci. 31: 901. of the Blade Coating Process. Project 3069. Progress
Carreau, P. 1. and Lavoie, P. A. 1993. Rheology of Report Four, The Institute of Paper Chemistry,
coating colors: A rheologist point of view. 2nd Appleton, WI.
Coating Fundamentals Symposium Proceedings. pp. Ginn, R. F. 1977. Fundamentals of flow. In Physical
1-18. Atlanta: TAPPI Press. Chemistry of Pigments in Paper Coating. Ed. C. L.
Chen, K. S. A. and Scriven, L. E. 1989. On the physics Garey. pp. 233-253. Atlanta: TAPPI Press.
of liquid penetration into a deformable porous Goodwin, 1. 1989. Coating Rheology. Technical
substrate. Coating Conference Proceedings. pp. Information Brochure. Bohlin Rheology.
93-106. Atlanta: TAPPI Press. Grankvist, T. O. and Sandas, S. E.1990. Measurement
Ciosset, G. P. 1986. Recent advances in coating of viscoelasticity at low and high shear rates.
References 671
Coating Conference Proceedings. pp. 473-478. Atlanta: Korpela, M., Palsanen, 1. and Pitkanen, S. 1986.
TAPPI Press. Practical experience with the short dwell coater.
Greener, Y. and Middleman, S. 1974. Blade-coating Wochbl. Papierfabrik. 114(8): 267-271.
of a viscoelastic fluid. Pol. Eng. Sci. 14(11): 791-796. Koseff, J. R. and Street, R. L. 1984a. Visualization
Guzy, C. 1. and Higgins, B. G. 1982. Viscous pressure studies of a shear driven three-dimensional recir-
across the nip of a blade coater and its effect on culating flow. J. Fluid Eng. 106: 21.
the final coated film thickness. Coating Conference Koseff, 1. R. and Street, R. L. 1984b. On end wall
Proceedings. pp. 63-71. Atlanta: TAPPI Press. effects in a lid-driven cavity flow. J. Fluid Eng. 106: 385.
Hassell, M. V. 1981. A study of coater blade wear. Paper Koseff, 1. R. and Street, R. L. 1984c. The lid-driven
read at the 2nd Latin-American Congress on Pulp cavity flow: a synthesis ofqualitative and quantitative
and Paper, June 22- 26, Torremolinos, Malaga, Spain. observations. J. Fluid Eng. 106: 390.
Hayward, G. 1973. Factors affecting the metering Kurath, S. 1965. A Fundamental Study of Coating of
characteristics of a blade coater. Coating Conference Paper and Paperboard: The Rheological Behavior
Proceedings. pp. 61-71. Atlanta: TAPPI Press. of Coating Colors. Project 2495. Progress Report
Hofmann, H. P. and von Raven, A. 1986. Production Two, The Institute ofPaper Chemistry, Appleton, WI.
problems from stalagmite-like buildups of coating Kurath, S. A. 1974. Viscosity of coating colors at high
colors on the blades of high-speed coaters. Wochbl. shear rates. Technical biformation Sheet 408- J. Atlanta:
Papierfabrik. 114(8): 261-266. TAPPI Press.
Holton, 1. E. and Klass, C. P. 1986. Simultaneous Kuzmak, J. M. 1986a. Bevelled-blade coating. Tappi
two-sided on-machine coating with billblade blade J. 69(2): 72-75.
coating. TAPPI Seminar Notes. pp. 115-124. Atlanta: Kuzmak,1. M. 1986b. Bevelled-blade coating: factors
TAPPI Press. affecting coat weight. Blade Coating Seminar Notes.
Hsu, T. 1986. The role of fluid viscoelasticity in the pp. 72-75. Atlanta: TAPPI Press.
dynamics ofbounded coating flows. PhD dissertation, Laun, H. M. and Hirsch, G. L. 1990. High shear
University of Massachusetts. rheometry of paper coatings including normal stress
Hsu, T. c., Malone, M. and Laurence, R. L. 1985. measurements. Coating Conference Proceedings. pp.
Separating forces in blade coating of viscous and 445-452. Atlanta: TAPPI Press.
viscoelastic liquids. J. Non-Newtonian Fluid Mech. Letzelter, P. and Eklund, D. 1993. Coating colour
18: 273-294. dewatering in blade coaters. Coating Conference
Huang, D. K. 1984. Blade streak and scratch problem. Proceedings. pp. 247-257. Atlanta: TAPPI Press.
Blade Coating Seminar. pp. 59-70. Atlanta: TAPPI Li, A. and Burns, 1. 1992. Effects of air entrainment
Press. on coat weight distribution with an enclosed pond
Huang, D. K. 1986. Blade streak and scratch problem. application. Proc. Tappi Coating Con! pp.173-184.
Blade Coating Seminar. pp. 175-186. Atlanta: TAPPI Atlanta: TAPPI Press.
Press. Luciani, M. and Galloni, M. 1978. Base paper and
Hwang, S. S. 1979. Hydrodynamic analyses of blade coating colour in the trailing blade process: depend-
coaters. Chern. Eng. Sci. 34: 181-189. ence of coating weight on raw stock smoothness.
Kahila, S. J. and Eklund, D. E. 1978a. Processes Paper. 190(4): 183-189.
occurring under the blade during blade coating: Lyons, A. V. 1993. A mathematical model of coat
(1) Theory. Wochbl. Papierfabrik. 106(17): weight control on the CLC. Presented at 2nd CLC
661-665. Users' Forum, May 2, Minneapolis, MN.
Kahila, S. J., Eklund, D. E. and Obetko, D. 1978b. Middleman, S. 1977. Blade coating. Fundamentals of
Processes occurring under the blade during blade Polymer Processing. Ch. 8, pp. 199-211. New York:
coating: (2) Practical tests. Wochbl. Papierfabrik. McGraw-Hill.
106(18): 709-714. Middleman, S. 1981. Elastic phenomena. Fundamentals
Kahila, S. 1. and Eklund, D. E. 1978c. Factors of Polymer Processing. Ch. 4, pp. 463-480. New
influencing the coat weight in blade coating with York: McGraw Hill.
bevelled blades: theory and practice. Coating Miura, H. and Aidun, C. 1992. Pressure fluctuations
Cmiference Proceedings. pp. 13-25. Atlanta: TAPPI and coat-weight nonuniformities in blade coating.
Press. Coating Conference Proceedings. pp.193-216. Atlanta:
Kartovaara, A. 1991. Lateral force under the blade TAPPI Press.
tip of bevelled-blade coating. Coating Cmiference Miyamoto, K. 1991. On the mechanism of air
Proceedings. pp. 13-21. Atlanta: TAPPI Press. entrainment. J. Industrial Coating Research. 1:
Kennedy, B. S. and Burley, R. 1977. Dynamic fluid 71.
interface displacement and prediction of air entrain- Miyamoto, K. and Scriven, L. E. 1982. Breakdown of
ment. J. Colloid Interface Sci. 62(1): 48. air film entrained by liquid coated on web. Presented
at the AIChE Annual Meeting, Paper 106F, Los lines. Paper read at the International Symposium
Angeles, CA. on Thin Film Coating, AIChE Spring National
Modrak, J. P. 1973. Effect of coating color rheology Meeting, Orlando, FL.
on the blade coating process. Tappi J. 56(10): Strenger, M. R., Secor, R. B. and Sramek, R. J. 1992.
70-73. Knife coating of a viscoelastic material. Presented
Nallasamy, M. and Krishna Prasad, K. 1977. On the at the 6th International Coating Process Science
cavity flow at high Reynolds numbers. J. Fluid and Technology Symposium, AIChE Spring National
M echo 79: 391. Meeting, March 30-April 2, New Orleans, LA.
Pan, F. and Acrivos, A. 1967. Steady flows in rectangular Sullivan, T. M. 1986. An experimental and computa-
cavities. J. Fluid Mech. 28: 643. tional investigation of rheological effects in blade
Pranckh, F. R. 1989. Elastohydrodynamics of blade coating. Doctoral dissertation, Department of
coating. PhD dissertation, Department of Chemical Chemical Engineering, University of California at
Engineering and Material Sciences, University of San Diego.
Minnesota. Sullivan, T. M. and Middleman, S. 1979. Roll coating
Pranckh, F. R. and Scriven, L. E. 1988. The physics of in the presence of a fixed constraining boundary.
blade coating ofdeformable substrate. Coating Co1!fer- Chern. Eng. Commun. 3(469).
ence Proceedings. pp. 217-238. Atlanta: TAPPI Press. Sullivan, T., Middleman, S. and Keunings, R. 1987.
Pranckh, F. R. and Scriven, L. E. 1990. Elasto- Use ofa finite-element method to interpret rheological
hydrodynamics of blade coating. AIChE J. 36(4): effects in blade coating. AIChE J. 33(12): 2047 - 2056.
587-597. Triantafillopoulos, N. G. and Aidun, C. K. 1990.
Ramp, J. W. 1983. The angles oftip tube blade coating. Relationship between flow instability in short-dwell
Coating Conference Proceedings. pp. 83-95. Atlanta: ponds and cross-directional coat weight nonunifor-
TAPPI Press. mities. Tappi J. 73(6): 127.
Richardson, C. A. 1957. The trailing blade coater. Triantafillopoulos, N. G. and Altug, N. 1991. A
Tappi J. 40(8): 155A. comparative study offorces in blade coating. AIChE
Roper, 1., III. 1990. Viscoelastic measurements and Forest Products Symposium Proceedings. pp.
their significance. Presented at the Coating Advances 141-158. Atlanta: TAPPI Press.
Seminar, August 11-14, Western Michigan University, Triantafillopoulos, N. and Grankvist, T. 1992. Vis-
Kalamazoo, MI. coelasticity and high-shear rheology in relationship
Roper, 1., III and Attal, 1. F. 1993. Evaluations of to short dwell coater runnability. Coating Co1!ference
high-speed runnability using pilot coater data, Proceedings. pp. 23-36. Atlanta: TAPPI Press.
rheological measurements, and computer modeling. Trist, A. R. 1945. US Patents 2,368,176, 2,593,074 and
Coating Co1!ference Proceedings. pp. 107-114. Atlanta: 2,796,846.
TAPPI Press. Turai, L. L. 1971. Analysis of the blade coating process.
Saita, F. A. 1984. Elastohydrodynamics and flexible Tappi J. 54(8): 1315-1318.
blade coating. PhD dissertation. University of Veverka, P. 1. and Aidun, C. K. 1991. Flow visualization
Minnesota. ofair entrainment and dynamic contact line instability
Saita, F. A. and Scriven, L. E. 1985. Elastohyd- in low-speed roll coating. Proc. TAPPI Engineering
rodynamics and flexible blade coating. Coating Co1!ference. 2: 719; Tappi J. 74(9): 203.
Co1!ference Proceedings. pp. 13-21. Atlanta: TAPPI Vodnick, J. L. et al. 1993. Weeping in blade coating.
Press. Part II: SEM and X-ray microanalysis of microtome-
Salminen, P. J. and Fors, S. 1992. Fundamental sectioned weeps. Coating Co1!ference Proceedings.
approaches for optimizing fibre coverage in blade pp. 87 -96. Atlanta: TAPPI Press.
coating. Coating Conference Proceedings. pp. 7-22. Windle, W. and Beazley, K. M. 1967. The mechanics
Atlanta: TAPPI Press. of blade coating. Tappi J. 50(1): 1-7.
Sandlis, S. E. and Salminen, P. J. 1987. Water penetration Windle, W. and Beazley, K. M. 1968. The role of
in coated papers. Coating Conference Proceedings. viscoelasticity in blade coating. Tappi J. 51 (8):
pp. 97-101. Atlanta: TAPPI Press. 340-348.
Sandis, S. E. and Salminen, P. J. 1991. The influence Young, T. and Fu, E. 1990. Associative behavior of
of the interaction between pigment and various cellulosic thickeners and its implication on coating
cobinders on rheology, dewatering and coating structure and rheology. Coating Conference Pro-
performance. Coating Co1!ference Proceedings. pp. ceedings. pp. 395-404. Atlanta: TAPPI Press.
51-59. Atlanta: TAPPI Press. Young, T. and Fu, E. 1991. Rheological behavior of
Schurz, 1. 1984. The rheology of paper coatings. LWC coatings containing associative cellulosic
Wochbl. Papierfabrik. 112(8): 275-280. thickeners. Coating Conference Proceedings. pp.
Scriven, L. E. 1982. How does air entrain at wetting 61-70. Atlanta: TAPPI Press.
FREE-MENISCUS COATING PROCESSES 13
P. Randall Schunk, Alan J. Hurd and C. Jeffrey Brinker
13.1 INTRODUCTION
When faced with depositing a liquid film on a
surface, laboratory scientists usually rely on free
coating by withdrawal or drainage, or so-called
free-meniscus coating. Simplistic and inexpensive, (al
these techniques abound in the research and
development of materials that can be deposited (bl
--
in the liquid state. What has come to be known
as dip coating (Deryagin and Levi 1964; Ruschak
1976; Scriven 1988), viscous lifting (Van Rossum
1958), or drag-out (Landau and Levich 1942)
begins with immersing a substrate in a vessel (el
filled with liquid. Withdrawal of the substrate
from the liquid, if managed properly, can result
in a thin coherent liquid film, as shown in Fig. (dl
t3.1b. Alternatively, the liquid in the vessel can

II; C UU111Cl:HUU. \UJ \..,UilllUI$ uY
be drained around the substrate. This technique substrate withdrawal. (c) Coating-by-drainage. (d)
is often referred to as coating-by-drainage (Jeffreys Continuous dip coating.
1930; Van Rossum 1958; Groenveld 1971) and
is diagrammed in Fig. 13.1c. Recourse is sometimes
taken from dip coating to coating-by-drainage continuous dip coating, and coating-by-drainage
when limited vertical space prevents substrate are essentially the same, differing only in the
withdrawal (cf Ashley and Reed 1984) or when frame of reference in which the flat part of the
coating small, short substrates to avoid the local meniscus is stationary.
thickening obtained when the trailing edge is . Even though dip coating is one of the oldest
withdrawn (Schroeder 1969). Although most coating techniques, it is still used widely to form
often dip coating is operated as a batch process, thin films. Dip coating is one of only several
it can be made continuous (Fig. t3.1d) when the means of coating irregularly shaped (non-planar)
substrate is a long, flexible sheet or filament and rigid objects. The others include spin coating
(Deryagin and Levi 1964; Scriven 1988); however, (cf Chapter 14), which can be used when the
sustaining a continuous coating process requires substrate is a disk, plate or bowl, and spray
the addition of make-up liquid to the vessel. In coating (cf Prescher 1991), which can be used
any case, the physics of batch dip coating, on objects of simple enough shape that adequate
London. ISBN 0 412 06481 2.
674 Free-meniscus coating processes
coverage be line-of-sight can be achieved. Thin used to predict the thickness and the stability of
films with remarkable uniformity (20 nm with the layer left behind. One example is the draining
thickness variations of less than 10%) can be films on the walls of nuclear reactors for which
achieved with spin coating over as much as proper cooling depends on the evolution of the
100 cm 2 • Spray coating is often adequate for film thickness (Bankoff 1983); another example
thick films but usually falls short of dip and spin is the amount of liquid left behind for accurate
coating in uniformity. measurement of liquid volume in burettes and
Free-meniscus coating techniques are encoun- pipettes, or in the accurate determination of
tered in many manufacturing applications. Before viscosity in capillary viscometers (Deryagin and
the discovery of pre-metered coating techniques Levi 1964).
dip coating was the most common way of In applications of free-meniscus coating the
producing photographic films (Deryagin and nature of the substrate can vary widely. For
Levi 1964). For decades, primers for protective instance, in the manufacture of membranes and
or cosmetic coatings were applied to automobile sensors the porous substrate is invaded by the
bodies by dipping and withdrawal (Rosenberg liquid during the process. In the manufacture of
and Beiter 1989). Wax replicas are dip coated multilayer laser reflectors, substrates are rigid
with a ceramic slurry to form mold housings in and flat but are often as large as 0.2 m 2 (Britten
the investment casting industry (Horton 1990). and Thomas 1992). In the automobile and casting
Solid reflective coatings on plate glass windows industries the substrate shape is highly irregular
are currently applied with dip coating followed and usually complex; clearly in these cases design
by heat treatment (Dislich 1988). Electron-beam considerations favor adequate coverage over film
resist is often applied to nonplanar objects with uniformity. Cylindrical shapes such as wire are
dip coating as a part of nonplanar lithography also common. In this case a high degree of
processes (e.g., Jacobsen et ai. 1991). The term uniformity can be managed but the range in film
'hot-dip galvanizing' implies dip coating iron or thickness is limited by the additional capillary
steel parts in molten zinc to protect against pressure due to the circumferential curvature
corrosion (Tuck 1983; Bablik 1950). In the recent when the radius is small. The substrate can be
surge of research on advanced materials, dipping long and flexible, as in most premetered coating
and withdrawal are not only used to apply the applications (Chapter 11). In most present-day
coatings, but also to control the final microstruc- applications for which film quality is crucial the
ture (e.g., Brinker and Scherer 1990). substrates are usually flat, rigid and short.
Flows that share many essential features with The physics underlying free-meniscus coating
free-meniscus coating can also be found in a great processes have seen heightened attention in the
variety of situations in technology. Dip coating last several years, stemming mainly from recent
is the essential part of a process known as interest in coatings of materials whose micro-
Marangoni drying, which is used to cleanse structure can 'be controlled with the process
microelectronics substrates of small particulates conditions. Tailored coatings of advanced ma-
(Leenaars, Huethorst and van Oeke11990; O'Brien terials for low-volume applications are the most
and Van Den Brule 1991). Arrangements that noteworthy; in particular, coatings with controlled
resemble dip coating are also encountered as part porosity (Brinker and Scherer 1990) and Lan-
of many high-speed self-metered coating tech- gmuir-Blodgett monolayers and bilayers
niques, such as 'roll-pan pick up' whereby a film (Tredgold 1987) are discussed in this chapter. We
is lifted out of a reservoir by an applicator roll emphasize the former, paying particular attention
and then split or stripped (Chapter 12). Together to sol-gels, materials for which a combined
with these situations, many applications beg for dip-coating flow and drying environment can be
the proper principles underlying coating-by- managed to form films for a wide variety of
drainage, especially when these principles can be applications. The dip-coating technique is key
Pre-metered versus self-metered coating flows 675
to the versatility in these applications because

the fluid mechanics and drying occur simulta- Film-EII/rainmenl Region
t
neously, unlike in most high-speed techniques. Slagnolion Line
Section 13.2 below puts free-meniscus coating
Sialic-Meniscus Region
into perspective by relating it to slot coating
(Chapter 11), its closest pre-metered relative.
Applications with varying effects of inertial and
capillary forces are categorized and described in
Section 13.3 together with the available theories
and experimental findings in the literature. The Figure 13.2 Dip coating.
highlights here are the several extensions of the
landmark analysis by Landau and Levich (1942).
These include both successful and unsuccessful two at the free surface (see Fig. 13.2). Between
attempts at theories for high speeds and low the two parts of the flow is a dividing streamline,
surface tension, as well as analyses ofevaporation hereafter referred to as the stagnation line. The
effects, with and without a condensed phase. liquid next to the substrate ends up in the final
Microstructure control with coating conditions film, whereas the liquid on the other side of the
is the subject of Section 13.4. There we introduce stagnation line is returned to the bath by gravity.
two applications which drive our research: In effect, this line is analogous to a metering
Langmuir- Blodgett film deposition and sol-gel element (blade, knife, roll) as discussed in Chapter
processing. Our discussion emphasizes sol-gel 12a. Its location can be linked to the thickness
processing, a subject more familiar to the authors. of the deposited film and is governed by several
The focus of Section 13.5 is the special methods factors (see Scriven 1988).
needed to analyze coating and drying. Drying of Clearly operating conditions such as coating
the multicomponent solvents peculiar to sol-gel speed and angle of substrate withdrawal affect
materials can drastically change the interfacial the meniscus shape and hence the position of the
physics of a film; components with different stagnation line. In addition, however, surface
volatilities and different surface tensions can lead tension lowers the pressure in the liquid underneath
to surface tension gradients that alter the flow. the concave meniscus, viscosity thickens the
These phenomena complicate conventional analy- boundary layer along the substrate, and inertia
sis of coating flows. In this section we generalize in the liquid boundary layer arriving at the
classical coating theories (cf Kistler and Scriven deposition region can potentially induce a
1983) to include two-phase flow with evaporation. significant hydrodynamic pressure. So the position
In Section 13.6 we consider a case study in sol-gel of the stagnation line and hence the coating
coating, to which we apply the complete two-phase thickness depend not only on the operating
flow theory. The results are validated against conditions, but also on the physical properties
measurements of concentration along the film of the liquid (Ruschak 1985; Scriven 1988).
and reveal potential effects of the drying envi- Unfortunately, coating at higher speeds leads to
ronment on the final film microstructure. thicker films, as does withdrawal at any angle
other than vertical (Tallmadge 1971). Hence, if
dip coating is the method ofchoice for a particular
application then coating speed may be a limiting
13.2 PRE-METERED VERSUS
factor to achieve thin films (e.g., for high
SELF-METERED COATING FLOWS
throughput). If the substrate is flexible and thin,
In dip coating the substrate is normally withdrawn i.e., capable of being wound into a roll for
vertically from the coating bath, entraining the compact storage, then dip coating is usually not
liquid in a viscous boundary layer that splits in a practical coating technique; in this case the
pre- and self-metered techniques covered in that begins to look like dip coating. In fact a
Chapter 11 are the methods of choice. slot coating arrangement can be operated in the
Of all the pre-metered coating techniques, slot usual slot coating mode or in the continuous dip
coating is the closest relative to dip coating, coating mode. The latter can be achieved from
especially when the coating web is drawn upward a slot coating arrangement simply by widening
against gravity (cf Figs 13.2 and 13.3). In slot the separation between the die lips and the
coating, a slot-fed die delivers liquid to a moving substrate (coating gap) and sealing the upstream
substrate at a prescribed volumetric flow rate lip against the substrate, as shown in Fig. 13.3
V*. The resulting film thickness is simply V*/WV, (Sartor and Scriven 1990). As the gap grows
where V is the substrate speed and W is the wider the film thickness becomes less sensitive
width of the coated layer. So simple adjustments to the inlet flow rate. Although this multimode
of the flow rate can be used to determine the capability may be impractical, it points out that
coating thickness, although that flow rate can a pre-metered coating apparatus can be managed
only be varied between limits, as is indicated in to coat in the free-meniscus mode, and that there
the lower part of Fig. 13.3. If the flow rate is too must be some parametric relationship between
high, the bead 'floods' if the lifting force imparted the steady operating states of both modes.
by the moving substrate is incapable of taking That such a relationship exists was demon-
away liquid at the same rate. If the flow rate strated by Sartor and Scriven (1990) and earlier
ceases, or becomes too slow, the coating bead by Ruschak (1976). The explanation hinges on
(liquid bridging the coating gap) starves and the the competition between several forces, the major
coating fails. The only way to regain a continuous players being capillary forces (surface tension),
coating is to lower the coating speed, a situation viscous forces, inertial forces and gravitational
forces. The competition between viscous forces
and surface tension is conveniently measured
with the capillary number, Ca == p.V/(J (Chapter
1). Here p. is the viscosity of the liquid, and (J is
the surface tension of the same. The competition
t
between inertial forces and viscous forces is
measured with the Reynolds number, Re == V Lp/p..
Here L is a characteristic length (usually taken
as the coating gap), and p is the density of the
liquid. The competition between gravitational
forces and viscous forces is measured by the
DIP COAnNG
WITH OBSTRucnON
Stokes number St == gL 2 p/p.V, and the compe-
tition between gravitational forces and surface
VI
SLOT tension forces is measured by the Bond number,
COATING
VI
W
Z
Bo == gUp/(J.
>0:
o Table 13.1 shows the values of these dimen-
~ DIPCOAnNG sionless numbers for slot coating of a glycer-
~
.:!
Ll.
ine-water solution (Sartor 1990) and for dip
coating of a sol-gel solution (Hurd and Brinker
1988), both under arbitrary but representative
eo «1 BO= 1 Bo> 1 operating conditions. It is noteworthy that gravity
is far less important than surface tension in slot
Figure 13.3 Slot-fed die taken from a pre-metered coating, as the Bond number is far less than
slot coating arrangement to dip coating flow (printed unity. A small capillary number also indicates
with permission from Sartor and Scriven (1990)). that surface tension dominates viscous forces. In
Pre-metered versus self-metered coating ft.ows 677
Table 13.1 Parameters for slot coating and dip coating coating configuration adjust in response to any
factor influencing AP, and adjust through the
Parameter Dip coating Slot coating film thickness H so that equation (13.1) is satisfied.
These factors include the flow rate and any
Length scale - 2 mm (capillary Coating gap
length) -O.5mm
applied pressure differential across the bead (cf
vacuum stabilization in Chapters 11a and 11 b).
U 0.5cm/s 13.3 cm/s Finite element calculations by Ruschak (1982)
Jl 0.001 Pas 0.029 Pas show that equation (13.1) fails at about Ca = 0.01.
h 10 11m 82 11m In fact, Ruschak (1976) showed that equation
(J 50mN/m 61 mN/m (13.1) is inapplicable if Bo Ca l/3 « 1.
Re 10 2.75 Gravity usually plays an important role in dip
Ca 10- 3 0.0632
St 784 0.762
coating. Correspondingly, the Bond number is
Bo 0.784 0.048 of order one or greater. The free surface sags and
is far from circular. Although this shape determines
the final film thickness uniquely, as was argued
above that at low Ca, the curvature cannot vary
dip coating, viscous forces can be important, in response to a pressure change within the
although in practice they are usually far less reservoir. In fact the surface shape is largely
consequential than surface tension, at least under determined by the hydrostatic pressure field, at
these operating conditions. Gravity also dominates least at small capillary numbers. Effectively this
viscous forces (St» 1) and has the same order means that the film thickness cannot be controlled
effect as surface tension (Bo '" 1). with forced flow into the reservoir, as it is in slot
One aspect that slot coating and dip coating coating. In fact, in most dip coating situations the
can have in common is the capillary number. At substrate is withdrawn from a quiescent bath
low capillary numbers the gross shape of the with no inflow at all. Sartor and Scriven (1990)
meniscus is controlled by surface tension and the studied the sensitivity of the final film thickness
hydrostatic pressure field. Landau and Levich along the parametric path from slot coating to
(1942) showed that the pressure jump across this dip coating, i.e., from low Bo to high Bo. Their
meniscus is related to the film thickness H by results show that a wider range in film thickness
is possible at low Bond numbers, where the
AP = 1.34Ca 2/3 (J/H (13.1)
meniscus shape can be controlled by external
This result can be derived by equating the pressure or internal flow rate. They also point
curvature of the meniscus in the film-entrainment out that at high Bond numbers, as in dip coating,
region (cf Fig. 13.2) to that in the 'static' meniscus this lack of control can be advantageous because
near that region (cf Landau and Levich 1942; of lower sensitivity of the final thickness to
Ruschak 1985). This equation was derived for external disturbances, even though the process
dip coating on the premise that the static meniscus may be self-metered. Their results are summarized
is an arc of a circle, and so it is useful only when in Fig. 13.3.
inertial and gravitational effects are negligible. Kheshgi, Kistler and Scriven (1992) recently
This is often the case in slot coating because the explored the effects of bath configuration. In
menisci are so small (Bo « 1) that gravity does what they call dip coating with an obstruction
not affect their shape. Under a two-sided balance they found that moving the far wall of a coating
between capillary forces and viscous forces, the bath close to the curved part of the meniscus
whole concept of pre-metered coating can be leads to a decrease in film thickness, all else being
explained on purely geometric grounds by the same. Deryagin and Levi (1964) exploited
equation (13.1) (Ruschak 1976): the curvature of this effect to allow faster coating at a given film
the upstream and downstream menisci in a slot thickness.
13.3 DIP COATING AND DRYING the low capillary number theory with the data
of Morey (1940) and found good agreement at
13.3.1 CLASSICAL THEORIES AND RELEVANT Ca less than 0.01 (Ruschak 1976). Countless
EXTENSIONS applications of equation (13.2) have been made
in research on coatings of advanced materials,
The velocity profile ofthe entrained film, regardless
which are normally applied at low Ca (Brinker
of the physical parameters, can be described by
and Scherer 1990; Hurd and Brinker 1990; Park
the classic Nusselt solution for gravitational flow
1991).
of a liquid film. That solution, however, is
White and Tallmadge (1965) were among the
parameterized by the flow rate, which can be
first to attempt the extension of equation (13.2)
determined only by analyzing the region around
to higher capillary numbers. They did so by
the stagnation line. Theories applicable at low
accounting for the effect of gravity in the film-
capillary numbers, i.e. less than 0.01, in this
entrainment region. Unfortunately, like many of
region were for a long time the only ones
their successors (Esmail and Hummel 1975; Spiers,
available, the most noteworthy arising from the
Subbaraman and Wilkinson 1974), their extension
analysis of Landau and Levich (1942). Invoking
was to add gravity, but gravity alone does not
the lubrication concept (Chapter 8), they treated
extend the validity of the theory of Landau and
the rising film as nearly one-dimensional and the
Levich, as Ruschak (1985) pointed out in his
meniscus near the reservoir as static. Using what
review on coating flows. Moreover, whenever
amounts to the classical lubrication equations
gravity or other higher-order effects are included,
together and the differential equation of hydros-
Landau and Levich's procedure for matching
tatics they identified a matching condition for
curvature breaks down (see Ruschak (1985) for
the film-entrainment and the static-meniscus
discussion). These analyses have nevertheless led
regions. For vertical withdrawal they used that
to reasonable agreement with the experimental
condition to get an expression for the film
data reported by Gutfinger and Tallmadge (1965)
thickness in terms of the key parameters and
and Spiers, Subbaraman and Wilkinson (1974)
physical properties:
- see for example the analyses of White and
(pD)2 /3 Tallmadge (1965); Spiers, Subbaraman and
H= C _ _ (pg)1/2 (13.2)
1 (11/6
Wilkinson (1974); Tekic and Jovanovic (1982);
The constant c 1 has been determined to be about and Nigam and Esmail (1980).
0.944 (Newtonian liquids). At about the same Successful attempts to solve the problem at
time Deryagin and Levi (1964) derived equation higher capillary numbers are now commonplace.
(13.2) independently, and then extended the Lee and Tallmadge (1974) solved the complete
theory to simple non-Newtonian liquids. Ruschak two-dimensional equations with the finite-differ-
(1974,1976) and Wilson (1982) have since shown ence method and explored the effect of inertia
that equation (13.2) is asymptotically valid as and of bath dimensions. They found good
Ca---+O. Experiments accord with equation (13.2) agreement with experimental data at capillary
provided that Ca is less than 10- 2 (see for numbers greater than one. Wilson (1982) used
example Spiers, Subbaraman and Wilkinson the method of matched asymptotic expansions
1974). to obtain a closed form solution for the film
Many have used the theory of Landau and profile applicable at all capillary numbers. Kheshgi,
Levich. Groenveld (1971) applied it to expand Kistler and Scriven (1992) integrated a film
Jeffrey's (1930) classical drainage result to include profile equation, similar to the one derived by
the effects of surface tension. Ruschak (1976, Ruschak (1978) and Tekic and Jovanovic (1982),
1985) extended it to explain the relationship to obtain first-order approximations of film
between pre-metered and self-metered coating profiles for a variety of situations geometrically
flows at low capillary number. He also compared similar to dip coating.
Dip coating and drying 679
Several theories, including the one of White which are discussed in Section 13.5, suggests
and Tallmadge (1965), are graphed in Fig. 13.4. c = 0.74.
The ordinate represents the film thickness in Early on, Morey (1940) made film thickness
units of (p.u jpg)1/2. In several of these theories measurements at capillary numbers as high as
an upper limit to the film thickness that can be 0.1. Van Rossum (1949) extended Morey's results
withdrawn is evident. The reason for this can be to Ca of order 1 and found that the maximum
explained by the diminishing effects of surface flux obtainable corresponds to c = 0.54. Non-
tension. When the liquid viscosity and the web- invasive capacitance probes were used by Spiers,
speed are high enough to straighten out the Subbaraman and Wilkinson (1974) to measure
curved meniscus, then the resulting film thickness film thicknesss over the Ca range 0.003 to 18.
is determined by the balance between viscous They built on the work ofTallmadge and Gutfinger
drag and gravity. As the drag force is proportional (1967) who used micrometer probes. The data of
to }J.U and the gravity force to pgh (for vertical Spiers et al. and Gutfinger and Tallmadge (1965)
withdrawal) the film thickness should follow show an upper limit in the film thickness parameter
h = C(}J.Ujpg)1/2, hence the choice of units on film to be about 0.8 (see Fig. 13.4), just above what
thickness in Fig. 13.4. Tuck (1983) and Homsey we predict. Experiments with roll-pan pickup
and Geyling (1977) believe that the proportionality were performed by Tharmalingam and Wilkinson
constant c should be unity if the capillary number (1978), Nigam and Esmail (1980) and others, with
grows large but inertial effects remain negligible. capillary numbers as high as 50 being achieved.
Experiments have shown that c is typically Modern computer-aided techniques of solving
around 0.8 (e.g., Spiers, Subbaraman and Wil- the complete Navier-Stokes system have made
kinson 1974). Our analysis, the details of full theories of free-meniscus coating possible
0.8
~ 0.6
:::J
:i.
OJ
Q. L..,
0 0.4
.r:::.
0.2
Figure 13.4 Variation of dimensionless film thickness with capillary number. Comparison of previous analyses
with that by the present authors.
(Chapter 9), as Fig. 13.4 makes plain. Besides the we can expect that vertically rising films are
analysis of Lee and Tallmadge (1974), Marques, always unstable. So how can dip coating oper-
Costanza and Cerro (1978) also used a finite- ations, even at slow speeds, produce acceptable
difference technique to solve the high capillary films? It may be that disturbance growth rates
number limit. Ruschak (1982) applied the finite- are small enough that growing waves can escape
element technique to approximate the two- detection, at least until the film is cured or
dimensional flow between partially submerged, solidified. It may also be that rising films in effect
counter-rotating rollers. He used that solution can carry wavy disturbances upward and away
to help determine the inlet boundary condition from the bath, as suggested by Kheshgi, Kistler
on an outer expansion of the film-entrainment and Scriven (1992).
region. The situation was basically free-meniscus It is perhaps relevant that film thickness
coating but did not include gravity. Tanguy, measurements at capillary numbers greater than
Fortin and Choplin (1984) also applied the about unity have been successful only for higher
finite-element method. Even though they con- viscosity fluids, i.e., at viscosities at least two-
centrated on capillary numbers above 0.01, their orders of magnitude greater than that of water,
analysis was extended to dip coating wires and or around 0.1 to 1 Pa s (e.g., Soroka and Tallmadge
cylinders and also to non-Newtonian liquids. 1971; Spiers, Subbaraman and Wilkinson 1974).
Unfortunately their resolution was not sufficient The stability of the solutions at Ca > 1 reported
in the region of high meniscus curvature, the in this work and represented in Fig. 13.4 were
necessity of which was made clear by Sartor attainable only if the viscosity was increased to
(1990). Sartor applied modern free-surface par- at least 0.1 Pa s, in which case the Reynolds
ameterization techniques (Chapter 9) to solve the number remained less than about 100.
complete set of operating states between slot
coating and dip coating.
13.3.2 DIP COATING WITH CONCOMITANT
As capillary number rises and entrained films
EVAPORATION
become thicker, the base flow state becomes
more susceptible to disturbances that can grow The potential for evaporation to affect the dip
appreciably. Falling liquid films on vertical surfaces coating process was first recognized by Slattely
are unstable to all two-dimensional disturbances and Givens (1933) while attempting to explain
(see Chapter lId), so it is likely to expect that discrepancies between Jeffrey's (1930) theory and
rising films akin to those in dip coating are experiment. Interest in evaporation effects has
unstable as well. Several attempts at measuring reappeared in sol-gel dip coating because the
entrained film thicknesses at high capillary fluid mechanics of film deposition overlaps with
numbers (Ca > 10) have been hampered by the evaporation, a phenomena that is prevalent
appearance of transverse waves that are falling when coating liquids are volatile and the coating
with respect to the bath (see Soroka and Tallmadge speed is small, i.e., low capillary numbers. An
1971; White and Tallmadge 1965). Some re- important feature observed in this case is the
searchers have discussed the stability of rising existence of a stationary 'drying line' at which
film flows (e.g., Tuck 1983; Kheshgi, Kistler and the volatile mass becomes exhausted. In this area
Scriven 1992), but the subject has yet to receive the most recent extension of the low Ca theory
direct theoretical treatment, so far as the authors of Landau and Levich (1942) was made by Hurd
are aware. Fortunately, the voluminous literature and Brinker (1990) who explored the effects of
available on the stability of falling liquid films evaporation on the entrained film, focusing on
is highly relevant (e.g., Benjamin 1957; Chapter a region within several millimeters of the drying
lId), mainly because the same forces are at play line. They found that the evaporation rate diverges
and the shape of the velocity profile takes on the at the drying line, although the amount of
same parabolic functional form. From this analogy vaporized mass remained finite (see Fig. 13.6).
Dip coating and drying 681
This led to a parabolic film profile of the form Jommg pressure. A review of this work by
Wayner (1991) makes plain that the ther-
h(x) = CX 1/ 2 (13.3)
modynamics of evaporation are important to the
Here x is the distance from the drying line understanding of the wetting of thin films. Where
measured along the substrate; the constant c applicable we consider his ideas in our develop-
depends on the diffusion coefficient of the vapor. ment of a complete free-meniscus coating theory
For single component systems, equation (13.3) (Section 13.5).
accords with ellipsometric measurements of film
thickness made by Hurd and Brinker (1988).
13.3.3 DRYING IN THE PRESENCE OF A
Below we summarize the theory (Section 13.6.1)
NON-VOLATILE PHASE
and experimental program (Section 13.4.2) de-
veloped by Hurd and Brinker (1988, 1990) before Several one-dimensional analyses of film drying
extending the theory and analysis to multicom- that pertain to dip coating downstream of the
ponent systems. two-dimensional metering region are available.
Leenaars, Huethorst and van Oekel (1990) The intent in most of these is to determine the
exploited the overlap of drying and deposition effect of increased volume fraction of non-volatile
in a process they contrived to clean substrates species on the drying rate (cf Poehlein, Vanderhoff
in microelectronics applications. The key to the and Wit-Meyer 1975). An analysis by Cairncross,
process is to create surface tension gradients Francis and Scriven (1992) considers the effect
along the meniscus while dip coating. They did of concentration- and temperature-dependent
so through the gas phase by feeding organic diffusivity, which tends to drop precipitously as
vapor down the film. The organic vapor condenses the residual solvent is drawn off. The most
on the film and creates a surface tension gradient common phenomenon in these studies is the
directed toward the water-rich reservoir. The concentration of the condensed phase near the
shear forces created by this movement are evidently free surface, or so-called skin effect (Bornside,
enough to dislodge unwanted particles from the Macosko and Scriven 1989; Tu and Drake 1990).
substrate. In fact, Cairncross, Francis and Scriven model
Some effort has been made to understand the the effect of a gelling front which originates at
detailed physics of thin drying films whose shape the free surface in the drying of silica sol-gel
is dictated by surface tension (low Ca and low films. Not surprisingly the same 'skinning'
Bo). Parks and Wayner(l987) analyzed a vertically phenomenon has been predicted in spin coating
extended meniscus wetting a stationary substrate, of polymeric and colloidal systems (Chapter 14;
a situation that shares many features with dip Bornside 1989; Rehg and Higgins 1992).
coating. Even though the substrate is stationary, Taking these analyses to two or three dimensions
flow is induced by evaporation and surface enables the assessment of drying stresses in a
tension gradients. To predict which force drives film. One of the few attempts at understanding
the most flow they measured the film thickness drying-induced stresses in gelled films was made
profile and then applied lubrication-type equations by Emery and Egolf (1989). With the finite-
to determine the mass flux induced by each element method they predicted moisture dis-
factor. For binary systems they found that tributions and moisture stresses in boehmite gels
concentration-induced surface tension gradients by solving the corresponding thermal-mechanical
tend to be more important than thermally induced stress problem. They found that thinner films
ones and that the former draws liquid into the have less propensity to crack. Tam, Stolarski and
thin film region if the less volatile species has a Scriven (1992) applied the finite-element method
higher surface tension. Their theory encompassed in two dimensions to study viscoplastic stress
extremely thin films which are controlled by development in a drying film. Their analysis
liquid-solid-vapor interfacial forces, e.g., dis- focuses on the warping of flexible substrates and
on the interaction of stress development around trace of solvent to evaporate. The mass fractal
neighboring defects. Significant theoretical work dimension d of a growing polymer relates its
towards understanding the drying stress problem molecular mass M to its characteristic radius r,
in sol-gels has also been made by Scherer (1988). i.e., M '" rd. The mass fractal dimension is by
Cairncross and co-workers (e.g. Cairncross et al. definition less than the imbedding dimension of
1995) have recently extended this analysis to space (Brinker and Scherer 1990).
examine the shrinkage and residual stress devel-
opment in gelled films with a complex theory of
13.4.1 LANGMUIR-BLODGETT (L-B) FILMS
partially saturated, deformable porous media.
Several other analyses exist, but they deal with Because L- B films are composed of one or
other materials (e.g., Cooper 1977; Lewis, Morgan several coherent molecular monolayers, they
and Thomas 1983). represent the ultimate challenge ofmicrostructure
control (cf Roberts 1990). These organic films
are formed layer-by-Iayer via the transfer of
13.4 MICROSTRUCTURE CONTROL WITH DIP
amphiphiles, i.e., molecules having polar hy-
COATING
drophilic ends and nonpolar hydrophobic chains,
It has recently become apparent that the micro- from a liquid surface to a substrate through
structure of films coated from solutions of multi-pass immersion and withdrawal (Fig. 13.5).
polymeric precursors or of nonreactive but The subphase liquid is usually a polar solvent,
flocculating particles can be controlled with such as water, and the molecules are surfactants
operating conditions in dip coating (Brinker et that are more-or-Iess insoluble in the liquid. As
al. 199Ib). That dip coating is amenable to such Fig. 13.5 makes clear, the Langmuir-Blodgett
control is a direct consequence of the concurrent (L- B) technique in its simplest form is free-
fluid mechanics of deposition and drying of the meniscus coating.
film, an effect that is usually absent from most Since the development of this technique
high speed coating techniques. Because advanced (Langmuir 1938; Blodgett 1935), it has been used
materials are often coated at low capillary number as a tool of surface chemistry for the study of
(Ca less than about 0.01), the coating rate is surface interactions and wetting (Swalen 1992).
usually slow enough for the fluid mechanics of L- B films have consistently attracted interest
the deposition process and the film-drying process because of their applicability in microelectronics,
to interact. Thin ceramic coatings of sol-gel
materials and Langmuir-Blodgett (L- B) coatings
of surfactants fall into this category. ...............
...............
In the following discussion we use several
I
...............
terms which need to be defined. The condensed SURFACTANT MOLECULES
phase of a sol-gel refers to the inorganic or metal
organic precursors which polymerize into a
network; in a Langmuir-Blodgett formulation
that phase is made up of amphiphilic molecules,
((I)
or surfactants (see Chapter 4), that form a
monolayer or bilayer structure. A drying line is
a mathematical line at which the meniscus formed
by the solvent recedes into the network of
polymers in a sol-gel. If no condensed phase is Figure 13.5 (a) Single- and (b) double-pass Lang-
present, the drying line is simply the line at which muir-Blodgett film deposition. (c) Computed streamline
the solvent disappears. In a Langmuir- Blodgett pattern for an inextensible surface (from Schunk et al.
film the drying line defines the locale of the last (1992)).
Microstructure control with dip coating 683
integrated optics, microlithography, and in a new to shearing of the monolayer (Chapter ltd;
field known as molecular electronics. The most Gaines 1966; TschoeglI962). The consequences
comprehensive reviews of these applications are of film viscosity to L- B film formation have been
given by Tredgold (1987) and Roberts (1990). studied by Malcolm (1985) and Daniel and Hart
Research activity in the last several years has (1985). Extensibility of the film is also important
heightened because one can now polymerize to the success of the deposition process. Several
L-B films of various materials to provide the mechanical properties that measure extensibility
long-sought mechanical rigidity and thermal are known to affect the fluid mechanics of depo-
stability (Agarwal 1988). sition. The most common is surface elasticity, or
The L- B process begins with a monolayer of the sensitivity of surface tension to the change in
surface active molecules (amphiphiles) initially surface area per molecule (Chapter 4; Chapter lId).
spread at the interface between water and air. In fact, the effects of large surface elasticities can
The support liquid is contained in what has be simulated in free meniscus coating with a steep
become known as a Langmuir trough, which is surface tension gradient directed from the bath to
equipped with a movable barrier for monolayer the film. Figure 13.5c shows the consequence: both
compression. After the monolayer is compressd substrate and surface forces draw liquid upward.
against the substrate to a desired surface pressure Were it not for evaporation or drainage, the
(this quantity is the reduction of surface tension entrained film would be much thicker than in the
below that of clean water; see Adamson 1982), absence of surface tension gradients.
as meaured by a Wilhelmy plate or du Nouy Film profiles during L-B deposition have
ring (see Chapter 4), the substrate is withdrawn been measured recently by Butler (1993). They
slowly, depositing the delicate monolayer on the used an imaging ellipsometry technique, similar
substrate. Subsequent immersion (down stroke) to the one described in Section 13.4.2, to probe
can then be used to deposit a second layer of the film thickness in the dynamic contact region.
molecules whose hydrophobic tails are directed Unfortunately, a quantitative comparison between
toward those of the first layer. During deposition those measurements and the available theories
it is important to keep the surface pressure and of film deposition (e.g., Park 1991) has not been
therefore the packing of the molecules constant completed.
so the microstructure of the film is uniform on Still an outstanding issue in dip coating
the substrate. Control of solvent drying is also monolayers is that of tolerable microstructure
crucial: rapid drying can lead to film defects (Sugi control. Because there is usually considerable
1985). In fact it is common practice to maintain interaction between the monolayer and the
slow withdrawal so that the subphase can drain substrate, the microstructure of L- B films of
away (Tredgold 1987). Speeds of 4 mm/min are simple straight-chained molecules is rarely the
typical, with maximum speeds of about one to same as the crystalline structure of the same
60 mm/min attainable. Capillary numbers for materials. Moreover, the L- B technique allows
these processing conditions are obviously small one to produce films of a precisely determined
('" 10- 5 ) despite the substantial reduction in number of layers, which ordinary crystallization
surface tension induced by monolayer compres- methods do not. Short-range order in the plane
SIOn. ofthe film can usually be achieved, but long-range
Controlling the mechanical response of the order in this plane has only been produced over
surface to the stretching motion of the deposition distances of 1 mm or less. Much of the current
process is important to film quality (Tredgold art of microstructure control hinges on the
1987). Several mechanical properties of the physics of dynamic contact (e.g., Petrov, Kuhn
monolayer have been used to characterize this and Mobius 1979) and the synthesis ofcompatible
response. Some investigators have ventured to amphiphile/solvent/substrate combinations (Petty
define a surface viscosity to measure the resistance 1987) of which little is known.
13.4.2 SOL-GEL FILMS membranes, semiconducting anti-static coatings,
Sol-gel films have also enjoyed recent attention electro-optic and nonlinear optical films, super-
because they are suitable for a wide range of conducting films, strengthening layers and fer-
applications (Brinker and Scherer 1990). This roelectrics. In most of these applications the sol
versatility stems from the wide range of porosity is applied to the substrate with dip coating or a
and thermophysical properties (e.g., dielectric related technique, and process conditions are
constant, conductivity, etc.) that can be tailored such that the capillary number is usually less
into thin films simply by varying the processing than 10 - 3. Several sections below are devoted
conditions. By 'sol' is meant a colloidal dispersion to the details of sol-gel dip coating.
of particles in a liquid; by 'gel' is meant a giant
aggregate or molecule that extends throughout 13.4.2.1 Factors influencing microstructure in
the sol. Typically the particles are inorganic or sol-gel films
metalorganic precursors which participate in a
polymerization (gelation) process. Processing a The dependence of molecular growth rate on
sol ~sually includes this gelation stage; dip solvent composition, extent of polymerization,
coatmg processes are no exception. and kinematic factors in the flow field provides
The most commonly used organic precursors the coupling between the physical chemistry of
for sol-gel film formation are metal alkoxides gel formation and the physics of the flow. The
(Brinker and Scherer 1990). Normally the alkoxides physical mechanism by which the polymerization
p~ocess affects the fluid mechanics of deposition
are dissolved in alcohol and hydrodlyzed by the
~ddition of water. This step produces hydroxyl
hmges on two phenomena. The first is the
lIgands on the precursors. Condensation reactions capacity of polymers which are growing in
involving these ligands result in polymers com- molecular weight to alter the rheological response
posed of metal-oxygen-metal or metal-hy- of the solution in which they are dissolved. The
droxyl-metal bonds plus, in most cases, the second is the impedance effect of a branching
by-products water and alcohol. Condensation polymer on the diffusive transport of a solvent
reactions that are reserved in the presence of (Cairncross, Francis and Scriven 1992), and
excess water or alcohol promote bond breaking hence the rate of evaporation, which itself can
and reformation processes that, if extensive, induce flow (cf Parks and Wayner 1987). Although
permit complete restructuring of the growing these effects are important to the thickness of
polymer. Thus controlling the growth of the films that can be deposited, far more important
inorganic polymers is a means of controlling its to the microstructural evolution in the film is
fractal dimension. This in turn provides a means the reverse effect, i.e., the effect of the flow on
of controlling density, and hence porosity, of the final microstructure. This mechanism has several
gels. parts (see Fig. 13.6).
The production of so-called 'monolithic' or Dip coating sol-gel materials involves more
bulk ceramics make up a portion of the appli- than a competition between viscous, capillary
cations of sol-gels, but thin films have shown the and gravitational forces; the mechanism which
greatest commercial success due to the ability to controls final film thickness and microstructure
control both chemical composition and micro- is considerably more complex. Film thinning by
structure in either single or multilayer depositions gravitational draining is assisted by vigorous
(Dislich 1988). Today sol-gel film coatings are evaporation, which also acts to decrease the
bei?g studied intensively for such diverse appli- amount of liquid that can be withdrawn. To
catIOns as protective and optical coatings, make matters even more complicated, sols are
passivation and planarization layers sensors typically formulated with several solvents, each
high or low dielectric constant films, 'inorgani~ of which usually differs in volatility and surface
tension. Sometimes the relative volatilities are
FILM COLLAPSE ANDIOR

FILM THINNED
BY EVAPORATION
~r PORE FORMATION
GELATION
j
AGGREGATION
FILM THINNED
BY DRAINAGE
AND EVAPORATION
ENTRAINED DILUTE SOL
Figure 13.6 Schematic of the steady state dip-coating process, showing the sequential stages of structural
development that result from draining accompanied by solvent evaporation and continued condensation reactions.
different enough to lead to differential solvent being coated on a planar substrate are parabolic
evaporation and ultimately to concentration in shape (equation (13.3)), and that c/> should vary
variations along the film. An exception here as (l/h) - X- 1/2 in the thinning film, x being the
might occur if there were an azeotrope-like distance in the upstream direction from the
composition at which the drying can be carried drying line.
out (Dillon 1977). Differential evaporation triggers The rapid concentration of the entrained
several other events at and beneath the liquid- gas inorganic species by evaporation is more evident
interface. First it may lead to concentration from consideration ofthe mean particle (polymer)
variations along the interface; these variations separation distance <r), which varies as the
make surface tension gradients which contribute inverse cube root of C/>, or <r) - X 1/6 • This
to the surface stress and alter the flow. Second, precipitous function implies that halfthe distance
differential evaporation leads to diffusion of the between particle (polymer) neighbors is traveled
volatile species towards the surface and non- in the last 2% of the deposition process, which
volatile ones away from the surface. translates to a time scale of about 0.1 s. The
As the film thins, entrained particles or molecules evaporation-driven transport may be viewed like
of the condensed phase are forced together (Fig. that induced by centrifugation or electrophoresis:
13.6). Above the stagnation point (cf Fig. 13.2), the centrifugal acceleration needed to cause an
where all entrained inorganic species are incor- equivalent rate of crowding is as much as 106
porated in the final deposited film, a simple times the acceleration of gravity (this assumes
balance of the flux of non-volatile material over that there exists no steric barriers to concentration;
a horizontal slice leads to often aggregation/network formation will interrupt
C:)
this dramatic compaction process).
h(x)c/>(x) = constant + terms of order Ofcourse as the film becomes more concentrated
in the condensed phase the rheological response
(13.4) of the liquid film changes from Newtonian to
shear-thinning (aggregated systems) or thixotropic
where c/>(x) is the volume fraction of solids and (ordered systems) and then to viscoelastic.
== (J1D/pg)1 /2. We see that c/> varies inversely
..1. Eventually the gelation extends throughout the
with h if h « ..1.. Hurd and Brinker (1990) found film and the material no longer yields, i.e., the
that film profiles of a single evaporating solvent film behaves as an elastic solid. As the residual
solvent is drawn off, receding mensci in the reduced fractal dimensionality of transport-
micropores of the film create extreme capillary limited aggregates should result in more porous
pressures which put the liquid into tension and network. (2) For sols composed of repulsive
the network into compression. This can have a particles, there is little time available for the
significant effect on the pore structure and pore particles to order as they are concentrated in the
volume of the final film. (See Caimcross et al. 1996.) thinning film. (3) There is little time available for
From the preceding discussion we expect that condensation reactions to occur. Thus gelation
the final film structure should depend on the may actually occur by a physical process rather
competition between such phenomena as evap- than a chemical process - in some systems, this
oration and capillary pressure, which tend to is evident by the rapid resolubilization after
compact the structure, and condensation reactions immersion in the solvent (Brinker and Scherer
and aggregation processes which tend to stiffen 1990). (4) As the gels in thin films are weakly
the structure. The fluid mechanics of the deposition condensed compared with bulk gels, they are
process influence the rate of evaporation and more easily compacted; first by evaporation and
hence the solvent composition when the structure then by the capillary pressure exerted at the final
is forming, thus influencing the compressive stage of the deposition (drying) process. Further-
capillary pressure (surface tension) forces once more the maximum capillary pressure is likely
the network has been formed. The chemistry of to be higher due to the collapse of the network
the polymerization and aggregation processes is (and hence reduction in pore size) that precedes
also influenced by the deposition process, as is the final stage of drying. Of course, greater
described next. capillary pressure also promotes greater com-
The short time of the deposition process paction of films, which leads to smaller pore size,
(-to s) and, more importantly, the extremely larger index of refraction, and higher mechanical
short time in which the inorganic species are in rigidity.
close proximity (<< 1s) establish important time All things considered, it is surprising that the
scales for the dip coating process and distinguish condensed phase has relatively little effect on the
sol-gel thin film formation from bulk (monolithic) deposition process, at least to within several
gel formation. For bulk systems, the gelling sol millimeters of the drying line. Figure 13.7 shows
is normally maintained at a constant concentra- the product of refractive index and film thickness
tion; the gelation, aging and drying stages occur versus the logarithm of the withdrawal speed U
sequentially. Often the time required to prepare
a monolithic bulk gel is on the order of days to
weeks. During dip coating the drying stage 10000,--------------------,
completely overlaps the gelation and aging stages.
The deposited sol remains rather dilute until the xw
o
final stage of the deposition process at which ~
point it is rapidly concentrated. There are only in
(/)
w
fractions of seconds available for condensation Z 1000
~
u SLOPE
reactions to occur. I • STRONGLY BRANCHED (0.62)
We anticipate several consequences for the t:. o WEAKLY BRANCHED (0.64)
short time scale of the film deposition process.
('J
o...J o PARTICULATE 95 NM (0.53)
• PARTICULATE 72 NM (0.58)
(1) There is little time available for networks to
1001------------:-:10~---------'
'find' low energy configurations. Thus (for reactive 1
systems) the dominant aggregation process LOG WITHDRAWAL RATE (inlmin)
responsible for network formation may change Figure 13.7 Product ofthickness and refractive index
from reaction-limited near the reservoir surface (proportional to film mass/unit area) versus withdrawal
to transport limited near the drying line. The rate (from Brinker et al. (1991b)).
for films prepared from a variety of silicate sols an analyzing polarizer. In null ellipsometry, the
in which the precursor structures ranged from analyzer and polarizer are adjusted until the
weakly branched polymers, characterized by a reflected beam is distinguished. With the polarizer
mass fractal dimension, to highly condensed angles measured as extinction, it is possible to
particles (Brinker et al. 1991 a). The slopes range calculate both the index n and the thickness H
from 0.53 to 0.64, values near those expected of a film covering a substrate with a known index.
from equation (13.2), i.e., slope of 0.5. This In initial attempts to study films, Hurd and
correspondence between the thicknesses of the Brinker found that the usual methods of narrow-
deposited films and a theory developed for low beam ellipsometry were unsatisfactory because
Ca coating of pure fluids (Landau and Levich of fluctuations in thickness. Furthermore, it was
1942) suggests that the entrainment of the desirable to obtain data from many points on
inorganic species has little effect on the hy- the film at once rather than one point per dip
drodynamics of dip coating, at least in the early coating. Hence, the laser was expanded and
stages of deposition where the entrained sol is collimated to illuminate approximately 1cm 2 of
quite dilute. Thus some insight into sol-gel film the film, thereby creating an image that can be
deposition may be gained by closer examination analyzed for thickness variations.
of the details of dip coating of multicomponent A thickness profile of an ethanol film obtained
systems without the condensed phase reactions. by imaging ellipsometry is shown in Fig. 13.8
together with a thickness profile of a pro-
panol-water film and that ofa toluene-methanol
13.4.2.2 Film profiling with imaging ellipsometry
film. Noteworthy in all cases is the blunt shape
and fluorescence spectroscopy
near the drying line, which turns out to be
It is evident from the discussion above that parabolic in functional form, in accord with
microstructure is influenced heavily by evaporative equation (13.3). The double parabolic shape is a
film thinning. However, quantifying this effect result of successive evaporation of more- and
experimentally is difficult; the refractive index less-volatile components and of surface tension
varies with composition and, under typical gradients directed toward the drying line. The
operating conditions, films are usually less than 'pile-up' feature in the methanol-toluene case is
5 Jlm thick in the regime in which the evaporative- an extreme example. The detailed mechanisms
thinning is most rapid. Moreover, there is a need are discussed later.
to measure the rate of film thinning, as well as A related approach is to use various fluorescent
the local thickness. In order to address these molecules, ions or crystals, as photophysical or
challenges, Hurd and Brinker (1988) developed
an imaging ellipsometer that allows acquisition
of spatially resolved thickness and refractive
index data over the entire area of the film being 2.0
deposited. E
A beam oflight from a surface suffers different
:::l. 1.5 IPURE ETHANOQ
<Ii
en
amplitude and phase shifts depending on its w
z 1.0
polarization. In ellipsometry these amplitude :I<:
U
I
and phase shifts are measured and interpreted t--
::;; 0.5
by an appropriate model (Azzam and Bashara ...J
ii:
1977). A typical ellipsometer consists of a laser,
a polarizer to prepare a known polarization, the
DISTANCE FROM APPARANT DRYING LINE. em
reflecting sample, a quarter-wave compensator
plate that resolves the polarization parallel to Figure 13.8 Thickness profiles measured with imaging
the surface and perpendicular to the surface, and ellipsometry (from Hurd and Brinker (1988)).
photochemical probes. The probes are entrained surface. Thus far pyranine has been used to sense
in the depositing film and transported to the the changing solvent composition (and hence
drying line along with the dispersed inorganic surface tension) in the binary fluid mixture due
phase. Acquisition of emission spectra as a to preferential evaporation of ethanol. Figure
function of height along the reservoir surface 13.9 (left) shows clearly in a plan view that the
provides structural information related to the concentration varies along the film, especially in
surroundings of the probe on a length scale of the last 20% or so of the entrained film. In that
one to several nanometers. This approach has portion concentration varies from that of the
been used successfully to characterize gelation, alcohol-rich liquid to that of nearly pure water
aging and drying in bulk systems (Kaufman and at the drying line. On the right we show a profile
Avnir 1986; Pouxviel, Dunn and Zink 1989) and view of the concentration variation as predicted
thin films (Nishida et ai. 1994). It benefits from from the theory described in Section 13.5. These
the availability of literally thousands of potential predictions appear to be qualitatively similar. At
probes that are sensitive to changes in viscosity, this time we have not quantified the concentrations
solvent polarity, local geometry, rigidity, pressure, in the experiments.
etc. Film thickness and concentration profiling
A sample result from the fluorescence spec- provide a means of determining the film thinning
troscopy experiments is shown in Fig. 13.9. A rate and film composition at the drying line; both
mixture of ethanol and water with a low factors are known to affect final microstructure;
concentration of organic optical molecules is but what about measuring film microstructure
coated at 0.4 cm/s. Using an argon laser excitation itself? We characterize the film microstructure
source with spatial resolution of about 100 Ilm, (surface area, percentage porosity, and pore size
fluorescence spectra are acquired from the probe distribution) in situ by analyzing N 2 adsorp-
molecules at various heights above the reservoir tion-desorption isotherms acquired using a surface
p.658
.0.576
ETHANOL LEAN/ p.493
WATER RICH
lOA I I
WATER LEAN/ ~).]2()
ETHANOL RICH
10.247
RESERVOIR
0.164
10.082
Figure 13.9 Plan view of concentration variations measured with fluorescence spectroscopy (left). Mass
fraction of ethanol predicted from theory (right, horizontal axis magnified 100 x).
Theory and analysis of free-meniscus coating and drying 689
acoustic wave (SAW) technique developed by our initial approach, but we found no available
Frye et al. (1988). The effects of processing theory for mass transfer in the gas phase to be
conditions on film microstructure determined adequate (Schunk, Hurd and Brinker 1992). We
with the SAW devices are reported elsewhere proceeded to analyze the coupled physics in both
(Brinker et at. 1991 b). There it is shown a striking phases, employing the proper interphase boundary
variation in film porosity achieved with silica conditions. This approach, although more costly,
based sols is created with simple adjustments to was far more successful at realistically predicting
the pH and pre-deposition aging. the sol-gel dip coating and drying process.
The development below pertains to the geometry
pictured in Fig. 13. lOa. How that geometry
13.5 THEORY AND ANALYSIS OF
relates to the pilot coating setup with which we
FREE-MENISCUS COATING AND DRYING
performed the benchmarking experiments is shown
The purpose of this section is to augment the in Fig. 13.10b. The diffusivities, viscosities and
theory developed in Chapter 2 to account for densities of both phases were taken as constant
coating of multicomponent solvents, differential and the process was considered to be isothermal.
evaporation and surface tension gradients, thereby The first assumption is appropriate because we
enabling the realistic analysis of sol-gel dip do not attempt to apply our theory to the regime
coating. Although most drying theories available in which the thermophysical properties depend
today include coupled momentum, energy and on concentration, i.e., we do not extend calculations
species transport in the liquid phase, they lump to high enough solid content where significant
the same theory for the gas phase into mass- dependencies are noticed. With regard to the
transfer coefficients (e.g., Emery and Egolf 1989; assumption of isothermal flow, evaporation can
Bornside 1989; Tu and Drake 1990). This was depress the temperature of the film and conden-
sation can raise the temperature of the film, but
initial calculations with the energy equation have
shown that the temperature variations (a fraction
of a degree K) do not have a significant effect
on the phenomena we investigate here (cf Schunk
hl.</ and Rao 1994). In fact Parks and Wayner (1987)
found that concentration effects on surface tension
«) were far more important than thermal effects in
a setup similar to dip coating.
13.5.1 TWO-PHASE MULTICOMPONENT

SYSTEMS
Fluid flow in the gas and liquid phase is governed
by the principles of conservation of mass and
momentum in the bulk liquid, i.e., the
SUBSTRATE
Navier-Stokes system (Chapter 2). This system
<al L...- -l (bl
is coupled to the convective diffusion equations
CONTROL CHAMBER for species transport in several ways. First,
Computational Domain Pilot Coating Setup diffusion itself can induce mass-average motion
and hence influence the balance of momentum,
Figure 13.10 Computational domain of pilot dip although this effect is not present with the
coating setup: (a) system geometry; (b) pilot coating constant densities we assume here. Second,
setup; (c) geometry around drying line. evaporation or condensation at a free surface
introduces fluid motion in the bounding phases; (cf Cairncross, Francis and Scriven 1992), but
this happens to be the basis of the boundary we see no major advantage gained in this case
condition on velocity we apply at the free surface. by doing so.
Third, species transport near a free surface of Boundary conditions on equation (13.5) and
each solvent component possessing a unique equation (13.6) are
surface tension can change the surface tension
of that surface. This upsets the balance of forces n'VX j = 0
at the free surface: local excursions of surface for i = 1,2,3, ... ,N at solid surfaces (13.7)
tension away from its equilibrium value affect
the normal balance of forces, and variations of n' [(us - U)X:pl - i:J = n' [(us - u)Xr p8 - irJ
surface tension along the surface affect the
tangential balance of forces. for i = 1,2,3, ... , N at free surfaces (13.8)
ij is the diffusive contribution to the total flux

13.5.1.1 Theory for mass exchange across N j == pXiu + ii of species i. ij is given by Fick's
interface law: i.e., i; == -pDjVX j. Us is the velocity of the
free surface. Equation (13.7) implies that all solid
In addition to the Navier-Stokes system and the
surfaces are impermeable to all species. Equation
energy equation outlined in Chapter 2, the
(13.8) allows bounding free surfaces to exchange
relevant convective diffusion equations are
species: the diffusive and convective flux of
·X·
a + u·Vp·XI = V'p.D~VX~ + q.
....f!..-..!.
component i normal to the surface in the liquid
at ' '" must be balanced by the diffusive and convective
flux of component i normal to the surface in the
for i = 1,2,3, ... ,N (13.5) gas. Of course we assume that the surface is
ap8x~
incapable of holding a significant inventory of
-at
- ' + U'Vp8X~' = V'p8D~VX~
I I
+ q.
I
any species in excess ofthe bulk, under equilibrium
conditions. That is, no species here is taken to
for i = 1,2,3, ... ,N (13.6) be surface active (see Chapter lId).
Analogous to the bulk equations above, we
where Xi is the bulk mass fraction of species i; replace one of the species kinematic conditions
D i is the pseudo-binary diffusivity of species i; (equation (13.8» with the overall kinematic
and qj is the rate of generation of species i. u condition (the sum of all equations in equation
represents the vector of mass-average velocity. (13.8» which states that the overall mass flux
The superscripts I and g denote liquid phase and across the surface be continuous, i.e.,
gas phase, respectively. Expressions for qi will
depend on the reactivity of species i with other n'(u - us)p· = n'(u - us)p8 (13.9)
species. For now we consider nonreactive systems
for which qj = O. When more than two species It is common practice to replace the right-hand
are included in any phase we ignore the effects sides of equations (13.8) and (13.9) with terms of
of cross-term diffusion coefficients and assume the form Ki(X i - X?) (see, for example, Bornside
every species behaves as if it were dilute (pseudo- 1989). Here Xi is the mass fraction i at the free
binary diffusion). The equation of local overall surface in the gas phase; under equilibrium
mass conservation, the continuity equation conditions it can be related to the mass fraction
V· u = 0, which is already a part of the at the same locale but in the liquid phase. X? is
Navier-Stokes system is used in place of one of a parameter used to set the overlying concentration
the components of equation (13.5) and equation of component i far away from the surface in the
(13.6). These equations can be rewritten in terms gas. K j is known as a mass-transfer coefficient
ofvolume fractions and volume-average quantities and is responsible for approximating the gas-
Theory and analysis of free-meniscus coating and drying 691
phase mass transport ofcomponent i. It is usually however. Raoult's law is satisfactory over a wide
deduced from experimental correlations or from range of concentrations (Denbigh 1981). We plan
simplified theories and models of flow and on taking this simplified approach here.
convective diffusion fields. To make equation (13.11) computationally
The mass-transfer modeling approach can be convenient we express Pi in terms of mole fraction
highly successful if the gas-phase fluid mechanics Xi in the gas phase with Dalton's law that
can be controlled or constrained to some known pdptot = xr·The remaining challenge is to convert
pattern (e.g.• Bornside 1989) and the overlying mass fractions into mole fractions so that equation
saturation can be controlled. Trouble arises (13.5) and equation (13.6) can be used together
when the pattern ofgas flow is changed. for which with equation (13.11). This can be accomplished
K i must be rederived or recomputed. Moreover, by solving all but one of the following definition
simplified theories that enable the evaluation of equations for the mole fraction of component i
K i are often inadequate approximations to the
true physics. In this case the only alternative is
to solve the complete theory in the gas phase as
well. which is often only accessible through
computer-aided mathematics. a route we chose fori=1.2.3 .... ,N (13.12)
in the work reported here.
The remaining boundary conditions needed together with the constraint that Xl + x2 +
on equation (13.5) and equation (13.6) must x3 + ... + xN = 1. Another alternative to solving
provide a datum which sets a concentration level equation (13.12) is to use a volume fraction-
for each species. In the dip coating case study based formulation. a situation for which Amagat's
below. that datum in the liquid phase is set at law states that volume fractions equal mole
an inflow boundary: fractions in ideal gases.
At each locale on the free surface the number
Xi = X? at inflow surfaces. (13.10) of unknown concentrations is twice the number
of volatile species - the species in the liquid that
Equation (13.10) might apply in the gas phase if are taken as nonvolatile and the species in the
gas of a known saturation of each component is gas that are taken as insoluble in the liquid are
being blown into the system. Otherwise the accounted for by no-penetration conditions of
datum is set at the free surface with the conditions the form N i = pXiu + t = O. To account for the
of vapor-liquid equilibrium. unknown concentrations we use the conditions
The solvent in the gas phase at the free surface given by equation (13.8) for the liquid phase and
is taken to be in equilibrium with the solvent in the vapor-liquid equilibrium conditions given
the liquid at the free surface. In this work we by equation (13.11) and equation (13.12) in the
approximate this equilibrium behavior with gas phase.
Raoult's law:
Pi = Piv Xi
-, h (13.11) 13.5.1.2 Coupling of momentum conservation to
species conservation
where pi is the saturation vapor pressure of
solvent component i at a specific temperature When mass exchange is allowed across the free
and Xi is the mole fraction of the same in the surface the velocity will be discontinuous there
liquid at the free surface. Pi is the partial pressure if the material undergoes a density change. In
ofcomponent i in the gas phase at the free surface. two dimensions. at a given locale on the free
Raoult's law usually holds for the component surface. there will be four velocity components
present in excess or when the system can be taken from a double-valued vector velocity. The
as thermodynamically ideal. For many systems, corresponding conditions at the free surface that
we use to determine these components are components in the liquid next to the surface, at
g
D' Til = D' Ti g + 2H(JD + D' pl(U I - us)(u1 - u ) least those components which affect the surface
tension. One approach is to use a linear mixing
+ Vs(J (13.13) law to describe the surface tension:
t'ul gas = t'Ulliquid (13.14)
(JIXl + (J2 X 2 + ... + (JmXm
pg(1 - X~ - X~ - ... - X~_l)D'(U - us) (J = (13.16)
X I +X 2 +"'+X m
= -pgDgV(X~ - X~ + ... + X~-l) (13.15)
Here m is the number of components which affect
Here t represents the unit tangent vector to the the surface tension. (Ji is the surface tension of
free surface; (J is the surface tension; 2H is the that component ifit were pure. A better constitutive
mean curvature of the surface; and T is the total equation arises from fitting a higher order function,
stress of a Newtonian liquid (cf Chapter 2). Each e.g., cubic, to the actual data. We used both
of these conditions is applied at the free surface, approaches and found the latter necessary to
the vector condition (13.13) being applied as a achieve qualitative agreement with the dip coating
boundary condition on Navier-Stokes' equations process. An example of this is shown below with
in the liquid phase and the scalar conditions ethanol-water as solvents.
(13.14) and (13.15) being applied as boundary
conditions on the Navier-Stokes' equations in
13.5.2 COMPUTATIONAL STRATEGIES FOR
the gas phase.
SIMULTANEOUS MOMENTUM AND SPECIES
Equation (13.13) is a vector condition that
TRANSPORT
balances the normal and tangential components
of the viscous forces and the hydrodynamic As we show in Section 13.6.1, the equations
pressure in both adjacent phases with the capillary simplified for single-phase, single-component flow
pressure, the vapor recoil by volume expansion, are analytically tractable when the meniscus
and the surface tension gradients at the interface. curvature is small and the film is thin and flat.
We neglect vapor recoil effects (the third term Predictions for multicomponent systems without
on the right) because it scales with the Reynolds the same restrictions come from solving the
number, which can be shown to make the term Navier-Stokes system (Chapter 2) together with
insignificant with the small evaporation rates equations (13.5) to (13.10) without approximation.
occurring here. Equation (13.14) enforces the We use the techniques based on the earlier work
continuity of tangential velocity at the interface. of Christodoulou and Scriven (1992) and Kistler
Equation (13.15) expresses the gas phase velocity and Scriven (1983) to accomplish this task. These
component normal to the interface as a function techniques, also described in Chapter 9, are based
of the gradients in concentration in the gas phase on subdomaining, Galerkin's method, and finite-
next to the surface. It is derived on the premise element basis functions. In order to cope efficiently
that the Nth component in the gas is effectively with the thin boundary layers in concentration
insoluble in the liquid and that all binary that develop next to the free surface we employed
diffusivities are equal. In this work the insoluble a two-level discretization scheme designed by
component is taken to be air. All other volatile Schunk (1989). The coarse subdomain structure,
components are allowed to enter the gas phase or mesh, was designed with the elliptic generation
by evaporation or re-enter the liquid phase by scheme described in Chapter 9 (Christodoulou
condensation. and Scriven 1992). This scheme produced meshes
In this work we have taken the surface tension that automatically conform to changes in bound-
(J to be a function of concentration alone. Future ary shape when physical parameters are varied.
work will include temperature. In any case what Figure 13.11 shows a sample of the two levels of
is needed is a constitutive equation that relates discretization used in this chapter. The fine level
the surface tension to the concentration of all was formed by algebraically 'sub-dividing' four
Case study: sol-gel thin films 693
times the rows of elements adjacent to the free 13.6 CASE STUDY: SOL-GEL THIN FILMS
surface on each side of the interface, with a
geometric progression in element thickness 13.6.1 ASYMPTOTIC ANALYSIS FOR
perpendicular to the free surface. The meshes SINGLE-COMPONENT SOLVENT
contain 1883 biquadratic elements in the coarse
The first attempt at a theory for dip coating and
grid and 2435 biquadratic elements in the fine
evaporation was made by Hurd and Brinker
grid, and gives rise to a total of 46 394 unknown
(1988) for single component (pure) solvents. They
coefficients for ternary systems in both phases.
assumed that the evaporation rate was constant
The resulting equations, which are a discretized
along the entrained liquid film and ignored the
version of the physical ones being solved, are
effects of surface tension. They predicted a linear
nonlinear and algebraic. They are solved as is com-
thinning thickness profile, i.e., wedge-shaped. In
mon practice in coating-flow analysis with Newton's
the same work, however, experiments indicated
method, as described in Chapter 9. The discon-
that the evaporation rate was not constant and
tinuities in velocity and concentration at the free
that the film profile was blunt (parabolic) in
surface, owing to the volume change on evaporation
shape, presumably due to an accelerated evap-
and vapor-liquid equilibrium, were handled with
oration rate near the drying line; see Fig. 13.8.
two nodes assigned to the same physical location
In the following development, we show that the
on the surface. One node was assigned to the
evaporation rate is controlled by the vapor
liquid-phase element and one to the gas-phase
diffusion away from the film.
element. Multiplying the weighted residual equa-
Because experiments show that the film is
tions applied at those nodes with the proper
much thinner than its width and breadth, and
liquid/gas property ratios makes for convenient
its slope is relatively small compared with the
cancellation ofthe diffusive flux contributions and
distance on the substrate over which it drys, the
eliminates the need to evaluate normal derivatives
flow is slow enough that a lubrication approxi-
at the boundary (see Schunk and Rao 1994).
mation applies in which flow is parallel to the
substrate and only gradients normal to the
substrate are important, i.e., U . Vu ~ O. The film
cannot be considered thin near the static reservoir,
however, but there the fluid mechanics is negligible;
in this region surface tension and gravity control
meniscus shape.
In the development we present here, the effects
of long-range molecular forces (e.g., disjoining
pressure) that can be important when films are
thin have been neglected. Here x measures the
distance along the substrate from the drying line
(bl toward the reservoir, and y measures the distance
normal to the film, as shown in Fig. 13.10. If the
substrate is pulled from the reservoir at an angle
IX from vertical, the Navier-Stokes equations,
duly simplified with the lubrication approximation,
can be integrated to give
u = - U[1 + j(x)h(x)y -1j(X)y2] (13.17)

Figure 13.11 Sample meshes for two-phase flow: (a) where the appropriate boundary conditions of
coarse mesh; (b) fine mesh. non-slip at the substrate surface (y = 0), vanishing
shear stress at the meniscus (y = h(x)), and the in the gas. Equation (13.22) is actually a statement
balance of capillary pressure with hydrodynamic of vapor-liquid equilibrium (Raoult's law, equa-
pressure at the free surface have been applied. tion (13.11)); equation (13.23) is a simplification of
f(x) is defined as equation (13.7) and ensures no evaporative source
JlU
dh - cosa ] -
f(x) == ( -pg ) [(slOa)-
.
dx
((1)
-
JlU
- dK
dx
above the drying line. Equation (13.21) is equivalent
to that for the electrostatic potential around a
conductive object, with the electric field at the
(13.18) surface being analogous to the evaporation rate
E(x) and the flux J(x) analogous to the charge
K being the mean curvature of the meniscus. An
equation for the steady state h(x) can be found accumulation between 0 and x. Near any sharp
boundaries, the electric field (evaporation rate)
by equating the mass flux J(x) == f~ pu dy to the
evaporative loss between x and the drying line, i.e., diverges but the charge (vaporized mass) must
remain integrable. This divergence in evaporation
x
accounts for the blunt profiles depicted in Fig. (13.8).
J(x) == - f E(x) dy (13.19)
As the profile h(x) is not known a priori,
o
equations (13.21) to (13.25) constitute a free
where the quantity E(x) is the local evaporation
boundary problem. However, because the en-
. h(x) (pg)l/2
rate at pomt x. If we take II == Ca l / 6 JlU ' trained film is very thin compared with its
breadth and length, the vapor sees it as infinitely
the equation for h(x) is thin; hence, its detailed shape is inconsequential.
d (
3
2
d II/de) . The solution for the potential and field near the
II de (1 + Ca2/3(dII/de)2)3/2 = 3[II - J(m edge of a thin sheet is (Jackson 1975):
X ~X 0 - a l r l / 2 sin (~) (13.24)
Here E=_DoXI = Da l x- l / 2 (13.25)

oy y=O
where (r, (J) are cylindrical coordinates defined
and by r sin (J = y and r cos (J = x. The constant a l
pg)l/2 must be determined from boundary conditions.
Ca- l /6 f E(x)dx
x
j(e) == (pU)-l ( - Equation (13.25) can be used to evaluate j(e) in
JlU 0
equation (13.20). Solving equation (13.20) near
The local evaporation rate E(x) is assumed to the drying line where curvature effects are assumed
be equal to the diffusive flux of vapor in the gas negligible and II and dII/de are small gives
phase at the interface, i.e., - D ~X I ' X being
uy y=h
(Hurd and Brinker 1990)
the mass fraction solvent in the gas phase. h(x) = Da l X l /2 (13.26)

pU
Neglecting convection effects in the gas phase,
Equation (13.6) becomes Fitting the pure ethanol data in Fig. 13.8 to the
2 form h = cx' yields v = 0.5 ± om, which holds
V X =0 (13.21)
over the experimentally observable range
X = X o at y = h(x) (13.22) 50 j,Lm < x < 1 cm. It can be shown that the
dX curvature is small in this range (radius ofcurvature
dy = 0 at y = 0 on dry substrate. (13.23) of order 1 km).
The singularity strength (exponent) in equation
Equation (13.22) and equation (13.23) represent (13.25) is sensitive to the geometry of the film.
boundary conditions on the solvent concentration For coating a cylinder of decreasing radius, the
strength passes from - 1/2 for large radius to

- 1 as the radius vanishes. This change can be E 70.------~----___,
demonstrated experimentally (Hurd 1992). Z
E 60
z 50
13.6.2 COMPUTER-AIDED RESULTS FOR o
MULTICOMPONENT ONE- AND TWO-PHASE en 40
SYSTEMS
z
UJ
l- 30
13.6.2.1 Base case UJ 20
U
The thermophysical properties and coater design «
u.. 10
specifications in Table 13.2 define a case of a dip a: °OL---------:~----~
:::> 0.5 1
coating process. This base case is of a typical en
process in sol-gel dip coating and is the one, with MASS FRACTION ETHANOL
the exception of the liquid composition and
substrate speed, for which the experiments Figure 13.12 Surface tension versus mass fraction
summarized in Section 13.4 were performed. A ethanol in water-ethanol solutions at several different
solution of ethanol in water with a dispersed, temperatures (from Weast (1974)).
non-volatile phase was chosen as a model system
because it is typical of a silica-based sol. The It is important to mention that the flow was
density and viscosity of the liquid were taken to observed to be essentially steady under the
be independent of the initial concentration of all conditions given in Table 13.2 (Hurd 1992).
species. The surface tensions at different ethanol Moreover, the flow was visibly stable when we
concentrations are shown in Fig. 13.12. We fit a varied the substrate speed and coating liquid
hyperbolic cosine function to the data in the composition, with the exception of a three-
range 0 to 0.2 mass fraction ethanol and a linear dimensional ribbing-like phenomenon that ap-
function in the range 0.2 to 1. The fit served as peared over a narrow range of coating rates
the constitutive equation for surface tension. (Hurd 1992). This instability was likely connected
to temperature- or concentration-induced surface
Table 13.2 Operating conditions and thermophysical tension variations because it rarely occurred in
properties for dip coating base case single-component systems. Interestingly, the
rib-like pattern decayed rapidly with distance
Operating conditions downstream, and the final film was apparently
Liquid composition 74:16:10 ethanol/water/
(volume %) non-volatile
flat and unaffected. Because our analysis is
Substrate speed 1.0cm/s strictly two-dimensional, we cannot pick up any
Substrate withdrawal evidence of this flow state. However, we are
angle Vertical confident that the flow and mass-transport fields
Reservoir width 7mm we report herein are stable with respect to
Reservoir depth 3mm two-dimensional disturbances, unless instabilities
arise because of lack of sufficient resolution in
Thermophysical properties
Liquid viscosity 3
2 X 10- Pas
the computational mesh.
Gas viscosity 1.8 x 10- 5 Pa s
Liquid density 1000kg/m 3
Gas density 1kg/m 3
13.6.2.2 Single-phase flow with mass
Surface tension transfer model
(equilibrium 20°C) 30mN/m The equations of overall mass, momentum and
Liquid diffusivity 1 x 1O- 8 m 2/s
Gas diffusivity 1.6 x 1O- 4 m 2/s species transfer were solved first for the liquid
phase only, with a mass-transfer model accounting
same for all species. It is noteworthy here that

K i is independent of the local interfacial concen-
tration, and that only through the mass transfer
model for the flux, i.e., ni = Ki(X i - X?), can it
depend on the local concentration X? in the 'far
---~
Uquid
Gas
t field' of the gas phase. Furthermore, the local
mass transfer coefficient decreases with increasing
distance downstream (Cussler 1984), which leads
to lower mass transfer rates. The simplified
lubrication approach in the previous section
suggests that the mass transfer rates must increase
with distance downstream to give rise to the
Figure 13.13 Similarity ofthe boundary layer between
parabolic-type meniscus shapes.
moving flat plate and free surface movement in dip
coating. Figure 13.14 shows the pattern of streamlines
for a composition of 82:8:10 by weight
for the fluid mechanics and mass transport in ethanol/water/non-volatile. Figure 13.15 shows
the gas phase. In this approach, equation (13.6) the corresponding mass fractions of ethanol
and the right-hand sides of equations (13.8) and (right) and of non-volatile species (left) in the film
(13.9) were replaced with expressions based on region above the level of the reservoir. The
the effective flux of each species across the pattern of streamlines shows a counterclockwise
interface (cf Bornside (1989)). These expressions recirculation driven by the upward movement of
were of the form N i = Ki(X i - X?), for each the substrate. Because we are approximating the
species i. The constants K i were estimated from process as steady, the balance of material leaving
classic boundary-layer theory of semi-infinite the system by evaporation and film entrainment
flow driven by a moving flat plate emanating (upper right) is the liquid brought into the
from a slit (Sakiadis 1961). Figure 13.13 suggests reservoir to maintain its level. The streamlines
that the boundary layer in gas flow that builds
up downstream of the stagnation point in dip
coating is of the Sakiadis type, except near the
stagnation point. In this boundary layer approach,
functional forms for the velocity components and
the concentration are substituted into the
Navier-Stokes system and the convective-diffu-
sion equations in the gas phase. The resulting
equation set is then solved for the rate of mass
transfer in terms of the concentrations (partial
pressure) of the transported species at the surface
of the film. The average gas-phase coefficient of
mass transfer K i resulting from this approach is
L.
(13.27)
Here Reg is the effective Reynolds number in the

gas phase based on the substrate speed and a
development length L along the film. SC g is the Figure 13.14 Pattern of streamlines predicted with
Schmidt number in the gas phase based on the boundary-layer type mass transfer model. Composition
binary diffusivity D i , taken in this case to be the is 82:8:10 by weight ethanoljwater/non-volatile.
0.987
0.925
HORIZONTAL
AXIS EXPA OED
30X
0.862
0.800
0.737
r
0.3 mm
0.675
~
0.612
0.550
0.487
MASS FRACTION NO -VOLATILE
MASS FRACTION ETHANO<
Figure 13.15 Concentration contours predicted with mass-transfer model. Left side shows the mass fraction
of non-volatile material and the right side shows the mass fraction of ethanol. The mass transfer coefficient
(given by equation (13.27)) is constant along the free surface. Conditions are the same as Fig. 13.14.
indicate this inflow along the substrate at the at the surface to vary. Unfortunately, the simplified
bottom of the reservoir. Near the free surface the boundary layer theory we employ does not
streamlines also indicate vigorous evaporation, account for variations in the overlying concen-
as several intersect the surface. The corresponding tration of volatile mass - X? must be set in an
concentration contours (Fig. 13.15) show stream- ad hoc fashion as a parameter - and possible
wise depletion ofthe more volatile species (ethanol) dependencies of the local mass transfer coefficient
and a boundary layer of ethanol lean (water-rich) K i on local concentration at the interface. With
solution along the free surface overlying the this approach we found no way to reproduce the
more-or-Iess static reservoir. That boundary layer experimentally measured profiles ofethanol- water
turns parallel to the film direction when the film films.
becomes so thin that the substrate presence To investigate the effect of a varying coefficient
influences the evaporation-induced concentration we presumed that the evaporation rate must
gradients. Streamwise enrichment of the non- increase with distance downstream, as suggested
volatile species is also evident in Fig. 13.15. The by the geometrical effects observed in experiments
concentration contours in this case are largely near the drying line. Correspondingly we tried
perpendicular to the interface where the film is thin. ramping in an ad hoc manner the mass transfer
The results in Fig. 13.15 show small variations coefficient with distance along the surface from
in volatile species concentration along the free the pool to the outflow plane. The results of one
surface. In fact, the ratio of water to ethanol case are shown in Fig. 13.16 (left). In this case
remains large (water-rich) and unchanged near the mass transfer coefficient was ramped from a
the surface, although the absolute mass fractions low value (one tenth of that given by equation
of these species change as the non-volatile species (13.27)) to the value used in Fig. 13.15 over the
concentrates along the substrate. This result is last 0.2 mm of the entrained film. Now not only
an artifact of the constant mass transfer coefficient is a substantial gradient of ethanol concentration
used so far, which outweighs any mechanism in apparent along the free surface, but the corre-
the liquid phase that might cause the concentration sponding surface tension gradient directed upward
30%
Relative
:c Saturation
f-
o of Ethanol
Z
U.1
....l
U
e>::
«
K·I
Figure 13.16 Effect of ramping of mass-transfer coefficient (left) and humidity change (right) on the contours
of constant mass fraction of ethanol. All other conditions are the same as in Fig. 13.14. The horizontal axis
is expanded thirty-fold.
is strong enough to create a small pile-up of effects of humidity and the effects of forced
water-rich liquid near the outflow plane. Instead convection, on the final film quality. Another
of varying the mass transfer coefficient we also advantage is that it circumvents the need to
tried adjusting the overlying concentration through search for gas-phase mass transport theories that
the parameter X? On the right of Fig. 13.16 we account for realistic variations in mass-transfer
show the effect of adding a 30% overlying coefficient, and especially theories that are
saturation of ethanol in the overlying gas, i.e., analytically tractable.
we set the partial pressure of ethanol to be 30% The base case here is the same as that outlined
of its equilibrium vapor pressure. In this case we in Table 13.2, but the coating composition is
notice that evaporation of ethanol is impeded taken to be 74:16:10 by weight ethanol/
enough to undermine the streamwise depletion water/non-volatile so that the water to alcohol
of ethanol. In fact, the mass fraction ethanol is ratio was identical to that in the experiments
still 0.4 at the outflow plane. performed by Hurd and Brinker (1988), cf Fig. 13.8.
The boundary conditions we impose on the
gas flow in the base case force dry air into the
13.6.2.3 Model of two-phase flow
domain over a portion of the artificial boundary
In place of the mass-transfer model in the gas that overlies the reservoir (see Fig. 13.17). The
phase we solved the mass-transport system given velocity of the gas over this portion was taken
by equations (13.5) to (13.12) together with the to be 0.1 cm/s, which is of the order of the
Navier-Stokes system in both gas and liquid substrate speed, i.e., 0.2 cm/s, but far less than
phases (see Schunk and Rao 1994 for details). the evaporation-induced velocity in the gas phase
This eliminates the need to lump the gas-phase (order of 10 cm/s). Over the same portion of the
physics into a single but variable coefficient, boundary we specified the incoming gas compo-
albeit at a greater cost. The problem then is no sition to be void of water and ethanol vapor.
longer one of the accuracy of the mass transport The remaining external boundary in the gas-phase
coefficient, but of the accuracy of the boundary we presumed to be free of viscous traction and
conditions on the gas flow. The advantage of this of diffusive flux of each species, except for the
approach, however, is that it enables us to solid boundary that contains the coating liquid
explore the effects of air conditioning, i.e., the on the left-hand part of the domain, which is
FORCED
GAS
INFLOW
0.1 mls
]
I
TRACTION I
FREE. r
DIFFUSION I
FREE II
IrI
/1/
]
nnnr I I
I I
I
1/
II/
1111
/ I "I
/
." 'It,
-, ,
GAS
LIQUID
(a) (b)
Figure 13.17 (a) Velocity vectors (absolute) in gas and liquid phases. (b) Pattern of streamlines. (c) Pattern
of streamlines in film region with horizontal axis expanded 100 times (multiplied by density).
taken to extend into the gas as well. Of course and the velocity vectors corresponding to the
these conditions are arbitrary, but they provide base case operating conditions. Between any two
a means of matching the experimental conditions. adjacent streamlines in Fig. 13.17b the volumetric
Unfortunately the experimental data we have flow rate is the same, but the mass flow rate
available for comparison were performed with changes if the streamlines cross the free surface.
no knowledge of the details of the gas flow. Both the pattern of streamlines and the velocity
Figure 13.17 shows the pattern of streamlines vectors in the gas phase show a clockwise vortex
driven by the counter-clockwise recirculation in however, the film becomes depleted in ethanol,
the liquid phase reservoir. This is a direct and the water concentration rises. But the water
consequence of no-slip boundary condition too is evaporating, and near the outflow plane
(equation (13.14)) at the interface. Over this (projected to be about 2 mm from the drying line)
portion of the free surface the evaporation rate the concentration of the non-volatile species
is relatively small, due mostly to a higher begins to concentrate precipitously. Figure 13.18
concentration of ethanol and water vapor in the shows the mass fraction non-volatile species
overlying gas. Far above the reservoir surface rising from 0.4 to 0.7 in the last 1cm of the
the streamlines indicate a strong vapor current film.
emanating from the thin liquid film. This When coating binary solvent-based sols, the
heightened evaporation is also evident in plots meniscus shape usually exhibits a peculiar plateau
of velocity vectors: the component of velocity region followed by a region of rapid thinning
normal to the surface, governed by equation and a second, lower plateau. This shape has been
(13.15), increases with distance downstream. Also observed in earlier experiments (Hurd 1992)
noteworthy is the viscous boundary layer in the where it was conjectured to be a consequence of
gas phase building downstream of the deposition differential volatilities and surface tensions of
region on the free surface (right side, bottom, of ethanol and water. This mechanism is corrob-
Fig. 13.17). It is this boundary layer on which orated by the theory and is shown in Fig. 13.18.
we based our model of mass transfer in the Below the second plateau (or 'foot') is a water-
previous section. The two-phase flow calculations rich phase that outlasts the alcohol due to a
indicate that the mass transfer model we employed lower volatility. This apparent 'phase separation'
earlier may be inaccurate in part because of appears regardless of the original composition
strong convection normal to the free surface. If of the liquid and over a wide range of drying
we extract the local mass transfer coefficient from conditions we tried through alterations in the
the two-phase flow calculations we find that the gas-phase boundary conditions. It even appears
portion of the evaporation rate coming from a at the evaporation-azeotrope composition at
vapor diffusion model is 20% too large (cf, which the mixture evaporates at constant con-
Schunk and Rao 1994). centration.
Vapor-liquid equilibrium conditions make Surface tension gradients between the alcohol-
for a concentration discontinuity at the free rich and water-rich portions of the film enhance
surface. This discontinuity defines the free surface, this plateau region. The velocity vectors in Fig.
as shown in Fig. 13.18. Unfortunately, owing to 13.17 (center inset) show a rapid acceleration of
relatively high vapor pressures, the concentration the surface, and hence of the underlying liquid,
levels of ethanol and water in the gas phase, and towards the downstream plateau, thereby
hence the gradients in concentration there, are lengthening the foot feature. With other solvent
too small to be resolved by the color spectrum. mixtures this can be even more dramatic. For
For that reason we show an overlay of the instance, in toluene-methanol systems the surface-
concentration contours in the gas phase. Note- tension-gradient driven flow is strong enough to
worthy is the gradient in concentration normal thicken the lower plateau, creating a pile-up of
to the free surface which grows larger with the less volatile toluene near the drying line (Fig.
distance downstream (i.e., contour lines in the 13.18). This phenomenon, which is known as
gas phase become more parallel to the film Marangoni flow (Sternling and Scriven 1959), is
downstream), and an apparent layer of relatively akin to the famous example of wine tears, where
high vapor concentration overlying the reservoir evaporation in the thin film region of a glass
(mass fraction 0.095). Correspondingly, we see induces a climbing motion of the wine (Walker
little enrichment of water and non-volatile species 1983). It has also been observed widely in spreading
near the reservoir. Well above the reservoir, drops of binary liquids, for which it is found that
O.6SS
0.576
OA93
OAII
0.329
0.247
0.164
0.005
0.015 O.OS2
0.025
0.035
0.045
OJ)55 - -
0.065
00.75
O.OSS
O.OC). L...-----".
(b)
HORIZONTAL
rI
AXIS EXPANDE9
(a) (e)
Figure 13.18 Concentration contours for two-phase flows: (a) mass fraction ethanol; (b) mass fraction ethanol
in film region; (c) mass fraction water in film region. The difference between one and the sum of (b) and (c) is
the mass fraction non-volatile at the same location.
when the more volatile component has a lower The shape of the drying film can have enormous
surface tension, the drop will spread faster due implications on the development of the gel
to surface tension gradients. A first-hand example microstructure. One way is through the kinematic
of this with ethanol-water systems was observed effect associated with rapid film thinning, which
by Pesach and Marmur (1987). brings gelling molecules close together at high
current numerical scheme is incapable of handling.
46
Experiments were performed at coating speeds
211 of 0.2 cm/s; the computations ranged from 2 cm/s
376
541
to 1 cm/s. Solutions to the theory at lower
706 coating speeds could not be attained due to
INTENSE SURFACE 71
[
TENSION GRADIENTS 1036
insufficient streamwise resolution in the com-
putational mesh needed to resolve film thickness
of less than 1 ~m.
Figure 13.20 shows a comparison between film
thickness profiles predicted with the theory and
the profile measured with imaging ellipsometry
(Fig. 13.8). Clearly there is a large discrepancy
STAGNATION
POINT REGION
IL 871
1036
1191
in the magnitude of the film thickness at any
point along the substrate. But the solution family
1346 parameterized by web speed shows that the
correct magnitude is apparently being approached
with a reduction in substrate speed, which is
Y
__ x
encouraging. Even more encouraging is that the
HORIZONTAL AXIS EXPANDED 100X shape features we predict are similar to the
experimental data, both of which show a rea-
Figure 13.19 Shear rates in dip coating, s- 1.
sonable fit to equation (13.26). It is also reasonable
to expect quantitative agreement had we been
able to match the conditions of the experiment.
rates (cf. equation (13.4». Another way is through We do not expect, however, to be able to predict
the large velocity gradients created by the the exact location of the drying line with the
variations in surface tensions. Figure 13.19 shows theory, unless we can match the drying environ-
that shear rates in the liquid underneath the ment in the experiments and account for the
rapidly thinning menisci can reach 1000 s - 1. microscale physics which dominate when the film
Locale where the rates are highest lie underneath
the stagnation point, which terminates a dividing
line between the entrained film and the draining THEORETICAL EXPERIMENTAL
74:16:10 WATER I ETOHI SILICATE 82:18 WATERI ETOH
film, and near the most intense surface tension
gradients between the water-rich region (down- 60.0 rTrr--:-;;:=:::::£!rEiBJ~~m==:;l
~ B.O
stream) and the ethanol-rich region (upstream). :zz

w
E
The two-dimensional model of sol-gel dip :::l. ~
z
coating described in Section 13.6 was tested gf 40.0 ....
3 o:-__-'-__----L_-=~.
w
against the experimentally measured film thickness z
:.:
Ii: 0 DiSTANCE ALONG SUBSTRATE. mm 30
profiles and concentrations profiles described u

J: 2cm/s
earlier. The coating liquid in the theory was ~
1.5 cm/s
::ii 20.0
composed of 74% by weight ethanol, 16% by ....I
u:: 1.2 cmls SUBSTRA TE SPEED
weight water, and 10% by weight non-volatile,
EXPERIMENTAL DATA
the last having a molecular weight of 46 g/mol.
In the experiments the liquid was 82% ethanol 0.0 1--':~o~.2~c~m/.~sS:~=:::::=::::~~~,-j.
10 20 0
and 18% water, so the ratio of ethanol to water DISTANCE ALONG SUBSTRATE, mm
was the same. The non-volatile component was
included in the theory to prevent complete Figure 13.20 Comparison between theory and ex-
evaporation of the film, a phenomenon the periment
Summary/future prospects 703
becomes thinner than 1ooA. Unfortunately, at case outlined in Table 13.2; however, the com-
the time ofthe experiments, little effort was made position in this case is 40:40: 20 water/ethanol/non-
to characterize the gas flow. This clearly leaves volatile. Figure 13.21 (right side) shows the
room for future work. velocity field under and around the air knife.
We also checked the predicted concentration Noteworthy is the stagnation-like flow pattern
profiles against those inferred from the fluorescence under the jet where it impinges on the film. In
spectroscopy measurements, as shown in Fig. this figure, the velocity vectors in either phase
13.9. Even though the experiments were performed are multiplied by density, and so they indicate
under different conditions, we see similar con- a mass flow rate, and the vectors on the free
centration variations along the film, with much surface in the gas phase have been magnified
of the ethanol depletion taking place well above over the others to show the heightened evaporation
the reservoir surface. Unfortunately, at the time underneath the air knife. Figure 13.21 (left side
of this writing, we had no quantitative measure- and top) shows the concentration contours of
ments available indicating the magnitude of these ethanol in both the gas and the liquid. Here the
variations. stagnant layer of gas with high ethanol vapor
content is visible beneath the air knife. Up on
the substrate the concentration of ethanol in the
13.6.2.4 Drying control
gas is much less and, correspondingly, the
Because the gas phase plays an important role evaporation rates are much higher. The highest
in the development of film microstructure, it is evaporation rates are directly under the air jet
possible to control microstructure with the drying at the film surface, as expected. Correspondingly
environment. There are several means of doing the volatile components are drawn off the film
so, the simplest being the humidity. With the more rapidly, with the film becoming virtually
more-simplified mass transfer theory for the gas depleted of the more volatile ethanol underneath
phase, humidity adjustments can be made to the the air jet. Of course this affects the final film
overlying saturation of species i, through X?, as thickness, but not as much as the metering effect
we demonstrated in Fig. 13.16. Unfortunately, through impingement pressure. This is also evident
changing the overlying saturation probably alters in Fig. 13.21 (top right) where there appears a
the flow field, and should be accompanied by depression in the film underneath the jet.
the appropriate adjustments to the mass-transfer Clearly the film concentration and thickness
coefficient K;. Rather than search for the evolution can be controlled with the air knife. If
appropriate coefficients we again opted for the two- the results we present here are not desirable to
phase flow approach, for which humidity control a coating practitioner, we are free to vary the
is a simple matter of conditions on the volatile concentration of the air blown through the jet.
species at the external boundaries of the domain. In fact it is possible to enrich the film in a
As an example, we show in Fig. 13.21 the effects condensable solvent component, as has been
of drying control with a narrow jet of dry air demonstrated in the Marangoni drying technique
impinging on the liquid film just above the level (cf Leenaars, Huethorst and van Oekel 1990;
of the reservoir. This approach, often termed O'Brien and van den Brule 1991).
air-knife control, is usually employed to accelerate
the drying of films after they are deposited and
13.7 SUMMARY/FUTURE PROSPECTS
solidified. Air knifes are also commonly employed
to control the film thickness, through an air- Dip coating epitomizes free-meniscus coating. Of
stripping process. Here an air knife was employed all coating processes covered in this book, it is
to achieve both effects. probably the oldest and most studied. It is also
The air is blown from a nozzle at 1m/s and perhaps the most unique. Film thickness can be
the coating conditions correspond to the base controlled only through substrate speed, angle
0..156
0.312
0.267
O.22J
0.178
O. J34
O.OS9 ···· .... .

O.(~
Figure 13.21 Drying control with air knife. Concentrations are mass fraction ethanol. Conditions are the
same as in previous figures.
of withdrawal, and various thermophysical coating even more unique: the mass transport
properties of the liquid to be coated (i.e., viscosity, by evaporation and the fluid mechanics of
density, surface tension, vapor pressure, etc.). deposition take place simultaneously, so any
Interestingly, even though ambient back pressure, realistic analysis must consider both phenomena.
inlet flow rate, and coating gap are crucial to We believe this chapter accomplishes this task,
the operability of most pre-metered techniques, and points out that there is still much to be
they have little or no effect on the range of film learned about dip coating and drying.
thicknesses that can be deposited successfully We attack dip coating in its simplest form and
with dip coating. with the complicating effects of concomitant
When substrate speeds are slow (i.e., low drying with a realistic theory based on conser-
capillary number) film drying makes free-meniscus vation laws of overall mass, momentum, species
References 705
and energy transport. The theory encompasses to be part of a much broader program on the
both gas and liquid phases, with proper interphase physical chemistry and processing of sol-gel
boundary conditions applied where appropriate. materials. However, quantitative predictions of
Multiple components and multiple phases together the effects of processing conditions on micro-
with the volume change accompanying evapor- structure are not accessible with the current
ation or condensation makes any numerical theory, at least at this stage. What is still needed
scheme of solving the governing equations more is a detailed model of the transport phenomena
challenging. The problem is that velocity and at the drying line, near which the fleeting formation
concentration are discontinuous (multivalued) at of the gel network will require a detailed model
the surface, the first being a result of a volume of the chemistry of gelation. We are currently
expansion that accompanies interphase transfer, addressing the detailed physics which dictate the
and the second stemming from vapor-liquid final pore structure of the film, but only once the
equilibrium laws. film has been deposited (Brinker et al. 1994 and
In this chapter we focus on the dip coating of Cairncross et al. 1995).
sol-gel materials, which are normally composed This chapter points out that it is perhaps
of a condensed phase component, which may unfortunate that our research is not more closely
polymerize or aggregate, and one or more solvent allied with that of Langmuir-Blodgett (L- B)
components which are volatile (e.g., alcohol and films, for the techniques and goals of both are
water). In most cases we solved the corresponding often similar. In fact many of the experimental
nonlinear differential equations with the Galer- techniques we use have also been applied in a
kin/finite-element method. We employed the best similar way to the dip coating ofL- B monolayers,
techniques available for mesh generation and free e.g., fluorescence spectroscopy (Ahlers et al. 1989).
surface parameterization. In L- B films the control of monolayer and
The solutions contain detailed information on bilayer microstructure is the ultimate goal, made
the meniscus shape and the distribution of solute challenging by the delicate nature of films only
in both phases during deposition. We found that one or several molecules thick. Even though our
differential volatility leads to surface concentration coverage of L- B films is sparse, we felt it deserved
gradients, and hence surface tension gradients attention because the same processing issues
that alter the flow and change the meniscus apply, e.g., drying control.
shape. For the first time the peculiar thinning
profile that has been observed experimentally in
alcohol-water systems is predicted theoretically. ACKNOWLEDGMENTS
Moreover, the predictions agree qualitatively The authors thank Terri Butler, University of
with profiles measured with imaging ellipsometry. Minnesota, for providing valuable advice on
Because of this agreement we extended the Langmuir- Blodgett coating. This research was
calculations to different drying environments to supported in part by the Basic Energy Sciences
demonstrate how drying strategies might be used program of the DOE and was performed under
to control sol-gel microstructure. This portion of the contract DE-AC04-76-DPOO789.
the analysis was made feasible only by extending
the theory to two-phase flow, rather than relying
on a mass transfer coefficient to account completely REFERENCES
for the gas-phase fluid mechanics and convective
Adamson, A. W. 1982. Physical Chemistry of Surfaces.
diffusion regimes.
4th edn. New York: Wiley.
Understanding how to control the microstruc- Agarwal, v. K.1988. Langmuir-Blodgett films. Physics
ture of sol-gel films with the drying regime is the Today. 40: 40-46.
main goal of the research program summarized Ahlers, M. et al. 1989. Specific recognition and formation
in this chapter. The work presented here happens of two-dimensional streptaridin domains in mono-
layers: application to molecular devices. Thin Solid Cairncross, R. A., Francis, L. F. and Scriven, L. E.
Films. 180: 93-99. 1992. Competing drying and reaction mechanisms
Ashley, C. S. and Reed, S. T. 1984. Sol-Gel-Derived in the formation of sol-to-gel films, fibers, and
AR Coatings for Solar Receivers. Sandia National spheres. 10th Anniversary issue of Drying Technology
Laboratories Internal Report SAND84-0662. J. 10(4).
Azzam, R. M. A. and Bashara, N. M.1977. Ellipsometry Christodoulou, K. N. and Scriven, L. E. 1992.
and Polarized Light. Amsterdam: Elsevier. Ch. 4. Discretization offree surface flows and other moving
Bablik, H. 1950. Hot-Dip Galvanizing. London: boundary problems. J. Compo Phys. 99: 39-55.
Spon. Cooper, A. R. 1977. Quantitative Theory of Cracking
Bankoff, S. G. 1983. Problems in interfacial instability. and Warping During the Drying of Clay Bodies, ed.
Proceedings of the 4th International Conference on L. Hench. pp. 261-276.
Physicochemical Hydrodynamics, ed. R. Pfeffer. 404: Cussler, E. L. 1984. Diffusion: Mass Transfer in Fluid
405-419. Systems. New York: Cambridge University Press.
Bein, T. et al. 1989. Molecular sieve sensors for Daniel, M. F. and Hart, 1. T. T. 1985. Effect of surface
selective detection at the nanogram level. J. Am. flow on the morphology of Langmuir-Blodgett
Chem. Soc. 111: 7640-7641. films. J. Mol. Electron. 1: 97-104.
Benjamin, B. T. 1957. Wave formation in laminar flow Denbigh, K. 1981. The Principles of Chemical Equi-
down an inclined plane. J. Fluid Mech. 2: 554-574. librium. Cambridge University Press.
Blodgett, K. B. 1935. Films built by depositing successive Deryagin, B. V. and Levi, S. M. 1964. Film Coating
mono-molecular layers on a solid surface. J. Amer. Theory. London: Focal Press.
Chem. Soc. 57: 1007. Dillon, P. W. 1977. Application of critical relative
Bornside, D. E. 1989. Spin coating. PhD thesis, humidity, an evaporation analog of azeotropy, to
University of Minnesota. Available from University the dryng of water-borne coatings. J. of Coatings
Microfilms International, Ann Arbor, MI. Technology. 49: 38-49.
Bornside, D. E., Macosko, C. W. and Scriven, L. E. Dislich, H. 1988. Thin films from the sol-gel process.
1989. Spin coating: one-dimensional model. J. Appl. In Sol-Gel Technology for Thin Films, Fibers,
Phys.66: 5185-5193. Preforms, Electronics, and Special Shapes, ed. L. C.
Brinker, C. 1., Frye, G. c., Hurd, A. J. and Ashley, C. Klein. pp. 50-79. Park Ridge, NJ: Noyes.
S. 1991a. Fundamentals of sol-gel dip coating. Thin Emery, A. F. and Egolf, 1. R. 1989. Finite element
Solid Films. 201: 97-108. analysis of the stresses and consolidation ofboehmite
Brinker, C. J., Hurd, A. 1., Frye, G. c., Schunk, P. R. gel during the constant rate drying period. Proceedings
and Ashley, C. S. 1991 b. Sol-gel thin film formation. of the 1989 National Heat Transfer Conference.
Centennial Memorial Issue of the Journal of the HTD-113: 95-104.
Ceramics Society of Japan. 99: 862-877. Esmail, M. N. and Hummel, R. L. 1975. A note on
Brinker, C. J. and Scherer, G. W. 1990. Sol-Gel Science. linear solutions to free coating onto a vertical
San Diego: Academic Press. surface. Chem. Eng. Sci. 30: 1195-1196.
Brinker, C. 1., Sehgal, R., Raman, N., Schunk, P. R. Frye, G. c., Ricco, A. J., Martin, S. J. and Brinker,
and Headley, T. J. 1994. Polymer approach to C. J. 1988. Characterization of surface area and pore
supported silica membranes. J. Sol-Gel Science and size of sol-gel films using SAW devices. In Better
Technology. 2: 469. Ceramics Through Chemistry III, eds C. J. Brinker,
Britten, 1. A. and Thomas, I. M. 1992. Large-area D. E. Clark and D. R. Ulrich. Materials Research
Sol-gel Multilayer Laser Reflectors Applied By Society, Pittsburgh. pp. 349-354.
Meniscus Coating. Technical Report, Lawrence Gaines,G. L.1966. Insoluble Monolayersat Liquid-Gas
Livermore National Laboratory, UCRL-JC-I09389. Interfaces. New York: Interscience.
Butler, T. L. 1993. Mechanisms of Langmuir-Blodgett Groenveld, P.1971. Drainage and withdrawal ofliquid
film deposition and the development of polymeric films. AIChE J. 17: 489-490.
film materials. PhD thesis, University of Minnesota. Gutfinger, C. and Tallmadge, J. A. 1965. Films of
Cairncross, R. A., Chen, K. S., Schunk, P. R., Brinker, non-Newtonian fluids adhering to flat plates. AIChE
C. J. and Hurd, A. 1. 1995. Recent advances in J. 11: 403.
theoretical modeling of deposition, drying, and Homsy, G. M. and Geyling, F. T. 1977. A note on
shrinkage in sol-gel coating processes. Proceedings instabilities in rapid coating of cylinders. AIChE J.
of the American Ceramic Society National Meeting, 23: 587-590.
Cincinnati, OH, 30 April-3 May. Horton, R. A. 1990. Investment casting. In Metals
References 707
Handbook 15. Prepared by ASM International. Marques, D., Constanza, V. and Cerro, R. L. 1978.
p.257. Dip coating at large capillary number: an initial
Hurd, A. J. 1992. Evaporation and surface tension value problem. Chem. Eng. Sci. 33: 87.
effects in dip coating. Adv. Chern. Series, No. 234. Morey, F. C. 1940. Thickness of a liquid film adhering
Colloid Chemistry of Silica, ed. H. Bergna (Amer. to a surface slowly withdrawn from the liquid. J.
Chern. Soc., Washington DC). Res. Nat/. Bur. Stand. 25: 385-393.
Hurd, A. J. and Brinker, C. J. 1988. Optical sol-gel Nigam, K. D. P. and Esmail, M. N. 1980. Liquid flow
coatings: ellipsometry of film formation. J. de over a rotating dip coater. Canadian J. Chem. Engng.
Physique. 49: 1017-1025. 58: 564-568.
Hurd, A. J. and Brinker, C. 1. 1990. Sol-gel film Nishida, F., McKiernan, J. M., Dunn, B., Zink, J. I.,
formation by dip coating. In Better Ceramics Through Brinker, C. J. and Hurd, A. 1. 1994. In situ fluorescence
Chemistry IV, eds B. 1. J. Zelinski, C. J. Brinker, imaging of sol-gel thin film deposition. J. Sol-Gel
D. E. Clark and D. R. Ulrich, Materials Research Sci. and Tech. 2: 477-482.
Society, Pittsburgh. pp. 575-581. O'Brien, S. B. G. and Van Den Brule, B. H. A. A.
Jackson, J. D. 1975. Classical Electrodynamics. New 1991. A mathematical model for the cleansing of
York: Wiley. silicon substrates by fluid immersion. J. Coli. Inter!
Jacobsen, S. C., Wells, D. L., Davis, C. C. and Wood, Sci. 144: 210-221.
1. E.1991. Fabrication ofmicro-structures using non- Park, C. 1991. Effects of insoluble surfactants on dip
planar lithography (NPL). Proceedings of the IEEE coating. J. Coli. Inter! Sci. 146: 382-394.
Workshop on Micro Electro Mechanical Systems, Parks, C. J. and Wayner, P. C. 1987. Surface shear
Nara, Japan, 31 Jan-2 Feb. near the contact line of a binary evaporating curved
Jeffreys, H. 1930. Draining of a vertical plate. Proc. thin film. AIChE J. 33: 1-10.
Camb. Phil. Soc. 26: 204-205. Pesach, D. and Marmur, A. 1987. Marangoni effects
Kaufman, V. R. and Avnir, D. 1986. Structural changes in the spreading of liquid mixtures on a solid.
along the sol-gel-xerogel transition in silica as Langmuir. 3: 519-524.
probed by pyrene excited-state emission. Langmuir. Petrov, J. G., Kuhn, H. and Mobius, D. 1979. Three-
2: 717-722. phase contact line motion in the deposition
Kheshgi, H. S., Kistler, S. F. and Scriven, L. E. 1992. of spread monolayers. J. Coli. Inter! Sci. 73:
Rising and falling film flows: viewed from a first-order 66-74.
approximation. Chem. Eng. Sci. 47: 683-694. Petty, M. C. 1987. Molecular engineering using the
Kistler, S. F. and Scriven, L. E. 1983. Coating flows. Langmuir-Blodgett technique. In Polymer Surfaces
In Computational Analysis of Polymer Processing, and Interfaces, eds W. 1. Feast and H. S. Munro.
eds J. R. A. Pearson and S. M. Richardson. Essex: pp. 163-185. John Wiley.
Applied Science Publishers. pp. 244-299. Poehlein, G. W., Vanderhoff, J. W. and Wit-Meyer,
Landau, L. D. and Levich, V. G. 1942. Dragging of a R. 1. 1975. Drying of latex films. Polym. Prepr., Am.
liquid by a moving plate. Acta Phys. Chim. URSS. Chem. Soc., Div. Polym. Chem., 16: 268.
17: 42-54. Pouxviel, J. c., Dunn, B. and Zink, J. I. 1989.
Langmuir, I. 1938. Overturning and anchoring of Fluorescence study of a aluminosillcate sols and
monolayers. Science. 87: 493-500. gels doped with hydroxy trisulfonated pyrene. J.
Lee, C. Y. and Tallmadge, 1. A. 1974. Dynamic Phys. Chem. 93: 2134-2139.
meniscus profiles in free coating III: predictions Prescher, M. 1991. Paint technology and the environ-
based on two-dimensional flow fields. AIChE J. 20: ment. EuroCoat. 1: 50-53.
1079-1086. Rehg, T. J. and Higgins, B. G. 1992. Spin coating of
Leenaars, A. F. M., Huethorst, 1. A. M. and van Oekel, colloidal suspensions. AIChE J. 38: 489-501.
J. J. 1990. Marangoni drying: A new extremely clean Roberts, G. 1990. Langmuir-Blodgett Films. New
drying process. Langmuir. 6: 1701-1703. York: Plenum Press.
Lewis, R. W., Morgan, K. and Thomas, H. R. 1983. Rosenberg, L. D. and Beiter, R. L. 1989. Impact of
The nonlinear modeling of drying induced stresses New Technology and Changing Practices in the
in porous bodies. In Advances in Drying, ed. A. S. Automotive Coatings Business. AIChE Symposium
Mujumdar. Washington: Hemisphere. Series on New Polymer Technology for Auto Body
Malcolm, B. R. 1985. The flow and deformation of Exteriors. pp. 33-36.
synthetic polypeptide monolayers during compres- Ruschak, K. J. 1974. The fluid mechanics of coating
sion. J. Coli. Inter! Sci. 104: 520-528. flows. PhD thesis, University of Minnesota.
Ruschak, K. J. 1976. Limiting flow in a pre-metered towards molecular electronics: a review. J. Malec.
coating device. Chem. Eng. Sci. 31: 1057-1066. Elec. 1: 3-17.
Ruschak, K. J. 1978. Flow of a falling film into a pool. Swalen,1. D. 1992. Molecular films. Ann. Rev. Mater.
AIChE J. 24: 705-709. Sci. 21: 373-408.
Ruschak, K. J. 1982. Boundary conditions at a liquid/air Tallmadge, J. A. 1971. A theory of entrainment for
interface in lubrication flows. J. Fluid Mech. 119: angular withdrawal of flat supports. AIChE J. 17:
107-120. 243-246.
Ruschak, K. J. 1985. Coating flows. Ann. Rev. Fluid Tallmadge, J. A. and Gutfinger, C. 1967. Entrainment
Mech. 17: 65-89. of liquid films. Ind. Eng. Chem. 59: 18-34.
Sakiadis, B. C. 1961. Boundary-layer behavior on Tam, S., Stolarski, H. K. and Scriven, L. E. 1992.
continuous solid surfaces. AIChE J. 7: 26-35. Defect-related stresses in drying films. AIChE Spring
Sartor, L. 1990. Slot coating: fluid mechanics and die National Meeting, New Orleans, LA, 1992.
design. PhD thesis, University of Minnesota. Tanguy, P., Fortin, M. and Choplin, L. 1984. Finite
Sartor, L. and Scriven, L. E. 1990. The progression element simulation ofdip coating, II: non-Newtonian
from dip coating to slot coating: What controls film fluids. Int. J. for Num. Meth. Fluids. 4: 441-457.
thickness? Presented at the Fall Annual Meeting of Tekic, M. N. and Jovanovic, S. 1982. Liquid
the AIChE, Chicago, IL. coating onto a rotating roll. Chem. Eng. Sci. 37:
Scherer, G. W. 1988. Aging and drying of gels. J. 1815-1817.
Non-Crystalline Solids. 100: 77-92. Tharmalingam, S. and Wilkinson, W. L. 1978. The
Schroeder, H. 1969. Oxide layers deposited from coating of Newtonian liquids onto a rotating roll.
organic solutions. In Physics of Thin Films 5, ed. G. Chem. Eng. Sci. 33: 1481-1487.
Haas. pp. 87-141. New York: Academic Press. Tredgold, R. H. 1987. The physics of Langmuir-
Schunk, P. R. 1989. Surfactant and polymer additives Blodgett films. Rep. Prog. Phys. 50: 1609-1656.
in coating and related flows. PhD thesis, University Tschoegl, N. W. 1962. Mathematical relations of
of Minnesota. torsion pendulum in study of surface films. Kolloid
Schunk, P. R., Hurd, A. J. and Brinker, C. J. 1992. Z. 181: 19-29.
Surface tension gradient effects in sol-gel dip coating. Tu, Y. and Drake, R. L. 1990. Heat and mass transfer
Presented at the 1992 Spring National Meeting of during evaporation in coating formation. J. Call.
the AIChE, 29 March-2 April, New Orleans, LA. Interf Sci. 135: 562-572.
Schunk, P. R. and Rao, R. R. 1994. Finite element Tuck, E. O. 1983. Continuous coating with gravity
modeling ofevaporation and condensation processes and jet stripping. Phys Fluids. 26: 2352.
in film and fiber formation. Int. J. Numer. Meth. Van Rossum, J. 1958. Viscous lifting and drainage of
Fluids. 18: 821-842. liquid. Appl. Sci. Res. A. 7: 121-144.
Scriven, L. E. 1988. Physics and applications of dip Walker, J. 1983. The amateur scientist: What causes
coating and spin coating. In Better Ceramics Through the 'tears' that form on the inside of a glass of wine?
Chemistry III, eds C. J. Brinker, D. E. Clark and Scientific American. May: 161-169.
D. R. Ulrich. pp. 717-729. Materials Research Wayner, P. C. 1982. The interfacial profile in the
Society, Pittsburgh. contact line region and the Young- Dupre equation.
Slatterly, J. and Givens, H. 1933. The shape of the J. Call. Interf Sci. 88: 294-295.
profile of a liquid film draining on a vertical, clean, Wayner, P. C. 1991. The effect of interfacial mass
wetted glass plate and the combined effects of transport on flow in thin liquid films. Call. Surf 52:
gravity, viscosity, surface tension and evaporation 71-84.
on the same. Trans. Roy. Soc. Canada. 27: 145. Wayner, P. c., Kao, Y. K. and LaCroix, L. V. 1976.
Soroka, A. J. and Tallmadge, J. A. 1971. A test of the The interline heat-transfer coefficient of an evapor-
inertial theory for plate withdrawal. AIChE J. 17: ating wetting film. Int. J. Heat Mass Transfer. 19:
505-509. 487-492.
Spiers, R. P., Subbaraman, C. V. and Wilkinson, Weast, R. C. 1974. Handbook ofChemistry and Physics.
W. L. 1974. Free coating of a Newtonian liquid 55th edn. CRC Press.
onto a vertical surface. Chern. Eng. Sci. 29: 389-396. White, D. A. and Tallmadge, 1. A. 1965. Theory of
Sternling, C. V. and Scriven, L. E. 1959. Interfacial drag out of liquids on flat plates. Chem. Eng. Sci.
turbulence: hydrodynamic stability and the Maran- 20: 33-37.
goni effect. AIChE J. 18: 231-240. Wilson, S. D. R. 1982. The drag-out problem in film
Sugi, M. 1985. Langmuir-Blodgett films - a course coating theory. J. Eng. Math. 16: 209.
SPIN COATING 14
Ronald G. Larson and Timothy J. Rehg
14.1 INTRODUCTION
such devices spin low-viscosity liquids to thick-
Spin coating is a batch process in which a liquid nesses from a few microns down to a few
film is spread by centrifugal force onto a rotating nanometers, with thickness nonuniformities no
substrate. A typical spin-coating system for greater than 1% (Moreau 1988). Solidified films
microelectronic wafers is shown in Fig. 14.1. At are obtained when the spin-coated liquid contains
high rotation speeds, around 1000 to 10000 rpm, polymers or colloids in a volatile solvent. Solvent
evaporation is greatly accelerated by the air flow
induced by high spinning speeds; thus, the
application, coating and drying of a film can be
accomplished in less than a minute. Because of
its speed, combined with its simplicity and low
cost, spin coating is unusual as a batch coating
process that is useful even in high capacity
industrial applications. Although in simplicity it
is bested by batch dip coating (see Chapter 13),
the uniformity and thinness of spin-coated layers
far exceed dip coated ones, and are unmatched
by all but relatively exotic methods such as
vacuum or Langmuir deposition. As a result,
spin coating has long been widely used in
high-volume production of advanced electronics
exhaust devices. It was used as long ago as the 1950s, for
example, to deposit phosphor onto the curved
glass surfaces of color television tubes.
Nowadays, spin coating is ubiquitous in the
microelectronics industry as a means ofdepositing
Figure 14.1 Schematic diagram of a typical spin polymer resist layers for photolithographic
coating system. Air is drawn down and across the processing of integrated circuits (Thompson,
wafer by centrifugal forces. The catch-cup traps solvent Wilson and Bowden 1983; Moreau 1988). Another
droplets that are flung from the wafer, preventing their
unwanted release into the environment. An exhaust
important modern application of spin coating is
suction assists this, and prevents droplets from the deposition of transparent or reflective inorganic
re-circulating and hitting the wafer, which would mar colloidal surface coatings on laser optical com-
the coating (from Bornside et al. (1993». ponents, such as highly reflecting mirrors.
London. ISBN 0 412 06481 2.
710 Spin coating
14.1.1 APPLICATIONS OF SPIN-COATED (Thomas 1986). This is especially true when the
FILMS desired optical component is a highly reflecting
mirror. In this application, one deposits layers
Polymeric resist films are used to delineate device of a material with high refractive index nh and
patterns of silicon wafers. Light is shone through thickness Dh = A./(4nh ), alternating with layers of
a mask to cast a fine pattern of light and shadow a material with low refractive index n. and
on the photosensitive resist that overlies a silicon thicknessD. = A./(4n.).In this way, Thomas (1986)
wafer. In areas exposed to light, photo-catalyzed has used spin coating to fabricate a laser mirror
reactions render the resist either more (positive (A. = 1.06 mm) with 18 alternating layers each of
tone) or less (negative tone) susceptible to colloidal alumina (D h = 185 nm, nh = 1.43)
subsequent removal during development. For and silica (D I = 217 nm, nl = 1.22), producing
positive resists, the exposed polymer is removed a surface with 99% reflectivity. Coating uniformity
during development by plasma etching or solvent is critical in these systems; small thickness
dissolution, while the unexposed polmer remains; variations induce losses that can accumulate to
the opposite is true for negative resists. In either significant levels when many optical components
case, the areas of substrate that remain covered are present. Unfortunately, the film uniformity
by resist are protected from attack by etchants, achieved by spin coating, although high, may
thereby transferring a pattern from the resist not always be high enough for these applications.
layer to the underlying silicon wafer. After the This topic is discussed further in Section 14.2.
etching is complete, the remaining polymer is The volatility of the coating solvent in which
removed, and the wafer is ready for further the resist polymer is dissolved or the inorganic
lithographic processing steps, including deposition colloidal particles are suspended is an important
of additional resist layers by spin coating. concern. Obviously, extremely low volatilities
Frequently, therefore, resist solutions must be are undesirable, because evaporation times will
spin coated onto wafers that contain micron- then be long. But very high volatility, which is
scale unevenness because of the previous etching usually correlated to low viscosity, can lead to
steps to which they have been subjected, a fact uneven film thicknesses, because of inertial or
with important consequences that we explore in evaporation-related Marangoni instabilities dis-
Section 14.3. cussed below. In addition, if the resist solution
Great attention is paid to the chemical design poorly wets the substrate, pinhole defects and
of resist polymers, as these must be highly other manifestations of dewetting can be the
photosensitive, so that a sharp change in the result; see Chapter 6.
reactivity ofthe polymer occurs with only a small
change in the total light exposure. With a highly
14.1.2 OVERVIEW OF THE SPIN-COATING
sensitive resist, sharp lines can be produced even
PROCESS
when the boundaries ofthe pattern ofillumination
are somewhat fuzzy, as is inevitably the case An excellent overview of the spin-coating process
because of wavelength limits to optical resolution. is provided by Bornside, Macosko and Scriven
Further discussion of the chemistry of these (1987) and by Scriven (1988). They describe four
highly specialized polymeric resists is beyond the stages of spin coating: deposition, spin-up, spin-
scope of this text. The motivated reader should off and evaporation. Deposition and spin-up
consult Moreau (1988). occur quickly; the substrate is usually brought
Inorganic colloidal coatings are important in to its final rotation rate within a fraction of a
high-energy laser-optic systems because they second. Spin-off and evaporation occur concur-
exhibit considerable resistance to laser damage rently; however, in the first seconds after spin-up,
and are less expensive to fabricate than are the the film is many microns thick, and the fluid has
traditional electron beam evaporated coatings little time to evaporate as it is flung from the
Spin coating over level substrates 711
disk into the 'catch-cup' shown in Fig. 14.1. why spin coating generally produces such highly
Later, the film (by then only a few microns thick) uniform films.
flows so slowly that it thins mainly by solvent Under the above assumptions, the equation of
evaporation. fluid motion in the film is simply
In the next section, we consider spin coating
over level substrates, both with and without av
2
2
-j1. ay 2 = pQ R (14.1)
evaporation, and discuss briefly the instabilities
that can degrade film thickness uniformity. In
where V is the radial velocity, j1. is the viscosity,
the subsequent section, the effects of substrate
and p is the density of the fluid; Q is the angular
unevenness on the leveling of the film profiles
rotation rate, and R is the radial position (distance
are discussed in detail, including the practical
from the axis of rotation). The boundary conditions
impact of departures from levelness on the
on equation (14.1) are V = 0 at Y = 0, and
performance of the film in microlithography.
av/ay = 0 at Y = H; these correspond, respect-
Finally, a brief summary of the state of under-
ively, to the no-slip and the stress-free boundary
standing of spin coating is offered.
conditions at the substrate (where Y = 0) and at
the film-air interface (where Y = H). Integrating
equation (14.1), we obtain a parabolic velocity
14.2 SPIN COATING OVER LEVEL profile:
SUBSTRATES
pQ 2 R 1 2
V = -j1.-(HY - "2 Y ) (14.2)
14.2.1 WITH EVAPORATION NEGLECTED:
THE ANALYSIS OF EMSLIE, BONNER AND The volumetric flow rate, Q, per unit length
PECK of film circumference is then
The earliest analysis of spin coating over a flat, H pQ2RH3
rotating, substrate is that of Emslie, Bonner and Q= Jo VdY= 3j1.
(14.3)
Peck (1958). Their analysis considers only
Newtonian fluids, and takes advantage of the The mass continuity equation, when combined
thinness of the film, compared with the radial with equation (14.3), implies that
expanse of the substrate, which justifies the use
ofa lubrication analysis. (For a detailed discussion aH = _ ~ a(RQ) = _ pQ2 ~ ~(R2 H 3)
of the lubrication approximation, see Chapter at R aR 3j1. R aR
8). Emslie, Bonner and Peck neglect all forces (14.4)
except centrifugal and viscous; thus, gravitational,
Coriolis, and capillary forces are ignored, as is Equation (14.4) describes the time evolution of
air-drag on the free surface. Neglect of these the film thickness profile over the spinning
forces is usually justified, except that for uneven substrate.
substrates capillary forces must be re-introduced; To solve equation (14.4), one must specify an
see Section 14.3. Less justifiable is the constancy initial film thickness profile. This initial profile
of fluid properties; this stipulation is derived is influenced by the method of dispensing liquid
from a completely unrealistic assumption: the onto the substrate, which is often not well
neglect of solvent evaporation. Neglect of solvent defined. Fortunately, however, Emslie, Bonner
evaporation renders the analysis unsuitable for and Peck showed that for Newtonian liquids,
quantitative predictions. Still, the model of Emslie, the initial nonuniformities tend to disappear
Bonner and Peck (1958) gives a simple estimate rapidly after start-up of spin coating. Figure 14.2
of the time required to spin a film down to a shows a typical calculation by Emslie, Bonner
given thickness and, more importantly, it shows and Peck: a film with an initially sinusoidally
712 Spin coating
1.2
T~O T~O
HJHo =1 + 0.15 cos aR 1.0 .05 0.05
0.8 0.2 0.2

tJ)
tJ)
0.8 C1l 0.6

C 0.8 0.8
0 0.3 ..><:
I~ t)
0.4
I 0.6 0.6 :.c
t-
3.2 3.2
0.4 0.2
~ n ~ 1.00 n ~ 0.95
LL 0
0.2 "0
C1l 1.2
~ 1.0
0 4 8 12 16 20
co
E
"-
aR 0 0.8
Z
0.6
Figure 14.2 Flattening of an initially oscillatory film
thickness profile during spin coating of a non-volatile 0.4
fluid at three different dimensionless times, (pQ2/31l)fj~ t.
In the above, H o(R) is the initial film profile, fj 0 is 0.2
n ~ 0.60 n ~ 0.20
the initial average film thickness, and IX is a constant a
controlling the wavelength of the initial film profile a 0.2 0.4 0.6 0.8 1.0 a 0.2 0.4 0.6 0.8 1.0
(from Emslie, Bonner and Peck (1958».
Normalized Radius
varying thickness quickly becomes nearly uni- Figure 14.3 Evolution of the thickness profile of an
formly thick as a result of spin-coating flow. The initially uniform, non-volatile, shear-thinning fluid
thickness becomes uniform, in part, because of film during spin coating. The solid lines are for the
the cubic dependence on thickness of the flow power law fluid, whose shear-rate-dependent viscosity
rate Qin equation (14.3). The flow rate in a locally is ,., = Ky·-I and y is the local shear rate, while the
thick area of the film is higher than in surrounding dashed line are for the Carreau model with
'1 = J.lo(l + ;'2 y2)(.-1)/2. Here n is the power law index.
thinner areas, and thus the thick area thins faster
The dimensionless time T is t/t char with
than the surrounding areas, making the film t char = (2n + 1)/n)(pQ2/C)R~-·H~+·)-I/•. The con-
thickness more uniform. Once the film has become stant C is given by K for the power-law model and
uniformly thick, with thickness Hi' the rate of by lloA·-1 for the Carreau model (reprinted with
uniform thinning of the film can be obtained permission from Jenekhe and Schuldt (1984); © 1984
from a simple solution of equation (14.4): American Chemical Society).
2
4 0 to the findings of Emslie, Bonner and Peck,
H = H.[1 + -p- H~t]-1/2 (14.5)
I 3Jl I Jenekhe and Schuldt found for Carreau fluids
that films that are initially nonuniform will tend
Acrivos, Shah and Petersen (1960) extended toward uniformity for long enough spinning
the analysis of Emslie, Bonner and Peck (1958) times. However, one important distinction from
to power-law shear thinning fluids. For a power- the work of Emslie, Bonner and Peck is that a
law fluid, the viscosity is infinite at the center of Carreau-type shear-thinning fluid that is initially
the disk (R = 0), because the shear rate there is uniform will tend to become nonuniform, and
zero; this leads to the unrealistic film thickness then return to a uniform state at long spinning
profiles shown in Fig. 14.3. Jenekhe and Schuldt times; see Fig. 14.3. The departure from uniformity
(1984) considered the more realistic Carreau occurs because the shear rate is higher near the
model for shear-thinning fluids, whose viscosity edge of the substrate than it is near the center.
approaches a constant at low shear rates. Similar Hence the viscosity near the edge is lower than
at the center, and thus the film will thin more et al. 1993). The flow pattern generated by an
quickly near the edge than at the center. As the infinite spinning disk was analyzed long ago by
film thins, however, the average shear rate von Karman (1921). Under laminar flow condi-
decreases, eventually becoming low enough tions, there is an Ekman momentum boundary
everywhere that the coating solution behaves as layer of thickness '" J(vG/fl) ~ 0.25 mm, for a
a Newtonian fluid, and Emslie, Bonner and typical rotation speed (2000 rpm). Here
Peck's prediction of a uniformly thick film is vG = 0.16cm 2/s is the kinematic viscosity of air.
recovered at long times. Above this boundary layer there is a
Whether a spin-coated film will in the end be downward axial flow of air with velocity ~
uniform depends on the magnitude of non- 0.89J(v GO) ~ 5 cm/s.
Newtonian effects, the time period over which Evaporation of the volatile solvent produces
the film returns to uniformity, and the time a mass-transfer boundary layer whose thickness
period over which the film ceases to flow due to is similar to that of the momentum boundary
solvent evaporation. Evaporative effects are taken layer, when, as usual, the Schmidt number
up in the next section. Sc == vG/D G is near unity, where DG is the
gas-phase diffusivity. Most importantly, the
thicknesses of these boundary layers, and hence
14.2.2 THE EFFECTS OF EVAPORATION
the evaporation rate, are predicted to be inde-
As discussed in the Introduction (Section 14.1) pendent of the radial position. In practice, of
to this chapter, the spin-coated liquid usually course, some nonuniformity arises from the finite
contains a volatile solvent that evaporates during radius of the substrate and the air flow pattern
spinning, leaving behind a thin solid film. As the in the catch-cup and the exhaust; see Fig. 14.1.
analysis of Emslie, Bonner and Peck does not However, a finite-element analysis ofthe air flow
account for evaporation, it cannot, for example, and mass transport in the typical spin-coating
predict the final dry film thickness. Hence, many device shows that the evaporation rate is
practitioners have relied on empirical correlations remarkably uniform and insensitive to the inlet
between final film thickness H f and spinning and outlet flow conditions (Bornside et al. 1993).
speed O. These correlations are typically expressed The high degree of uniformity in the evaporation
as a power law, rate leads to a similar uniformity in the final film
thickness. This uniformity, as mentioned in the
(14.6) Introduction (Section 14.1), is responsible for the
The exponent b usually is close to 0.5 (Sukanek commercial usefulness of spin coating.
1985), but higher values are sometimes found; Nonuniformities of the film thickness can
the most notorious example is that of Daughton occur very near the edge of the spinning wafer,
and Givens (1982), who report b = 0.81 for a because of higher evaporation rates caused by
solution of polyamic acid. The correlation (14.6) the combination of both radial and axial flow
is adequate for many practical purposes, although (Bornside et al. 1993). The result is a thick edge
the constants K and b must be measured for each bead confined to a few unused millimeters at the
spin coating fluid. wafer's edge, which is of no practical consequence.
Solvent evaporation couples liquid-phase The uniformity of the evaporation rate across
diffusive transport ofsolvent toward the interface the substrate was intuited years ago by Meyerhofer
to gas-phase convective mass transport ofsolvent (1978). He modified the analysis ofEmslie, Bonner
away from the interface. The solvent vapor is and Peck by including in the mass balance
carried away in a stream of air that is drawn equation (14.4) a uniform solvent evaporation
down axially toward the surface and driven rate E; thus
radially across it by a 'centrifugal pumping oH = _ 2p02 H3 _ E (147)
action' created by the spinning substrate (Bornside ot 3J1. .
714 Spin coating
where E is in units of solvent volume evaporated and /10 == /1(C o)' The concentration is assumed
per unit area per unit time. If the film is thin to depart from Co only after the transition to
enough that the solvent concentration remains evaporation-dominated thinning.
uniform throughout the depth of the film, a mass If the air flow remains laminar, and if the
balance equation for the nonvolatile solute can vapor-phase resistance to mass transfer dominates
be written: the liquid-phase resistance, then for an infinite
2H3 disk, the von Karman boundary-layer flow implies
~(CH) = -C~ o(RQ) = _C 2 0. (14.8) that the rate of evaporation E depends on the
at R oR 3/1(C) spinning speed 0. as
where C is the volume fraction of solute in the
E = k0. 1 / 2 (14.11)
film, and equation (14.4) has been used.
Equations (14.7) and (14.8) can be solved for where k is a constant determined by the coating
a specified initial condition to obtain C(t) and solvent; an equation for it is given shortly. With
H(t), if the dependencies of the viscosity /1 and equation (14.11), equation (14.10) becomes (Meyer-
evaporation rate E on concentration Care hofer 1978):
known. Meyerhofer measured the dependence
/1(C) and fit this dependence to a simple power Hf = (~r/3 k 1/ 3 Co(1- CO)-1/3p-l/3/1~/30.-1/2
law; and he took E to be constant in time as
well as position. Then, integrating equations (14.12)
(14.7) and (14.8), Meyerhofer found, not surpris-
ingly, that for typical spinning speeds, the final The proportionality H f OC 0.- 1/2 predicted by
film thickness H f is nearly independent of the equation (14.12) agrees with Meyerhofer's data,
initial film thickness as long as the initial thickness and data obtained in other spin-coating studies
is at least as large as 100 Jlm or so. The numerical with polymer resist solutions (Sukanek 1985).
solution confirmed that early in spin coating, Also agreeing with experiments is Meyerhofer's
when the film is thick, flow dominates the prediction that the time to dry the film is
thinning, but later, when the film has become proportional to 0. -1. The measured concentration
much thinner, flow is slow, and thinning is due dependence of H f also agrees with that predicted
mainly to evaporation. by equation (14.10), if the measured concentration
Meyerhofer showed that if there is an abrupt dependence of the viscosity is used, and the
transition from thinning by flow only, to thinning constant k is obtained by a fit to one of the data
by evaporation only, then the final film thickness points.
can be estimated analytically. The transition Bornside, Macosko and Scriven (1991), see
point is defined to be the point at which the rate also Lawrence (1988, 1990), recently showed that
of solvent loss by flow equals the rate of loss by k can be computed a priori; for vapor-phase-
evaporation; that is limited mass transfer, k is given by
(1 - C)20.2pH~
-'-----------:--.:.---..:: = E
3/1
( 14.9) k= C~op)(p;~) (14.13)
where H 0 is the film thickness at the transition where p* is the vapor pressure ofthe pure solvent,
between flow-dominated and evaporation- M is its molecular weight, Do is its diffusivity in
dominated thinning. This gives for the final film the vapor phase, and c ~ 0.55 is a prefactor with
thickness: a weak dependence on Sc. Equation (14.13) yields
a value of k ~ 10- 5 cm s - 1/2 for typical spin-
3/1oE )1/3 coating solvents. Using equation (14.13), Bornside,
H f = CoHo = Co ( 2(1 _ C )p0.2 (14.10)
O Macosko and Scriven were able to predict a
where Co is the initial concentration of solute, priori the film thicknesses measured by Flack et
al. (1984) for PMMA polymer films spun from 10-3

chiorobenzene solvent.
The assumptions underlying Meyerhofer's IrJl
10-4
rJl
simple prediction, equation (14.12), then, are that Ql
c
10.5
there is an abrupt transition from flow with "'"
0
:c
negligible evaporation, to evaporation with I-
negligible flow, and that this occurs when the ~ 10-6
u:: (a)
rate of thinning by flow is equal to the rate of
10-7
thinning by evaporation; and that the rate of
10- 1
evaporation is controlled by vapor phase con- Ul
vective mass transport, which, in turn, is set by ~ 10-3
Ql
a boundary-layer mass transport rate proportional iii
a: 10-5
to 0 1 / 2 • The qualitative correctness of these Ol
c
idealizations, and hence of equation (14.12), for 'cc 10-7
spin coating ofresist solutions, has been supported :c
I- 10-9
not only by experiments, but also by analyses ~
more complete than that of Meyerhofer. Ohara, u:: 10- 11
Matsumoto and Ohashi (1989), in particular, 10°
start from the general two-dimensional equations c
.Q
for axisymmetric fluid, mass and energy transport ~
C
in the laminar liquid and gas phases, and reduce Ql
0
these to one-dimensional equations by using the c 10- 1
0
0 At the Bottom
von Karman similarity forms for the radial Ql
dependencies. They then integrate the one- :5

(5
(f) (c)
dimensional equations numerically to obtain the
10-2
time-dependent concentration and velocity fields 10-3 10-2 10-1 10° 10 1 102
in the liquid and gas phases. Figure 14.4 shows Time(s)
the numerical predictions obtained by Ohara,
Matsumoto and Ohashi (1989) for an experiment Figure 14.4 Simulated spin coating of a solution of
of Flack et aI., in which a 9% solution ofPMMA PMMA in chlorobenzene onto an infinite disk spinning
in chiorobenzene was spin-coated at 2000 rpm. at 2000 rpm. (a) The film thickness H and the solute
portion of that thickness H sl ' (b) The contributions of
Figure 14.4b shows that at early times, convective
convection and evaporation to the overall rate of
thinning dominates, while the rate of evaporative thinning. (c) The solute concentration at the free
thinning is roughly constant, and much slower surface of the liquid and at the bottom of the liquid
than convective thinning. Roughly 10 s after the film, next to the substrate (from Ohara, Matsumoto
start of spin coating, the rate of convective and Ohashi (1989».
thinning quickly drops below the rate of evap-
orative thinning, and only after this occurs does
the solute concentration rise greatly above its scaling, He OC 0- 1 / 2 , obtained in Meyerhofer's
initial value. Thus, these simulations support simple anlaysis. The numerical simulation therefore
Meyerhofer's assumption that the transition from supports the assumptions of Meyerhofer's sim-
flow-dominated to evaporation-dominated trans- plified analysis in almost every respect. The only
port is abrupt. Note from Fig. 14.4c that the significant failure of the simple theory is a failure
solute concentration remains roughly uniform that is shared by the numerical solution of Ohara,
throughout the liquid film, as assumed by Matsumoto and Ohashi, namely that when the
Meyerhofer. Ohara, Matsumoto and Ohashi initial polymer concentration is low, 2% or so,
also found for their numerical solution the same the theory does not predict an experimentally
716 Spin coating
observed deviation from the scaling hr oc 0 -1/2 solutions of polyamic acid. Thus, if spinning is
(Meyerhofer 1978; Ohara, Matsumoto and Ohashi continued until evaporation becomes the dominant
1989). cause of thinning, Meyerhofer's simple analysis
Other analyses ofspin coating of resist solutions is typically qualitatively, and frequently quanti-
have been presented; in rigor and complexity tatively, correct for spin coating of polymer resist
these are intermediate between Meyerhofer's solutions.
simple analysis and the numerical solution of For the spin coating of colloidal suspensions,
Ohara, Matsumoto and Ohashi. Notable are the Meyerhofer's analysis is not as satisfactory. Rehg
analyses of Sukanek (1985) and of Lawrence and Higgins (1992) examined the spin coating of
(1988,1990). These, like the study of Ohara, colloidal suspensions with a model essentially
Matsumoto and Ohashi, show that typically the identical to that of Ohara, Matsumoto and
rate of thinning by flow drops quickly during Ohashi (1989). This numerical study shows that
the early stages of spin coating so that the for colloidal suspensions Meyerhofer's analysis
cross-over to evaporation-dominated thinning is has two significant failings; namely, it under-
abrupt, and when it occurs, the solute concen- predicts the coated film thickness, and it does
tration is still near its initial value. As the volume not capture the correct concentration dependence
of solute that is finally trapped on the substrate of the film thickness. These departures are
is largely determined at this cross-over point, especially evident for colloidal suspensions with
more complex effects that might occur later in solids loading near the solidification point. Near
spin coating have little or no effect on the final this point, the viscosity is strongly concentration
film thickness. dependent, and even the small increase in solids
When spinning is terminated before the cross- concentration that occurs early during spin coating
over, and the film is allowed to dry in the absence changes the solution viscosity enough to invalidate
of spinning, both Lawrence and Zhou (1991) and Meyerhofer's assumption of a constant viscosity
Yonkoski and Soane (1992) have found theor- during the convection-dominated stage of spin
etically that equation (14.6) still holds for the coating. As a result of this early increase in film
final film thickness, although with b > 0.5. viscosity, the convective thinning decreases more
Yonkoski and Soane conducted experiments rapidly, and becomes small compared with
with a polyimide photoresist to verify this predicted evaporation sooner than predicted by Meyer-
result. They found best fits to equation (14.6) for hofer's analysis; the result is a film that is thicker
b = 0.5 when the spinning time was 200 seconds, than Meyerhofer's analysis would predict. Never-
and b = 1 when the spinning time was reduced theless, the numerical analysis does support
to 30 s. Skrobis, Denton and Skrobis (1990), in spin- Meyerhofer's assumption that thinning abruptly
coating experiments with a thermosetting isoimide changes from being convection to being evapor-
oligomer, found that the exponent b depended on ation dominated and the oft observed scaling
the coating-solution solvent when the spinning H r oc 0- 1 / 2 is still obtained. In fact, equation
time was held to 15 s. They found b = 0.57 for (14.12) should still hold if one could somehow
the solvent of highest volatility, and b = 0.97 for determine the solution viscosity }.t at the transition
the one with lowest volatility. For the latter, between convection and evaporation dominated
spinning was terminated well before the film had thinning, and use this to replace the initial
become thin enough for the cross-over to solution viscosity }.to.
evaporation-dominated thinning to occur. These When the viscosity of the coating solution is
findings suggest an explanation for the Daughton both shear thinning and sensitive to small changes
and Givens' exponent of 0.8 mentioned earlier, in concentration, the coated film will be susceptible
as in their experiments spinning was terminated to nonuniformities. This is shown theoretically
after only 20 s. A larger spinning time might well by Sukanek (1991) and Britten and Thomas
have yielded an exponent closer to 0.5, for their (1992), who extended Sukanek's Newtonian model
(Sukanek 1985) to the cases of shear-thinning concentration in the film decreases only slightly
polymer solutions and shear-thinning colloidal during the stage when convective thinning is
suspensions, respectively. Britten and Thomas important (see Fig. 14.4). Thus, it is likely that
(1992) also demonstrated this phenomenon in a shear-thinning film will have enough time to
spin-coating experiments with colloidal become uniform before the flow stops (see Fig.
alumina/water suspensions. Their experimental 14.3), a conclusion supported by the universal
results, shown in Fig. 14.5, are in qualitative experience of practitioners that spin-coated
agreement with the theory; the film thickness polymer films are uniform. Indeed, Lawrence and
decreases from the substrate center to the edge. Zhou(1991), in an extension of Lawrence's earlier
The observed quantitative disparity is due to work (1988,1990) on shear thinning systems,
some uncertainty in the shear-rate-dependent conclude that nonuniformities will be present
solution viscosity. Despite this disparity, the only for large Deborah numbers De, where De ==
qualitative agreement supports the inference drawn 0. 2 PH 0 R s / J.lo Yc' and R s is the substrate radius
in Section 14.2.1 that nonuniformities in film and yc is the shear rate above which shear
thickness occur when the cross-over from non- thinning is important. For large Deborah numbers,
Newtonian to Newtonian flow occurs too late De » 1, shear thinning will persist even at very
to restore uniformity to the film before evaporation low shear rates and the film may not become
renders the film immobile. (As mentioned in Newtonian before convective thinning becomes
Section 14.2.1, a cross-over to Newtonian flow unimportant.
occurs when the film is thin enough that the As non-Newtonian effects are generally not
shear rate drops into a regime of Newtonian important, unless the coating solution is close in
behavior). concentration to a gel or solidification point, the
For polymer solutions, the absence of experi- equation for film thickness, equation (14.12),
mental data showing nonuniformities induced should be adequate in many commercial settings
by shear thinning is consistent with our earlier to predict the spinning speed required to obtain
discussion. The analysis of Ohara, Matsumoto a desired film thickness. This is rarely done,
and Ohashi (1989) shows that the solvent however. In practice, it is usually easier to
spin-coat resist solutions at various spinning
speeds and measure the resulting film thicknesses
1.00 to obtain a correlation between film thickness
.,
U'l
U'l
and spinning speed for a given spin-coating
C x solution, as discussed at the beginning of this
'"u'" 0.95 0 0 0
F + + + -7
section.
E +
0.90
C Meas. Calc.
-0
750 RPM 14.2.3 STABILITY OF SPIN COATING
0
N
x
~ 0.85 0 1500 RPM Spin coating is, by and large, a rather stable
...E
0 + 2500 RPM
coating method, as evidenced by the reliably
;Z uniform films that it produces. However, as most
0.80
0 0.2 0.4 0.6 0.8 1.0 practitioners of spin coating have experienced,
ormalized Radiu
surface features are occasionally exhibited in the
spin-coated films that render the coatings unac-
ceptable for their intended purpose. Three principal
Figure 14.5 Calculated and measured film thickness features characterizing these unacceptable films
profiles, normalized with the film thickness at the
substrate center, for coatings deposited from colloidal are cloudiness (Lai 1979), orange peel (Lai 1979),
alumina/water suspensions (from Britten and Thomas and radial striations (Elliott and Hockey 1979).
(1992)). Cloudy films, characterized by an opaque ap-
718 Spin coating
pearance, seem to be produced when the solvent 1979; Leers 1981; Frasch and Saremski 1982),
and polymer are only partially miscible; when and colloidal films (Rehg 1992; Rehg and Higgins
solvent evaporation drives the polymer concen- 1992). Figure 14.6 shows a photograph of a
tration up beyond its solubility limit in the colloidal coating exhibiting radial striations. At
solvent, the polymer precipitates. Selection of a the substrate center, where the magnitude of the
solvent/polymer system that does not phase bulk flow is small, the striations give way to
separate is an obvious remedy for this condition. polygonal cells. Daniels et al. (1986), Rehg (1992),
Lai suggests that, even with this remedy, water and Rehg and Higgins (1992) all postulate that
contamination of the coating solution can lead the striations and polygonal cells are a consequence
to phase separation, especially if the solvent has of a Marangoni flow, the secondary flow driven
a higher volatility than water. Preferential by temperature-induced surface-tension gradients
evaporation of the solvent will concentrate the that spontaneously develop in volatile thin films
residual water, thereby increasing the likelihood (see Chapters 2, 6 and 13). The structures of both
of phase separation. Therefore, selection of a
solvent in which water is insoluble is also a
valuable strategy for avoiding cloudy films.
Both orange peel and radial striations are less
well understood. To date, formal stability analyses
have not been undertaken that clearly indentify
the mechanisms that induce these surface features.
Nevertheless, experimental studies have identified
means of avoiding them. Orange peel, which is
characterized by randomly positioned surface
dimples, has been examined by Lai (1979) and
Spangler, Torkelson and Royal (1990) in spin
coating. (Orange peel also occurs in other coating
flows, as discussed in Chapter 6). Both of the
above experimental studies find that the incidence (a)
of orange peel correlates with solvent volatility;
nonuniform spin coated films are obtained when
the solvent has a low boiling point, and uniform
films are obtained for high-boiling solvents.
Spangler, Torkelson and Royal also found that
uniform films can be spin coated if the coating
solution is viscous enough, even for highly volatile
solvents. In practice, photoresist producers use
relatively high boiling solvents in their formula-
tions, such as xylene (Tb = 140°C) and orange
peel is rarely observed nowadays in commercial
spin coating. It is noteworthy that water is only
sparingly soluble in xylene and that the vapor
pressure of water is nearly four times that of
xylene at room temperature; thus xylene-con-
Figure 14.6 Surface topography of a film spin coated
taining coating solutions can also be expected from a colloidal silica/ethanol suspension. (a) Radial
to avoid precipitation-induced cloudiness. striations on a wafer, 6cm in diameter. (b) Polygonal
Radial striations have been observed in both cells on an enlarged view of the substrate center (from
spin-coated photoresists (Elliott and Hockey Rehg (1992)).
the polygonal cells and the striations are consistent ratio of surface-tension-gradient forces to viscous
with those of other film flows that are susceptible stabilizing forces; see Chapter 6. Here IX is the
to Marangoni convection. Quiescent films have thermal diffusivity, and d T is the temperature
long been known to develop polygonal convection difference between the top and the bottom of the
cells (Pearson 1958), and films draining under film.
gravity have been observed experimentally - and For low-viscosity films, the Marangoni number
predicted theoretically - to develop longitudinal is initially large, so the absence of striations in
roll cells (Ludviksson and Lightfoot 1968), due this case (see Fig. 14.7) is less obvious. Rehg
to the Marangoni instability. In spin coating, (1992) and Rehg and Higgins (1992) argue,
surface-tension gradients can be induced either however, that films spin coated from low-viscosity
by temperature gradients (due to evaporative coating solutions are stabilized by the dynamic
cooling) or by concentration gradients. thinning process; that is, even if a film is initially
Rehg (1992) demonstrated experimentally that susceptible to Marangoni convection, the striations
the spin coating conditions that give rise to do not grow to appreciable magnitude before
striations are consistent with Marangoni con- the flow is restabilized by viscous forces that
vection. In particular, he found that films spin become increasingly important as the film rapidly
coated from both high- and low-viscosity poly- thins. Thus, while J.l is small, the film thickness
methylmethacrylate/l,2-dichloroethane coating H quickly becomes small, and the Marangoni
solutions are free of striations; Figure 14.7 shows number is therefore small as well. An analogous
that only solutions of intermediate viscosity phenomenon, the formation and suppression of
show the striations. The lack of striations for inertial circumferential waves, has been predicted
high-viscosity solutions is readily explained: when by Reisfeld, Bankoff and Davis (199Ia). Using a
the coating solution viscosity is large, Marangoni linear stability analyis, they show that spin-coated
convection is suppressed by large viscous forces. films are unstable to formation of these waves
Recall that the Marangoni number, defined as during spin up; but as the film thins, the stabilizing
Ma == (oa/oT) HdT/(J.lIX), is a measure of the effect of viscosity intervenes, and the disturbance
decays. (To the authors' knowledge, in experi-
mental spin-coated films, circumferential waves
4500 r - - - - . . , . . . . - - - - - - - - , - - - - - - - , have never been observed. For further discussion,
see Reisfeld, Bankoff and Davis (199Ia,b).)
Other data have been reported that support
3500
Rehg's (1992) and Rehg and Higgins' (1992)
assertion. Rehg (1992) performed spin coating
E
g experiments with a colloidal silica/ethanol coating
c: 2500 solution known to produce polygonal convection
cells in quiescent films and striations in films spin
coated at 2000 rpm. Films free of striations were
1500 ' - ~ ____' obtained when the substrate rotation rate was
0.0001 0.001 0.01 0.1 10 increased to 3500 rpm. When the substrate
~Io (Pa s) acceleration rate was reduced from 10 krpm/s to
1 krpm/s and the rotation rate was maintained
Figure 14.7 'Operability diagram' showing conditions at 3500 rpm, striations were again evident in the
where films free of striations can be spin coated from coated film. Slow rotation and slow acceleration
polymethylmethacrylate/l ,2-dichloroethane solutions are known to reduce the rate of film thinning
whose viscosities 110 were controlIed by varying the
(Rehg and Higgins 1988), thereby prolonging the
polymer concentration. The unshaded region represents
conditions yielding striation-free coatings (adapted spinning time before the stabilizing effect of
from Rehg (1992». viscosity bcomes important and dampens Maran-
720 Spin coating
goni flow. High spinning speeds (Leers 1981; Measures that lend themselves to rapid thinning
Daniels et al. 1986) and high acceleration rates of the film (low viscosity, rapid acceleration and
(Frasch and Saremski 1982) are both found to rapid rotation rate) are not universal in their
attenuate striations in experiments with photo- ability to eliminate the instability. When large-
resists although, for the particular solutions used viscosity coating solutions are used, only thick
by Leers and Daniels et ai., the striations could films are attainable. Health and flammability
not be completely eliminated. Both Elliott and issues make it undesirable to saturate the
Hockey (1979) and Daniels et al. (1986) report atmosphere with solvent, unless shields (Chiba
that striations can be suppressed by saturating et al. 1982), air barriers (Peugh and Ward 1984),
the ambient environment with the spin-coating or other means of gas handling are used; these,
solvent. This reduces the rate of solvent evapor- of course, add complication and expense. Sur-
ation and therefore decreases both the temperature factants have been shown to suppress striations
and concentration gradients that can induce (Daniel et ai. 1986; Rehg 1992), as shown in Fig.
Marangoni convection. Rehg and Higgins (1992) 14.9, but their use may not be tolerable in all
used their model to reproduce qualitatively these applications (i.e., high energy laser systems).
experiments and found, without exception, that Fortunately, practitioners have found that a
the instantaneous Marangoni number is reduced more robust solution than these is simply to use
when the spin coating conditions are changed to solvents with low volatility. Such solvents generate
those less likely to yield striations; see Fig. 14.8. less severe temperature and concentration gradi-
While these results give insight into the ents that can drive the Marangoni secondary
mechanism that induces striations, they do not flow. In photoresist systems, at least, striations
provide practical means of avoiding them. are rarely observed nowadays.
06
05
A
II~c
r-
04
<l :l.
~I~ 0.3
II
ro
::?: 0.2
01
0
0 20 40 60 80 100
1: =tQ
Figure 14.8 The calculated instantaneous Marangoni number versus dimensionless spinning time. The coating
solution is a suspension of colloidal hard spheres in ethanol. Spinning conditions: (A) Q = SOOO rpm, 2000 rpmfs
substrate acceleration, 0% saturation of the atmosphere with ethanol; (B) Q = SOOO rpm, impulsive substrate
acceleration, 0% saturation of the atmosphere with ethanol; (C) Q = 10000 rpm, impulsive substrate
acceleration,O% saturation of the atmosphere with ethanol; (D) Q = SOOO rpm, impulsive substrate acceleration,
30% saturation of the atmosphere with ethanol (from Rehg (1992), and Rehg and Higgins (1992); (A), (B) and
(D) reproduced with permission ofthe American Institute ofChemical Engineers, © 1992 AIChE. All rights reversed).
Spin coating over uneven substrates 721
uniformity of the spin coated film to be only

'\ slightly degraded by the presence ofthese Ekman
.
4~4\()
" '.
: ~, no surraClant spirals; the coating thickness only varied by 0.2% .
/) \/\
,
When larger substrates are coated, such as in
the spin coating ofmagnetic media, film uniformity
...
" '
is degraded by the gas-phase turbulence; the spin
coated film exhibits 'spokes' that prevent the
surfaCI~Ult magnetic head from traveling as smoothly over
'. , the medium as is desirable (Scheffel and Richter
.
,, , 1978; Hagan, Maloy and Wilke 1986). Scheffel
\: and Richter discovered that the inertial instability
is suppressed if a co-rotating, parallel plate,
:!OO 400 600 >«Xl 1000 12(XI
called an air barrier, is positioned 2 to 20 mm
position (rnicrons)
above the substrate to be coated. With this
method, the gas flow between the air barrier and
Figure 14.9 Film profiles with and without surfactant
in a 14 wt% polymethylmethacrylate/l,2- the substrate is then largely just solid-body
dichloroethane solution spin coated at 2000 rpm. The rotation. Hagan, Maloy and Wilke found that a
coating solution that produced the striation-free film stationary barrier, combined with the use of
contained O.14vol.% FC-431 surfactant The film helium as the ambient gas adequately suppresses
profiles were measured by profilometry (3M Corpor- the 'spoke' instability. Because the kinematic
ation; from Rehg (1992». viscosity of helium is nearly eight times that of
air, the Reynold's number is approximately eight
Another instability, usually a fairly benign one, times smaller, and the critical radius beyond
occurs when the vapor phase Reynolds number which the gas flow is turbulent is correspondingly
Re == R 2 Q/v G becomes high enough, Re ~ lOS, larger, when helium is substituted for air.
that laminar air flow is no longer mantained
(Wahal et al. 1993, and references therein). For
14.3 SPIN COATING OVER UNEVEN
a large disk spinning at high speed Q, there will
SUBSTRATES
then be a critical radius R 2 such that for R > R 2 ,
the flow is turbulent Laminar flow is found only As discussed in Section 14.2, spin coating tends
whenR < RI.ForintermediateR,R I < R < R 2 , to produce films of uniform thickness; hence on
laminar flow is disrupted by three-dimensional flat substrates the films are flat On an uneven
spiral vortices known as Ekman spirals. Experi- substrate, however, the surface of the film tends
ments have shown that the Reynolds numbers not to be flat, but to conform to the topography
corresponding to R 2 and R I are Re 2 ~ 3 x 10 5 of the underlying substrate; see Fig. 14.10.
and ReI ~ 105 . Unevenness in the film-air surface can have
In a spin-coating apparatus using smoke trails deleterious consequences in further lithographic
for visualization, Wahal et ai. (1993) observed processing steps. In particular, as the sizes of
Ekman spirals, with negative winding direction; microelectronic devices are decreased to submicron
i.e., opposite direction to those reported by dimensions, sharper focusing is required of the
others. For these spirals of negative winding, radiation that exposes the resist film; and this
WahaletaUoundRe l ~ 0.2 x 105. Linear stability cannot always be obtained over all areas of an
analyses predict that spirals of either winding uneven surface. Thus, it is desirable to level, or
direction can occur, but the negative ones should 'planarize' the polymer film, either during, or
occur at the lowest Reynolds number, namely at after, spin coating (Rothman 1980; Burggraf
Re ~ 0.037 x 105 . In their experiments with 6-inch 1986; Stillwagon, Larson and Taylor 1987).
diameter substrates, Wahal et ai. find the Leveling is also required in the etch-back
722 Spin coating
SOLVENT
Adams and Capio 1981; White 1985a). However,
EVAPORATION the degree of planarization depends not only on
TOPOGRAPHIC
>50fLm SPIN
COAT ~RESIN
$ CONFORMAL
the width and shape of the feature, but also on
the surrounding topography; closely spaced
SUBSTRATE~ RESIN COATING features are planarized better than isolated ones,
other things being equal (Rothman 1980, White
BAKE
1985b; Stillwagon and Larson 1988). White (1985b)
~SOLID found that planarization and the film profiles
over line features depend on the orientation of
Figure 14.10 Stages of spin coating over uneven
those lines on the wafer, and on their distance
substrates. Thin conformal coatings are produced over from the center of rotation. Planarization also
features that are wider than about 50 11m; these depends strongly on the properties of the resist
coatings can be leveled to some extent during the solutions; in particular, leveling is enhanced by
baking step, which melts the polymer film, allowing increasing the concentration of polymer in the
reflow to occur (from Stillwagon and Larson (1990)). resist, and by decreasing the polymer molecular
weight (Rothman 1980; Bassous et al. 1983;
method (Sato et al. 1973; Adams and Capio 1981) LaVergne and Hofer 1985; Stillwagon and Larson
for producing inter-layer metal contacts. In this 1992). Resist solutions whose viscosity increases
scheme, metal pillars are formed on top of a rapidly as solvent evaporates;' planarize poorly.
metal layer. Onto this uneven substrate, a Thicker films planarize better. However, for a
conformal dielectric layer is spin coated. A resist given thickness of resist, better planarization is
overlayer is then added, which is planarized. The achieved if the resist is applied in several layers,
resist and dielectric are then etched by a process rather than just one (Bassous et al. 1983).
that removes both at the same rate, eventually White (1985a,b) has developed an empirical
exposing the metal pillars, which then jut out method for estimating film profiles and planar-
from the remaining dielectric layer. These are ization of resist films over complex topologies.
used to form contacts to a second metal layer, He measures the degree of planarization of a
which, apart from the contacts, is insulated from given resist over a series of isolated features of
the first layer by the dielectric. In this etch-back varying widths under given spinning conditions,
process, both a conformal and a planarized layer and thus obtains an empirical correlation between
must be deposited; to achieve both types of film thickness and feature width. Using a Fourier
coverage over the same substrate (i.e., a conformal transform, he then represents the uncoated surface
layer in one step of the process and a planarized topography by a two-dimensional series of sine
one in another step), the factors that control the and cosine functions. The coated surface top-
leveling process must be understood and con- ography is obtained by multiplying the coefficients
trolled. of the sine and cosine functions in the series for
In a series ofspin-coating experiments, Rothman the uncoated substrate by weighting parameters;
(1980) and others (Adams and Capio 1981; White these weighting factors are obtained from the
1985a,b; Bassous et al. 1983; Wilson and Piacente measured correlation between film thickness and
1986; Stillwagon and Larson 1988, 1990; Peurrung feature width. The weighting factors for long
and Graves 1991) have established the conditions wavelength topographies (> 50 Jlm) are near unity,
required to maximize planarization. These ex- implying little planarization for these features,
periments, reviewed by Stillwagon (1987), show while for short wavelengths, the weighting factors
that fine features are more readily planarized; are close to zero, implying nearly perfect planar-
films over features smaller than 10 Jlm are typically ization. In this way, White obtains reasonable
planar, while those over features wider than predictions of the film profiles over complex
50 Jlm are often quite conformal (Rothman 1980; two-dimensional topographies. This empirical
method needs to be supplemented, however, by solvent volume fraction of perhaps 10-20%

a more fundamental approach to provide under- remains in the film (Shaw, Frisch and Dill 1977).
standing of the role of resist fluid properties or The residual solvent is driven off by a baking
processing conditions on planarization. step after spin coating; the film shrinkage produced
A more fundamental approach to planarization by solvent evaporation during the baking step
is taken in the lubrication analyses of Stillwagon can reduce the degree of planarization of the film
and Larson (1988; 1990; 1992), of Brodkorb, (Schiltz, Abraham and Dechenaux 1987). The
Kohler and Ritzel (1989), and Sukanek (1989). baking step can also melt the dried polymer and
These theories, which follow earlier work on the enhance planarization (Fig. 14.10), as will be
leveling of paints by Smith, Orchard and Rhind- discussed in more detail below.
Tutt (1961) and by Overdiep (1977), account for If, as often assumed, the rate of evaporation
the influence of capillary forces on the film remains uniform across the wafer, then thicker
profile. The theories have several simplying portions of the film, such as those over planarized
approximations in common, including the as- features, will retain higher solvent concentrations
sumptions ofuniform rate ofsolvent evaporation, than surrounding thinner portions. Although
film properties that are uniform throughout the vertical diffusive equilibrium is maintained
film thickness, isothermal conditions, neglect of throughout most of the spin coating process
gravitational forces, and a shear-rate independent because of the thinness of the film, the degree to
viscosity. which lateral diffusion occurs over a feature
Most importantly, these analyses use the depends on the width of the feature. For features
lubrication approximation (see Chapter 8), which less than about 10 Ilm wide, lateral diffusion
is valid when WID;;;: 10, where W is the width cannot be neglected. Bornside (1990) has simulated
of the feature and D its depth. Under the both lateral and vertical diffusion over fine
lubrication approximation, the vertical component grooves with W ~ 5 Ilm during the late stages of
of velocity is taken to be negligible; and the spin coating when flow is assumed to have
profile oflateral velocity versus height is parabolic. stopped. He used a two-dimensional finite-
While this approximation is expected to fail near element analysis, and showed that lateral polymer
the steep walls of sharply defined features, the diffusion can enhance planarization of small
wall regions are a small fraction of the total features by transporting polymer into the fluid
substrate area, and the lubrication approximation over the grooves, so that less shrinkage occurs
can provide a good estimate of the average during the final, drying stage of spin coating.
resistance to flow experienced by a fluid particle. This analysis neglects capillary driven flow, which
The assumption that fluid properties (such as is reasonable when the film becomes very viscous.
viscosity) are constant throughout the thickness However, because the diffusivity decreases as the
ofthe film requires that the rate ofpolymer-solvent solution viscosity increases, it is not clear that
interdiffusion be fast enough compared with the there will be a regime in which lateral diffusion
rate of thinning of the film that the polymer is important, while viscous flow is negligible.
concentration remains vertically uniform. Because As discussed above, lateral concentration
of the thinness of the films, this assumption seems gradients exist late in spin coating, at least for
to be a good one, except late in spin coating wide features. As a result, if the surface tension
when the polymer concentration becomes high depends on concentration, surface tension gradi-
enough that the polymer-solvent binary diffusion ents can develop; and this induces bulk flow by
coefficient becomes small. At this point it is the Marangoni effect. Sukanek (1989) investigated
possible that a skin of concentrated polymer theoretically the effect ofsurface tension gradients,
forms on the surface of the film, which blocks using a hypothetical relationship between surface
evaporation of the remaining solvent. Typically, tension and polymer concentration, and found
at the conclusion of spin coating, a residual that the gradients of surface tension promote
724 Spin coating
conformal films and therefore reduce leveling. forces, is given by

The magnitude of the effect is sensitive not only
to geometric parameters such as film thickness j.lW 4
tL "'-- (14.14)
and feature width, but also to the concentration uH:
dependence of the surface tension, however, and
this dependence is often unknown. Overdiep where W is the width of the feature, j.l is the film
(1977) found that Marangoni flow is important viscosity, He is a characteristic film thickness
in the leveling of paint brushmarks. However, and u is the liquid-air surface tension of the film:
the characteristic unevenness of these brushmarks Notice that the leveling time increases rapidly
was around 4 mm, and the paint film thickness with increasing feature width W Essentially the
was around 100 Jlm; these length scales are much same equation was obtained years ago by Smith,
larger than the corresponding length scales for Orchard and Rhind-Tutt (1961), and generalized
spin coating. The effect of Marangoni flow on by Overdiep (1977), for leveling of paint brush-
leveling during spin coating has yet to be assessed. marks over flat substrates.
The above discussion indicates that many Often a feature of an uneven substrate is much
phenomena can in principle occur during spin wider than it is deep; i.e., WID;;;: 10, where D is
coating over uneven substrates, including capillary the depth of the feature. When this is the case
flow, Marangoni flow, and vertical and lateral a lubrication analysis accurately predicts the fih~
diffusion of polymer and solvent. Understanding thickness during the leveling process. From the
the interplay ofall ofthem would require numerical lubrication analysis, one obtains a differential
simulation of the coupled multidimensional equation describing the evolution of the film
transport equations for mass and momentum. thickness profile over an uneven substrate. For
However, experiments described below show a two-dimensional groove or trench, we define
that, in many cases, a satisfactory understanding the depth of the trench D, its width u-; and the
of the leveling process in spin coating can be local heights of the substrate S(X) and the film
achieved using a simplified lubrication analysis <l>(X); see Fig. 14.11. The local thickness of the
that neglects Marangoni flow and lateral diffusion film, H(X), is <l>(X) - S(X). The lubrication analysis
of polymer. The leveling that is simplest to then gives (Stillwagon and Larson 1988; Brodkorb,
understand is that which occurs during the Kohler and Ritzel 1989)
baking step at the conclusion of spin coating,
and we consider it first. (14.15)
where X is the coordinate in the plane of the

substrate that is perpendicular to the trench.
14.3.1 LEVELING AFTER SPINNING
When a spin-coated wafer is heated, the polymer
layer can be liquified and can level itself by y
capillary flow; see Fig. 14.10. Some materials,
such as polyimides, undergo reactions during the
heating step that increase the viscosity of the film
<1>(X) ---r-
Hf
and limit the time available for leveling of the
.....-........_ - S(X) ---l...-
film (Stillwagon and Larson 1990). Even for films
that are not subject to such reactions, leveling
~fter spin coating is only feasible economically x
if the time t L for capillarity to induce the required
leveling is not too large. This leveling time t , Figure 14.11 Sketch ofa film profile over a trench-like
which is set by a balance of capillary and visco~s feature (from Stillwagon and Larson (1990».
When we make equation (14.15) dimensionless thins as fluid is drawn from the center toward
using x == X/He, h == H/H e, s == S/H e, and the walls to reduce the large film curvature that
<p == $/He, where He is the initial uniform film initially exists near the walls. The thinning of the
thickness, we obtain film near the center of the feature quickly comes
to an end, however, and for T ~ 5 X 10- 3 , the
(14.16) film thickens with time over the entire trench by
drawing fluid in from the plateau regions that
neighbor the trench.
Here T is the dimensionless time given by
We here define the degree of planarization P
tuHe3 to be the film thickness in the center of the trench,
T= __ (14.17) normalized so that it is zero when the film is of
- W 4 J.L
uniform thickness (and hence conformal) and
For simplicity, we take the viscosity J.L to be a unity when the film is level. For an isolated
constant. Figure 14.12 shows film profiles cal- feature, this definition is equivalent to
culated from equation (14.16) for leveling over
an isolated trench with steep walls represented P == H(O) - H(oo)
(14.18)
by arctangent functions. In these calculations, D
the film is assumed to be initially uniformly thick;
i.e., perfectly conforming to the underlying where H(O) is the film thickness in the center of
substrate. Notice in Fig. 14.12 that near the the feature and H( 00) is the film thickness far
center of the trench at x = 0, the film initially from the isolated feature.
2.0
FILM PROFILES OVER AN ISOLATED TRENCH
1.5
•
of
iC
0
1.0
~ 5
§
f.i:
0.5
SUBSTRATE
0.0L.0---===--l....-----lL.0-----1.J....5------=2-'-::.0~---~2
5
Po ·ilion, x
Figure 14.12 Predictions from the lubrication equation of the film profile over an isolated trench-like
feature, at various dimensionless times T after cessation of spin coating of non-volatile liquids. At the end of
spin coating, the film profile is assumed to be conformal. The width W of a typical feature is much greater
than its depth D, and thus for the typical feature, the vertical scale of this figure is greatly expanded compared
with the horizontal scale (from Stillwagon and Larson (1988)).
726 Spin coating
----
100 as discussed earlier. Not surprisingly, the planar-
---- ----- ization is greater for a hole than for a trench
80 /
,. ,. Periodic Trenches
with width Wequal to the hole's diameter. Also,
/
/
planarization is more rapid for periodic trenches
/
~ 60
~
/ Isolated Holes than for isolated ones, as the former can be
I
c
0 I leveled in part by the erosion of the neighboring
Isolated Trenches
~ 40
N
plateaus. For periodic trenches, the definition of
.~ 2/1
P in equation (14.18) is changed by replacing
c
ctJ H( 00) by the film thickness at the center of a
a:: 20
plateau. For photoresists, during the baking step
0 the viscosity can change with time because of
evaporation of residual solvent, or because of
-20
curing reactions. This effect is not included in
0 0.05 01 015 02 Fig. 14.13, but is accounted for in the equations
T of Brodkorb, Kohler and Ritzel (1989).
The most important conclusions from these
results are that the time required for a film of
Figure 14.13 The degree of planarization P versus
dimensionless time T after cessation of spinning of thickness H r to level a feature of width W is
non-volatile liquids. Data are given for isolated proportional to W4/H~. Hence, for typical films
trench-like features with HelD ratios of 1 and 2 (closed of thickness 2 ~m or less, impractically long times
and open circles, respectively). Theoretical predictions can be required to level features with widths of
from the lubrication equation (14.22) are given for 200 ~m or more. And as the degree of planarization
isolated trenches with HelD of 1and 2, and for periodic is negative at small T, attempts to level features
trenches and isolated round holes with HelD = 2 (from
that are too wide may, in the leveling time
Stillwagon and Larson (1988».
available, actually make the film less level. Thus,
if planar films are desired, wide features should
Figure 14.13 shows the predicted dependence be avoided.
of the degree of planarization on dimensionless
time T after cessation of spin coating for isolated
14.3.2 LEVELING DURING SPINNING
trenches, for isolated circular holes, and for
periodic trenches separated from each other by If the feature is narrow enough, the film over it
plateaus of width equal to the trench width W can be partially leveled during spin coating; this
The ratio HriD ofthe film thickness to the feature might in some cases eliminate the need for
depth has a small effect on the degree of leveling after spinning. Above, we stated that
planarization; this is shown in Fig. 14.13 for partial leveling could typically be expected for
isolated trenches. Figure 14.13 also shows feature widths of about 50 ~m or less. We can
interferometry measurements of the degree of justify this estimate, and calculate film profiles
planarization versus dimensionless time for the during spin coating, by extending the lubrication
leveling ofnon-volatile silicon liquids over isolated analysis presented above to include the centrifugal
trenches. In these and similar experiments, the forces generated by the rotation of the substrate,
feature width ~ the film thickness H r, the fluid and to include the influence of the solvent
viscosity IJ., and the feature depth D were all evaporation that occurs during spinning.
varied; Fig. 14.13 shows that the data agree well For the moment, let us neglect the solvent
with the lubrication theory. At very small T, the evaporation, while acounting for centrifugal forces.
degree of planarization is negative, because in Hwang and Ma (1989), Brodkorb, Kohler and
the initial stages of leveling, fluid is drawn from Ritzel (1989), and Stillwagon'and Larson (1990)
the center of the feature toward the steep walls, derived lubrication equations for film flow over
a spinning substrate with axisymmetric uneven- from the feature. Then we integrate the right side
ness. For a trench-like feature whose long of equation (14.19), and obtain
dimension is perpendicular to the radial coordi-
nate, and whose width W is much smaller than
the distance R o of the trench center from the (14.22)
center of rotation of the wafer, the lubrication
equation is The profiles computed from equation (14.22)
are in excellent agreement with experimental
profiles; see Fig. 14.14. These experimental profiles
were measured by spin coating non-volatile
(14.19) epoxy liquids that were cured during spinning
by exposure to ultraviolet radiation; the height
where
profiles ofthe hardened films were then determined
by mechanical profilometry (Stillwagon and
(14.20) Larson 1990). Notice that the profiles are
asymmetric; at high 0)2 there is a bump just
and upstream of the trench. The mechanism by which
WJl these asymmetries occur is discussed by Peurrung
(14.21) and Graves (1991).
t == H;pfl R o
e 2
The degree of planarization P was obtained
is the 'centrifugal time' required for the local film from these profiles; Fig. 14.15 is a plot of P versus
profile to come into local equilibrium with the dimensionless spinning speed 0)2, for HelD = 2.
centrifugal force. In equations (14.20) and (14.21), The relationship between P and 0)2 predicted by
He is a characteristic film thickness that we define equation (14.22) is also plotted in Fig. 14.15;
below. Except for the centrifugal term, equation agreement with the measurements is good. In
(14.19) neglects inertia. From equation (14.5) of the experiments, a universal curve was obtained
Emslie, Bonner and Peck (1958), we find that far when data for varying fl, Jv, R o' He and D were
from any topographic features the relative rate plotted against 0)2 with HelD held fixed. One of
of thinning of the film, namely -(dH/dt)/H, is the dashed lines in Fig. 14.15 was obtained by
roughly given by W/(Rot e ). Thus the rate, l/t e , linearizing equation (14.22) about a conformal
at which the film profile over a feature comes profile; the other was obtained by linearizing
into local equilibrium with the centrifugal force about a level profile. These two linearizations
is faster by a factor Ro/W than the rate at which are accurate, respectively, at large 0)2 and at small
the film thickness far from the feature changes 0)2. As in leveling after spinning, during spinning
by a significant percentage. Thus, except for large there is also a dependence of the degree of
features that are close to the center of the planarization on the ratio HelD (Stillwagon and
spinning wafer, a local equilibrium approximation Larson 1990). Figure 14.15 shows that some
is warranted. As the film thins, the profile over planarization is achieved at 0)2 = 10, and almost
the feature can then be assumed to adjust rapidly complete planarization is obtained for 0)2 = 0.1.
to changes in the film thickness far from the In typical applications, fl ~ 4000 rpm, p =
feature, so that at each instant, the film profile 1 g/cm 3 , R o = 2.5 em, (J = 30 mN/m, and
is in a state of local equilibrium between the He = 1 J.1m. We then find from the definition of
surface tension and the centrifugal forces. 0)2 in equation (14.20) that planarization improves
With this local equilibrium approximation, the significantly as W decreases from 50 J.1m to 10 J.1m;
time derivative in equation (14.19) can be dropped, this agrees with the conclusions drawn from
and the characteristic film thickness He can be other experiments, mentioned earlier.
defined as the slowly changing film thickness far Because of the neglect of solvent evaporation,
728 Spin coating
Fig. 14.15 does not directly apply to most spin

coating applications. Solvent evaporation increases
the viscosity of the liquid film and slows its flow
1.0 r-nn" across the substrate. When the film viscosity
- becomes high enough, flow becomes negligible,
and the thinning of the film occurs mainly by
0.6 - evaporation. However, earlier in spin coating,
the film is mobile and we expect that its thickness
profile will achieve local equilibrium over a
0.2 - feature, just as it does when evaporation is
absent. Peurrung and Graves (1991) have measured
u}=O.11 film profiles during spin coating by using
-0.2 +- I I I
stroboscopic interfereometry, and have shown
that local equilibration does occur during the
early stages of spin coating. Figure 14.16 shows
1.0 the film profiles that they measured over long
-& rectangular bumps during spin coating. Solving
~ equation (14.22), these profiles can be predicted.
is0 0.6
s..
Note that w 2 changes with time, as w 2 depends
Q.. on He' and He decreases continuously during
-e
~
0.2 spinning. If w 2 at each instant in time is computed
from the instantaneous value of He' then from
w2 = 0.88 equation (14.22) the film profile at that instant
-0.2 can be accurately computed. Thus equation
(14.22) holds not only for non-volatile fluids, but
also for volatile liquids early in spin coating.
1.0 A two-dimensional version of equation (14.22)
for features of arbitrary two-dimensional shape
has been presented recently by Peurrung and
0.6
Graves (1993). They solved the two-dimensional
equation by a finite-element method for features
0.2 such as the square bump shown in Fig. 14.17,
and obtained iso-thickness profiles during spinning
w2 = 9.6 that have the same shapes as those seen in their
-0.2 experiments. Note in Fig. 14.17 the peculiarly
-4 -2 0 2 4 shaped 'wake' in the film profile downstream of
Position, x the square bump; the predicted and measured
wake profiles are similar, showing that the two-
dimensional lubrication theory is valid.
During spin coating, so much solvent eventually
Figure 14.14 Film profiles over isolated trench-like evaporates that the film becomes very viscous,
features during spin coating of non-volatile liquids. and its flow rate becomes so low that it thins
The points are experimental data for epoxy liquids mainly by evaporation. When this point is reached,
suddenly solidified during spin coating and the lines
the film profile can no longer equilibrate over
are the predictions of the lubrication equation (14.22).
The vertical scale of this figure is greatly expanded the feature, and equation (14.22), and its two-
compared with the horizontal (from Stillwagon and dimensional analog, no longer hold. Accurate
Larson (1990). analysis of leveling during the final stages of spin
100
0
0
D_~
80
~
.fl
~,
0
~
z 60 l:!.' ~o
0
~
<t 40
N
a:
<1
z 20
<1
..J
ll.
0 0
-20
0.01 10000
,
oo·
Figure 14.15 Plot of the degree planarization versus dimensionless spinning speed w 2 • The data points are
for epoxy liquids hardened during spin coating with HriD = 2, and J¥, R o , H f , D, and n all varied. For (0),
H f and D are fixed; for (~), n is fixed; and for (0), n, H f , and D are fixed. The solid line is from the lubrication
equation solved numerically, and the dashed lines are for linearizations of the lubrication equation, valid at
large and small w 2 (from Stillwagon and Larson (1990».
coating therefore requires that evaporation and solvent is finally evaporated, the final degree of
its effects on film viscosity be accounted for. planarization P 2 over that feature is given by
Brodkorb, Kohler and Ritzel (1989) and
Stillwagon and Larson (1990; 1992) have developed
P2 = C 1P 1 (14.23)
methods of estimating the effects of solvent where, as mentioned above, C 1 is the polymer
evaporation. Stillwagon and Larson, extending volume fraction at the instant the film becomes
the ideas of Meyerhofer (1978) and LaVergne immobile. Equation (14.23) neglects any effect of
and Hofer (1985), assume that the film profile lateral diffusion of polymer during the final
remains in local equilibrium until the polymer evaporation stage. As the film profile is assumed
volume fraction C in the film reaches a value C 1 to remain in local equilibrium over the feature
high enough that the film can be assumed to be until the polymer concentration reaches C l' PI
immobile. At this point, fluid rearrangement over is given by the theory for non-volatile solvents.
the feature is assumed to stop, and the film Thus P 1 is a function of wi and H tiD, where H 1
thereafter changes its local thickness only by is the film thickness far from the feature when
evaporation. Thus, during the final evaporative the polymer concentration reaches C 1 , and
stage, the film thickness merely shrinks because wi == pQ 2 W 3 RolaH 1, is defined as in equation
of the loss of solvent. Thicker portions of the (14.20), except that H f is replaced by H 1 •
film, such as those over narrow, leveled features, Stillwagon and Larson (1992) estimated the
contain more solvent and hence shrink more concentration at which the film is effectively
than thinner parts of the film. Thus the leveling immobilized to be that at which the rate of
achieved earlier in the spin coating process is thinning of the film by evaporation equals lite'
partially lost. As a result, if P 1 is the degree of the rate ofequilibration given by equation (14.21).
planarization achieved over a feature at the The rate of evaporation has been estimated by
moment the fluid becomes immobile, and if all Lawrence (1988, 1990) and by Ohara, Matsumoto
730 Spin coating
WET FILM PROFILE
50-11m feature
18 -r---~------------,
'"
1 = 2.5 S, llJ2 =1.6
c
eu 12
.§
.;
'"
c'"
-'" 9
.!:! (b)
-=
.5
ti:
I =lOs, llJ2 =3.6
1
6
• ~
B g
3-'-
I 1 = 16.5 S, llJ2 =4.5
o ..,.-----+---.......
-200 100 200
Po ilion, micr ns
Figure 14.17 (a) Interferometric picture of a 100 11m

square bump over which fluid is flowing during spin
Figure 14.16 The data points are film profiles over coating, with flow from top to bottom. The picture
a 50 11m rectangular bump measured by interferometry was taken 1 s after start-up of spin coating. (b) The
during spin coating ofan evaporating polymer solution. dashed line shows the shape of the square 'bump', and
The lines are the predictions ofthe lubrication equation, the solid lines are the predicted iso-thickness contours
where w 2 varies with the instantaneous film thickness of the film over the bump. The contours represent
He (from Peurrung and Graves 1991; reprinted by thickness variations of 0.1 57 11m, which also corresponds
permission of the publisher, The Electrochemical to the thickness variations between dark fringes on
Society, Inc.). the left (from Peurrung and Graves 1993; reprinted
with permission from IEEE Trans. Semicond. M anuf
6(1): 72-76; © 1993 IEEE).
and Ohashi (1989), who assumed that the slow
step in evaporation is the diffusive transport in
the vapor phase; see Section 14.2.2. In this way, relationship between viscosity and polymer
Larson and Stillwagon estimated C 1 and found concentration. P 1(wi, H 1/D) is obtained by solving
that it depends on the initial polymer concentration equation (14.22). A dependence of C 1 on W/R o
Co, on the concentration-viscosity curve for the exists because wide features near the center of
solution, and on the ratio W/R o . Thus the wafer have a longer equilibration time t c than
smaller features that are farther from the center
P2 ~ C 1(C o ,M, W/R o ) x p 1 (wi, Ht/D) (14.24)
of the wafer; see equation (14.21).
Here M is the molecular weight of the polymer; Stillwagon and Larson spin coated a series of
for a given type of polymer, M controls the polystyrene solutions of varying concentration
Summary and perspective 731
and molecular weight and generally found that, uniform shrinkage during the evaporation stage.
by adjusting C I, the measured dependence of Evidently, the assumption of a sharp transition
degree of planarization P 2 on spinning speed w~ between flow and evaporation stages of the level
could be fit to equation (14.24); see Fig. 14.18. process, with perfect equilibration before the
Here w~ is defined in terms of the measurable transition and no flow and no transverse diffusion
final film thickness H 2' and is related to wi by after the transition, does not lead to completely
wi = W~CI' Qualitative agreement was found satisfactory predictions.
between predicted and measured dependencies
of CIon Co, M and WjR o' However, the
predicted values of CI were higher than the 14.4 SUMMARY AND PERSPECTIVE
experimental values; quantitative agreement could Using intuition and seat-of-the-pants engineering,
only be attained by introducing a large arbitrary early practitioners made the process of spin
parameter, as suggested earlier by Brodkorb, coating thin films over level substrates into a
Kohler and Ritzel (1989). Also, Peurrung and highly successful industrial coating method over
Graves (1991) found that the shapes of the dried fifty years ago. Modeling and analyses developed
film profiles were significantly different in some since then have served mainly to explain why
cases from the shapes predicted by assuming spin coating works so well, and to delineate some
theoretical limits, should future generations find
reason to optimize the technique to obtain even
100 better performance. Simple analyses ofspin coating
~
PS51 assume uniform evaporation, and then predict
- film thicknesses. These analyses, backed by
.S2 80 o 16.8%
experimental results, show that thicker films are
~
N
t>23.3%
II
60 + 28.9% obtained by increasing the viscosity of the solution,
x 33.8%
~ the solids concentration, or the volatility of the
Z solvent, or by decreasing the spinning speed. As
0 40
~ long as overly volatile solvents are avoided (these
-
<
CI:
<
20
t>
can lead to Marangoni-type instabilities), the
uniformity of the film thickness in spin coating
is usually less than a percent and often a few
Z 0 o
<:
..l tenths of a percent or better. If for some reason
Q.
·20 one wants to improve on this uniformity, spinning
0.01 0.1 10 100 1000 10000 conditions must be carefully controlled so that
wi = w~ C, other instabilities, including Ekman spirals, are
avoided.
Often, because of previous microelectronic
Figure 14.18 Planarization PI = P 2/C I versus processing, the wafers over which solutions are
dimensionless spinning speed Q)~ for 16.8%, 23.3%,
28.9% and 33.8% solutions of polystyrene (molecular spin coated contain topographic unevenness of
weight = 51000) in the solvent propylene glycol a micron or so in depth, and tens to hundreds
monomethyl ether acetate. The degree of planarization of microns or more in width. Most of the factors
P2 and the spinning speed Q)~ were calculated using that control leveling or planarization of the
the final film thickness H 2' PI and Q)~ are defined in spin-coated film over such topography both
terms ofthe film thickness HI = H 2/C I at the transition during and after spin coating are well understood,
from the flow to the drying stage of spin coating. By
thanks to recent efforts, and can be predicted, at
adjusting C I , the data for all concentrations collapse
onto a single curve that agrees with the line given by least qualitatively, by simple lubrication models.
the lubrication theory for H tiD = 4 (from Stillwagon The most important of these factors is the width
and Larson (1992)). W of the topographic feature underlying the fluid
732 Spin coating
film. Some planarization during spinning can be 1987. On the modelling of spin coating. J. Imag.
expected if W;5 50 Jlm, and a high degree of Technol. 13(4): 122-130.
planarization can be expected if W ;5 10 Jlm. The Bornside, D. E., Macosko, C. W. and Scriven, L. E.
1991. Spin coating of a PMMA/chlorobenzene
degree of planarization depends also on the solution. J. Electrochem. Soc. 138(1): 317-320.
topography surrounding the feature; isolated Bornside, D. E., Brown, R. A., Ackmann, P. W., Frank,
features are more poorly planarized than closely J. R., Tryba, A. A. and Geyling, F. T. 1993. The
spaced ones. The properties of the resist solution effects of gas phase convection on mass transfer in
affect planarization. Solutions that remain fluid spin coating. J. Appl. Phys. 73(2): 585-600.
at high solids concentration planarize best; also Brodkorb, W., Kohler, J. M. and Ritzel, O. 1989. A
the higher the initial polymer concentration, the hydrodynamic model for the spin-coated substrate
topography. Parts I and II, Experimentelle Technik
better the planarization. Following spin coating, der Physik. 37(6): 519-525; 527-540.
further planarization is possible if the film's Burggraaf, P. 1986. Wafer steppers and lens options.
mobility can be restored by heating. The depend- Semicond. Int. 9(3): 56-63.
ence on feature width of the time to achieve Chiba, K., Tsunoda, T., Kato, Y., Endo, M., Sawase,
leveling by surface-tension-driven flow is so F., ltikawa, K. and Imamura, M.-A. 1982. Coating
severe, however, that it may be impractical to method. US Patent 4,353,937.
level features wider than about 200 Jlm, unless Daniels, B. K., Szmanda, C. R., Templeton, M. K. and
Trefonas III, P. 1986. Surface tension effects in
the film is unusually thick. microlithography - striations. Proc. SPIE. 631:
In sum, spin coating over level substrates is a 192-201.
well developed practice, which usually produces Daughton, W. J. and Givens, F. L. 1982. An investigation
satisfactory results for commercial purposes. The of the thickness variation of the spun-on thin films
most serious problems seem to occur when the commonly associated with the semiconductor
substrate is uneven, and the spin-coated film is industry. J. Electrochem. Soc. 129(1): 173-179.
sometimes not level enough for high-resolution Elliott, D. J. and Hockey, M. A. 1979. One micron
range photoresist imaging: a practical approach.
photolithography. The conditions under which Solid State Technology. 22(6): 53-60.
such problems will occur are now well understood Emslie, A. G., Bonner, F. T. and Peck, L. G. 1958.
and can be modeled semi-quantitatively. Future Flow of a viscous liquid on a rotating disk. J. Appl.
work might be directed toward making these Phys. 29(5): 858-862.
predictions completely quantitative, or toward Flack, W. W., Soong, D. S., Bell, A. T. and Hess,
developing new ways to improve leveling. D. W. 1984. A mathematical model for spin coating
of polymer resists. J. Appl. Phys. 56(4): 1199-1206.
Frasch, P. and Saremski, K. H. 1982. Feature size
REFERENCES control in IC manufacturing. IBM J. Res. Develop.
26(5): 561-567.
Acrivos, A., Shah, J. J. and Petersen, E. E. 1960. On Hagan, J. A., Maloy, J. E. and Wilke, M. A. W. 1986.
the flow of a non-Newtonian liquid on a rotating Magnetic disk coating method and apparatus. US
disk. J. Appl. Phys. 31(6): 963-968. Patent 4,589,139.
Adams, A. C. and Capio, C. D. 1981. Planarization Hwang, J. H. and Ma, F. 1989. On the flow of a thin
ofphosphorus-doped silicon dioxide. J. Electrochem. liquid film over a rough rotating disk. J. Appl. Phys.
Soc. 128(2): 423-429. 66(1): 388-394.
Bassous, E., Ephrath, L. M., Pepper, G. and Mikalsen, Jenekhe, S. A. and Schuldt, S. B. 1984. Coating flow
D. J. 1983. A three-layer resist system for deep u.v. of non-Newtonian fluids on a flat rotating disk. Ind.
and RIE microlithography on nonplanar surfaces. Eng. Chem. Fundam. 23(4): 432-436.
J. Electrochem. Soc. 130(2): 478-484. Lai, J. H. 1979. An investigation of spin coating of
Bornside, D. E. (1990). Mechanism for the local electron resists. Polym. Eng. Sci. 19(15): 1117-1121.
planarization of microscopically rough surfaces by LaVergne, D. B. and Hofer, D. C. 1985. Modeling
drying thin films of spin-coated polymer/solvent planarization with polymers. Proc. SPIE. 539: 115.
solutions. J. Electrochem. Soc. 137(8): 2589-2595. Lawrence, C. J. 1988. The mechanics of spin coating
Bornside, D. E., Macosko, C. W. and Scriven, L. E. of polymer films. Phys. Fluids. 31(10): 2786--2795.
References 733
Lawrence, C. 1. 1990. Spin coating with slow evapor- interconnection technology utilizing polyimide.l EEE
ation. Phys. Fluids A. 2(3): 453-456. Trans. Parts, Hybrids Packaging. 9: 176-180.
Lawrence, C. 1. and Zhou, W. 1991. Spin coating of Scheffel, D. and Richter, V. 1978. Apparatus for the
non-Newtonian fluids. J. Non-Newt. Fluid Mech. manufacture of the magnetic coatings of magnetic
39: 137-138. ' disks. US Patent 4,073,262.
Leers, D. 1981. Investigation of different resists for far Schiltz, A., Abraham, P. and Dechenaux, E. 1987.
and deep UV-exposure. Solid State Technology. Improvement of photoresist planarization properties
24(3): 90-92. by thermal cure. J. Electrochem. Soc. 134(1): 190-194.
Ludviksson, V. and Lightfoot, E. N. 1968. Hy- Scriven, L. E. 1988. Physics and applications of dip
drodynamic stability of Marangoni films. AIChE J. coating and spin coating. Mat. Res. Soc. Symp. Proc.
14(4): 620-626. 121: 717-729.
Meyerhofer, D. 1978. Characteristics of resist films Shaw, 1. M., Frisch, M. A. and Dill, F. H. 1977.
produced by spinning. J. Appl. Phys. 49(7): 3993-3997. Thermal analysis of positive photoresist films by
Moreau, W. M. 1988. Semi-Conductor Lithography. mass spectrometry. IBM J. Res. Dev. 21(3):
New York: Plenum Press. 219-226.
Ohara, T., Matsumoto, Y. and Ohashi, H. 1989. The Skrobis, K. 1., Denton, D. D. and Skrobis, A. V. 1990.
film formation dynamics in spin coating. Phys. Effect ofearly solvent evaporation on the mechanism
Fluids A. 1(12): 1949-1959. of the spin coating of polymeric solutions. Polym.
Overdiep, W. S. 1977. Levelling of paints. In Engng Sci. 30(3): 193-196.
Physicochemical Hydrodynamics, ed. D. B. Spalding. Smith, N. D. P., Orchard, S. E. and Rhind-Tutt, A. J.
New York: Advance Publications Ltd. pp. 683-697. 1961. The physics of brush marks. J. Oil and Colour
Pearson, J. R. A. 1958. On convection cells induced Chemists Assoc. 44(9): 618-633.
by surface tension. J. Fluid Mech. 4(5): 489-500. Spangler, L. L., Torkelson, J. M. and Royal, 1. S. 1990.
Peugh, H. V. and Ward, A. W. 1984. Coating thickness Influence ofsolvent and molecular weight on thickness
and wedge geometry control for magnetic disks. US and surface topography of spin-coated polymer
Patent 4,485,758. films. Polym. Eng. Sci. 30(11): 644-653.
Peurrung, L. M. and Graves, D. B. 1991. Film Stillwagon, L. E. 1987. Planarization of substrate
thickness profiles over topography in spin coating. topography by spin-coated films: a review. Solid
J. Electrochem. Soc. 138(7): 2115-2124. State Technol. 30(6): 67-71.
Peurrung, L. M. and Graves, D. B. 1993. Spin coating Stillwagon, L. E., Larson, R. G. and Taylor, G. N.
over topography. IEEE Trans. Semicond. Manu! 1987. Planarization of substrate topography by spin
6(1): 72-76. coating. J. Electrochem. Soc. 13(8): 2030-2037.
Rehg, T. 1. 1992. Spin coating ofmonodisperse colloidal Stillwagon, L. E. and Larson, R. G. 1988. Fundamentals
suspensions: evidence of evaporative convection. of topographic substrate leveling. J. Appl. Phys.
PhD dissertation, University of California Davis. 63(11): 5251-5258.
Rehg, T. J. and Higgins, B. G. 1988. The effects of Stillwagon, L. E. and Larson, R. G. 1990. Leveling of
inertia and interfacial shear on film flow on a thin films over uneven substrates during spin coating.
rotating disk. Phys. Fluids. 31 (6): 1360-1371. Phys. Fluids A. 2(11): 1937-1944.
Rehg, T. J. and Higgins, B. G. 1992. Spin coating of Stillwagon, L. E. and Larson, R. G. 1992. Planarization
colloidal suspension. AIChE J. 38(4): 489-501. during spin coating. Phys. Fluids A. 4(5): 895-903.
Reisfeld, B., Bankoff, S. G. and Davis, S. H. 1991a. Sukanek, P. C. 1985. Spin coating. J. Imag. Technol.
The dynamics and stability of thin liquid films 11(4): 184-190.
during spin coating. I. Films with constant rates of Sukanek, P. C. 1989. A model for spin coating with
evaporation or absorption. J. Appl. Phys. 70(10): topography. J. Electrochem. Soc. 136(10): 3019-3026.
5259-5266. Sukanek, P. C. 1991. Dependence of film thickness on
Reisfeld, B., Bankoff, S. G. and Davis, S. H. 1991 b. spin speed in spin coating. J. Electrochem. Soc.
The dynamics and stability of thin liquid films 138(6): 1712-1719.
during spin coating. II. Films with unit-order and Thomas, I. M. 1986. High laser damage threshold
large Peclet number. J. Appl. Phys. 70(10): 5267- 5277. porous silica antireflective coating. Applied Optics.
Rothman, L. B. 1980. Properties of thin polyimide 25(9): 1481-1483.
films. J. Electrochem. Soc. 127(10): 2216-2220. Thompson, L. F., Willson, C. G. and Bowden, M. 1.
Sato, K., Harada, S., Saiki, A., Kimmura, T., Okubo, 1983. Introduction to Microlithography. Washington,
T. and Mulai, K. 1973. A novel planar multilevel DC: American Chemical Society.
734 Spin coating
von Karman, Th. 1921. Ober laminare und turbulente White, L. K. 1985b. A model study of superficial
Reibung. Z. Angew. Math. Mech. 1: 233-252. topography for improved lithography. J.Electrochem.
Wahal, S., Oztekin, A., Bornside, D. E., Brown, R. A., Soc. 132(12): 3037-3041.
Seidel, P. K., Ackmann, P. W. and Geyling, F. T. Wilson, R. H. and Piacente, P. A. 1986. Effect of circuit
1993. Visualization of a gas flow instability in spin structure on planarization resist thickness. J.
coating systems. Appl. Phys. Lett. 62(20): 2584-2586. Electrochem. Soc. 133(5): 981-984.
White, L. K. 1985a. Approximating spun-on, thin film Yonkoski, R. K. and Soane, D. S. 1992. Model for
planarization properties on complex topographies. spin coating in microelectronic applications. J.
J. Electrochem. Soc. 132(1): 168-172. Appl. Physics. 72(2): 725-740.
INDEX
Page numbers in bold refer to figures and page numbers in italic refer to tables.
Accommodation coefficient 196 outer expansion 256 finite-element implementation

Adams-Bashforth/trapezoid-rule outer region 256,266-7,272 305-6
integration 332 stretching transformation 256, free 303
Adhesion 105-6 267,273 inflow 303
Adsorption, see Surfactants see also Domain perturbation natural 305
AerosolOT 117,121,127 methods; Regular outflow 303
Air entrainment 6, 86, 89,93-5,93, perturbation methods; traction 302
446-7,473-4,482-4,482,485, Singular perturbation methods see also Dynamic boundary
486-8,486,546,641,657-9 Auto-focus method 227 condition; Interfacial
air bubble collapse 87,483 boundary conditions;
air film 87,92,473,483,658 Back-diffusion 108 Kinematic boundary
empirical correlations 88-9,446 Bandedness 429,456 condition; No-slip condition
hydrodynamic theories 92 Bank rolling 157,311,558-9,559, Boundary-element methods 352
hysteresis 90, 487 630-1 Boundary integral techniques 351-2
macroscopic flow effect 86-7 instabilities 602-3 Boundary layer 28
macroscopic regime 658-9 Barring 418,421,448-51,479 in curtain coating bead 31,473,
microscopic regime 92,473,483, Bead coating, see Slide coating 480,484,485-6,488,747-8
658 Beer's law 222 in overflow cell 414,515
models for entrained air film Bent blade, see Flexible blade for mass transfer 713
90-2,93-5 coating in slide-coating bead 434,435
molecular-kinetic theory 93 Bernoulli's equation 29-31 in surfactant concentration 522,
prevention with hydrodynamic Beveled blade, see Flexible blade 524
assist 89-90,463,484 coating Box-Wilson method 742
sawtooth-shaped wetting line Bifurcation 655 Bright-field detector 231
86-7,87,88,95,658,659 symmetry-breaking 653 Brushmarks 187
shear-thinning viscosity effect 89, Bifurcation points 320, 328, 338, Bubbles 547,596
484 655 as defect source 458, 546
surface roughness effect 89, 658 Bingham model 155,337 static pressure inside 39
surface wettability effect 89 Bingham-Carreau-Yasuda model trapped in recirculation 472,547,
see also specific coating methods 375 596, 755
Amphiphilic molecules 682 Blade coating 10,637,637 see also Maximum bubble pressure
Analysis of variance (ANOYA) 215 inverted 638 method
Arbitrary Lagrangian-Eulerian of paper 138 Buckingham theorem 215
(ALE) methods 299-300, 305, see also Flexible-blade coating; Bury-Hartley model 499
348-9 Knife coating; High-speed
comparison with standard finite- blade coating Calcium carbonate 141
element method 349 Blasius boundary layer 490 Calendering 12
Architectural paints 138, 141, Bleeding 444, 665-6 Capacitance gauging
160-75 Blotchiness 458 of film thickness 225-6
see also Latex coatings Bond number 48-51,51,52,53,63, of thin gaps and slots 226
Arnoldi's algorithm 322 73,86,406,676-7 Capillary forces 63, 71
Asymptotic methods 8,72,251-94 limit of low values 50,51,676 Capillary hydrodynamics 7, 19-62
composite expansion 256 limit of high values 50,676 Capillary length 52, 63, 272
inner expansion 256 Boundary conditions 31-47,301-3 Capillary number 48,64,71,75,80,
inner region 256,267-9,273 complete set 47,301-3 81,85,86,91-4,158,303,406,
linearization 257 distinguished 306, 315 439,481,511,543,574,620,
matching 256,269, 275, 279-80 essential 305 676,679,679,680
770 Index
Capillary number (contd) design configurations 370, 403-4, Coating processes
limit of low values 48-52, 573 404 dynamics, or transient behavior
limit of high values 52-3,336, design considerations 8, 383-96, 739-40
475,484 752 industrial and academic research
Capillary pressure 38, 686 design optimization 752-61,752, 4
Capillary rheometer 147, 662 760 overview 3-15
Capillary rise 66-7, 112 dimensional analysis 379-82 robustness 741,746
for measuring surface tension 112 distribution chamber, see Manifold yield 763
Capillary statics 66-8, 272 elastic deflections 381 Coating rheology 66, 137-82,
Capillary tube, meniscus advancing feed 372-3,373 659-63
in 72-3 center 372 interplay among components and
Carrier layer 447 end 372 application performance
Carreau(-Yasuda) model 154,213, multiple 372 157-77
337,375 recycle 395 numerical analysis 337-9
Cascade instability 563-6, 564, 566 inertial effects 386, 387 rheometric flows 142-9
Casson model 154 internal flow analysis 372-83,393 Coating technology, origins 4
Cavitation 157,557,596 manifolds 369,372,645 Coating uniformity 216-22,334,
Cavity design, see Coating dies multilayer manifold design 394-5 546-7,641,659,741,756-9,757,
CCD arrays 237, 238 nonuniformity from flow-split in 758
CCD video camera 217,230 center-fed cavity 373 control 740-1
Characteristic equation 276,279 pressure distribution 374-5, 376 power spectrum 220
Chatter 457 reactive flows 395 quality index 218
see also Barring secondary cavities 393-4,752 quantification 216-22
Chebichev methods 318 sensitivity to small variations in threshold level as quality
Chemical potential 192 slot 379, 395, 760 criterion 222
Chemical vapor deposition 3 slot 369 Coating window 745-6
Chevrons 457 accuracy 760 computational analysis 327-8,
Choker bar 369,371 slot flow analysis 376-9,377 331
Clay 141, 175 streak formation 755 feasibility 745
Cleaning-in-place (CIP) system 736 temperature uniformity 395, 760 quality 329-30, 745
Cloudy films 717-18 thermal analysis 382-3 sensitivity 745
Coagulation, shear induced 395 viscoelastic effects 381 see also specific coating methods
Coathanger manifold 369, 386-93, see also Choker bar; Coathanger Coextrusion 323, 394
388 manifold; T-manifold; Wall interfacial instability 323
design 386-93, 390--2 shear stress critical layer thickness ratio 394
tapered 386, 387 Coating equipment, see Pilot coating Cohesion 105-6
Coating bead 73,401-2,427, equipment Colloidal coatings 710,716,718-20,
450-1,542,744 Coating failure 6 720
see also Slide coating; Slot Coating flows 42,301-8,501-5 shear-thinning 717
coating album of 243-6, 243 see also Sol-gel films
Coating by drainage 673, 673 boundary conditions 31-48, Compressive flow 144
Coating by withdrawal 673, 673 301-3 Computational methods 8,
Coating defects, see Defects governing equations 20-31, 297-366
Coating dies 25 301-3 Cone-and-plate rheometer 147, 164
adjustment mechanisms for slot mathematical analysis 252-4 Conjoining pressure 190, 193
369,371,741 Coating fluids 137 see also Disjoining pressure
capacitance gauging of slot test fluids for pilot experiments Conjugate gradient (CG) method
uniformity 226 212-13 318,351
cavity 369 Coating formulations 139-42 Conjugate residual (CR) method
cavity cross-sections 370 effect on defects 184 318
cavity flow analysis 29, 373-5 Coating gap 401 Conservation laws 20-31,301-3
cavity-slot transition analysis control 422 linear momentum 22-4,691-2
378-9 Coating industry 4 mass 20-2, 32-5
contamination and cleaning 547, Coating irregularities, see Defects solute 43
753-5 Coating machine species 43, 690
control of cross-web uniformity capacity 763 Consistent pressure Poisson equation
741 change-over time 764 (CPPE) 308
cross-web flow distribution 379, down time 764 Constitutive equations 24, 38, 138,
380, 756-9 specific machine utilization 763 154-5
Index 771
generalized Newtonian fluid 24, Crazing 199,201 sheet-forming flow 467-72,468,
154,337 Creeping flow 28 469
interfacial stress 37 Critical micelle concentration 101, see also Teapot effect
Newtonian fluid 154,302 499,501,516,519 sensitivity to disturbances 475-80
surface 504 Critical pigment volume start-up 490-1,492
surface excess concentration 498 concentration (CPVC) 141 steady-state operation 466-75
viscoelastic fluid 155 Critical point tracking 327-8 variable coating width 492
see also Bingham model; Carreau Critical points 320, 327 Curvature
model; Cusson model; see also Bifurcation point; Hopf mean radius 38,39,40, 64
Maxwell model; Power-law point; Turning point principal radii 39
model Curling 199 sign 39
Constitutive theory 7 Curtain coating 9,463-93,
Contact angle 45-7,45,64,64 744-52 Deborah number 660
advancing 46, 65, 66 air boundary layer 747 Defects 6,7,13,199,201,217,219,
apparent 45, 65 air current effect 479 221,448-51,456-8,544,546-7,
boundary condition 70, 302 air entrainment 463,473,482-4, 616,663-8,755
hysteresis 46,65-6,335,467,526 482,485,745 classification 231
interface between miscible fluids with heel 486-8,486 frequency spectrum of 221
335 air shields 479,491-2 healing 184-9
measurement 65 applications 463,466 on-line detection 230-2
receding 46,65,66 applied pressure difference 477, prevention 735, 736
submicroscopic, or actual 45 479 quantification 216-22
surfactant effects 526-8 barring 479 see also Laser scanners
see also Dynamic contact angle; coating window 481-2,481,486, Deflection number 381
Static contact angle 745, 745, 750 Deformable roll coating, see Squeeze-
Contact line, see Wetting line curtain flow zone, see Fal1ing roll coating
Contamination 100,547,753-5 curtain Deformation rate, see Rate of strain
airborne particles 58 dynamic contact angle 69,70 Degassing 596,735,755
particulate 547 dynamic wetting 336, 473-5, Derivative, substantial 23
Continuation procedures 319-21, 744-52 Dewatering 664
327 edge effects 471-2,478,490 Dewetting 8, 104, 189-96,191,
arc-length 320, 350 see also Edge guides 192
around singular points 320 equipment configurations and Die coating, see Extrusion Coating;
first order 319 design 464-6; 464, 465, 466 Slot coating
step-size control 320 finite-element analysis 473,474, Die design, see Coating dies
Continuity equation 22, 259-60, 475,476,481,483 Die swell 467
302,307,654 fixed-grid simulation 344 Differential/algebraic equation
see also Conservation law frequency response 479,480 system (DAES) 331
Control 11,735-41 heel formation 310,310,473, numerical solvers 332
algorithm 739 474,476,485-8,746-7 Diffusion 43, 108-10, 125, 197-9,
feed-back 735-6, 735 impingement angle 484 497-8,508,690,713
feed-forward 735-6, 736 impingement flow 70, 463, 473-5, interlayer in multilayer coating
loops 740 475,476,746,748,749,751 450
on-line 740 impingement velocity 463, 472 within surface or interface 42,
Control volume 21,286,433 linear finite-element stability 497
Convective diffusion 497-8, 504 analysis 326 see also Convective diffusion;
in computational methods lip design 488-9 Surfactants
339--40,692-3 maximum coating speed 466,487, Diffusion coefficient, see Diffusivity
equations for 690 488 Diffusion limited transport 197-8,
Corduroy, see Ribbing minimum coating thickness 326, 713
Corner 463,481 Diffusivity 108,109,125,127,127,
flow in 245 multilayer 464, 465 128,198,497
wetting-line mobility at 66 non-Newtonian effects 484 concentration and temperature
see also Gibbs' inequality operability 481-8 dependence 681
Couette flow 582 optimization dynamic of wetting for micelles 500
Couette-Poiseuille flow 442 line position 744-52 in polymer films 198
Couette rheometer 147 puddling (vortex in heel) 745 pseudo-binary 690
Cracking 199,201 pulled-film phenomenon 473, 474, Dimensional analysis 48
Craters 192,458 481-2,483,484 see also Buckingham theorem
772 Index
Dip coating 11,271,673,675, pressure inside 64 Dynamic uniaxial extensional
678-82 sessile 115 viscosity (DUEV) 146, 150,
applications 673-4 Dry patches, growth or healing 151, 178-9
batch 673,673 194-6 Dynamic wetting 6,7,63-4,69,
continuous 673, 673 Drying 7, 13, 196-9,678-82 90-3,95
evaporation, concomitant 680-1 air flow effects 503 driving force 76
film thickness 678-80, 678, control 703, 704 failure 63, 70, 86-95
686--7,686 defects 197 finite-element analysis 335-7,
film uniformity 674 at die lips 422, 466 473,484,526,528
irregularly shaped substrates 673, diffusion limited in film 197-9 hydrodynamic models 71,76--8,
674 gas-side mass transfer resistance 79
Landau-Levich analysis 678 196--7 laser-Doppler velocimetry 87,235
non-Newtonian liquids 680 microstructure control 703 line, see Dynamic wetting line
obstruction to control film in presence of non-volatile phase matched asymptotic analysis 77
thickness 677 681-2 molecular forces 72
operating parameters 677, 695 of sol-gels 682 molecular-kinetic theory 76
stability 547, 566, 680 of suspension of solid particles 20 I multi-valued velocity 336, 526
stagnation line 675 of thin films 681 slip models 77,335
two-phase flow model 689-92, two-phase flow model 689-93 speed limit 63
698-703,699,701 see also Dip coating; Evaporation; stick-slip behavior 83, 85
versus slot coating 676--7,676 Spin coating; Stress buildup in three-phase zone 87
see also Rising film flow coatings with surfactants 335-6, 525,
Discontinuous velocity condition Drying line 680-1, 682, 694 526
531 film profile at 681 wedge flow 76
Discretization, see Finite-element Du Nou}' Ring 112,117-19,533, see also Air entrainment
method; Mesh 683 Dynamic wetting line 9,69-75,531,
Disjoining pressure 43-5, 43, 64, Dye-injection as flow tracer 240-1, 532,659,744-52
72, 190, 192-3, 194 576, 579 optimization of position 484,
profiles 193 Dynamic boundary condition 744-52
Disperse polymer coatings, see 36--41,38,43-5,47,49,55, sawtooth-shaped 86--7,87,88,95,
Polymer dispersions 58-60 658,659
Dispersion Dynamic contact angle 45,69,69, singularity 71
particle intersections in 149 70,71-3,73,75-86,92,95,531
viscosity of 149 apparent 70 Edge effects 490, 491
Dissipation, viscous 382 boundary condition 70, 73 Edge guides
Disturbance 283 empirical correlations 80 slide coating 457
amplitude 283 experimental data 79-86 curtain coating 471-2,489-90,
phase speed 283 flow geometry effect 85 489
wave number 283 hydrodynamic bending and Edge of liquid film
Divergence theorem 21,23,33,305 shaping of meniscus 70,335, meniscus shape 68
for surfaces 36, 306, 325 336 retraction of 189
Doctor blade 540 hydrodynamic theory 76--8 necking-in in coating bead 457
Doctoring 546, 629-30 low-speed effects 80 Eigenvalue problem 283, 322,
coating thickness 629,630 maximum value of 1800 69,70 325
in gravure coating 629-30 measurement techniques 69 computational solution 322-3
regimes 629 molecular-kinetic theory 76 Ekman layer 713
reverse-angle 546, 612 speed dependence 46,69,70, Ekman spirals 721
trailing-blade 546 79-86 Elastic layers 604-5
Domain perturbation methods surface roughness effects 85 elastic foundation approximation
256--8,271-80,282 Dynamic mechanical measurements 605
for slowly varying deformation 144, 151 plain-strain theory 604-5
257-8,258,271-80 Dynamic surface tension 99-100, spring model 605
for small deformation 256--7,257, 108-10,109,131,132-3,132, Elastic-viscous split stress (EVSS)
275 495, 513, 514, 524 formulation 339
see also Asymptotic methods measurement 100, 111-16, Elasticity number 605,607,620
Drag-out, see Dip coating 513-14 Elastohydrodynamic coating
Drop parametric simulation 110 methods 10,599-635
in flow with surfactants 502 for selected surfactants in water / common types 599-603,600
pendant 115 gelatin 124-8 mechanical instabilities 603
Index 773
Elastohydrodynamics 599,603-7 surfactant effects 477 stability of multiple layers 290-2,
finite-element analysis 340-1, 604 trajectory 467,470,472 448-50
traction boundary conditions 607 see also Curtain coating; Teapot surfactant effects on stability 449,
see also Elastic layers; Plates effect 502-3,505-12,505,511,512
Electrostatic assist 6, 12,404,447, Falling curtain method, for diffusion 450,507, 508
480,484 measuring dynamic surface micellization 508-9,509
Electrostatic charges 12 tension 113-14 phase lag 510-12
Elongational viscosity 146, 150, Falling film flow 271-80,271 surface elasticity 507-8, 507
151, 169-73 domain perturbation analysis viscosity stratification 291,
Electrostatic double layer 193 271-80 448-50
Emcol5130 117,122,123 film profile equation 275,278 visualization of slot-exit flow 243,
Empimin LSM30 117, 123,125, inertia effects 277 244-5
125, 126 standing waves 278, 280, 432 visualization of merging of two
Empirical model building 215 surfactant effects on stability layers 243,245
Emulsion polymerization 140 502-3 Film pressure 106
Endoscope 237,243 see also Film flow down an incline Film profile equation 193,275-9,
Energy absorption as a function of FAN method 376,378 472
film thickness 222 Fast Fourier transfonn (FFT) 217, Film profile evolution equation 711,
Entrance length 27 220,238 724-5
Entrance pressure loss 378 Fat edges 457 for uneven substrate 727
Equations of motion 49,265 Feathering 665 Film profiling, see FI uorescence
see also Conservation laws; Fenopon CO-433 117,123 spectroscopy; Imaging
Navier-Stokes equations Fiber coating ellipsometry
Euler-Newton continuation 319 stability of coated film 289 Film retraction, at edge 189
Eulerian-Lagrangian methods, see see also Dip coating Film rupture 189,191, 192-4,
Lagrangian-Eulerian methods Fiber plunging into pool 195
Euler's equation 28,29, 30 dynamic contact angle 78, 79 Film split, see Forward roll coating
Evaporation 289,680-1,693,713-17 Fibonacci search method 742 Film thickness measurement 222-6,
concomitant with coating 680-1 Fick's law 43, 198,690 224
differential 684-5 Filamentation 157, 173, 557 Film thickness unifonnity, see
singularity at drying line 694 Film, see Liquid film Coating uniformity
see also Sol-gel film dip coating Film flow, see Falling film flow; Film Film thinning 190-2,191
Evaporation cooling 197 flow down incline; Rising film surface-tension gradient driven
Evolutionary operation (EYOP) 743 flow; Rotating film flow 191-2, 192
Experimental methods 8,209-50, Film flow down incline 280-92,281, Film viscosity 193
297 429,448-50,505-12 Filtercake formation 652, 664
see also Statistical experimental critical Reynolds number for long- Filtering
design wave instability 285 of fluids 735, 755
Explicit interface tracking 345-6 density stratification 291 of signals 217,219
Explicitly elliptic momentum equipment vibrations 289 Finger strain tensor 604
equations 338 film thickness 429 Finite-difference method
Extension number 381 flow around local disturbance 352 for free-surface flow 350-1
Extensional flow 144 frequency response of two layers for time discretization 332
Extension-thickening behavior 338 530 Finite-element method 299, 300,
Extrusion coating 401,402 heated from below 289 301-34, 543
flow visualization 414 linear stability analysis 281-92, basis function 304
505-7 discretization 304-8
Factorial design 215,742 as method for measuring dynamic discretization errors 308
Falling curtain 472-3,475-8 surface tension 119-20, 120 impostion of boundary
capillary waves 113 mechanisms controlling stability conditions 305-6
finite-element stability analysis 285-8,509-13 steady-state problems 315-19
326,478 nonlinear stability 288 three-dimensional problems 304,
film-profile equations 472 non-Newtonian effects on 308,490
free-fall velocity 31,113,472 stability 289,291,339,449 see also Frequency response
low-flow limit 326,464,475-8, slot exit design 244-5,428-9, analysis; Linear stability
477, 745 429 analysis; Mesh; Mesh
Marangoni effect 479 solitary waves 289 generation; Transient
sensitivity to disturbances 463 stability of single layer 280-9, analysis
stability 326,475-8,745 281,351,352 Finite-volume method 350-1
774 Index
Fixed-grid methods 341-6 Fluorescence spectroscopy 687-8, Free-volume theory of polymerI
dynamic boundary conditions 687 solvent diffusion 198
343-4,345 Foams 104, 501 Frequency-response analysis
local grid adjustment 345 Foil bearing 617 329-30
Flexible blade coating 599,600, Force balance 23,40,51, 52, 55,57 three-dimensional disturbances
601,607-13,608,637,638, Force singularity 71 330
646-53 Forces two-dimensional disturbances
bent-blade regime 609,611,637, centrifugal 709,727 329-30
639,647 body 22 Front capturing methods 342
beveled-blade regime 609, 611, surface 22 Front tracking 341-6
637,639,639,647 Forward kiss coater, see Tensioned- Frontal method 317
blade design 608 web roll coater Frumkin equation of state 107
blade-tip geometry 611 Forward roll coater 10, 157-60, Function optimization 742
elastohydrodynamic action 601, 261-71,539,540,541,552-9
609-12,647-8 air entrainment 538 Galerkin weighted residuals 253,
feed systems 608 applications 540 305,306,307
film-splitting instability 647 asymmetric 264, 555 Gauss elimination 317
inertia effects 609,647,650 bank, rolling 157,558-9,559 Gaussian quadrature 306
loading configurations 608 deformable rolls, see Squeeze-roll Gel slugs 458
low-angle-blade regime, see Bent- coating Gelatin 213,452,483,484,487
blade regime cavitation 157, 557 subbing layer of 65
operability 613 filamentation 157, 173, 557 surface activity 125
operating response 609,610, 611, film-split flow 553-5 Gellation 685
612-13,647 film-split instabilities 157, 556-8 Generator lines 308
on porous substrate 612 film-split location 264, 266, 268, Gibbs' adsorption equation 41,107
puddle-flow instabilities 613 269,270,554-5 Gibbs' complementary condition, see
ribbing 602, 613 finite-element analysis 554-5,554 Gibbs' inequality
streaks 609, 613 inlet flooded 552, 573, 574, 576 Gibbs' inequality 66, 246, 302, 432,
trailing-blade regime, see Beveled- inlet starved 553, 558, 573, 574 467,470,471
blade regime see also Meniscus roll coating in finite-element algorithms 320,
see also High-speed blade linear stability analysis 556-7 334
coating lubrication analysis 261-5, 553-5 Gibbs' surface elasticity 110-11,
Flow 20-47 metered film thickness 270, 555, 498,504,507-8,507
fully developed 27 555,574 Giordano roll coater 139,139, 159
laminar 27 misting 157 Glass transition 200
turbulent 27 non-Newtonian effects 555,555, Glassy regime 153
see also Flow visualization; Fluid 557-8 Glycerin, aqueous solution 213,452
velocity measurement; Free- operability window of 559,560 Graetz number 382
surface flow pressure distribution 157, 262, Gravity number 380
Flow forming zone 303 263,553,553 Gravure coating 10,600,601,
Flow separation, see Recirculation primary flows 540, 542 627-32
Flow visualization 90,237,238, ribbing 157,556-8 bank 630-1
239-43,240,241,576-9,721 singular perturbation analysis cell patterns 627, 628
examples - album of coating 265-71, 543 cell transfer, or pick-out 630-2
flows 216-22,243 string to eliminate ribbing 556-7 differential 627
high-speed 174,666 Fountain applicator, see Jet fountain direct 600, 627
illumination 238 applicator elastohydrodynamic action 602
tracers 238, 240, 242, 576, 579 Fourier analysis 217 feed systems 629
inside vortices 241 Fractal dimension 682, 686 forward 600, 627
Fluid Free-boundary parameterization offset 627
definition 18 308-15,341-6 reverse 627
incompressible 22 Free-falling flow 31,472 ribbing 631
see also Newtonian fluid; Non- Free-Lagrangian methods 347 stability 631
Newtonian fluid Free-meniscus coating 673-705 volume factor 627
Fluid delivery 12 see also Coating by drainage; see also Doctoring
Fluid handling 736 Coating by withdrawal; Dip Grid, see Mesh
Fluid preparation 12,736 coating Grid reconnection 347
Fluid velocity measurement 232-6 Free-surface flows 5,31-45,47-62,
Fluorad FX-I002 117, 123 300-1,301 Hamaker constant 44
Index 775
Hansen approximation 109 Hydrophobically-modified, alkali- Interfacial rheology 503
Healing of coating irregularities swellable emulsion (HASE) Interfacial tension, see Surface
184-9 thickeners 175 tension
Hele-Shaw flow 326 Hydrophobically-modified, Interferometry
Hertz contact theory 621 ethoxylated urethane 173-5, mechanical, see Moire technique
Hexagonal cell pattern, see Orange 178 optical 228, 728
peel Hydroxyethyl cellulose 161,163, stroboscopic 728, 730
High-speed blade coating 637-69 167,168,172,557,664 Internal design of coating dies, see
air entrainment 641,657-9 Hysteresis Coating dies
applicator configurations 640--6, contact angle 46,65-6,335,467, Invariant, of tensor 154
641 526 Inventory control, of product 762-3
blade design and loading 639 hydrodynamic 469 Inverse iteration 322, 325
blade tip geometry 639,651,667 Inverse problem 333-4
blade wear 651-2 Igepal CA-720 117,123 Inverted blade coater 638,641
computational analysis 650 Image processing 238-9,244,576, Irrotational flow 29
cross-web uniformity 334,641, 579 Iterative matrix solvers 318-19
659 Imaging ellipsometry 687, 687 Iterative solution of non-linear
control 740--1 Imaging techniques 236-9, 579 problems 315-17,742
defects 663-8 Immersion of substrate 673
deformation rates 640, 660, 662 Immobilization solids 664 Jacobian,ofmapping 313
equipment 640--6,641 Implicit time integration 331 Jacobian matrix 316,320,322
inertia effects 650 Implicit volume tracking 342-5 Jet coater, see Jet fountain applicator
instabilities 641, 653-6 Inclined film flow, see Film flow Jet fountain applicator 642,644-5
lubrication-flow models 648-9 down incline manifold design 645
maximum coating speed 641,657 Inclined plane method for measuring
metered coating thickness in 646, dynamic surface tension Kelvin-Helmholtz instability 342,
648,650-1,651 119-20,120 483
metering action 646-53 Infrared absorption, for measuring Kinematic boundary condition
non-Newtonian effects 652-3, film thickness 222-3,223,224 32-5,47,55,60,61,302,690
660--3 Initialization of finite-element Kirchhoff-Love assumptions 606
normal forces 661-2 calculation 321 Kiss coating, see Tensioned-web roll
streaks and striations 658,659 Instabilities 6, 403 coating
time scales 639, 660 centrifugal 654-6 Knife coating 258-61,539,540,
viscoelastic effects 660, 661 free-surface, or wave 285 541,547-552
see also Flexible-blade coating; interfacial wave 290,394 applications 540
Inverted blade coater; Jet long-wave 285 configurations 540
coater; Paper coating process; shear-flow 284 finite-element analysis 548
Pressurized pond coater; spatial 283 flooding 552
Puddle-type blade coater; Roll symmetry-breaking 323, 653-4 lubrication theory 258-61,
applicator; Short-dwell coater temporal 283 547-9
High-speed photography 174,237 Integral methods 253,289,351-2 metered film thickness 260,
High-speed video 237,576,579 Interface 19,31-45,47-62 548-9,549,646,660--1
High-solids coatings 141 air-liquid 44,48, 54, 59, 61 non-Newtonian viscosity effects
Hookean solid 341 fluid-fluid 31,34,35,47 548-9
Hopf points 320, 323, 327 impermeable 34, 497 operability window of 552, 552
Hopkins flow-separation condition Interface reconstruction 343 pond-feed 540
262, 263, 270 Interfacial boundary conditions pressure profiles 260,549,550,
Hopper, see Coating die 31-45,301-3,302,690--2 551
Hydraulic radius 375 mass balance 32 ribbing instability 549, 551, 661
Hydrodynamic assist 89-90,463, normal stress balance 48-54 separating forces 660--1
484 species balance 42,690,691-2 starvation 552
Hydrogen bubble method 235-6, stress balance 36-41,58-62,302, steady flow 547-9
240,241,241 691 streaks 552
cathodic corrosion protection tangential stress balance 54-61 viscoelastic effects 548-9, 552,
241 see also Dynamic boundary 660--1
Hydrostatic pressure distribution condition; Kinematic Krebs units 149,161,162,164
25 boundary condition
Hydrophile-hydrophobe balance Interfacial phenomena 19-62, Ladyzhenskaya-Babuska-Brezzi
(HLB) number 101 107-11,498,689-92 (LBB) condition 304
776 Index
Lagrangian methods 343, 346-50, Leveling of coated films 8, 53, 54, failure of ad-hoc boundary
354 159-60, 184-9, 185, 186,724-6 conditions in 264, 270-1, 554
Lagrangian mesh generation 315 capillary-pressure driving force for flow in liquid film 274
Lagrangian-Eulerian methods 305, 54,186 for flow between solid surfaces
345,348-9 film-thickness effect 161 259-60,542-3
see also Arbitrary Lagrangian- inverted films 188 requirement for success 260, 264
Eulerian (ALE) methods mechanisms 185-6 two-dimensional 376, 377
Lambert's law 222 Newtonian liquids 185-8 see also Reynolds equation
Landau-Levich theory of film non-Newtonian liquids 167-9,
deposition 279-80,583,677, 188, 339 Magnetic assist 404, 414
678 surface-tension gradient effects Manifold design, see Coathanger
Langmuir adsorption isotherm 108 188-9 manifold; Coating dies; Jet
Langmuir constant 107, 108, 115, time requirements 186-7,188, fountain applicator; T-manifold
498 725 Mapping 308
Langmuir-Blodgett films II, 682-3, uneven substrates 724-6 iso-parametric 305, 349
682 viscosity effect 162 sub-parametric 305, 313
applications 682 see also Spin coating Marangoni drying 674
deposition 503, 683, 709 Lid-driven cavity 580, 654 Marangoni elasticity III
microstructure control 683 cellular flow 655 Marangoni flow 38,41-3,54-62,
Langmuir-Knudsen equation 196 stability of flow 654-5 103,104,192,477,699,700,
Langmuir-von Szyszkowski Light-guide rod collector 231 718-20, 723--4
equation 107, 108,498 Light-sheet illumination 238 Marangoni instability 198,718-20
Laplace equation, see Young- Limit points, see Turning points Marangoni number 198,719,720
Laplace equation Linear equations, solution of Marker-and-cell (MAC) method
Laser-Doppler anemometry (LDA), 317-19 344
see Laser-Doppler velocimetry Linear stability analysis 280-92, Mass transfer 196-7, 688-91,
Laser-Doppler velocimetry 87, 300,321-7,505-7,654-5 695-91,695-8,695-8,713
233-6,233,490 finite-element method 321-7 Mass transfer coefficient 197, 690,
applications to coating flows method of normal modes 283, 696
234-5,490 505 Mass transfer models 197,690,696
Laser scanners 230-2, 230, 232, quasi-static condensation method Material function 138, 142
735 325 Matrix solver 317-19
Laser strobe 238 three-dimensional disturbances direct 317-18
Latex 140-1 323-7,654-5 iterative 318-19
as additive to alter rheology 213 two-dimensional disturbances Maximum bubble pressure method
coalescence 140 321-3 114
preparation 140 see also Orr-8ommerfeld equation Maxwell model 155,156,339
stability 141 Liquid film Membranes, see Plates
Latex coatings 160-75 fate after coating 183-205, 183 Membrane coating 10,599,600,
drying 200 healing of irregularities 184-9, 601,613-17
effective volume fraction in 163, 185, 186, 194-6 air entrainment 616
164 non-wetting 192, 196 coating uniformity 616-17
elastic response 167-9 rupture 189, 191, 192-4,195 configuration and components
elongational viscosity 169-73, see also Film flow down incline; 600,614
172 Leveling; Thin liquid film defects 616
formulation 141, 160-1,161 Lodyne S-103 117, 123 elastohydrodynamic action 60 I,
leveling of 167-9 London-van der Waals dispersion 614-15
particle size effect on rheology forces 44, 107 finite-element analysis 615-16
161,163,163,164 Lot size, of coated product 762 minimum film thickness 617
roll spatter 169-73,171 LV factorization 317,319,320,328 operating response 615-16,615
thickener effect on rheology Lubrication approximation 93, 94, pool-flow instabilities 616
161--4,173 259,376,603,693,723 shearing action 613
transient structure formation and Lubrication flows 30,258-61,542-3 Meniscus
rheology 167,167,168,169, Lubrication theory 258-62, 542-3, advancing in capillary tube 72-3
176 582-4,603-4 asymptotically planar, against flat
viscosity at high shear rates boundary conditions 260, 262, plate 50-2,51, 67-8, 68
161-2,162 553--4, 583 in coating bead 73--4, 73, 74, 402,
viscosity at low shear rates derivation with domain 524
164-7,163 perturbation method 265-7 dynamic 68-9
Index 777
hydrodynamic bending near Microstructure of coatings 11, 674 Newton's iteration method 306,
dynamic wetting line 71,77 control in dip coating 682-7 316-17,319,320,327,340,350
between parallel plates 48-52,48, control through drying 703 basin of attraction 319
53 see also Langmuir-Blodgett films; derivation of Jacobian matrix
quasi-static 50 Sol-gel films 316
rise in capillary tube 66-7, 67, Microwave absorption Niaproof #4 117, 121,122, 123,
112 for measuring film thickness 127
static 64, 66-8 223-4 Nodes 304,308
for tape entering pool 69,75 for measuring moisture content spacing 309,312
Meniscus roll coating 11, 540, 223 Non-Newtonian fluids 6,7,85,89,
573-97 Misting 157, 643 137,138,158,289,445,668-9
applications 540 Mixed-mapping method 313 examples 149-53
back jet 588,589,590-2,590,591 Modulus in finite-element computations
bead break-up 595 complex 145 337-9
bead collapse 596 dilational surface elasticity 110 Nonwetting liquid 191
bubble generation 596 dynamic 146 Normal stresses 142,374,380-1,
capillary action 574 loss 145, 152 661-2
double eddy structure 585, 586 shear 604 Normal stress coefficient 144, 381
feed condition effect 586-90 storage 145, 152 Normal stress number 381
finite-element analysis 584-5, see also Young's modulus Normal vector to surface 34
584,585 Moire technique 227-30, 228,229 No-slip condition 35,47,60,61,71,
film thickness ratio 584, 585, Moisture content, measurement with 302
587 microwave absorption 223 Nucleation, of dry patches 189-90,
flow structure 576-85 Molecular forces 43,65,190,193 191
flow transitions 585-92 Molecular-kinetic theory 76, 80, 93 Numerical methods, see
flow visualization 575,576-7, Molecular-rate process, see Dynamic Computational methods
578,579 wetting Nusselt film flow 281-2,431,678
lubrication theory 582-3 Momentum equations 305, 307
parameter ranges 592-5,592 see also Conservation laws; Original equipment manufacturer
pressure distribution 580, 584, Navier-Stokes equations (OEM) coatings 138, 139,
587 Momentum number 380 141-2,159,175-7,177
pressure field transformation Mottle 197,221,458 formulation 141-2
585-6 Mudcracking 201 rheological response 175-7,177
small-flux model 580-4 Multidimensional scaling method Olin lOG 117,123,124,124
snaking fluid stream 574,575, 217 Operability analysis
581,588,590,591,592 Multilayer coating three-dimensional disturbances
start-up 573-4 choice of surface tensions 106, 328
streamline pattern 580,581,586 458 two-dimensional disturbances
ultra-starved inlet 588, 592 edge control 520-1,525 327
zero-flux model 580 leveling 189 Operability window, see Coating
Mesh 308 surfactant optimization 130-3, window
boundary conforming 299, 308 520-1 Operations model 762,765-6, 766
two-level 340,692-3 see also Coating dies; Curtain Optical sectioning technique
unstructured 349 coating; Film flow down 239-42,240,241
Mesh generation 308-15,341-6 incline; Slide coating; Optimization 11,215, 333-4,736-8,
adaptive 313 Tensioned-web slot coating 741-66
algebraic 308-11 Multiphase flow 20,20 algorithms 741-4
elliptic, or differential 311-15 applications 744-66
generalized algebraic 309-10 Nahme number 382 constraints 741
Lagrangian pseudo-solid 315 Navier-Stokes equation 24, 25-6, criteria 741
Mesh orthogonality 312 93,259,265,269,302,304,315, macro 738
Mesh refinement 72,313,692-3 432,505,605,607,654,691-3 micro 737
Mesh smoothness 312 Navier's slip boundary condition of operability latitude 334
Methyl methacrylate 140 302 of shape of coating device 334
Micelles 101,102,496,499-501, Nekal BX 117,123,127 Orange peel 198,718
500,508-9,509 Neural networks 744 Ordinary differential equations
kinetics of formation 499-501 Neutral curve for stability 284, 556 system (ODES) 331
Microscopes 237 Newtonian fluid 24,154,157,158, Orr-Sommerfeld equation 283,
mono zoom 242 185,188,235,288,603 448-9,503,505,506,505-7
778 Index
Oscillating jet method for measuring Picard iteration, see Trial method Predictor-corrector scheme 319,332
dynamic surface tension PID control system 740 Pre-metered coating methods 9,
112-13, 118-19 Pigment 140 399-536
Oscillatory shear flow 144,151 chemical pretreatment 173--4 versus self-metered methods
Overflow cell for measuring dynamic extender 141 675-7
surface tension 114-115, Pilot coating equipment 211-13, Premicelles 499
513-19 211,212 Pressure-Poisson equation (PPE)
cell design 513 Pinholes 194 formulation 304,307-8,319,
computational analysis 514-19 Pinning of wetting line, see Static 333
flow structure 515 wetting line Pressurized pond coater 643, 645-6
Pipe flow, pressure drop 26 air entrainment 646, 656, 659
Paint, see Latex coating Pivoting 317 cellular flow 655-6, 657
Paper coating process 138-9, Plates 605-7 centrifugal instabilities 646,
637--69 bending stiffness 606 655-6
air entrapment in pores 650 clamping conditions 607 see also Short-dwell coater
compressibility effects 649,650 contact between 607 Priming layer 12
liquid penetration into pores 175, finite-element analysis 606-7 Primitive-variable formulation
652 linear elastic theory 605-7 304-7
non-Newtonian effects 659-63 Plug flow 30, 31 Process capability 739
porosity effects 649, 650 PMMA solution 715,715 Process control, see Control
relaxation after application 666 Poiseuille flow 26-7,29,30,582 Process optimization, see
roughness effects 649,650 Poisson equation 313 Optimization
swelling 652 Poisson's ratio 201,604 Product
see also High-speed blade coating Poly(ethylene oxide) thickeners 164, application 3
Paper coatings 141, 175,638,640 173 design 333--4
formulation 141,640,659,664 Poly(ethylene theraphtalate) 69,81, fabrication 737,738
rheology 175,659--60,661-3,666 81,82,83,84,87,87,88,88,89, specifications 739
runnability 175,663 92,484 unit-cost minimization 761-6,
structure formation and break-up Polymer additives for enhanced 765
175,660,662 coating performance 137, Profilometry 226-7
transient rheological response 160-77,213,448 Property parameter 467
176,660 Polymer dispersions 141-2,175-7 Puddle coater 638,638,641,642-3,
Parallel-plate rheometer 147 Polymer melts, dynamic uniaxial 642,644
Parameter optimization 741 extensional viscosity 150 air entrainment 642
Particle defects 458 Polymer solutions misting 643
Particle image velocimetry (PlY) dynamic mechanical analysis 152 one-sided 642
236,576,580,582 clastic effects 564 two-sided 642, 644
Particle-in-cell (PIC) method 343, rheology during gellation 145 II-theorem, see Buckingham
349 shear-thinning 564,717 theorem
Particle methods, computational Polymeric resist films 710
349-50 Polymers Quality assessment 216
Particle orientation 665 amorphous 152 Quality control 217
Particle tracking 238, 242, 576, 579 cross-linked 152 Quality index (QI) 220
Peclet number 328 grafted 140 Quasi-Newton methods 316
Peeling of coating 199 water reducible 141-2,175-7
Penalty function formulation 307 water-soluble 140,163-7,173-5 Radial striations 718-20
Perturbation methods, see Polystyrene solutions 730-1, 731 Rakelines, see Ribbing
Asymptotic methods; Domain Poly(vinyl alcohol) 165,452,664 Raoult's law 691
perturbation methods; Regular Polyurethane based, aqueous Rate of deformation, see Rate of
perturbation methods; Singular dispersion (PURAD) 176-7, strain
perturbation methods 178 Rate of strain 142
Petrov--Galerkin method 306 Potential flow 29 in coating processes 160,337,
Phase inversion method 198 Powder coating 189,189 565,640,660,662
Phase separation in coatings 198-9, Power-law index 154,213,375 extensional 144
199,717-18 Power-law liquids 375,376,384, shear 143
Phonographing, see Ribbing 385 Rate-of-strain tensor 24, 154
Photolithographic process 709,710 Power-law model 154,337,375 Rayleigh instabilities 159
Photography 237,238 Preconditioning of matrix 318-19 Rayleigh-Taylor instability 342,
high-speed 174, 237 Precursor film 65, 77 483
Index 779
Real-time imaging 239 Ribbing 157, 158,158,159,323-7, dip coating in 547, 566, 679
Recirculation, or flow separation 418,447-8,480,544-6,544, falling film flow 271
defects generated by 546, 549-52,556-8,557,563,566,613 feed systems 540, 541
755 critical capillary number 158, gap setting 577,578
Refractive index 233,686-7,686 545,549,556-7,557 primary flows 540-2, 542
Regular perturbation methods effect of divergence of flow sensitivity to process
254-5,255,268,282,285 channel 549 perturbations 546
Relaxation time 153,155,660 elongational viscosity effect 159, scraping, or wiping 579,612
for micellization 500 557-8 starvation 546
Repellency spots 458 frequency as a function of see also Forward roll coating;
see also De-wetting capillary number 158 Meniscus roll coating; Reverse
Residence time 385 inertial mechanism for 545-6 roll coating; Squeeze roll
Resins, see Water-reducible resin interface profiling 227 coating; Tension-web roll
coatings interfacial force balance 544-5, coating
Reverse roll coating 10,539,540, 544, 551 Rotating film flow
547,559-67 linear finite-element stability inertia-driven ribbing instability
air entrainment 566 analysis 326, 420, 448 545
applications 540 lubrication theory 326 Rotating flow, stability 546
cascade instability 563 mechanisms 157-8,544-6,563 Rubber
coating window 566, 567 with static wetting line 549 deformable roll cover 341
configurations 541 string to eliminate 556-7 dynamic mechanical analysis 152
dynamic wetting line position surfactant effects 447 Rubbery regime 153
560,561 three-dimensional analysis for Runnability 175,663
failure modes 563-5, 563-6 large amplitudes 557,558
finite-element analysis 561 viscoelastic effects 410 Sakiadis boundary layer 480, 484,
lubrication-flow analysis viscous mechanism 545 485-6,488,696,748
559-60 see also Forward roll coating; Slide Sagging 188,450
metered film thickness 562, 562 coating Schmidt number 197, 497
metering gap flow 559-63,560 Rising film flow 271-80,271, Schur-Wielandt deflation 322
nip-transfer flow 565-6 678-80 Scraper blade 579
non-Newtonian effects 563,564, domain perturbation analysis see also Doctor blade
565 271-80 Scratches 663, 664
pressure profiles 565 film profile equation 275,278 Screening experiments 215
primary flows 540, 542 stability 547, 566 Search methods 742
ribbing instability 563, 565 see also Dip coating Segregated solver approach 319
Reynolds equation 260, 341, 380, Rivulets 418,595 Self-metered coating methods 10,
542,565,583,603,605,607, Robin boundary condition 303 537-672
609,621 Robustness of coating process 751, versus pre-metered methods 574,
derivation of 259-60 759 637,675-7
Reynolds number 27,28,30,48,49, Roll Sensitivity analysis
185,197,284,285,287,289, applicator 540,541,638 three-dimensional disturbances
303,326,406,439,467,473, back-up 540,541 330
481,507,510,511,543,654, cantilevered 577, 577 two-dimensional disturbances
721,758 fountain 540, 541 329-30
Reynolds transport theorem 20, 23, impression 540, 541 Sensors 213-14
33 metering 540,541 Shape factor 374, 375, 754
Rezoning 347 transparent 243,416,578 Shear-thinning behavior 24, 151,
Rheology 137-79 Roll applicators 154,165,166,337,375,376
see also Rheometry for high-speed paper coating 641, Shell theory, cylindrical 605-7
Rheometers, see Capillary rheometer; 642,643-4 in finite-element method 341
Cone-and-plate rheometer; limitations 644 Sherwood number 197
Couette rheometer; Parallel-plate for paints (mat roll) 138 Short-dwell coater 639,641,643
rheometer Roll coating 539, 540, 541 defects 646
Rheometric flows 142-9 air entrainment 546 maximum speed 646
Rheometry 7,137,662 auxiliary flows 542 see also Pressurized pond coater
simulation of deformations in common configurations 539-42, Silicone oils 579,579
coating processes 160,662 541 Simple-line interface calculation
see also Dynamic mechanical dampening of inlet disturbances (SLIC) 343
measurements 546-7 Simplex method 740,742, 743
780 Index
Singular points, see Critical points viscosity stratification 328 Sol-gel film dip coating 680--1 ,
Singular perturbation methods velocity field near dynamic wetting 684-7,685,688-93
255-6,256,265-71,272 line 336 asymptotic analysis for single-
Skinning 197, 681 vortex formation, or recirculating component solvent 693-4
Skips 663 flow 429,439-43,440 compaction 686
Slide coating 9,427-62,519-31 wetting meniscus in 73, 74, 74 computational analysis 694-703
air entrainment 446-7 see also Film flow down an incline single-phase flow 694-7
application angle 246,430--1,430 Slip model, see Dynamic wetting; two-phase flow 698-703
applicability 428 Navier's slip boundary evaporation effects 684-7,
approximate force-balance condition 693-703
analysis 433-40 Slot coating 9,401-26,401 film thickness 686
barring 448-51 air entrainment 422 film profiles 685,687-9,687,689,
bead 450--1,432,433 applications 402 702
bead break-up 74, 546 barring 418,421 flow and concentration fields
bead oscillations 451 critical point tracking 328 696--8, 701
bead vacuum effect 245-6, 428 coating window 403,417,419, gellation 685,700
bleeding, or weeping 444 420,421 mass transfer model 696-7
coating gap 431 die-lip design 403-4, 404, 406 microstructure control via drying
coating window 330,331,451-3, die-lip edge sharpness 404, 422 703
451,452,453 dynamic behavior 421 microstructure formation 684-7,
critical point tracking 328 equipment 403-5 685, 700--2
die-lip design and its effects 246, flow regimes 407 multi-component systems
431-2,448 flow visualization 414-16 694-703,695-9,701-2
die-lip edge sharpness 431-2 maximum gap 419 ribbing-like instability 695
dynamics, or frequency response meniscus shapes 406,408-10, shear rates 701,702
329,454-6,455,479,480, 677 Sol-gel films II, 683-8
529-31,530 multilayer 402, 404-5, 405, 415 applications 684
flow 314,314,443,521-31,521 numerical simulations 411-13, drying stresses 682
flow visualization 243-6, 243, 440 411,412 formulation 684
free-surface profiling 227 operating modes 409 polymerization 684
finite-element mesh 314 operating parameters 406-7, 677 pore structure 686
largest gap 73 operating limits 328,408-10, refractive index versus withdrawal
linear finite-element stability 416-20 speed 686--7, 686
analysis 323, 324, 326 pressure control with vacuum SOLA-VOF method 344
maximum bead vacuum, or box 404,412 Solidification 13
pressure difference 328 pressure distribution 408-9,412 Solvent evaporation 684,713--17,
minimum film thickness, or low- recirculating flow 410,412,413, 728-30
flow limit 246,437-8,438, 415 Solvents
446-7 ribbing instability 410,418,420 multiple 684
multilayer 427,445 rivulets 418 organic 141
non-Newtonian effects 445 sensitivity to small disturbances Spattering 169,171
operating limits 328, 445-53 420 Spectral element method 351
quality window 330, 331 starvation, or low-flow limit 418, Spectral methods 283,339,351
ribbing 326, 328, 447-8, 528 419 Spin coating 11,709-34,709
sensitivity to disturbances 453-6 steady-state operation 405-16 air flow effects 713, 721
slide inclination 431 streaks 421 air flow instabilities 721
slide length 431 surfactant effects 503 air flow visualization 721
slot-exit design 244-5,428-30, versus dip coating 676-7,676 applications 709,710
429 viscoelastic effects 411, 415 circumferential waves 719
standing waves 432 weeping 418 colloidal suspensions 716,
static wetting line position 431-2 see also Tensioned-web slot 718-20,720
steady-state operation 432-45, coating concentration-dependent viscosity
521-9 Slot flow 376-8 effects 716-17
streaks 429, 456 pressure drop 29 defects 717-21
surfactant effects 445,521-31 Slurry coatings 198 deposition 710
surface-age time scales in 103, Soap film, axisymmetric 40 diffusive transport 713
103 Sodium dodecyl sulfate (SDS) 117, edge effects 713
three-phase zone 87 121,123,128,130,506,516-17, evaporation 710,711,713-17,
vacuum box 432 516--18,522,524 728-9
Index 781
film profile evolution 712, 728, Starry night defect 201 Subbing layer 12
730 Start-up of coating process 330, swelling of 65
final film thickness 714,715 402,430,456,490,573-4 Subdomain 304
mass transfer 713,714 Static contact angle 45, 64-5, 75 Substrate 12
Marangoni effect 718-20,723-4 see also Contact angle; Gibbs' air supported 480
Meyerhofer's analysis 714-16 inequality primed 12
non-Newtonian effects 712, Static surface tension 99-100, subbed 12
716--17 107-8,131 uneven 721-31
operability diagram 719 measurement methods 111-16 wavy 480
planarization 721-4,725,726, for selected surfactants in water Successive overrelaxation (SOR)
727-9, 729, 731 and gelatin 120-3,121 319
processing stages 71 0-11 Static wetting 7,63,64-8 Successive substitution, see Trial
spin-off 710 finite-element analysis 334-5 method
spin-up 710 with flow 64, 68 Surface acoustic wave (SAW)
stability 717-21 Static wetting line 9, 66--8, 744 technique 688
surfactant effects 720, 721 mobility 66 Surface-active agents, see
thinness of coatings 709 of interface between miscible Surfactants
uneven substrates 721-31,722 fluids 335 Surface age 120,124,515-16
leveling after spinning 724-6 pinning at corner 66, 404, 422, for surfactants in water 129
leveling during spinning 431-2,467,552 Surface compression 103, 111,497
726--31 Statistical clustering analysis 743-4 Surface convection 497
rectangular bump 728, 730 Statistical experimental design Surface diffusion 497
rectangular trench 724, 724, (SED) 214-16,742 Surface dilation, or expansion 103,
725 Statistical process control 739 111,115,497
uniformity of coatings 709, Steady-shear flow 142-4 Surface elasticity, see Gibbs' surface
712-13,716--17 Step-in-strain flow 145 elasticity
Spines 309 Step-in-strain-rate flow 146 Surface elasticity number 498, 511
Splice passage 330,431,456,480, Stick-slip behavior, see Dynamic Surface heterogeneity 65
745 wetting Surface porosity 65
Spray coating 12, 674 Stochastic fluctuations 741, 756 Surface pressure 683
Spreading 105-6,458,502 Stock of coated material, flow Surface profilometry 226--7
halting at corner 66 through 762-3,763 mechanical 226
see also Dynamic wetting; Gibbs' Stokes flow 277,580,582,584 optical 227
inequality Stokes number 53, 303, 380 Surface roughness 65, 85, 89
Spreading coefficient 106 Stormer viscometer 149, 162, 164 Surface tension 38,44,51-4,54,55,
Spreading pressure 107 Strain hardening 150 56, 105, 106
Sputtering 3 Streaks, or streaklines 219,421, concentration dependence 41,
Squeeze-roll coating 10,600,601, 429,456,471,547,552,608, 56--8,108,120-3,122,123,
620-7 613, 663, 666, 755 498,692,695
dry contact limit 621,622,624 Streamlined finite-element method in interfacial stress tensor 38
elastohydrodynamic action 602, 309,349 measurement techniques 111-16
620 Streamline visualization 238-40 of mixtures 692
elastic roll models 604-5,621-2 Stress 23-4, 142 molecular contributions to 106--7
film-split patterns 602, 603 anisotropic, at interface 43 temperature dependence 41,54,
gap profiles 623, 624 interfacial 36 59
metered coating thickness 625, normal 142,374,380-1,661-2 see also Dynamic surface tension;
626 shear 142,145,155 Static surface tension
nip adjustment or loading 620 yield 155,660 Surface-tension-gradient driven flow,
nip mechanics 620-2 Stress buildup in coatings 8, see Marangoni flows
operability limits 626--7 199-201,681-2 Surface tension gradients 41-3,55,
pressure profiles 623, 624 defects from 199 57,59,191,198,531,681
rigid-roll limit 621,624 Stress growth function 145,153 Surface topography measurement
sensitivity 627 Stress overshoot 153 226-32
viscoelastic roll models 622 Stress ratio 381 Surface traction vector 23
Stability 280-1 Stress tensor 23, 24 Surface viscosity, dilational 110-11,
spatial 283 Striations, see Bandedness; Radial 504
temporal 283 striations Surfactant actions 495-533
see also Linear stability analysis Stroboscope 238 coating aid for defect control
Stalagmites 665, 666 Styrene 140 104,495
782 Index
Surfactant actions (contd) meniscus 69,75 semi-implicit scheme 333
in coating flows 42, 101, 102-3, Teapot effect 467-71 time discretization 332, 348
103,496-501,496,501-5, deflection of liquid stream 467, time-step size control 332
519-21,521 470 Transport theorem 32, 33
at dynamic wetting lines 525, 526 finite-element analysis 469,470, see also Reynolds transport
finite-element analysis 339-40, 471 theorem
504-5 fixed-grid simulations 344 Traube's Rule 42
immobilization of interface 42 hydrodynamic hysteresis 467 Trial method 315-16,350
influence on product properties lip design to prevent 471 Triangulation method 227
104-5,495 Tensioned-web roll coating 10,600, Triton 770 116, 123
promotion of wetting and 601,617-18 Triton X-I00 121
spreading 99, 502 elastohydrodynamic action 601, TritonX-102 117,121
retardation of interfacial flow 56 617 Triton X-200 117,123
stabilization of coating flows film-split patterns 602 Turning points 320,323,327,338,
505-13 finite-element analysis 310, 311 470,471
stabilization of coating solutions foil-bearing theory 617
104,495 forward mode 310,311,600,617 Upwinding 306, 338
theory 496-501 operating limits 618 Uniformity index 383
time scales, characteristic 99, 103, reverse mode 617
124-8,508,509,514-15,522-8 reverse wiping with complete liquid Van Karman similarity solution
see also Surfactants transfer 618 714,715
Surfactant modified water-soluble rolling bank 311 Video techniques 237
polymers 173-5 stability 618 high-speed 237,576,579
Surfactants 7,99-136,289,445, web instabilities 617 Vinyl acetate 120
494-536,682-3 Tensioned-web slot coating 10,405, Viscoelasticity 24,66,167-9,175,
adsorption at interface 99, 101, 600,601,618-20 188,660,667,669
107-8,498,523--4 control of coating bead stability constitutive models 154
adsorption density 109 620 deformable roll covers 341
adsorption isotherm 107 die-lip design 619 die-internal flows 374,381
aggregate structures 101, 102 elastohydrodynamic action 601, in finite-element computations
classification 101, 116 619 338-9
convective diffusion 339-40, multilayer configuration 619 linear regime 144
497-8 Texapon K-12 117,122,123,123, Viscosity 24, 143
convective redistribution 42 128 complex 144
diffusion 108-10,508 Thermocapillarity 59,59,289 concentration-dependent 756
diffusion-controlled adsorption Thick edges 457 dynamic 144-5
108-10, 125 Thickeners 163-9,172-5,213,740 elongational 146, 150, 151,
in emulsion polymerization of see also Latex coating 169-73
latex 140, 161 Thin air film 87,92,473,483,658 planar elongational 144
excess surface concentration 41, finite-element analysis 336-7 shear 143
107,497 lubrication-flow model 93-6 temperature dependent 382,756
inventory in solution and micelles molecular forces 71-2 uniaxial elongational 144, 150,
496,499-501,500 stability and collapse 45, 483 178-9
patents 105 see also Air entrainment Viscous number 380
properties of 101-2, 116,506,522 Thin liquid film 43-5, 44 Volume factor 627
recruitment depth 506 drying 681 Volume-of-fluid (VOF) method 343,
selection criteria 128-30 left behind after dewetting 194 344
see also Critical micelle molecular forces 71-2 Vortex instabilities 566,567
concentration; Langmuir- overlap of interfacial regions 44 Vortex methods 350
Blodgett films; Micelles; stability and break-up 45, 192-4
Surface elasticity Thixotropy 146,660 Wall shear stress 375,389,395,
Swift-Steiber conditions 262-3,270 Titanium dioxide 141 753-5, 754
Total linearization method (TLM) Water bell 477
T-manifold 369,380 317 Water retention 664, 665, 666, 668
design 383-6, 386 Transfinite interpolation 310-11 Water-reducible resin coatings
Tape plunging into pool 69,75 Transient analysis 330--4 141-2,175
air entrainment 86-9,87,88,446 implicit scheme 331 formulation 141-2
dynamic contact angle 69,69,75, large amplitudes 301 rheological response 175-7
75,81-4 linearized equation set 329 Wave instability 285
Index 783
Wave number 283 Weissenberg number 153 Wilhelmy plate 111-12,119-20,
Waves Wetting 63-97, 105-6 533,683
damped, with surfactants Ill, complete 64, 105 Winkler approximation 605
502, 505-13 kinetics 658 Wire coating, see Fiber coating
interfacial 290 maximum speed of 86,641, WFL equation 382
solitary 289 658
standing 278, 280, 432 partial 105 X-ray fluorescence (XRF) 224-5,
Web cleaning 12 see also Air entrainment; 224,225
Web handling 12 Dynamic wetting; Static
Web instabilities 617 wetting Yield stress 155, 376, 660
Weber number 48,49,64,70,73,75, Wetting lines 63,64 Young-Laplace equation 29,38,
85,86,90,93,326,477-8,543 in computational analysis 64-9,77, 112, 114
Wedge, flow in 245 334-7 disjoining pressure contribution
Weeping stationary with flow 64, 68 to 44
paper coating 175, 665-6 see also Dynamic wetting line; Young's equation 64,70,105, 106,
slide coating 444 Static wetting line 526
slot coating 418 Wicking 66-7 Young's modulus 145,201,604
CONTROL AND OPTIMIZATION OF 15
COATING PROCESSES
Peter M. Schweizer
15.1 INTRODUCTION - THE IMPORTANCE

OF CONTROL AND OPTIMIZATION
control systems are only necessary, if something
went wrong upstream in the process.
One of the focal points of many quality improve- The use of feed-back control is most meaningful
ment programs is the concept of prevention. during the learning and development phase of a
Within the scope of coating, prevention refers to new process. One could even postulate that the
'do it right the first time' by avoiding defects purpose of feed-back control is 'to put itself out of
which would have to be removed at a later stage business' by constantly climbing up along the
of the manufacturing process. For example, learning curve, and by changing know-how into
producing coating solutions free of bubbles and know-why, thereby approaching the philosophy of
other unwanted particles would make it unnec- prevention. An example is a laser scanner installed
essary to install excessive and costly de-gassing downstream from the coating point In its traditional
and filtering systems. Furthermore, the finishing use, it is an element of a feed-back control loop.
operation would be more efficient because there However, it can also be used to help eliminate
would be no need to cut out spot defects. defect-generating processes upstream from the
In its purest sense, the manufacturing philosophy coating point by searching for the problem causes
of prevention excludes the use of traditional every time the scanner signal exceeds its threshold
process control with feed-back loops, because, level. In this fashion, fewer and fewer defects will
as is schematically shown in Fig. 15.1, such occur until, ideally, there will no longer be a need
for a scanner on the coating machine.
An alternative control algorithm which is
more in line with the concept of prevention is
feed-forward control. In contrast to feed-back
control where the effect of a disturbance is
Control
Valve
measured downstream from the process and the
corrective action is fed back upstream to ma-
nipulate a suitable variable, the disturbance in
______________Jt
feed-forward control is detected upstream from
Control Signal
the process, and the corrective action is imposed
L Set Point
on a suitable variable also upstream, so as to
avoid a disruption of the process by the disturb-
Figure 15.1 Schematic diagram of feedback control. ance, see schematic diagram in Fig. 15.2.
736 Control and optimization of coating processes
parameters, where feed-back or feed-forward

Output Variable
loops have reached a high degree ofsophistication,
include the web speed and flow rate in premetered
coating processes, as well as the speed ratio
,, between two rolls and the load on the doctoring
,, blade in self-metered coating operations. The aim
,,
,, of control is to maintain variations of relevant
1 , _
process parameters at 'acceptable' levels. This

will result in significant benefits such as more
Figure 15.2 Schematic diagram offeedforward control.
consistent product uniformity, increased produc-
tivity, raw material savings, timely and accurate
process information, improved operator and
Feed-forward control may be encountered in management decision tools, etc. (Swain 1985;
the fluid preparation and handling system of a Shapiro 1993).
coating plant. Here, vessels, delivery lines, pumps, In contrast to process control, optimization
filter housings, etc. are often cleaned with provides a framework for implementing the
automated CIP systems (CIP = Cleaning In concept of prevention by designing effects of
Place). In the photographic industry, for example, parameter variations out of the system. The
cleaning efficiency is high, if the temperature of purpose of process optimization is to determine
the detergent solution is kept at an elevated level, (optimize) those operating conditions (independent
thus requiring a recirculation loop which comprises process parameters, pJ which lead to extreme
a heat exchanger and a reservoir. The challenge values (maxima or minima) of selected dependent
for the control loop is to maintain a constant variables Xj' With regard to coating processes,
detergent temperature under conditions of ran- one might ask to set values for the viscosity,
domly and suddenly changing detergent flow surface tension, roll diameters, roll separation,
rate, depending on the cleaning needs throughout roll speed ratio, etc. of a roll coating operation
the system. Upon increased cleaning demand, such as to maximize the web speed without
warm detergent from the reservoir is mixed with entraining excessive air, or such as to minimize
fresh, cold detergent, and heated by the heat the level of a particular defect, ribbing lines, for
exchanger before entering the system. A temporary example (see Chapter 12).
drop in detergent temperature after the heat As will be shown below, process optimization
exchanger can now be prevented, ifa temperature can be carried out whether or not a functional
sensor installed upstream from the heat exchanger relationship between independent process par-
detects the forthcoming load change for the heat ameters and dependent variables is known.
exchanger and triggers an increase of the heat However, optimization works particularly well
source supplied to the heat exchanger. The ifthe process can be represented by one or several
amount of source increase is controlled by a analytical expressions which are either derived
second temperature sensor installed downstream from theoretical models, or from empirical fits
from the heat exchanger, which compares the toexperimentaldata,i.e"xj = f(Pi),j = 1,2 . .. ,m,
detergent temperature with a given set point. i = 1,2, ... , n. For the simple case of j = 1, the
Process control in feed-back or feed-forward optimum value for the parameter Pk' which will
mode is justified in many instances, for example lead to an extreme value of x, can be found by
where process sensitivity to small changes of forcing
parameters is very high. In addition, prevention
can rarely be applied to its full extent, because aX =0 (15.1)
many processes are too complex to be perfectly °Pk
optimized. Important examples of controlled It can be concluded from the sign of the
Introduction - the importance of control and optimization 737
second derivative of x, whether the extreme value the impinging curtain and the web, but the
corresponds to a minimum (iPx/op; > 0), a vicosity should also be high to prevent cross-
maximum (02 X / Op; < 0), or an inflection point waves on the hopper slide, particularly because
(02 X / Op; = 0). Moreover, process optimization is the onset of such waves is promoted by high flow
typically subject to boundary conditions and rates needed for high coating speeds.
constraints, and extreme values of dependent As a consequence, process optimization will
variables may only be valid locally, within the most likely result in some kind of a compromise
parameter range considered. In any case, opti- with regard to assigning values to process
mization by mathematical modeling is generally parameters. How to arrive at this compromise
the last step in the development cycle of a specific is not a trivial task; it is determined by how the
process because it requires much basic knowledge optimizations of the sub-processes are weighted
of the process. against one another, which in turn depends on
Optimization of coating processes can be the particular application. For simple systems,
carried out on different levels with increasing the optimization compromise may occasionally
degrees of scope and complexity. The simplest be eliminated, if parameter conflicts caused in
scenario is called micro-optimization which focuses subsequent sub-processes by the optimization of
on a single detail of a coating flow, for example the first sub-process can be eliminated or shifted
optimizing the geometry for flow around a corner forward to the next sub-process (domino syn-
inside a coating die to prevent flow separation, drome). Success of this procedure necessitates
or optimizing the position of the dynamic wetting manipulating only those parameters which are
line in curtain coating to reduce the sensitivity particular to the sub-process under consideration,
with regard to external disturbances, see Section and which do not occur in subsequent sub-
3.3.1. processes (often geometrical parameters).
The next level of optimization addresses specific The primary objective of an industrial coating
process steps, sub-processes, or unit-operations. company, however, is not to design and optimize
Any coating process encompasses several of manufacturing processes, but to produce coated
them, including fluid preparation and handling, goods with a high value for the consumer. The
web handling, formation of a liquid film (die or making of a coated product can schematically
hopper design, see Section 3.3.2 and Chapter 10), be described as the simultaneous interaction
film flow along solid surfaces, unsupported flow between a recipe (e.g., prescription of how raw
in short coating beads or extended liquid curtains, materials have to be combined to form the
dynamic wetting, flow after coating (leveling), coating solution), one or several processes that
setting, curing, drying, etc. act on the coating solution, one or several
Expanding the scope of optimization to the machines on which the processes are implemented,
next level, the purpose then becomes to maximize and a set ofever-present disturbances, see Fig. 15.3.
the efficiency and effectiveness of the entire Many of the available parameters are simul-
coating process (or, more generally, the entire taneously fundamental to the recipe, to the
manufacturing process) by simultaneously opti- process, and to the machine. The available
mizing each of the relevant sub-processes. parameters can be grouped into the following
Invariably, the complexity of such a system will three classes:
lead to conflicting requirements for the quantitative
levels of the process parameters which are available. • physical fluid or material properties, such as
These include viscosity, surface tension, flow rate, rheological properties, surface tension, density,
coating speed, various geometric variables and contact angles, modulus of elasticity, etc;
others. In slide and curtain coating, for example, • operating parameters, such as coating speed,
the viscosity should be low to promote high meniscus vacuum, film thickness, fluid tem-
coating speeds without entraining air between perature, etc.;
Production
Figure 15.3 Relevant elements of the making of coated products.
• geometrical parameters, such as die dimensions, via manipulating geometrical parameters only,
roller diameter, nip or gap dimension, slide for that is the only class of parameters which is
angle, application angle, curtain height, etc. completely separated from the product recipe,
see Fig. 15.3. In such a scenario, the design
It is apparent that the development of new freedom for the formulation chemist is unlimited.
products is most effectively carried out as an The above described scenario is called combined
interactive team effort between chemists who process/product optimization. Its complexity is
typically design new raw materials and formulate still higher, because it aims at optimizing product
recipes, chemical engineers who typically are properties (including chemical and physical
responsible for coating and other unit processes, properties, etc.) and not only the recipe, or the
and mechanical engineers who typically design process, or the machine itself.
and construct the machinery necessary to run From an industrial viewpoint, process opti-
the processes. mization ultimately must be carried out with
Sometimes, circumstances require that priorities respect to economics. The goal of this level of
in product design are directed toward the recipe optimization (macro-optimization) is to minimize
in order to affect chemical, physical, or mechanical the total manufacturing costs in terms of product
product properties other than film thickness or unit costs. Here, the scope includes not only
film thickness uniformity, for example, speed and coating and manufacturing processes, but also
contrast of photographic products, surface texture stock levels, and costs associated with distribution,
and tribological parameters of magnetic recording marketing, research, engineering, administration,
media, electrical conductivity of polymeric etc. Obviously, optimizations at lower levels
coatings, etc. This can be achieved by manipulating affect the outcome of higher level optimizations,
physical fluid properties and operating parameters. and ideally, they should be carried out first.
Consequently, these two parameter classes may However, high level optimization may still be
no longer be available for the engineers to meaningful, even if not all technical aspects of
optimize the coating and other processes, as well coating flows are known and optimized in detail.
as machinery. Now, if possible at all, the challenge The above description of optimization takes
for the engineers lays in optimizing product us back to the essence of this book, which is the
design by optimizing processes and machinery explanation of scientific principles of coating
Control of coating processes 739
processes, as well as their technological (and on a few issues where process control is specifically
economical) implications. The knowledge of these applied to coating processes.
principles, and the associated theoretical models Process control can be divided into two main
provide a useful framework for optimizations on areas. The first one, statistical process control,
the first and second level (micro-optimization is a method to learn and visualize when a process
and optimization of specific process steps or unit is in control. Being in control means the process
operations). is operating within its capabilities, and any
Section 15.2 addresses process control as it is fluctuations are due only to sources of natural
applied to coating processes. In particular, variability. Statistical process control also allows
statistical issues and the time-dependent behavior one to detect sources of abnormal variability
of the process are discussed. Section 15.3 is which cause the process to go out of control.
devoted to process optimization. In the first part, Should the variability exceed a given, statistically
optimization criteria and an overview of opti- calculated limit, then corrective actions are
mization algorithms are presented in general required to bring the process back into control.
form. In the second part, three examples of Process capability must be distinguished from
optimizing specific aspects of coating technology product specifications. If product specifications
are provided. Section 15.3.3.1 exemplifies micro- are tighter than process capabilities, the product
optimization by focusing on the position of the is not properly manufacturable. An acceptable
dynamic wetting line for premetered coating product can only be obtained through intensive
processes. Section 15.3.3.2 illuminates the opti- inspection after the fact by the quality control
mization of die or hopper design, which is a department. Naturally, these activities and the
specific process step. In Section 15.3.3.3, a modest accompanying loss of yield will not only lead to
attempt is presented to model and optimize an excessive manufacturing costs but also to customer
entire coating company in order to minimize dissatisfaction, because there is no guarantee that
product unit costs. Finally, the chapter closes a defective product will not slip through the
with an outlook on interesting issues yet to be inspection process.
conquered. These and many more issues relevant to coating
Owing to the professional experience of the processes are disussed in detail by Frost and
author, the view points and examples presented Gutoff (1991).
in this chapter tend to be biased toward the The second main area of process control
photographic industry and toward premetered concerns the dynamics, or the time-dependent
coating processes. However, it is hoped that the behavior, of a process. The objective is to develop
reader will be able to recognize the essential controllers (software algorithms and hardware)
statements and apply them to other specific which will allow the process to run automatically
coating situations. As will become clear below, and in the desired way. Apart from the above
there is no 'once and for all' solution to control mentioned feed-back and feed-forward controllers,
and optimization problems, as they are always process stability is of major concern, both for
subject to a particular set of boundary conditions open-loop processes with no controllers in the
and constraints. system, and for closed-loop processes with
controllers present. Luyben (1973) is just one
example of a textbook where these issues are
described in detail.
15.2 CONTROL OF COATING PROCESSES
The determination of optimum values for the
Process control in general is part ofany engineering parameters of the control algorithm can easily
education, and its essence has been published in be achieved, so long as the process can be
a great number of books and scientific papers. described by a mathematical model. However, it
Therefore, the discussion in this chapter is focused is not an easy task if the process is too complex
to be modeled, and if the optimization of the an out-of-spec viscosity value is often limited to
controller must be accomplished during normal adding solvent. This is a one-way procedure
process operations. For such cases, Skormin because solvent removal is usually not possible.
(1991) has developed an experimental procedure Furthermore, on-line changes ofsolvent quantities
employing a signal analyzer and a simplex will affect the performance of the coating process,
optimization algorithm to adjust the parameters the wet film thickness, and the set points for the
of industrial PID control systems. In particular, setting and drying zones, all of which will make
he used the speed control of a coating machine overall control and operation of the coating
as an example. Variation of web speed can be process very difficult. Finally, the addition of
associated with certain types of coating defects, polymers in the form of thickening agents may
which explains the rather strict requirements on not be a viable option to control viscosities.
the accuracy of web speed controllers (typically Thickeners often increase the viscosity only in
<0.1%). Hence, optimum controller parameters the low shear rate range, while in the high shear
can be obtained by minimizing the variance of rate range they promote non-Newtonian rheologi-
web speed with respect to the parameter values. cal behavior, particularly shear thinning. In
Today, modern coating plants in all areas of pre-metered coating methods, the viscosity in the
the coating industry are highly automated and boundary layerjust downstream from the dynamic
comprise a great number of control loops. Many wetting line may actually become lower if a
applications are found in the coating solution thickener is present than if the solution contained
preparation area and in the delivery system, but no thickening agent. This feature may not be
also in the coating machine itself as well as in desirable, because it may move the position of
the dryer. Parameters to be controlled include the dynamic wetting line too far upstream,
coating speed, flow rate, fluid temperature, fluid thereby increasing the sensitivity of the coating
pressure, concentrations of various chemical process with respect to external disturbances, see
compounds added to the coating solution, air Section 15.3.3.1.
temperature and humidity, web tension, web For some coating processes, it is possible to
direction, coating gap, roll separation distance, control the chief parameter, the lateral film
roll separation force, etc. These parameters may thickness distribution. In blade coating during
not only be held constant, but they may follow paper manufacturing, for example, cross-direc-
complex curves as a function of time, for example tional film thickness nonuniforrnity is often caused
temperature profiles during the making of silver- by blade beam deflection, which in nearly all
halide emulsions for photographic products, or cases is a result of nonuniform blade beam
speed profiles during acceleration and deceleration temperature. The problem is amplified during
of the coating machine. the start-up phase (unsteady non-isothermal
One area that provides some degree ofdifficulty operating conditions), and by the very large
is the on-line control of fluid properties like coating widths (> 8 m) that are encountered in
viscosity or other rheological parameters and the paper industry. To solve the problem, Sollinger
surface tension. This is unfortunate because fluid (1990) presented a control mechanism called
properties directly affect the coating process. thermal anti-deflection system. Ducts for water
While several manufacturers of viscometers offer flow are installed at the inside front and rear
equipment for on-line measurements, the relevant surfaces of the blade beam which is of square
geometrical dimensions are not always suitable cross-section. The front and rear ducts are
for mounting in small-diameter delivery lines. connected to separate water circulation systems.
Hence, on-line viscosity measurement is preferably The duct adjacent to the coating roll is operated
used in solution preparation and holding vessels, at a constant temperature. In contrast, the water
the location of which is typically far upstream temperature of the second duct is controlled by
from the coating point. In addition, correcting a laser measurement, which determines the blade
Optimization of coating processes 741
beam deflection in the center of the beam. With the coating speed, to maximizing the film thick-
such a system, any blade beam deflection vertical ness uniformity, to minimizing the product unit
to the blade level can be adjusted, regardless of costs, etc. While the choice of these kinds of
whether the coater is in operation or at rest, and variables may be meaningful and justifiable in
regardless ofthe temperature of the coating fluid. many instances, we propose also to consider the
In another example, the control offilm thickness following optimization philosophy.
uniformity for slot or extrusion coating is described Process optimization shall be aimed at defining
by Braatz et at. (1992). In this case of single-layer and quantifying as many constraints for inde-
pre-metered coating, the height of the metering pendent process parameters and dependent process
slot of the coating die, which is crucial for variables as appropriate. The parameter to be
determining the final film thickness distribution, optimized shall not be a particular dependent
can be adjusted by means of several equally variable such as those described above, but
spaced bolts. Cross directional control is aimed rather the sensitivity of such a dependent process
at maintaining a uniform thickness profile by variable to process fluctuations. The sensitivity
measuring the film thickness downstream from is a measure for the robustness of the process.
the coating point (after the dryer in this case), Minimizing the sensitivity is equivalent to a flat
and by feeding a corrective signal back to adjust response surface around the optimum. Moreover,
the position of (some of) the bolts. The math- sensitivity is related to stochastic effects which
ematical model used to design the control are natural and always present in any process;
algorithm does not account for detailed process hence, they are predestined to be minimized.
dynamics because ofthe large time delay between Appropriate constraints for independent and
change of slot height and the resulting change dependent process parameters will vary from
in film thickness, owing to the large distance application to application. Consider the fictive
between coating die and sensor. Even though example, where, for coating speeds of up to 2 mls
this control can only eliminate slowly acting and viscosities of up to 1()() mPa s, the cross-
disturbances, it was shown experimentally to directional film thickness nonuniformity is required
perform well within the given constraints, for to be < 1%. The variable to be minimized is not
example, as related to excessive mechanical stress the film thickness nonuniformity itself, but the
in the coating die. sensitivity of the film thickness nonuniformity to
stochastic fluctuations of parameters such as
viscosity, coating speed, coating gap, wetting
15.3 OPTIMIZATION OF COATING
characteristics of the web, etc.
PROCESSES
Two specific examples of this optimization
Before embarking on optimization, the variable(s) philosophy are presented in Sections 15.3.3.1 and
to be optimized must be defined, boundary 15.3.3.2.
conditions and constraints must be identified,
and optimization strategies and algorithms must
15.3.2 OPTIMIZATION ALGORITHMS
be chosen. These aspects are discussed below.
The selection of an optimization strategy, i.e. the
choice of the most appropriate optimization
15.3.1 OPTIMIZATION CRITERIA
procedure that leads to the desired result, depends
The choice for variables to be optimized is large. on whether or not a relationship between the
As explained in Section 15.1, it depends on the variable to be optimized and the process par-
level of optimization one is interested in, and it ameters is known, and on whether the optimum
ranges from minimizing the corner radius of a values for independent variables of a stationary
flow boundary, to minimizing the residence time process are sought (parameter optimization), or
for liquid flow through a coating die, to maximizing optimum functions for control variables of a
non-stationary process are to be determined optimization is the determination of a low-order

(function optimization), Wilde (1964), Wilde and polynomial regression model over specified ranges
Beightler (1967). for the independent parameters, and the calculation
Function optimization requires a functional of response surfaces (equidistant lines) that
description ofthe process by differential equations. represent constant values of the variable to be
Most often, it is applied to unsteady processes optimized. At the same time, the experimental
such as chemical reactions. Apart from drying efforts to carry out this procedure shall be
and curing, coating operations are steady processes minimized by utilizing suitable experimental
and therefore, parameter optimization is the design plans.
method of choice. The distinction here is between The choices are numerous. A sure method for
one-dimensional (the optimum depends on only finding an optimum is the full factorial plan,
one variable) and multidimensional (the optimum where the entire parameter space is covered with
is determined by several independent variables) a sufficiently fine grid and the variable to be
systems (Bandermann 1972). optimized is determined at each grid point. As
As mentioned above, for simple cases where the number of required experiments is given by
the process can be modelled with analytical n P with p being the number of independent
functions, the method of differential analysis parameters and n being the number of levels for
according to equation (15.1) allows one to optimize each parameter, the method is rather laborious
the process quite easily. and not recommended for problems with p > 3.
Perhaps more important than differential A far more superior and efficient approach is
analysis are iterative strategies or so-called search the Box-Wilson method (Box and Wilson 1951;
methods, which are applicable without analytical Box and Hunter 1957). The procedure is sche-
models, as long as the goal parameter assumes matically depicted in Fig. 15.4 for the case of two
an optimum within the range of interest for the independent variables. The starting point is an
process variables. The location of the optimum arbitrary area on the response surface of the
must be evaluated experimentally, and so the variable to be optimized that is sufficiently
design of suitable experimental plans is of major remote from the optimum. It is assumed that in
importance, see Chapter 7 in this book, as well this region, the response surface can be approxi-
as Montgomery (1984) and Reklaitis, Ravindran
and Ragsdell (1983) for general references.
In the case of one-dimensional systems, sim-
ultaneous search methods are called for if all P2
experiments should be carried out at the same
time, for example in order to save time (Wilde
1964). However, more efficient search methods
can be applied, if the experiments can be carried
out sequentially, because experience gained in
earlier trials can be used to design later ones. A
well known and very efficient procedure in this
class is the Fibonacci search method (Wilde and
Beightler 1967; Denn 1969). With only ten
experiments, the size of the interval in which the
optimum is located can be reduced by a factor
of 100 compared with the initial interval size.
In the field of coating technology, parameter
Figure 15.4 Schematic diagram of the simplex method:
optimization within multidimensional systems is well designed local experimental plans (-e-) define
a more realistic scenario. Here, the purpose of the path toward the optimum.
mated by first-order polynomial models. The ments. However, due to imperfect scale-up
coefficients of these models are determined by conditions, the optimum ofa full-scale production
way ofexperiments (orthogonal first-order designs process may not exactly be located at the same
like full or fractional factorial plans). parameter values as in the laboratory. To overcome
The above models provide the basis for applying this discrepancy, Box developed a method called
the method of steepest ascent (steepest descent, Evolutionary Operation (EVOP), allowing the
if a minimum is searched). It is a procedure for routine plant operating conditions to be moved
moving sequentially toward the optimum, along toward the optimum (Box and Draper 1969). The
the path with maximum increase in response. In principle of the method consists of carrying out
this context, a frequently used procedure is the a 22 or at most a 23 factorial design plan around
simplex method, which is a first-order orthogonal the current operating point. The variations of
experimental design (Spendley, Hext and Him- the independent parameter levels must be chosen
sworth 1962; Neider and Mead 1965). A simplex to be small enough so as to prevent excessive
is a regularly sided structure with k + 1 vertices loss of yield, but large enough so that the
in k dimensions. Thus, for k = 2 the simplex is variation of the variable to be optimized exceeds
an equilateral triangle, while for k = 3 it is a its standard deviation, and that the potential
regular tetrahedron. If well designed, the simplex improvements in process performance will event-
method adapts itself to the local landscape and ually be discovered. EVOP is continually applied
changes step size according to the local slope of in the direction of presumed process improvement
the landscape. Most importantly, it generates the until the experimental costs exceed the expected
coordinates for a new local experimental plan gains in revenues.
along the path toward the optimum, see Fig. 15.4. The main advantages ofEVOP include process
The sequence of local experimental plans is optimization based on routine operating data,
repeated until the goal parameter is sufficiently while the risk of excessive losses is kept low.
near the optimum. Here, the response surface Moreover, the optimization can be carried out
can no longer be accurately described by first-order by the operating personnel if the procedure is
models; its curvature requires the use of high- well specified. The main disadvantage, on the
order polynomial functions which can account other hand, is the rather slow progress toward
for parameter interactions. The associated ex- the optimum. This is particularly cumbersome
perimental design plans must have at least three if, for some reason, such as changes in equipment
levels for each independent variable so that the or process, the location of the initial optimum
model parameters can be estimated. The preferred has been changed to other operating conditions
class of second-order response surface designs is during the course of the experiment, rendering
the class of composite and rotatable designs. previously obtained results practically useless.
These designs are based on 2P factorial or Some of the disadvantages can be avoided,
fractional factorial plans, which are augmented however, by using the above described sequential
by 2p axial points and by several center points. simplex method (Nelder and Mead 1965).
Rotatable refers to the feature that the variance There may be situations where neither the
of the predicted response at some point x is only method of differential analyis nor experimental
a function of the distance between the point x search procedures are applicable for process
and the design center, but not a function of the optimization, due either to a large number of
direction between the design center and the point independent process parameters or to the un-
x (Montgomery 1984). Frequently used composite availability of accurate and automatic sensing
designs include names like central composite, devices. In such cases, statistical clustering analysis
star, Box- Behnken, etc. (Box and Behnken 1960). of process variables may be ofhelp. This technique
In practice, optimization techniques as described is widely used in pattern recognition. A pattern,
above are often applied in laboratory environ- described by a large number of continuous
variables, can be recognized as a pattern of a 15.3.3 APPLICATIONS
particular class by analyzing only a small group
Examples of optimization, such as the ones
of dominant variables. These groups of key
presented below, cannot provide general solutions
variables, associated with particular classes, can
to particular types of optimization problems, e.g.
be selected by processing available statistical
die design, because inherently, a process can only
data. Skormin and Siciliano (1991) have formulated
run at an optimum for exactly one set ofoperating
such a clustering procedure for the film coating
conditions (combination of physical fluid or
process, aimed at analyzing the product quality
material properties, operating parameters, and
as well as optimizing and controlling the
geometrical parameters). As soon as one of these
process.
parameters is changed, another parameter should
Skormin and Siciliano chose the quality of the
be adjusted for the process to remain at the same
coated film as the variable to be optimized: it
optimum. Alternatively, a process may be opti-
was visually assessed by the quality inspector
mized such that the variable to be optimized
and assigned a value 'A' if acceptable, or a value
remains within a specified but acceptable range,
'B' if unacceptable. Unacceptable quality may be
while some of the operating conditions vary
caused by typical defects like streaks, repellencies,
within specified limits.
mottle, bubbles, cross bars, remelt, etc. Clustering
Section 15.3.3.1 on the position of the dynamic
analysis provides a means for identifying relevant
wetting line belongs to the former type of process
process parameters like viscosity, surface tension,
optimization, and Section 15.3.3.2 on die or
emulsion temperature, etc., which determine the
hopper design represents an example of the latter
product quality. Moroever, if distinct clusters are
class of optimization problems. Finally, Section
present, lines separating acceptable from unac-
15.3.3.3 shows how product unit costs can be
ceptable quality can be identified and quantified.
minimized by modeling an entire coating company.
These lines form the boundaries of the operating
The purpose of presenting these applications is
window. In that sense, clustering analysis of the
to layout procedures, and to stimulate interest
available experimental data provides a reliable
by offering new ideas about optimizing certain
alternative to the direct investigation of particular
aspects of coating processes.
defect mechanisms.
Another diagnostic method that has become
popular in recent years is called artificial neural
15.3.3.1 Position of the dynamic wetting line for
networks. Based on available data, neural networks
premetered coating processes
are able to 'learn' nonlinear but empirical
relationships (regression models) between input The class of premetered coating processes
and output parameters of a process. They have comprises the slot or extrusion, slide and curtain
been applied to process and product design, as modes. Many of their characteristic features have
well as to process operations (optimization) and been described in the scientific literature (Sartor
control. More specifically, networks have been 1990; Christodoulou and Scriven 1989; Kistler
used to design fixed or adaptive model-predictive 1983) and in Chapter 11 of this book. In spite
control systems, to diagnose process faults and of distinct differences with regard to operating
to identify patterns and root causes, and to conditions and operating behavior between the
monitor process trends in order to assess process three main modes, they have in common the fact
performance or product quality. Applications of that ultimately a liquid sheet or bridge or 'bead'
this method to coating processes, however, are is formed by the coating head and subsequently
not known to the author. An overview ofartificial deposited onto the moving substrate. This liquid
neural networks and an entry point to the vast . bridge is characterized by two static contact lines
literature can be found in the paper by Bakshi located on the coating head where the two free
and Stephanopoulos (1993). surfaces of the bridge are formed, and by a
dynamic contact line situated where the bridge A coating window such as the one in Fig. 15.5
first touches the web. The three coating modes applies only when all other relevant process
distinguish themselves in the design of the coating parameters are held constant, e.g. viscosity, surface
head, i.e. in the way the liquid bridge is formed tension, bead vacuum and geometric parameters
(be it single- or multilayered), in the length of such as curtain height, slide angle, impingement
the liquid bridge, and in the orientation of the angle, application angle, lip design, etc. Any
liquid bridge with respect to the gravity vector. variation of any of these parameters will alter
In practice, operating conditions of coating the shape and location of the coating window. If,
processes are depicted in so called coating windows for example, the fluid viscosity is increased, the
where the multidimensional space of relevant onset of air entrainment will shift towards lower
process parameters is divided into a region (or speeds, thereby decreasing the size of the parameter
regions) ofadmissible operating conditions leading space leading to acceptable product uniformity.
to acceptable product uniformity, and into a Operating conditions of industrial coating
region of unacceptable product uniformity. The processes are often situated close to one of the
boundaries of the coating window, which separate boundaries of the coating window. While such
areas of high and low uniformity, are associated conditions may offer advantages, for example
with the onset of flow instabilities or other with respect to the maximum attainable coating
undesirable features, which in turn cause coating speed for a given set of parameters, they may
defects. The names of these defects depend on also be characterized by increased process
the selected parameters of the coordinate system uncertainty, i.e. any small deviation in any of the
of the coating window and on the descriptive process parameters from its nominal value may
colloquialisms which vary from company to shift the operating point across the boundary of
company. Figure 15.5 shows a schematic example the coating window into an area of decreased
of a coating window for the curtain mode (Kistler product quality. Such a shift is particularly
1983; Blake, Clarke and Ruschak 1994). In the objectionable if the coating defect depicted by
plane of coating speed versus volumetric flow that boundary is characterized by an 'on/off'
rate per unit width, the area of admissible mechanism which will immediately lead to
operating conditions is bound by the onset of catastrophic process failure.
air entrainment, the onset of 'puddling' (vortex The goal of modern product design is not only
in curtain heel), and the loss of stability of the to build in distinct product features but also to
falling curtain. make the product manufacturable. In the context
of a coating window, this means the operating
conditions should not be located close to one of
\ the boundaries but should be sufficiently remote
\\ from it so as to guarantee maximum process
certainty.
Chen and Scriven (1992) expanded on the idea
of the coating window by introducing the feasibility
window (range of relevant parameters within
which a coating can be established and maintained
Curtain Stability at all), the sensitivity window (range in which
the coating is stable to spontaneous disturbances
Coaling Speed
triggered by process fluctuations and unresponsive
to forced disturbances from the surroundings),
Figure 15.5 Schematic diagram ofthe coating window and the quality window (range without vortices
for curtain coating. Air entrainment: - - detected in the flow field, and static wetting lines pinned
on undisturbed web; - - - - - detected after splice passage. to straight edges on the coating die). The idea
of relating quality or optimum flow conditions momentum of the liquid will cause the jet to split
to the absence of vortices was already mentioned into two symmetrical films which will flow away
by Schweizer (1988). In Fig. 15.5, the line depicting from each other along the surface. If the surface
the onset of 'puddling' is a boundary of the is set in motion, it will exert a force (viscous drag)
quality window, while the solid line depicting the on the liquid which tends to pull all the fluid in
onset of air entrainment belongs to the feasibility the direction of the motion. The above arguments
window as long as air entrainment is detected are also valid if the jet impinges on the surface
on the undisturbed web by simply increasing the at an angle other than perpendicular. Successful
coating speed. If, on the other hand, the onset coating, that is the formation of a dynamic
of air entrainment is measured after the passage wetting line, is achieved if the net change in
of a splice beneath the impinging curtain momentum parallel to the substrate is balanced
(representing a large disturbance), the respective by the sum of all forces acting on the fluid in
boundary of the coating window moves left the plane of the substrate, and if the net change
towards lower coating speeds (dashed line). This of momentum in the direction perpendicular to
boundary is now part of the sensitivity window. the substrate exceeds a critical value in order to
The purpose of this section is to show how avoid air entrainment. Depending on the position
one may go beyond the onion-like structure of of the dynamic wetting line relative to the front
coating windows, which define increasing product or back surface of the falling curtain, the
uniformity from outer layers toward the center, impingement zone may assume a heel form.
by identifying a single operating point inside the Referring to Fig. 15.6, the integral form of the
coating window (schematically indicated as point momentum equation in X- and Y-directions,
P in Fig. 15.5), for which the product uniformity using a control volume whose boundaries coincide
becomes optimum for a given set of relevant with a cut through the curtain upstream of the
process parameters. Conversely, the goal of such impingement zone, with the substrate surface,
an optimization strategy is also to shift the and with a cut through the fully developed film
coating window around a desired operating downstream of the impingement zone, can be
point such that the resulting product uniformity written as
will be the best possible.
The variable to be optimized was chosen to pQ(U - Vsina) = IF x (I5.2a)
be the position of the dynamic wetting line which, pQVcosa = IF y (15.2b)
beside gross defects such as air entrainment,
'puddling', ribbing lines, vortices, etc., has been
shown - according to our experience - to affect H y
the uniformity of the coated product. In particular,
product uniformity decreases if the wetting line
is not straight across the coating width but wavy
V!----'----L-L..L-'-I
as a result ofperturbing forces occurring randomly
and sporadically. The optimum position of the
dynamic wetting line is characterized by maximum
x
robustness with respect to ambient disturbances.
The arguments are presented by using the
impingement zone of a falling curtain as an
exemplary flow field. The results, however, will
only be qualitative and can, in principle, also be
applied to slide, slot, and extrusion coating.
Consider a plane liquid jet (curtain) impinging Figure 15.6 Schematic diagram of the curtain im-
on a horizontal and stationary surface. The pingement zone.
where p is the density, Q is the volumetric flow is of particular interest. To a first approximation,
rate/width, U is the coating speed, V is the curtain the change of X -momentum there is mainly
impingement velocity, H is the thickness of the balanced by viscous drag and pressure forces.
impinging curtain, H is the final wet thickness
C1) According to the boundary layer theory of
of the coated film, and oc is the impingement Sakiadis (Sakiadis 1961; Kistler 1983), the viscous
angle. The types of forces F x and F y in the X- drag force, or wall shear stress, is strongly
and Y-direction include viscous drag, pressure negative in the vicinity of the dynamic wetting
(static, dynamic, or pressure difference across the line and increases in the downstream direction
curtain), gravity, surface tension and molecular with (X)-1/2. For large heels, i.e., when the
interactions. In addition, there are force disturb- dynamic wetting line is located far upstream of
ances stemming from variations of any of the the rear curtain surface, therefore, the change of
relevant process parameters. One important viscous drag force underneath the curtain is
disturbing force, that is generally not included weak. Pressure forces, on the other hand, arise
in the model of equation (15.2), is the pressure from a transformation of the Y-momentum into
exerted on the curtain heel by the air boundary a stagnation pressure, qualitatively characterized
layer carried along by the uncoated web. The air by a bell-shaped distribution whose maximum
boundary layer cannot be avoided and, according value is proportional to the square of the normal
to our experience, it is characterized by some component of the curtain impingement velocity,
degree of nonuniformity across the coating width. (V cos ocf (Blake, Clarke and Ruschak 1994), and
The resulting effect on the coating uniformity is whose tails approach low values far downstream
manifested by broad bands in the down-web from the impingement zone and far upstream,
direction, whose occurrence is random in time particularly if the heel is large. Evidence to
and cross-web position. Another important support this statement is given by the special
disturbance is given by the passage of the tape case of a jet impinging onto a horizontal and
used to splice subsequent rolls of substrate. stationary wall (Deshpande and Vaishnav 1982),
Often, the tape is fastened over both ends of and by Kistler (1983) and Gilbert, Wagner and
adjacent rolls, thus providing a step-like sudden Nagel (1994) who both used the finite-element
change of web thickness, which has been observed method to compute the flow field (pressure field
to trigger excessive air entrainment. Both of the in terms of isobars) in the curtain impingement
above described disturbances act in the coating zone.
direction. On the grounds of the above arguments, it can
Any change of any force in the coating (X-) be stated that the distribution of X-momentum
direction and near the plane of the substrate will along the substrate surface and beneath the
result in a small displacement of the dynamic impinging curtain is mainly characterized by the
wetting line. The sum of all forces in the X- bell-shaped curve of the stagnation pressure,
direction can only be balanced within certain particularly if the curtain heel is large. In the
limits. These limits mark the boundaries of the light of this flow characteristic, the optimum
coating window, i.e., air entrainment. position of the dynamic wetting line is postulated
Equation (15.2) describes the conservation of as the point P of maximum slope on the upstream
momentum for the fluid contained in the defined tail of the pressure (momentum) distribution
control volume. As the equation is written in curve, see Fig. 15.7. At that point, any disturbing
integral form, it does not suffice to accurately force in the coating direction, such as those
reveal the details of the momentum and force described above, tending to displace the dynamic
distributions along the boundaries of the control contact line by the distance ~X can only do so,
volume. With regard to the optimum position if it overcomes the resisting change of X-
of the dynamic wetting line, the distribution of momentum ~M.
the X-momentum near the plane ofthe substrate Owing to the bell-shaped momentum distribu-
terized by a heel. The relative wetting line

position has been shown to depend on all
relevant process parameters, and it can be
approximated by way of an analytical expression
based on the Sakiadis boundary layer theory
(Blake, Clarke and Ruschak 1994). While this
expression is quite successful in predicting the
u position of the dynamic wetting line, it does not
provide the necessary information to determine
x
the optimum value for I according to the above
criterion, because the Sakiadis-type boundary
Figure 15.7 Schematic diagram of the curtain im-
layer model does not account for shock-like
pingement zone with a bell-shaped momentum dis- disturbances from splice tapes nor for the effects
tribution along the substrate. The optimum location, of the impinging air boundary layer, both of
P, for the dynamic wetting line is postulated to be at which are considered important here.
the steepest slope of the momentum distribution. In order to validate the above described
optimization criterion, numerical solutions of the
tion curve, the resistance to contact line displace- full Navier-Stokes system and associated inves-
ment will be smaller if the wetting line is located tigations of the sensitivity of the flow to ever
at any other point than P. The only exception present disturbances are needed. The most
is a point P' located at the steepest slope on the attractive and efficient method is frequency
downstream branch of the momentum distribu- response analysis, see Chapter 9. This tool was
tion. P' will be the optimum position only if the not available to the author at the time of writing.
disturbing force is acting in a direction opposite One piece of evidence, however, was provided
to the web motion. However, it is assumed that by Stadler (1993), who used the finite-element
such forces are less important compared with method to calculate an exemplary flow field of
disturbances stemming from the splice passage the curtain impingement zone, see Fig. 15.8.
and from the air boundary layer, which are both Flow parameters were chosen such that the
acting in the direction of web motion. resulting curtain heel became relatively large
A convenient way of quantitatively describing (l = 3.10). This situation is not considered
the position ofthe dynamic wetting line has been optimum, because the dynamic wetting line is
presented by Blake, Clarke and Ruschak (1994). located too far upstream in a region of relatively
They defined a dimensionless ratio I, called the low pressure and low pressure gradients. This
relative wetting line position (see Fig. 15.6). fact is indicated in Fig. 15.8 by the pressure
distribution along the substrate surface (bell-
1= L~ shaped curve superimposed on the streamline
(15.3)
(Hlcos ex) pattern). The left tail of the curve clearly starts
to flatten toward the dynamic wetting line.
where L~ is the length of the boundary layer Notice that the pressure curve is not shown all
which originates at the dynamic wetting line and the way to the dynamic wetting line due to
extends to the front surface of the curtain, and unreliable values owing to the singularity there
H Icos ex is the thickness of the curtain as projected (see Chapter 9), and that the origin ofthe pressure
onto the plane of the substrate. Consequently, for coordinate system coincides with the substrate
I = 1, the dynamic wetting line is in the plane of plane. The maximum point on the pressure curve
the rear curtain surface. For I < 1, the wetting corresponds to a pressure of 1982 Pa for a curtain
line is located further downstream, while for impingement velocity of 2 mis, which is nearly
I> 1, the curtain impingement zone is charac- equal to the stagnation pressure p V 2 /2 if the
o
1.57
3.10
o
Figure 15.8 Streamline pattern of the curtain impingement zone with superimposed pressure distribution
along the substrate. The value for the relative wetting line position is 3.17. The point P of steepest slope on
the pressure curve (= proposed optimum position for the dynamic wetting line) is located at 1.57 curtain
thicknesses behind the front surface of the curtain. Reynolds number Re = 4.5, capillary number Ca = 18.0,
speed ratio VjU = 1.8, curtain impingement angle IX = 0·, apparent dynamic contact angle 0 0 = 170· (from
Stadler 1993).
density assumes a value of l000kgjm 3 . The slope of the pressure curve is fairly constant in
steepest slope on the pressure curve is located the range of about 1 < I < 2, indicating a flat
at point P, which is considered the optimum optimum. Therefore, the robustness of the flow
position for the dynamic wetting line, resulting system is not expected to change much, as long
in maximum robustness with respect to disturbing as the dynamic wetting line is located in that range.
forces acting on the heel in the coating direction. Blake, Clarke and Ruschak (1994) have pres-
For the flow parameters chosen in this example, ented a different criterion for the optimum wetting
the value for the dynamic wetting line position line position by stating that the pressure generated
at P equals 1.57. Barring qualitative results for by the impinging curtain can act as a load and
other values of the flow parameters, it is assumed thus may delay the onset of air entrainment.
that the optimum value for the relative wetting Based on a balance of the Y-momentum, they
line position lies in the range of 1 < I < 2. found a maximum for the impingement pressure
Moreover, it is apparent from Fig. 15.8 that the if 1=2. Referring to Fig. 15.8, the maximum
impingement pressure on the substrate occurs (Carreau model). In addition, the surface tension
inside the projection of the curtain (I < 1), at has been set to 40 mN/m, the application angle
least for the parameters chosen in the example. to 0·, and the curtain height to l00mm. Three
While the maximum pressure criterion proposed locations ofconstant relative wetting line position
by Blake, Clarke and Ruschak presents an are depicted inside the area of acceptable coating
optimum with respect to the impingement pressure uniformity. The values for I are of order unity,
which presumably is related to the maximum hence the symbols drawn in the figure point
speed of wetting, it is not an optimum with toward the area of maximized process robustness
respect to the stability of the wetting line position inside the coating window. The I-values have
in response to perturbations which in turn is been calculated with the formula based on the
related to the uniformity of the coated film. For Sakiadis boundary layer model and published
either argument, however, optimum conditions by Blake, Clarke and Ruschak (1994). As I
seem obtained if the relative wetting line position increases beyond a value of about 2 a sizable
I is of order unity. This opinion is also shared heel is formed, indicating a departure from
by Van Abbenyen, Mues and Goetmaeckers optimum operating conditions and an approach
(1992) who postulate optimum conditions for of the 'puddling' boundary of the coating window
I = 1. Their argument is based on the maximum (Fig. 15.5; Kistler 1983). On the other hand, if I
speed of wetting, but they do not present a refined decreases below a value of 1 the air entrainment
physical model to support it. boundary is approached, and operating conditions
As an illustration, the air entrainment boundary in this area of the coating window are characterized
of the curtain coating window has been drawn by increased process uncertainty.
from the data published by Blake, Clarke and With the help of the optimization criterion
Ruschak (1994), see Fig. 15.9. The curve is valid described above (dynamic wetting line located
for an aqueous gelatin solution characterized by at steepest slope of pressure distribution), the
a low shear viscosity of 30 mPa s, a power law area inside the coating window has been reduced
index of 0.85, and a time constant of 0.000 248 s - 1 to a line of preferred operating conditions,
described by I = constant and of order unity. A
further reduction to a single optimum point as
depicted in Fig. 15.5 is achieved by selecting the
thickness of the coated film. In the Q- U plane
~ (Fig. 15.9), conditions of constant film thickness
1:-:. are given by straight lines through the origin.
- '" ~
'" It may occur that curtain coating suffers from
~
excessive air entrainment, even if the operating
'"
y
--
po
A A
parameters are chosen such that the relative
;: ~ A
wetting line position is at its optimum value near
,g 1
C1l unity. Under these circumstances, it may be
E
::> possible to clear the air entrainment by changing
<> the operating parameters such that the relative
> 0
o 2 3 4 5 6 7 wetting line position moves from the steepest
Coating speed, U (m/s) slope to the maximum ofthe pressure distribution.
This approach satisfies the optimization criterion
Figure 15.9 Curtain coating window in the U-Q put forth by Blake, Clarke and Ruschak (1994).
plane showing experimental data (-.-) of the It is, however, associated with a reduction of
boundary for air entrainment (from Blake, Clarke and process robustness according to the arguments
Ruschak 1994). Legend: Calculated values for the
position of the relative dynamic wetting line, 1 = 1.0 presented above. We propose a different approach
A; 1.5 ~; 2.0 :+:. to clear the air entrainment by changing the
operating parameters such that 1 remains at the (free falling curtain versus liquid film on an
location of steepest pressure gradient, but that inclined plane and flow in a narrow slot).
the pressure at the location of steepest slope In slide and slot coating, the position of the
increases sufficiently by increasing the curtain dynamic wetting line can effectively be controlled
impingement velocity (increasing the curtain by the pressure difference across the bead.
height). As is shown in Fig. 15.10, the result is However, the relationship between wetting line
not only a higher peak of the momentum position and pressure difference is difficult to
distribution curve, but also a steeper maximum obtain experimentally and is often not known
slope, and therefore an increased resistance to for real applications, which is why these coating
wetting line displacement in response to stronger modes are characterized with a great deal of
perturbations. uncertainty.
Thus, the robustness of premetered coating Furthermore in slide and slot coating, the
processes can be increased by increasing the position of the dynamic wetting line determines
Y-momentum while keeping the wetting line at the separation angle of the lower meniscus from
its optimum position. High impingement speed, the die lip which in turn controls the migration
high coating speed at constant thickness of the of the static wetting line along the face of the lip
coated film, and a liquid jet impinging perpen- (Kistler 1983). A flow field where the static
dicularly onto the substrate (a = 0°) favor such wetting line is not pinned to a sharp corner of
conditions. One might even question the existence the die lip but wets the die face is similar to a
ofan absolute maximum speed of wetting, because large heel in the impingement zone of a falling
there is no conceivable limit to increasing the curtain. The optimum position of the dynamic
Y-momentum. wetting line (l near unity) promotes a static
In this light, curtain coating is superior to slide wetting line that is pinned to the die lip, thereby
and slot coating, because the momentum of the reducing the chances for streaks and lines.
impinging jet is one order of magnitude larger In summary, the optimum position of the
dynamic wetting line for premetered coating
processes is postulated to coincide with the
steepest slope of the pressure distribution in the
plane of the substrate. It is argued that the
process robustness with respect to random
disturbances in coating direction is maximized
at this operating point. Today, computer-aided
methods of frequency response analysis exist to
calculate the optimum position of the dynamic
wetting line in the presence of process perturba-
tions for any arbitrary combination of process
u parameters. However, such methods were not
available to the author. Nevertheless, it is reasoned
x that the optimum value for the relative dynamic
wetting line position, I, must be of order unity.
For curtain coating, 1 can be calculated with
Figure 15.10 Schematic diagram of the curtain sufficient accuracy for practical applications using
impingement zone with bell-shaped momentum dis- a model based on the Sakiadis boundary layer
tributions along the substrate. The steepest slope of theory and published by Blake, Clarke and
the momentum distribution increases with increasing
momentum, M, of the impinging curtain, and so does Ruschak (1994). Finally, the optimization criterion
the robustness of the location P for the dynamic can qualitatively also be applied to slide and slot
wetting line with respect to external disturbances. coating. For slide coating in particular, operating
conditions with I near unity promote a flow field and outer metering slots, H 1 and Hz, the shapes
with the static wetting line pinned at the corner of the inner and outer cavities, Pl and pz (see
ofthe die lip, which in turn reduces the propensity Chapter to, and Liu 1983), the sizes (cross-
for lines and streaks in the down-web direction. sectional area) of the inner and outer cavities,
A1,o and A z , the size of the inner cavity at X = Jv,
A1,l' and the taper index, m, indicating how the
15.3.3.2 Die design
size ofthe inner cavity cross-section changes with
The physical principles and specific details of die X, see equation (15.7). L is the total length of
or hopper design are presented in Chapter 10 the fluid flow portion of the die element; for given
and in the scientific literature, for example Durst, values of the above defined parameters, L is not
Lange and Raszillier (1994) and Ruschak and independent.
Weinstein (1996). The purpose of this section is The goal of die design optimization is to assign
to complement Chapter 10 by showing how these numerical values to each of these geometrical
principles can be combined with some of the parameters such that die performance criteria
optimization procedures described above to meet given quality specifications.
determine which die performs the best for a set The purpose of a die is to create a liquid film
of boundary conditions. by redistributing fluid from a point source (feed
The issues considered here are confined to pipe) to a line source (metering slot). The
internal die design and to simple applications. performance of a die has traditionally been
In particular, liquids will be Newtonian, and fluid judged by only one quality criteria, namely the
flow will be modeled analytically for ease of uniformity of the film thickness. While this is
presentation. These issues, however, can easily undoubtedly an important criterion, it is insuf-
be applied to more complex and realistic situations ficient because the overall die performance must
such as non-Newtonian rheology, but at the be assessed by several criteria, depending on the
expense of more complex modeling efforts. particular application. The list of performance
A side-fed dual cavity hopper will serve to criteria may include:
discuss the examples below. As sketched in Fig.
15.11, the fluid flow portion of such a hopper is • uniformity of film thickness;
defined by twelve geometrical parameters, namely • tendency to form streaks in the liquid film;
the coating width or cavity length, Jv, the bar • tendency to contaminate the die;
thickness, T, the lengths of the inner and outer • ease of die cleaning;
metering slots, L 1 and L z , the heights of the inner • residence time characteristics;
• fluid pressure at die inlet;
• weight of die.
y
In addition to judging the performance, these
criteria may also serve as boundary conditions
-' or constraints during die design optimization by
requiring them to respect specified values. In any
case, the criteria must first be translated into
suitable parameters that are accessible to fluid
flow modeling, before they can be used for quality
assessment or as guidelines for design optimization.
Further arguments about die design optimization
o w x and design constraints have been presented by
Winter and Fritz (1986), Sartor (1990) and Durst,
Figure 15.11 Schematic diagram of a side-fed dual Lange and Raszillier (1996).
cavity die. Left: front view; right: cross-section. The following discussion will be restricted to
the first four criteria in the above list. The issue Prevention of contamination and cleaning of
of residence time has been addressed in the flow systems, particularly the removal of particles,
literature (Winter and Fritz 1986; Lee and Liu is not specific to flow inside dies. In fact, the food
1989). However, according to our own experience, processing industry has been dealing with these
models relying on one-dimensional fluid flow are issues for many years, and the scientific literature
insufficient for accurately predicting residence is extensive, for example Grasshoff and Reuter
time parameters. Fluid pressure at the die entrance (1983); Barouch, Wright and Matijevic (1987);
will only become important if gravity is the Dussan v. (1987); Yeckel and Middleman (1987);
driving force to feed the coating solution, or if Sharma et al. (1992). Some important results of
the pressure is high enough to mechanically these investigations can be summarized as follows.
deform the die.
• The complex process of removing particles
Die contamination and cleaning from a solid surface can be simplified and
characterized by a single parameter, namely
Many dies are being used to coat not only one the wall shear stress.
but several different products, each of which is • There is an exponential relationship between
characterized by its own specific chemical and the wall shear stress supplied to the flow
physical properties. Thus, it is imperative to system and the diameter of the smallest particle
prevent cross-contamination during product that can be removed from the flow system.
change-over by removing and purging any residual
material of the previous product. Better yet, From the die designer's point of view, the wall
cleaning would not be necessary if the die did shear stress is a convenient parameter, because
not contaminate in the first place. it is not only proportional to the product of
Contamination of solid surfaces may take viscosity times flow rate, but it also depends on
several forms such as thin films of soluble or geometrical parameters of the flow configuration,
insoluble liquid or gel, crusts and particles adhering and thus it can be controlled by appropriate die
to the surface (Schweizer 1992). Here, the word design. Consequently, cleaning efficiency and
particle is used as a generic name, standing for prevention of contamination can be optimized
solid particle, gas bubble, droplet, or any other by choosing appropriate dimensions for distribu-
inhomogeneous part in the otherwise homogene- tion cavities, metering slots and cavity inlets.
ous coating solution. What exactly these dimensions should be depends
The mechanisms by which contamination builds on the level of wall shear stress required to
up on a solid surface are complex. Factors that remove any particles of unacceptable size which,
affect contamination include surface roughness, for a given range of viscosities and flow rates, is
flow state (laminar or turbulent), kinetic forces determined by the quality requirements of the
(form and friction drag), surface tension forces, particular products under consideration.
adhesion forces, etc. In any case, ifcontamination In summary, a design constraint relating to
is present, the forces making it stick to the surface cleaning efficiency or prevention ofcontamination
are greater than the forces required to remove it can be formulated and requires that the wall
from the surface. Conversely, if contamination shear stress, t w must exceed a minimum value
is to be prevented, the removing forces must be to anywhere inside the die at all times ofoperation.
greater than the adhesive forces everywhere (15.4)
inside the hopper and at all times of operation.
If a die is not taken apart during a product Based on experience, the value of to is on the
change, mechanical devices such as brushes cannot order of 1.0 N/m 2 for many coating applications.
be used to scrub dirty surfaces. The only cleaning The flow in the inner cavity is predominantly
aid available are kinetic forces supplied by the along the cavity axis, and the average wall shear
coating fluid itself or by a cleaning solution. stress can be approximated by equation (15.5)
which was derived by Miller (1972) and is valid the cavity entrance can be calculated to give
for ducts with arbitrary cross-section.
~)
(1 _b:')3m/2
'. (1 - (158)
(15.5)
!w,o =
Here, J1. is the viscosity, {3 is the shape factor For b = 1.0, as is the case in our example, the
(Liu 1983), V· is the volumetric flow rate which wall shear stress distribution depends on m, and
decreases linearly along the cavity axis according to several curves are drawn in Fig. 15.12. Obviously,
the design criterion (15.4) is satisfied for m ~ 2/3.
V· = Vo• (1 - ~) (15.6)
The condition b = 1.0 has been chosen to keep
the mathematics simple. In reality, b > 1.0,
resulting in enhanced three-dimensional flow
where Vo• is the volumetric flow rate at the cavity and, as known from experience, in reduced
inlet (X = 0), S is the wetted perimeter and A cleaning efficiency near the cavity end. Conse-
the cross-sectional area of the duct. In the most quently, equation (15.8) is not accurate near
general case, S and A are both functions of X, X/W = 1.0.
and A is conveniently represented by Turning the attention to the metering slots,
the flow there is mainly along the Y-axis, and
the wall shear stress can be calculated according to
(15.9)
with b being defined as A1,o/(A1,o - Al,l)' In
our example, both cavities have circular cross- where Qo is the uniform volumetric flow rate per
sections, and b is chosen to be 1.0 (A1,1 = 0.0). unit width, and H is the slot height. Owing to
To satisfy the design criterion (15.4), both the the characteristic dimensions, the wall shear
wall shear stress at the cavity entrance, !w,o, and stress in the slot is typically one to two orders
the wall shear stress distribution along the cavity ofmagnitude higher than in the inner distribution
axis must be above the minimum value, to' for
all products to be considered. Thus, for a given
range of viscosities and flow rates, the shape and 2.0 r - - - - - - - - - - - - r - - - - - ,
size of the cavity cross-section must be chosen 1.8 t---------------;l'"-------4I

~
accordingly. Values of the shape factor {3 have 1.6 t----------:;JIf-------f-j
~
been published by Liu (1983), and Liu and Hong ~ 1.4 t - - - - - - -.....- - - - - - - - - y - j
iii
(1988) for several duct shapes. {3 is largest for a 1.2-!----;;;:;>II........-------::=--....-::.-.--I
t~~~~::~;::::==:j
circular cross-section ({3 = 1/8n); corners with ~ 1.0
small included angles are disadvantageous. It ~g 0.8
follows that by choosing a cavity shape, the size ."
~ 0.6 i--------""""'-::~""""<;;<----=-..::::s.....;;o--1
of the cross-section at the cavity inlet (X = 0) is ~
E 0.4 t----------=~~----""~'X:"--_llti
determined by equation (15.4). The choice of o
z 0.2 i------------===-::::-"'"<:::;-i\,
shape may be a compromise between high wall 0.0 .f-~-...,.._~:__~----::':_":"':""_:_':----.--.,........="W.
shear stress and manufacturing costs of the die, 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
bid Coordinol., X/W [-)
0.8 0.9 1.0
because symmetrical shapes (e.g., circle) require

more machining time than nonsymmetrical shapes Figure 15.12 Wall shear stress distribution for a
(e.g., semi-circle). tapered cavity of circular cross-section. Legend: taper
For circular cross-sections, the wall shear index m = 1.0 - . - ; 0.75 -f---; 0.66 -lj-; 0.5 '-8-;
stress distribution normalized with the value at 0.25 -x-; 0.0-'-.
cavity, and therefore, equation (15.4) is much particle, possibly accompanied by phase separ-
easier to satisfy. ation. However, if a particle is trapped inside a
Finally, the flow in the outer distribution vortex, it will spin around owing to the circular
cavity is also mainly in the Y-direction. In or helical fluid motion, thereby disturbing the
general, the wall shear stress here must be liquid, and particularly interfaces and/or free
computed numerically. Except for shallow cavities, surfaces outside the separating streamline. In any
the level of wall shear stress will be low due to case, the strength and the optical appearance of
suddenly expanding or at least diverging flow the streak in the coated film depends on the
boundaries. In spite of this disadvantage, the relative size and type of particle and particle trap,
presence of an outer cavity may be useful to and on the distance of the particle trap from the
dampen excessive flow nonuniformities produced coating point. Moreover, particles alone, or
by the inner cavity/metering slot, see Chapter 10. particle traps alone, will not cause streaks, only
In summary, the shapes and sizes of both the interaction of both will.
distribution cavities are determined by the criterion To prevent streaks, the number of particles
(15.4). In contrast, the criterion is easily satisfied must be reduced, and the particle traps must be
by a wide range of slot heights, so they are still eliminated. While the first item is an operational
open for optimization. In general, high wall shear problem related to filtering and de-gassing of
stress is obtained by small characteristic dimen- fluids (Schweizer 1992), the second item is of
sions of the flow domains which, as will be seen interest to the die designer. Since vortices are
below, is in conflict with achieving uniform film characterized by a separating streamline that
thickness distributions. The solution to this begins and ends at separation points on the flow
conflict is high-viscosity fluids. boundary where the wall shear stress is zero,
equation (15.4) is a suitable design criterion for
the prevention of streaks. Requiring the wall
Streak formation
shear stress to be above a minimal level does not
Based on experimental work involving flow allow particles to adhere to the flow boundary,
visualization (Schweizer 1988), the origin ofstreaks and it implies the absence of stagnation points,
can be attributed to an interaction between a and hence the absence of vortices.
'particle' and a 'particle trap' anywhere in the The issue of vortices is also closely related to
flow field. Here, streaks refers to narrow lines in the design ofthe entrance to the inner distribution
the downweb direction with steep property cavity. Many dies are of the 'coat-hanger' type
gradients along their sides; it does not include where the cavity is center-fed. Consequently, the
broad diffuse streaks nor flow instabilities such liquid arriving in the feed section must turn a
as ribbing lines. sharp corner of 90· to enter both arms of the
The word 'particle' is used in a generic sense; cavity. If the feed section has a small cross-
it can be a solid particle, a gas bubble, a nick in section, the resulting peak from the stagnation
a corner of the flow boundary, or any other pressure may translate into a local thickness
inhomogeneous part in the otherwise homogene- nonuniformity in the coated film, and the liquid
ous coating fluid. The principal types of 'particle may separate from the cavity wall downstream
traps' have been identified as flow separation or of the corners. If, on the other hand, the feed
vortices, areas oflow wall shear stress, and rough section has a large cross-section or is diverging,
surfaces. the wall shear stress may be too low, even to the
The mechanism which leads to the formation point where vortices start to form in the feed
of streaks depends on the type of particle trap. section. All of these features are potential causes
If a particle is stuck to a rough surface or if it for hopper contamination and streak formation.
adheres to an area of low wall shear stress, then They can be reduced by feeding the cavity from
a streak is the result of the wake behind the the side.
Film thickness uniformity deterministic and a stochastic contribution. Let
us take term (3) and give more details. In many
The thickness uniformity ofthe liquid film exiting
applications, the viscosity of the coating fluid is
the die is a measure of the die performance. It
a function of concentration, C, and temperature,
may also be a measure ofthe quality ofthe coated
T, both of which may vary with time, t, and along
film further downstream. In any case, it is
the cavity axis, X. Hence, term (3) must be
strongly determined by the outer metering slot
replaced by its total derivative according to
which is the last of a series of flow elements
comprising the die. More specifically, the film
thickness uniformity at the slot exit is proportional term (3) = - -1 {Ojl
- -dX + -dt
[OC OC]
jl oC oX ot
to the volumetric flow rate distribution across term (5)
the coating Width, which in turn is proportional
to the pressure profile in the outer cavity according Ojl [OT dX OT]}
to + oT OX + -at dt
term (6) (15.12)
AP(X)H(X)3
Q(X) = 12jl(X)L(X) (15.10)
Terms (5) and (6) describe the variation of
viscosity due to spatial and temporal variations
Equation (15.10) is valid for Newtonian fluids. of the fluid concentration and temperature,
Q is the volumetric flow rate per unit width, AP respectively. The partial derivatives Ojl/oC and
is the pressure difference across the slot, and L ojl/oTare fluid properties; they are always positive
and H are the length and height of the slot, and can be determined experimentally (Schweizer
respectively. If the pressure at the slot exit is 1992). oT/oX stands for the variation of fluid
arbitrarily set to zero then AP reduces to the temperature along the cavity axis. If the fluid
pressure, P, at the slot entrance which is temperature at the die entrance is not equal to
approximately equal to the pressure distribution the die temperature, then the die will act as a
in the outer distribution cavity. The other variables heat exchanger, and consequently, the fluid will
have been defined above. change its temperature as it flows along the
An analytical expression for the variation of distribution cavity. This temperature difference
Q, and hence for the variation of the film between fluid and die at the inlet is a measure
thickness, is obtained by taking the total derivative for the degree of isothermal operating condition.
ofequation (15.1O)(Schweizer 1992). If normalized Any values other than zero will cause flow rate
with the uniform volumetric flow rate per unit nonuniformity. The effect and its resulting
width, Qo, the derivative takes the form contribution to nonuniformity is called systematic
dQ(X) dP(X) 3dH(X) djl(X) dL(X) or deterministic, if the temperature distribution
--=--+ ------ T(X) can be calculated (Schweizer 1992).
~ P H jl L
On the other hand, even if the process is
term (1) (2) (3) (4) nominally operated isothermally, there will still
(15.11) be a slight contribution to nonuniformity due to
random or stochastic fluctuations of the fluid
Term (1) represents the lateral variation of the temperature as a result of imperfect process
pressure drop across the metering slot, term (2) control capabilities (oT/ot in term (6) will not
is the variation of the slot height, term (3) is the exactly be zero). Similar arguments can be
variation of the viscosity and term (4) indicates developed for term (5), which is related to mixing
the variation of the slot length. and fluid homogeneity.
As has been shown by Schweizer (1992), equation In summary, deterministic contributions to
(15.11) can be further expanded to visualize that nonuniformity arise when the die is operated at
each of the four terms is composed of a nominal conditions other than those assumed as
boundary conditions during the design stage.

Their effects are additive. In contrast, stochastic
contributions express process control capabilities,
and their combined effect is equal to plus or
minus the square root of the sum of their
variances. The variance is the square of the
standard deviation, s, and the value of the
standard deviation can be multiplied by 1, 2, or
3, depending on the desired confidence level (Box,
Hunter and Hunter 1978). It follows that the flow
rate nonuniformity (equation (15.11)) is more
correctly expressed as
where OJ and Sj denote, respectively, any deter-

ministic and stochastic contribution which can
be quantified, e.g., terms (I) to (6) in equations
(15.11) and (15.12). Notice that in contrast to
equations (15.11) and (15.12), the flow rate
nonuniformity is now expressed in terms of a
finite difference and not as a differential in the
true mathematical sense (i.e., dQ -> 0), because
variations ofindependent process parameters are
finite in reality.
As shown in Fig. 15.13, the widthwise flow
rate uniformity (or nonuniformity) is thus char-
acterized by a band whose overall slope is
determined by the sum of the deterministic
,,-----------------,
~
o
08
)( 06
O'
------
III " - - - - ' - - - ! ! ! !

o 01 02 03 O. 05 06 07 08 00
Coordinate along Cavity Axis, XIW H
Figure 15.13 Cross profile (widthwise flow rate

nonuniformity) characterized by band with overall
slope stemming from deterministic contributions (---.-)
and width stemming from stochastic contributions
( +s; - . - -s).
gravitational effects are neglected. The flow in (deterministic contribution to the cross profile).
both metering slots and also in the outer cavity For an optimized die according to the present
is assumed to be purely viscous. Both cavities example, terms (1) and (4) in equation (15.11) can
have circular cross-sections, and the dimensions thus be reduced to
of the outer system (cavity plus slot) remain
constant across the coating width. The situation dQ(X) _ [ 1 aLl dX] (15.15)
is schematically depicted in Fig. 15.14. ~- - Ll ax (Re-Re
D
)
The essence of the model is to solve a differential

equation for the length of the inner metering slot As from term (3), there will also be a stochastic
according to equation (15.14), and the result is contribution to the cross profile from terms (1)
the so called taper function for the inner slot length. and (4). Even if the die is operated at Reo, the
Reynolds number will still slightly vary with
dL l _ H~ (dP) (15.14) time, either from batch to batch due to incon-
dX - 12J.lQ dX cavity sistencies in solution preparation, or within a
batch due to pump rate fluctuations or due to
Within the scope of the model, the taper viscosity variations as a result of hardener addition
function for L l , and therefore the flow rate which triggers a chemical reaction in the coating
distribution at the slot exit depend on the fluid, for example.
Reynolds number defined as Re = pQo/J.l Machining imperfections during the manufac-
(p = fluid density), as well as on additional turing of the slot height and slot length will also
parameters such as power law index and time contribute to the cross profile. The effect is
constant, if the fluid is non-Newtonian (Lee and particularly serious for the slot height because it
Liu 1989). The value of the Reynolds number is is multiplied by a factor of 3 owing to the 3rd
associated with a particular fluid being pumped power ofthe slot height in equation (15.10). While
through the die, and as such it is indicative of slot height and length can be measured and hence
fluid properties and operating conditions. As is should be treated as deterministic contributions,
typical for many companies, coating equipment we chose to consider them as stochastic, because
including dies is used for manufacturing not only appropriate data is not always readily available.
one but several different products. However, it Additional deterministic contributions to the
is expected that the equipment perform equally cross profile can be attributed to the terms (2)
well for all products. This requirement presents and (4) in equation (15.11). If, for example, an
a problem in die design, because optimized existing die is operated at significantly different
geometrical parameters are generally fixed and conditions from those assumed during the design,
cannot be adjusted iffluid properties or operating the resulting strongly nonuniform cross profile
parameters vary from one product to the next. can be empirically reduced by appropriately
It is therefore necessary to define optimum die modifying the length ofthe outer metering slot (re-
performance by allowing a small but acceptable tailoring). Another case refers to multi-slot dies
amount of nonuniformity which is not to be which are often used in the photographic industry
exceeded by any operating conditions within the for simultaneously coating multilayer films (see
specified range. Chapters 11b and 11c). Here, the individual die
Ll(X) will be optimum (and the flow rate bars can be considered as cantilevered plates
distribution Q(X) will be perfectly uniform) for which deflect under the action of fluid pressure
exactly one Reynolds number, namely Reo, the in the metering slots on each side of the bar. If
Reynolds number for which the die has been a die is used for coating more than one product,
designed. For any other Reynolds number, L l (X) the resulting pressures in the slots will differ from
should have been designed differently, and this product to product, and so will the bar deflection
discrepancy will cause a flow rate nonuniformity and the modification of the nominal slot height.
As is discussed above, any die will always etot is the same for both bands and consequently,
produce some small amount of nonuniformity. they are acceptable as long as equation (15.16)
Let its magnitude be etot ; it is obtained by is satisfied. One band is wide and has a negative
integrating equation (15.13) over the coating slope of small magnitude, indicating small
width, W etot can be looked at as the difference contributions from deterministic factors and large
between the highest and the lowest point of the contributions from stochastic factors. In contrast,
nonuniformity band, see Fig. 15.13. Conversely, the band with the larger but positive slope is
etot can be used as a boundary condition for the narrower, indicating large deterministic contri-
optimization procedure by forcing it to be less butions but small contributions from factors
than a given maximum value related to random process fluctuations. It is this
(15.16) feature which is most important, because the real
cross profile, which lies inside the band, will be
much more consistent within a batch, and from
Optimization procedure batch to batch, and product consistency is crucial
So far, it has been shown that the values of some to customer satisfaction.
of the geometrical die parameters can be optimized The optimization criterion can thus be for-
based on wall shear stress requirements. The mulated analytically as
values ofthe remaining geometrical die parameters
are obtained by applying an optimization pro- minimum [ ~sJ ] (15.17)
cedure for which an optimization criterion must
first be defined. Clearly, all geometrical parameters Note that the criterion corresponds to the
affect the magnitude of etot which in turn must second term on the right-hand side of equation
satisfy equation (15.16). (15.13). The optimum design is now found via
The criterion ofchoice optimizes the robustness one of the search procedures described above. If
of the cross profile with respect to random analytical fluid flow models are used, computa-
process fluctuations. In Fig. 15.14, two sigmoidal tional efforts will not be excessive, and a full-
nonuniformity bands are drawn schematically; factorial plan is convenient. More specifically,
each of the remaining geometrical parameters to
be optimized (inner and outer slot height, inner
and outer slot length, position ofthe outer cavity,
bar width) is taken at several levels within a
predetermined range, and equation (15.13) is
calculated for each case subject to equation
(15.16). The parameter configuration that satisfies
equation (15.17) will be taken as optimum design.
Three partial results of this multidimensional
optimization process are presented below. The
purpose is to visualize qualitatively how certain
geometrical parameters affect the cross profile.
o Coaling Width W
Therefore, details on the other parameters are
less important and are not given. Similar examples
have also been published by Sartor (1990).
Figure 15.14 Comparison ofcharacteristic cross profile
bands. - - - small deterministic and large stochastic In Fig. 15.15, etot is shown to depend on the
contributions leading to a wide band susceptible to ratio of inner to outer slot length, L t /L 2 . This
disturbances. - -- - - large deterministic and small ratio indicates the position of the outer cavity
stochastic contributions resulting in a narrow but for a given value of L (see Fig. 15.11). Depending
robust band. on the ratio of inner to outer slot height, H dH 2'
4.0 7,----------------------,
~ ~
:.. 3.5 :::: 6 +-------------~'---:_:A'
o §
~ 3.0 'U; 5 H:------------:;;~~,?--__;
~ ~
~ 2.5
~.
u.: 4 i--"'\---------~~---:::;;----__;
~
~
.E 2.0 ~3 'iI.---="*=::;:-----;-:;vJI.'=----...c..------_;
.E
~ ~
E 1.5 E 2 i--~-------;;1'~-------__;
o 0
c c
Iii 1.0
~
Iii 1
~
r---=:::::::..:E:;:;::;;:::;:;:~:::::::=====1
0.5 -I--~~~~-r-T..,...,...----~~~~~__r_M
0.1 1 10 200 300 400 500 600 700 800
Ratio of inner slot to outer slot length H Outer slot length, H2 (micron)
Figure 15.15 Optimum position of outer cavity. Figure 15.16 Optimum outer slot height. Total flow
Total flow rate nonuniformity as a function of ratio rate nonuniformity as a function of outer slot height.
of inner to outer slot length. Legend: ratio of inner to Legend: -a- inertia effects; -+- mechanical slot
outer slot height H 1/H 2 = 0.25 -a-; 0.43 -+-; accuracy; - . - sum of both effects (total flow rate
1.0 -k-. nonuniformity).
the optimum value for LdL 2 is either on the inadequate, resulting in considerable nonisother-
edge or inside the allocated range for the search mal operating conditions. In any case, owing to
procedure. the absence of perfection in real manufacturing
The effect of the outer slot height, H 2' on the processes, there will always be some amount of
cross profile, Stot, is shown in Fig. 15.16. For
small values of H 2' the damping capability of
the outer cavity/metering slot is large, resulting
in small deterministic contributions (only inertia 6.------------------,
effects are accounted for in this example), but
the random effects of machining accuracy (3/lH 2/ ~ 5 -t-------------7':7::'"'"
g
H 2 ) are considerable. For large values of H 2 , on
the other hand, these trends are reversed, leading ~4+-------------:::,-:==-~=-----l
w
to a distinct minimum for Stot and hence, to an
optimum value for H 2' In both examples, the
·f 3 ±=::::;;;::::::::e-~~~~=------___j
nonuniformity curves in the vicinity of the .E

'§ 2 F O - . . = ' - - - = , . . . . ; ; - - - - - - - - - - - - - - - - I
optimum point are fairly flat, indicating low c
o
sensitivity to changes in the respective independent .::lb~--------------___l
<0
parameters. ~
Finally, the effects of random process fluctu- O.j..-~-~-~~-~-~~-~-~--l
o 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
ations on Stot are shown in Fig. 15.17. In particular, Temperature control, +/- 6T (0C)
the effects ofslot accuracy (machining capability),
/lH, dominate the effects of temperature fluctu-
Figure 15.17 Process control capabilities. Total flow
ations, /l T, for small values of /l T, i.e., for
rate nonuniformity as a function of temperature
operating conditions that are nearly isothermal. control capability. Legend: mechanical slot accuracy
On the other hand, slot accuracy becomes less IlH = 0.0 J.1m -a-; 0.5 J.1m -+-; 1.0 J.1m -.-;
important, if temperature control capabilities are 1.5J.1m -8-.
flow rate nonuniformity, even if all deterministic in the distribution chamber by recirculating
contributions are zero. liquid through the sides of the die (Tanaka and
In reality, it may often be difficult if not Chikamasa 1983), and adjusting the height of the
impossible to reach the optimum design because metering slot via choker bar (Anderson et al.
of conflicts among various design constraints. A 1975) or flexible hopper plate (Nissel 1976), see
typical scenario is described as follows. The also Section 15.2, and Braatz et at. 1992.
demand for high wall shear stress requires the
inner and outer cavity cross-sections to be small,
15.3.3.3 Minimization of product unit costs
which in tum makes the die sensitive to changes
in Reynolds number and rheological properties, In a competitive market one ofthe foremost goals
particularly if the operating ranges of these two of any manufacturing company is to minimize
parameters are large. The sensitivity can be product unit costs, because it provides maximum
diminished if the heights of the metering slots leverage in terms of product pricing, profit
are small or the lengths of the slots are large. margins, and control of market share. For a
The slot heights cannot be too small, in order typical coating company the total product unit
to prevent mechanical slot accuracy from be- costs, Cu, expressed as currency per unit area,
coming important. If the slot lengths are too e.g., ($/m 2 ), depend on several factors according to
large, deformations of the slot heights may occur
due to excessive bending of the die bars as a (15.18)
result of differential pressures in adjacent slots where Cp are the production costs, CL are called
of a multilayer hopper. The bending can be logistics costs, and Co denote overhead costs or
controlled by making the bars thicker, which operating expenses. All three cost factors on the
leads to a longer hopper slide and hence promotes right-hand side of equation (15.18) depend in
the inception and growth of slide waves, if the tum on a variety of other factors, namely
multilayer film is ill designed. Moreover, in slide
coating, for example, long metering slots may Cp = CRM + Cc + CFP + CQC (15.19a)
not be acceptable, if the new hopper must fit on
the old hopper mount, and if it must meet the (15.19b)
coating roll at the same application point as the
old hopper.
Such a situation either requires solutions to Where CRM are raw material costs which, as in
newly formed problems such as well designed the photographic industry, may include costs for
multilayer films that are not susceptible to slide operating elaborate manufacturing processes,
waves even on a long slide, or it necessitates a e.g., the making of light-sensitive silver-halide
compromise (departure from the optimum design) emulsions; Cc are coating costs; CFP are finishing
by relaxing one or several design constraints. and packaging costs; CQC are quality control
Which constraints can best be relaxed depends costs; Cs are stock or inventory costs for raw
on the particular application and must be materials, intermediates and finished goods; CDT
considered carefully. Often, the choice is between are distribution and transport costs; CMR are
high wall shear stress and good flow rate maintenance and repair costs; CCI are capital
uniformity, with the uniformity taking priority. investment costs; CRD are research and develop-
To close this section, it must be mentioned ment costs; CMS are marketing and sales costs
that film thickness uniformity is not always and CA are administration costs such as finance
controlled by design as described above, but and control, human resources, senior management,
often by choosing appropriate operating pro- etc. Production costs, Cp , are sometimes divided
cedures, particularly if the coating width is large. into direct and indirect costs, whereby direct
Such procedures include varying the flow rate costs are generated by work directly related to
the manufacturing process, e.g., operating a (i.e., minimum product unit costs) instead of just
coating machine, and indirect costs are associated local sub-optima.
with activities not directly related to the manu- The level of resolution of the operations model
facturing process, e.g., supervision, laboratory can be developed to any desired degree. As soon
services, etc. Furthermore, many of the cost as the dependency of a particular process and
factors are split into fixed and variable costs, the its related costs upon independent parameters is
latter one indicating a dependency of the cost known from data, or at least can be described
factor on the production volume. Variable costs in words, it can also be translated into an
may include direct labour costs, energy costs, etc. analytical expression which will become part of
For many coated products the production volume the operations model. The degree of uncertainty
is expressed as 'area coated per unit time.' associated with model predictions is directly
The income of a coating company is generated related to the basic understanding ofthe processes
by the sale of consumables, i.e. coated products. to be modeled: if a process is well understood,
Therefore, the sum of production and logistics it can be modeled with high resolution and the
costs are often called 'cost of sales.' Overhead confidence in predicted results will be high.
costs, on the other hand, are not directly related It is beyond the scope of this chapter to
to the manufacturing process, but they must be develop the complete operations model to any
absorbed by it. Typically, overhead costs are practical degree of resolution, because such a
allocated to the unit costs of the product or task strongly depends on the boundary conditions
products being manufactured in indirect propor- of a particular application. Nevertheless, the
tion to their respective production volumes. power of the model is demonstrated below by
As will be shown below, production and way of a selected example.
logistics costs, and to a lesser degree overhead Suppose that the annual net sales volume (in
costs, depend on a great number of independent terms of coated area) of a particular product is
parameters which are characteristic of the A O•net ' It is then interesting to know how many
manufacturing processes, of the products being times this product has to be coated per year such
coated, and of other relevant activities being that its unit costs are minimized. The net lot size
carried out in the company. In this light, equations per coating, AL.nel' and the number of annual
(15.18) and (15.19) provide the basis for what we coating events (coating campaigns), N, are related
call a comprehensive operations model. Such a to A O•net according to
model is a powerful management tool for the
decision making process, because it relates specific A _ A O•net
(15.20)
L.net - N
product and operation parameters to the product
unit costs. Consequently, it allows one to test It is known from our experience that at least
various manufacturing scenarios and, most raw material, coating and stock costs depend on
importantly, to minimize product unit costs by the size of the coating lot, see equation (15.19).
choosing appropriate values for the independent If we succeed in modeling these three cost factors
parameters. The model is also invaluable, for it as a function of independent parameters, we can
allows one to take a view of the entire manufac- predict the dependence of the unit costs upon
turing process and of how it affects the product the lot size.
unit costs by simultaneously considering all Focusing on inventory, a typical stock cycle
relevant parameters. This feature is significant, is depicted in Fig. 15.18. When discussing issues
because without such help most human mental of inventory it is useful to measure time in units
capabilities are unable to process the very large of (day). At time to a lot ofthe product is coated,
number of independent parameters (several tens, finished, packaged, and put in stock. However,
or more) contained in the model. Moreover, the the product may not be sold yet during the time
feature is essential for finding the global optimum period t T , because its properties must first be
(15.21)
•
N
.s
Q>
A<
i- to = AL,net = 365. AL,net
Rs AO,net
(15.22)
E
:0
"0
The area underneath the saw tooth curve in
> Fig. 15.18 represents the material value of the
:x:
g
iii stock. This value itself does not contribute to the
As product unit costs. What does contribute, however,
is the interest of the money equivalent to the
stock value that either is lost because the money
I, I, I, is temporarily blocked in inventory instead of
Time [day] invested elsewhere, or must be paid to a bank
in exchange for a loan.
Figure 15.18 Material flow through stock: typical
Product unit costs accounting for the variable
periodic sawtooth-like dependency of finished goods portion of stock costs can thus be modeled by
inventory upon time.
A Lnet ( 1:+2"
. FG [ 365+-'-
CS=Z'C tT
A O.net
1)] ( 15. 23)
tested, or because the product properties have where C FG is the specific value of the (finished)
not yet reached equilibrium. In addition, the goods residing in stock, e.g., the sales price
Marketing Department may request to maintain expressed in currency per unit area, and i is the
a certain volume of safety stock, AS •net ' in order annual interest rate.
to absorb unpredictable sales fluctuations. It Turning the attention to the coating process,
may be appropriate to link the safety stock to the product unit costs associated with the coating
the size of the coating lot according to process, Cc, and with the costs of raw materials,
As,net = 1:' AL,nel' with I: being a small suitable C RM , may be approximated by
number. The safety stock is equivalent to another
1
waiting period of length t s' Now, at time t 1 , the Cc = -·SMU·CCM (15.24)
product can be sold at a rate R s , which is YB
measured in units of (m 2 /day), and which can
approximately be quantified by R s = AO.net /365. C RM = ~[CB + L (M{ C
YB j YMj
Mj
)] (15.25)
After the period to the stock is depleted and the
product cycle comes to an end at t 3 • The next where YB is the base yield of the coating process
lot must not becoatedatt 3 ,butatt 2 = t 3 - tT - ts with respect to the uncoated raw base. The
in order to account for the waiting periods variable SM U is called specific machine utilization;
associated with testing and safety stock. Conse- it is defined as the time required to coat a certain
quently, product flow through stock is charac- volume of a particular product, and it is expressed
terized by a periodic, saw tooth-like curve whose in units of (h/m 2 ). SMU is equal to the inverse
cycle time is equal to the product depletion time of the coating machine capacity, CMC; SMU is
to, see thick line in Fig. 15.18. The first (start-up) preferred over CMC, because the analytical
cycle is different from all subsequent (equilibrium) expression of its model is slightly simpler. CCM
cycles, because no product can be sold during in units of ($jh) is called the specific costs or
the waiting period between to and t 1· hourly rate of the coating machine. For the sake
The time periods t s and to are calculated of simplicity, CCM is held constant in this example.
according to In reality, however, CCM is a complex variable
that depends on the annual gross coating volume by various coating defects. YB can accurately be
and on fixed and variable cost contributions determined by measuring the area of raw base
from labor, energy and auxiliary materials entering the coating machine and the area of
consumption, etc. Materials costs that apply to finished goods entering the stock. However, YB
the coating solution must be summed over the can also be modeled by identifying and quantifying
number of components, j, that comprise the as many events as possible that lead to a loss of
coating solution. M is the specific material raw base (loss of yield). In particular, base losses
coverage (kg/m 2 ), CM is the specific material cost occurring once per coating campaign and losses
($jkg), and YM is the material yield, i.e. the ratio occurring on every coated roll can be distinguished.
of coating solution being coated onto the web In the present example, we shall account for start
to coating solution being initially prepared in losses, ASt = t St ' V· J¥, where t St is the time period
the kettle. CB is the specific base cost ($/m 2 ). between the coating start and the establishment
Apart from the coating time proper, the variable of an acceptable coating uniformity; losses in
SMV also contains nonproductive times that can length direction stemming, for example, from
be allocated to a specific product, such as sample taking, coating defects, splice passage,
change-over time, time for coating trials during etc., A L = LL' J¥, where LL is the total length of
product development, down time due to machine full-width cuts per roll, and the number of rolls
breakdowns or operating errors, time for pre- per campaign is equal to AL,gross/(LR ' W) with
ventive maintenance, etc. The specific machine 4 being the length of the roll; and losses in
utilization can be calculated according to cross-web direction resulting from uncoated
margins on the substrate, coated edge regions
SM V = n [_1_ + (tchange-Over + tnonproductive)] with unacceptable thickness profile, etc.,
W' V AL.gross
A c = WQ ' LR , where WQ is the total width of
(15.26) wasted edge area per roll. With the above
information the following approximate model
Here, n is the number of coating passes for the
for the base yield can be constructed
product, W is the coating width, V is the coating
speed, averaged over the number of passes
according to YB = 1 _ n(tSt ' V· W + LL + WQ )(15.29)
AL,gross LR W
lin 1
- =- L- (15.27) Similar arguments as presented for the base
V n i=l Vi yield YB apply also to the material yield JiM' Its
If n > 1, YB in equations (15.24) and (15.25) must value is less than 1.0 due to losses during start,
also be averaged over the number of passes. finish and unplanned machine breakdowns. A
tchange-over is the change-over time between suc- simple model for YM relates Vu the total lost
cessive products or coating passes, and tnonproductive volume of coating solution, to the volume of
is the balance of nonproductive time per product solution coated onto the web (H 00 is the final
or pass. Note that in contrast to the above wet thickness of the coated film)
discussion about inventory, the lot size to be
VL
coated per campaign is taken as gross value, YM = 1 - ---=---- (15.30)
AL,gross' whereby
H 00 • A L,gross
AL,net = YB' AL,gross (15.28) Finally, the specific base cost CB may also
depend on the lot size, if the base supplier offers
Typically, the value of the base yield YB is less
price reductions for buying large base quantities.
than 1.0, because not all of the raw base that is
In the present example a linear dependency is
moved through the coating machine will be
assumed according to
covered with usable product due to edge losses,
start and finish losses, as well as losses caused (15.31)
where CB,o is a reference base cost and R D is the According to the operations model and the
rate of discount offered by the supplier. values of the independent parameters as defined
Equations (15.18) to (15.31) form a simple in Table 15.1, coating costs decrease hyperbolically,
operations model whose main intent is to study stock costs increase linearly, and raw materials
the effect of the coating lot size upon the product costs decrease at varying rates with increasing
unit costs. The model as it stands contains a total gross size of the coating lot. The resulting
of 22 independent parameters, relating raw product unit costs reach a minimum for
2
material costs, product properties, process ca- AL,gross = 200000m iftheannualproductvolume
2
pacity, process capability, and financial parameters equals 4 mio. m • However, the optimum coating
to the resulting product unit costs. Investigating lot size decreases to 70000m 2 if the annual
such complex multidimensional relationships product volume is only 1 mio. m 2 . In general, the
without the help of a mathematical model clearly minimum of the product unit costs is shifted
exceeds human capabilities. towards larger coating lot sizes if coating and
Figure 15.19 shows the product unit costs, C u , materials costs dominate stock costs; the opposite
as well as coating, stock and raw materials costs is true if inventory costs are larger than coating
as a function of the gross coating lot size, AL,gross and raw material costs. The minimum of the
for the arbitrary chosen parameter values in product unit costs curve as a function of the
Table 15,1. The number of components in the coating lot size may not be very pronounced as
coating solution, j, has been set to 1 for reasons is the case for the conditions in Fig. 15.19.
of simplicity. Some of the units do not conform Nevertheless, a small reduction in production
with the SI norm, but they work well in practice costs may just make a significant difference in
or match other units in a particular equation. terms of sales volume and market share in a
highly competitive market place, and the oper-
ations model is the necessary tool to realize this
difference. In practice, the realization of the
3.5 minimum product unit costs may be hindered
./ by additional boundary conditions such as
N'3.0
E
~2.5
I'.... / restricted product shelf life, particularly if the
rate of sale is low and the operations model calls
(J) ~ H-_ ~ .£>SI for large coating lot sizes.
~ 2.0 In summary, an operations model has been
()
'§ 1.5 presented that provides a means for relating

t5 financial business parameters (product unit costs)
-5 1.0 to technological performance criteria of the
e
a.. 0.5 ~
-----
manufacturing process. In particular, issues of
a
~'
.. raw material costs, process capacity and process
10 100 1000
capability have been accounted for. The latter is
Gross size of coating lot (1000 m2 ) a measure of the ability of the process to produce
products of acceptable quality. The ability to
avoid various coating defects, edge nonunifor-
Figure 15.19 Product cost contributions as a function mities, and other losses is implicitly contained
of the gross size of the coating lot. -a-: coating in the yield factors for raw base and coating
costs; -k-: stock costs; -+-: raw materials costs; solution. The description of the current state of
-8-: resulting product unit costs for an annual
basic knowledge of coating processes and their
product volume of AO,net = 4 mio.m 2 ; -<><l-: resulting
product unit costs for an annual product volume of technological implications necessary to achieve
2
AO,net = 1mio.m ; the values of all other independent high process capacities and capabilities is the
parameters are given in Table 15.1. subject matter of this book.
Table 15.1 Values of independent parameters for operations model
AOinet CB,o CCM CFG CM e HOC)

(m ) ($/m 2 ) ($jh) ($/m 2 ) ($jkg) (- ) (/lm)
4'10 6 1 1000 20 4 0.25 100
LL LR M n RD tC-D
(- ) (m) (m) (kg/m 3 ) (- ) ($/m 4 ) (h)
0.12 20 3500 0.1 2 1'10- 7 1
t _
N P t SI tT U VL W WQ
(h) (s) (day) (m/s) (m 3 ) (m) (m)
1.5 30 10 2.5 0.2 1.5 0.1
The strength of the operational model lies in 15.4 OUTLOOK

its ability to evaluate various action scenarios in
terms of return on investment, i.e., it allows one Optimization and control is the ultimate goal of
to quantify the necessary efforts of a particular any research and development activity. The basis
scenario (human and financial resources) to for optimization is a fundamental and complete
achieve the resulting reduction in product unit understanding ofthe process under consideration,
costs. which in turn requires well educated scientists
The operations model also visualizes the and engineers who have the best tools available
relationship of various levels of optimization. to carry out experimental and theoretical inves-
While equations (15.18) and (15.19) may be tigations. Consequently, process optimization
regarded as a model for optimization on the can only be realized effectively if the underlying
macro-level, parameters such as the specific technology has reached a fairly mature stage. In
machine utilization (equation (15.26» point to coating technology, this stage has been reached
the level of optimization of entire sub-process, in certain areas as is evidenced by the examples
e.g., the coating process, and still other parameters above, but additional interesting problems remain
such as the base and material yields (equations to be investigated.
(15.29) and (15.30» can be maximized by opti- Another consequence of optimization is the
mization on the micro- and unit operation level. difficulty of presenting results in a general and
The practical usefulness of the operations broadly useful form. The reason is the complexity
model can and should be increased by modeling and the multidimensional aspects of most coating
additional terms in equations (15.18) and (15.19), flow problems, and thus the inability ofexpressing
and by increasing the resolution of some of the the relationship between relevant process par-
terms modeled above. The latter can easily be ameters in analytical form. Moreover, optimum
achieved for the prediction of the yield factors operating conditions can only be maintained for
by accounting for other contributions to the loss exactly one configuration of process parameters.
of yield. Another example is the hourly rate of Optima for one particular application are therefore
the coating machine. In reality, this parameter of limited use if one or several parameters are
depends on the total annual production volume changed for another application. There is a need
of the coating machine; it is therefore tied to for scientists to develop or acquire the necessary
parameters such as manned machine time, direct knowledge and tools to perform the optimization
and indirect labor costs, boundary conditions in for local boundary conditions and constraints.
union contracts, etc. The possibilities appear In some applications, there is a need to depart
unlimited. from the true optimum by allowing a certain
References 767
amount of imperfection, so that the optimum Kinetics of particle detachment. J. Colloid Interface
can be reached from within a specified range of Science. 118(2): 473-481.
the input parameters. There is also a need to find Blake, T. D., Clarke A. and Ruschak, K. J. 1994.
Hydrodynamic assist of dynamic wetting. AIChE
flat optima, where the sensitivity of the optimum J. 40(2): 229- 242.
with respect to changes of process parameters is Box, G. E. P. and Behnken, D. W. 1960. Technometrics.
small. 2(4): 455.
Optimization is an expression ofthe philosophy Box, G. E. P. and Draper, N. R. 1969. Evolutionary
of prevention. It is the foremost means ofachieving Operation. New York: John Wiley.
and improving manufacturing productivity; one Box, G. E. P. and Hunter, 1. S. 1957. Multifactor
of the most important challenges of industrial experimental designs for exploring response surfaces.
Annals of Mathematical Statistics. 28: 195-242.
companies. For scientists and engineers, optimiz- Box, G. E. P., Hunter, W. G. and Hunter, J. S. 1978.
ation is an opportunity to strive for the best by Statistics for Experimenters. New York: John Wiley.
fulfilling the ultimate goal ofcoating flow research. Box, G. E. P. and Wilson, K. B. 1951. J. Roy. Statist.
More yet: the development of useful optimization Soc. Ser. B (London). 13: 1.
models in terms of economic parameters such as Braatz, R. D., Tyler, M. L., Morari, M., Pranckh, F.
the one expressed by equation (15.18) goes beyond R. and Sartor, L. 1992. Identification and cross-
coating technology. It requires interdisciplinary directional control of coating processes. AIChE J.
38(9): 1329-1339.
collaboration among experts from all aspects of
Chen, K. S. A. and Scriven, L. E. 1992. Two views of
manufacturing technology and logistics. The multilayer coating beads. I ndustrial Coating Research,
attempt to model and optimize a manufacturing Volume 2.
company as an entity is a formidable task that Christodoulou, K. N. and Scriven, L. E. 1989. The
still awaits successful completion. fluid mechanics of slide coating. J. Fluid M echo 208:
321-354.
Denn, M. M. 1969. Optimization by Variational Methods.
ACKNOWLEDGMENT New York: McGraw-Hill.
Deshpande, M. D. and Vaishnav, R. N. 1982. Submerged
I am grateful to J. Bruce Ikin, Ken A. Ridley, laminar jet impingement on a plane. J. Fluid Mech.
and David J. Nield from ILFORD Limited, and 114: 213-236.
John DeNewtown formerly from ILFORD Durst, F., Lange, U. and Raszillier, H. 1994. Optimiz-
Limited for challenging discussions and valuable ation of distribution chambers of coating facilities.
Chem. Eng. Sci. 49(2): 161-170.
contributions to the issue of optimum wetting Durst, F., Lange, U. and Raszillier, H. 1996. Minimiz-
line position. I am also thankful to Dirk Stadler ation and control of random effects on film thickness
from the Lehrstuhl flir Str6mungsmechanik at uniformity by optimized design of coating die
the University of Erlangen for his spontaneous internals. Proceedings of the 1st European Coating
willingness to calculate an example of the flow Symposium on the Mechanics of Thin Film Coatings,
field of the curtain impingement zone. eds P. H. Gaskell, M. D. Savage and J. L. Summers,
pp. 85-94. Singapore: World Scientific Publishing.
Dussan V., E. B. 1987. On the ability of drops to stick
REFERENCES to surfaces of solids. Part 3. The influences of the
motion of the surrounding fluid on dislodging
Anderson,1. W., et al. 1975. US Patent 3,884,611. drops. J. Fluid Mech. 174: 381-397.
Bakshi, B. R. and Stephanopoulos, G. 1993. Wave-net: Frost, P. J. and Gutoff, E. B. 1991. The Application
a multiresolution, hierarchical neural network with of Statistical Process Control to Roll Products. 2nd
localized learning. AIChE J. 39(1): 57-81. edn. PJ Associates, 1001 Marina Drive, Suite 605E,
Bandermann, F. 1972. Optimierung chemischer Reak- North Quincy, MA 02171.
tionen. In Ullmanns Encyklopiidie der technischen Gibbs,1. W. 1906. The Scientific Papers of J. W Gibbs.
Chemie, Band 1. pp. 362-418. Weinheim: Verlag Volume I, Thermodynamics. Longmans, Green. pp.
Chemie. 314-331.
Barouch, E., Wright, T. H. and Matijevic, E. 1987. Gilbert, N., Wagner, H. G. and Nagel, P. 1994. Slot
extrusion coating in the presence of magnetic forces. Schweizer, P. M.1992. Fluid handling and preparation.
Paper 2c, presented at the 7th International Coating In Modern Coating and Drying Technology, eds E.
Process Science and Technology Symposium of the D. Cohen and E. B. Gutoff. pp. 23-61. New York:
AIChE Spring National Meeting, Atlanta, GA, VCH Publishers.
April 17-21, 1994. Shapiro, S. I. 1993. Process control. Chapter 5 in The
Grasshoff, A. and Reuter, H. 1983. Untersuchungen Coating Processes, ed. J. C. Walter. pp. 175-180.
zum Reinigungsverhalten zylindrischer Totriiume. TAPPI Press, Atlanta, Georgia.
Chemie-Ingenieur- Technik. 55(5): 406-407. Sharma, M. M., Chamoun, H., Sita Rama Sarma, D.
Kistler, S. F. 1983. The fluid mechanics of curtain S. H. and Schechter, R. S. 1992. Factors controlling
coating and related viscous free surface flows with the hydrodynamic detachment of particles from
contact lines. PhD thesis, University of Minnesota. solids. J. Colloid Interface Science. 149(1): 121-134.
Available from University Microfilms International, Skormin, V. A. 1991. A simple solution to PID
Ann Arbor, MI. controller adjustment. Power Conversion and Intel-
Lee, K-W. and Liu T. J. 1989. Design and analysis ligent Motion (PCIM). 17(3): 44-48.
of a dual-cavity coat-hanger die. Polymer Engng Sci. Skormin, V. A. and Siciliano, R. J. 1991. Novel
29(15): 1066-1075. approach to quality analysis, optimization, and
Leonard, W. K 1985. Inertia and gravitational effects control of film coating process. J. Coatings Technol.
in extrusion dies for non-Newtonian fluids. Polymer 63: 103-110.
Engng Sci. 25(9): 570-576. Sollinger, H. P. 1990. New developments in coating
Liu, T. J. 1983. Fully developed flow of power-law CD profile control. TAPPI Proceedings of the 1990
fluids in ducts. I&EC Fundamentals. 22: 183-186. Coating Conference. pp. 347-351.
Liu, T. 1. and Hong,C.-N.1988. The pressure drop/flow Spendley, W., Hext, G. R. and Himsworth, F. R. 1962.
rate equation for non-Newtonian flow in channels Sequential application of simplex designs in opti-
ofirregular cross-section. Polymer Engng Sci. 28(23): mization and evolutionary operation. Technometrics.
1559-1564. 4: 441.
Luyben, W. L. 1973. Process Modeling, Simulation, Stadler, D. 1993. Lehrstuhl fUr Stromungsmechanik,
and Control for Chemical Engineers. New York: University of Erlangen/Niirnberg, Cauerstrasse 4,
McGraw-HilI. D-91058 Erlangen, Germany. Finite element calcu-
Miller, C. 1972. Ind. Eng. Chem. Fundam. 11: 524. lations of the curtain impingement zone performed
Montgomery, D. C. 1984. Design and Analysis of specifically for this work.
Experiments. 2nd edn New York: John Wiley. Swain, R. H. 1985. Closing the loop on coating
Neider, J. A. and Mead, R. 1965. A simplex method processes. Paper, Film and Foil Converter (6,7).
for function minimization. The Computer Journal. Tanaka, Y. and Chikamasa, H. of Fuji Photo Film
8: 308-313. Co. Ltd. US Patent 4,416,214.
Nissel, F. R. 1976. Thickness control system for an Van Abbenyen, W., Mues, W. and Goetmaeckers, B.
extrusion die. US Patent 3,940,221. 1992. Explaining limits of operability in curtain
Reklaitis, G. V., Ravindran, A. and Ragsdell, K M. coating. Paper 43a, presented at the 6th International
1983. Engineering Optimization: Methods and Ap- Coating Process Science and Technology Symposium
plications. New York: John Wiley. of the AIChE Spring National Meeting, New Orleans,
Ruschak, K 1. and Weinstein, S. J.1996. Mathematical LA, March 29-April 2, 1992.
modeling of fluid dies. Proceedings ofthe I st European Wilde, D. 1. 1964. Optimum Seeking Methods. Englewood
Coating Symposium on the Mechanics of Thin Film Cliffs, NJ: Prentice-Hall.
Coatings, eds P. H. Gaskell, M. D. Savage and Wilde, D. J. and Beightler, C. S. 1967. Foundations of
J. L. Summers, pp. 3-22. Singapore: World Scientific Optimization. Englewood Cliffs, NJ: Prentice-Hall.
Publishing. Winter, H. H. and Fritz, H. G. 1986. Design of dies
Sakiadis, B. C. 1961. Boundary layer behaviour on for the extrusion of sheet and annular parisons: the
continuous solid surfaces: II. The boundary layer distribution problem. Polymer Engng Sci. 26(8):
on a continuous flat surface. AIChE J. 7: 221. 543-553.
Sartor, L. 1990. Slot coating: fluid mechanics and die Yeckel, A. and Middleman, S. 1987. Removal of a
design. PhD thesis, University of Minnesota. viscous film from a rigid plane surface by an
Schweizer, P. M. 1988. Visualization of coating flows. impinging liquid jet. Chem. Eng. Comm. 50: 165-175.
J. Fluid Mech. 193: 285-302.

Liquid Film Coating

Uploaded by

Copyright:

Available Formats

Liquid Film Coating

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Liquid Film Coating

Uploaded by

Copyright:

Available Formats

LIQUID FILM COATING

JOIN US ON THE INTERNET VIA WWW, GOPHER, FTP OR EMAIL:

SPRINGER-SCIENCE+BUSINESS MEDIA, B.V.

© 1997 Springer Science+Business Media Dordrecht

ISBN 978-94-010-6246-6 ISBN 978-94-011-5342-3 (eBook)

Library of Congress Catalog Card Number: 96-86101

® Printed on permanent acid-free text paper, manufactured in

Colour plates appear between pages 228 and 229

List of contributors vii

1. Coating science and technology: an overview 3

Part One: Physics and material interactions of coating processes 17

3. Wetting: static and dynamic contact lines 63

4. Surfactants: static and dynamic surface tension 99

5. Coating rheology: component influence on the rheological response and

6. The fate of thin liquid films after coating 183

Part Two: Methods of investigating coating processes 207

7. Experimental methods 209

8. Asymptotic methods for the mathematical analysis of coating flows 251

9. Advances in computational methods for free-surface flows 297

Part Three: Theory and practice of coating processes 367

10. Analysis and design of internal coating die cavities 369

Chapter II: Premetered coating processes 399

Chapter 12: Self-metered coating processes 537

13. Free-meniscus coating processes 673

14. Spin coating 709

15. Control and optimization of coating processes 735

CYRUS K. AIDUN J. EDWARD GLASS

KIMIAKI MIYAMOTO L. E. (SKIP) SCRIVEN

MICHAEL D. SAVAGE WILLY VAN ABBENYEN

P. RANDALL SCHUNK HANS-GUNTER WAGNER

PETER M. SCHWEIZER STEVEN J. WEINSTEIN

M mass matrix V(t) volume of fluid body [m 3 ]

K dimensionless curvature of curve Young's equilibrium contact angle

LIQUID SUPPLY COATER DRYER/CURE

Figure 1.1 Simplified schematic of industrial liquid film coating process.

unresolved issues in coating flow hydro- never-ending refinements in experimental and

PHYSICS AND MATERIAL INTERACTIONS

F10~ Fluid 1 Fl~

2.3 MATHEMATICAL DESCRIPTION OF

To help motivate the mathematical treatment of 2.3.1.1 Conservation of mass

Material Body Fluid

Figure 2.2 Arbitrary material body for a single fluid phase.

As by definition, all the fluid points always

remain in the material body, it is clear that

integration domain. This form, however, is often

way to achieve a constant volume as a material

:t f pU dV= f tsdS + f pgdV (2.13a)

VxU=O (2.30a) (2.31)

2.3.2.1 The kinematic boundary condition:

The kinematic boundary condition provides a

different from that of the local fluid velocity.

in phase 2. Using this definition, we note that

hand, the interface A(t) is not in general a

Thus, (2.40) simplifies to the intuitively pleasing

U 2 °0 1 = Va at the surface, (2.4ta) In component form, the kinematic boundary

f °1'(T 2 -T 1)dS+ f m'T"dC=O (2.53)

• Note that the left-hand side of equation (2.52) is negligible (2.58a)

: / - - - ~- - - - - OscUiaiingCirCle Ly~£in P'ane2 - - - - /- - "7

this problem is that the soap film is exposed to

where 'II = 1 + 2(JlA/PV)3/2 + 3(JlA/p vf; Jl is given by (Brown 1961):