Goriely 2017

Interdisciplinary Applied Mathematics 45
Alain Goriely
The Mathematics
and Mechanics
of Biological
Growth
Interdisciplinary Applied Mathematics
Volume 45
Editors-in-chief
S.S. Antman, University of Maryland, College Park, MD, USA
L. Greengard, New York University, New York, USA
P.J. Holmes, Princeton University, Princeton, NJ, USA
Series editors
R. Durrett, Duke University, Durham North, NC, USA
L. Glass, McGill University, Montreal, QC, Canada
Alain Goriely, University of Oxford, Oxford, UK
R. Kohn, New York University, New York, NY, USA
P.S. Krishnaprasad, University of Maryland, College Park, MD, USA
J.D. Murray, University of Oxford, Seattle, WA, USA
C. Peskin, New York University, New York, NY, USA
S.S. Sastry, University of California, Berkeley, CA, USA
J. Sneyd, University of Auckland, Auckland, MI, USA
Problems in engineering, computational science, and the physical and biological
sciences are using increasingly sophisticated mathematical techniques. Thus, the
bridge between the mathematical sciences and other disciplines is heavily traveled.
The correspondingly increased dialog between the disciplines has led to the
establishment of the series: Interdisciplinary Applied Mathematics.
The purpose of this series is to meet the current and future needs for the
interaction between various science and technology areas on the one hand and
mathematics on the other. This is done, firstly, by encouraging the ways that
mathematics may be applied in traditional areas, as well as point towards new and
innovative areas of applications; and secondly, by encouraging other scientific
disciplines to engage in a dialog with mathematicians outlining their problems to
both access new methods as well as to suggest innovative developments within
mathematics itself.
The series will consist of monographs and high-level texts from researchers
working on the interplay between mathematics and other fields of science and
technology.
More information about this series at http://www.springer.com/series/1390

Alain Goriely
The Mathematics
and Mechanics of Biological
Growth
123
Alain Goriely
Department of Mathematics
University of Oxford
Oxford
UK
ISSN 0939-6047 ISSN 2196-9973 (electronic)

Interdisciplinary Applied Mathematics
ISBN 978-0-387-87709-9 ISBN 978-0-387-87710-5 (eBook)
DOI 10.1007/978-0-387-87710-5
Library of Congress Control Number: 2016963741
Mathematics Subject Classification (2010): 74L15, 74B20, 92C10, 92Bxx, 92C30, 92C50, 92C80
© Springer Science+Business Media LLC 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
Printed on acid-free paper
This Springer imprint is published by Springer Nature

The registered company is Springer Science+Business Media LLC
The registered company address is: 233 Spring Street, New York, NY 10013, U.S.A.
This book is dedicated to the 3Z:
Zébulon, Zakkai, and Zéphyr.
Not as an acknowledgement of their patience or help,
as they have none and gave me little,
but for the daily joy and chaos that they create
and for the fabulous examples of growth processes
that they generated over the last seventeen years.
My interest in growth conveniently coincided
with their birth and childhood.
Preface
To the nature lover, there is a distinct feeling of awe and beauty when observing the
gradual development of a child, the slow growth of trees, the fine structure of a
seashell, or the opening of a flower. Throughout human cultures and civilizations,
philosophers, artists, and scientists have marveled and pondered at the cycle of life,
the changes from an embryonic form to a newborn, the maturation of the newborn,
and the constant physiological renewal of the adult. All these processes can be
summarized by a single concept: growth. Growth provides an organism with the
ability to adapt and control its environment through its life and through time.
Growth is at the very core definition of life itself.
The problem of growth has been traditionally central to all aspects of biological
research but of marginal interest to physicists, engineers, and mathematicians.
However, in the last thirty years with the rise of medical bioengineering,
biophysics, and mathematical biology, the problematic of describing and under-
standing growth quantitatively has become a main topic of multidisciplinary
research.
Writing a book in an active field, spanning centuries of knowledge and covering
multiple disciplines, is a risky proposition. The idea for this monograph came to me
more than ten years ago when I realized that the general topics of mathematics of
growth was becoming a central theme of research for many scientists in different
communities. There was a clear need to bridge different concepts and ideas origi-
nating from multiple communities and, in particular, create a common language to
ix
x Preface
describe phenomena appearing in different scientific disciplines. This monograph is

an attempt in this direction.
Following my own interests and limited abilities, there is a strong bias in the
choice of topics presented in this monograph.
First, most of the descriptions are at the continuum level, essentially from tissues
to organs with very little discussion on cellular processes responsible for growth.
Whereas much is known at the cellular level, our understanding of transduction
mechanisms, linking cellular processes to tissue and organ growth, is still in
infancy.
Second, the emphasis is on physical and mechanical aspects of growth at the
level of organs and organisms but not at the population level. The mathematics of
evolving populations of cells or individuals, and their coupling with chemical fields
is well developed. It can be found in classic textbooks of mathematical biology and
will not be repeated here.
Third, the theory is developed around modern concepts of solid mechanics and
illustrated through the use of reduced simplified models that can be analyzed by the
methods of applied mathematics. Unavoidably, the concepts may be advanced but
the models are often simple. The hope is that these models provide some insight
into the mechanisms governing growth and the interplay between growth,
mechanics, and geometry. More realistic models would typically require both an
extensive discussion of the underlying biological system and extensive computa-
tional analyses. I leave these tasks to the experts in these different fields.
Fourth, the emphasis is on the consequences of growth rather than on its origin.
The discussion is mostly restricted to the analysis of tissue and organs made out of a
single elastic component rather than the more general theory of mixtures that takes
into account the coupling between fluids and various tissue components. These
advanced theories for growth and remodeling have been used to develop realistic
models but cannot be easily analyzed mathematically. They also require a more
general computational framework that is still in development. My general philos-
ophy is that little progress can be made for models with multiple components unless
we have a thorough understanding of the simpler problems studied here.
Fifth, whereas I try to provide general introductions to different topics and key
references to many authors, most of the topics presented here have come about
through my own research projects. I have worked on these with various collabo-
rators over the last twenty years. Therefore, this monograph is not an exhaustive
review of the field as much as my personal views on the subject. I do not believe
that it is the only approach or even that it is superior to other points of view. I would
like to encourage other researchers to provide alternative, complementary, or
contradictory approaches as it will only enrich the debate and help develop a
general theory of growth. While I have tried to be thorough in citing relevant works
in the literature, I have undoubtedly missed important references and, I can only
apologize to the colleagues that I have offended in the process.
This book is designed to be at the quadruple interface of mathematics, biology,
physics, and mechanics. Life at the interface is particularly rich and exciting as it
takes advantage of ideas, concepts, and methods from different fields. It is also
Preface xi
particularly dangerous as it is the ideal ecological zone for highly specialized

predators. I expect that biologists will find the biological modeling over-simplistic
and focussing on questions of little interest to them. I believe that many mathe-
maticians will find the mathematical description too informal and lacking the rigor
expected in various well-established disciplines ranging from partial differential
equations to differential geometry. Some engineers may lament at the lack of
finite-element simulations and detailed mechanical measurements. And, I fully
expect that many physicists will view the treatment of mechanics as being too
technical and unnecessarily complicated. These criticisms are all valid. It is the
curse of interdisciplinarity to always fall short of the expectations required by
disciplinary purity. But, it is only when these opprobriums will be bestowed on me
that I will know that I have managed to reach different communities and that I may
have attained some measure of success.
A Reader’s Map
This book was conceived to be read at different levels, depending on the reader’s
interests and background. The difficulty is that a mathematical and mechanical
theory of growth naturally combines aspects of biology, mathematics, and
mechanics. The bio-mechanician with a good grip of solid mechanics may not
always be familiar with some methods of applied mathematics. Similarly, many
applied mathematicians and physicists, while often well trained in fluid mechanics,
are not typically exposed to advanced concepts of solid mechanics. For the bio-
logically trained but mathematically inclined readers, mathematics and mechanics
may be appealing but may present technical difficulties. Accordingly, topics are
presented in order of conceptual and mathematical difficulties.
Inspired by the structure of the excellent textbook “Nonlinear Dynamics and
Chaos” by Steven Strogatz, I organized this monograph according to the dimen-
sionality of the problem, starting in dimension one before considering problems in
dimension two and only then presenting the general theory in three dimensions.
Indeed, the coupling of growth and mechanics can be illustrated in simplified
geometries where the fundamental concepts can be easily understood. Once these
concepts are understood, they are progressively generalized.
Part I presents a general introduction to growth, hopefully accessible to all
readers. It presents basic aspects and classification of growth processes and, more or
less, use historical developments and abundant examples from biology and phys-
iology to introduce key concepts relating biological growth to physical cues.
Part II was specifically developed for this monograph, both to introduce basic
mechanical ideas such as elasticity, viscoelasticity, and plasticity; but also to
illustrate the interplay between growth processes and mechanics. In the first chapter,
I discuss the simplest instances of growth by restricting deformations along a line.
In the process of writing this book, I realized that there was no general theory of
growth for filamentary structures. With Derek Moulton and Thomas Lessinnes,
xii Preface
we filled this gap and showed how to generalize the theory of elastic rods to include
the effects of growth and remodeling. These ideas are used to model many inter-
esting systems, mostly taken from the world of plants.
Part III further generalizes these concepts in simple two-dimensional geometries
with applications to accretive growth problems such as seashells and microbial
systems exhibiting tip growth. Most of the discussion of two-dimensional elastic
surfaces is restricted to axisymmetric membranes and shells. The general problem
of deriving a general theory of morphoelastic shells would require a few more
chapters and only a short introduction to the general problem is given.
Part IV presents a general theory of growth for three-dimensional bodies based
on the twin concept of evolving reference configuration and the multiplicative
decomposition of the deformation gradient. This part starts with a brief description
of the classic theory of nonlinear elasticity so that readers not versed in the language
of large deformations mechanics can learn the basic tools. An extensive discussion
on the kinematics of growth viewed as evolving configurations is presented. It is
followed by a general discussion on growth laws, dynamics, and stability. The two
last chapters are devoted to detailed examples and applications in spherical and
cylindrical geometries.
Rather than providing a final conclusion to a field that is still blooming,
I conclude, in Part V, with a list of ten challenges. It is my hope that these
challenges will motivate other researchers and help move the field forward.
Oxford, UK Alain Goriely

2016
Acknowledgements
I am deeply indebted to a number of friends and collaborators. First, it is with great

pleasure that I acknowledge the influence of Michael Tabor on both my profes-
sional and intellectual life. Twenty years ago, tired of playing with problems in the
theory of integrability, Michael and I decided to look at problems related to growth
and mechanics. At that time, we knew next to nothing about rods, membranes,
nonlinear elasticity or growth. All we had were a few questions that we thought
deserved to be investigated and a few methods from applied mathematics that
people did not seem to be using in mechanics. We taught each other the basics and
tried to model simple systems, inching our way to a better understanding. Many
of the basic ideas and applications presented in this book were born out of these
discussions.
Another major influence on this work was my collaboration with Martine Ben
Amar. During a sabbatical in Paris, we learned the foundations of rational
mechanics and developed together the theory of growth-induced instabilities for
nonlinear elastic systems. It was an exciting time of scientific discovery and it
remains one of the best academic years of my life.
I would also like to acknowledge the many striking contributions of Derek
Moulton through our years of collaboration. In his own work, Derek has been
exploring the consequences of growth in many different settings and, doing so, has
systematically propelled the field forward. We have been partners in crime for many
years, shared the meager rewards of our malfeasance, and paid the price for our
idiosyncratic interests. I hope that his punishment will be lighter than mine.
Many of the ideas discussed in this monograph are also the product of many
years of work, discussions, reflection, and meetings with my collaborators. In
particular, I have tremendously benefited from my interactions with Jon Chapman,
Régis Chirat, Michel Destrade, Marcelo Epstein, Krishna Garikipati, Ray
Goldstein, Andrew Hausrath, Ellen Kuhl, Thomas Lessinnes, John Maddocks,
Angela Mihai, Sébastien Neukirch, Giuseppe Saccomandi, Dominic Vella, Sarah
Waters, John Wettlaufer, and Arash Yavari. I take this opportunity to let them know
that their intellect and friendship are precious to me.
xiii
xiv Acknowledgements
Central to all academic endeavors is the training and interactions with students.
I have been very lucky to be able to supervise several bright students whose hard
work features prominently here: Alexandre Erlich, Georgina Lang, Tyler McMillen,
Andrey Melnik, Joe McMahon, and Rebecca Vandiver.
It is with great pleasure that I acknowledge the courage of many helpful readers
to climb mountains of typos, to dig into incoherent statements, to sort out incon-
sistent notations, and to uncover logical flaws. In particular, I am indebted to: Nita
Goriely, James Kwiecinski, Angela Mihai, Oliver O’Reilly, Thomas Woolley, and I
am especially grateful to Nicola Kirkham for her outstanding work and to Achi
Dosanjh and Nick Valente at Springer for their patience and support. Their feed-
back and help are gratefully acknowledged and, obviously, the remaining mistakes
are all mine.
Finally, none of it would have happened without the constant help of my wife,
Nita. I will say no more, as the rest is between the two of us.
Contents
Part I Introduction: Where It All Starts

1 Basic Aspects of Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1 Classification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.1 Tip Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.2 Accretive Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.1.3 Volumetric Growth . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2 The Scaling of Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Relative Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.4 The Kinematics of Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2 Mechanics and Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.1 Growth is Influenced by Stress . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.1.1 The Growth of Stems . . . . . . . . . . . . . . . . . . . . . . . . 30
2.1.2 The Growth of Axons . . . . . . . . . . . . . . . . . . . . . . . . 30
2.1.3 Thoma’s Law for Arteries . . . . . . . . . . . . . . . . . . . . . 32
2.1.4 Woods’ Law for the Heart . . . . . . . . . . . . . . . . . . . . . 32
2.1.5 Wolff’s Law for Bones . . . . . . . . . . . . . . . . . . . . . . . 34
2.1.6 Davis’ Law for Soft Tissues . . . . . . . . . . . . . . . . . . . 36
2.1.7 Tumor Spheroid Growth . . . . . . . . . . . . . . . . . . . . . . 38
2.2 Stress is Influenced by Growth . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.2.1 Tissue Tension in Plants . . . . . . . . . . . . . . . . . . . . . . 40
2.2.2 Residual Stress in Physiology . . . . . . . . . . . . . . . . . . 43
2.3 The Theory of Morphoelasticity . . . . . . . . . . . . . . . . . . . . . . . . 45
2.4 A Short History of Growth Modeling . . . . . . . . . . . . . . . . . . . . 46
2.5 A Review of Reviews . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3 Discrete Computational Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.1 On-Lattice Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.1.1 Cellular Automata . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.1.2 Cellular Potts Models . . . . . . . . . . . . . . . . . . . . . . . . 52
xv
xvi Contents
3.2 Off-Lattice Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

3.2.1 Center Dynamics Models . . . . . . . . . . . . . . . . . . . . . . 54
3.2.2 Vertex Dynamics Models . . . . . . . . . . . . . . . . . . . . . 56
3.2.3 Advantages and Drawbacks . . . . . . . . . . . . . . . . . . . . 58
Part II Filament Growth: A One-Dimensional Theory

4 Growing on a Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1 Example: A Growing Rod in One Dimension. . . . . . . . . . . . . . 66
4.2 Purely Elastic Deformations . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.3 Growth Without Elastic Deformations . . . . . . . . . . . . . . . . . . . 69
4.3.1 Example: Tip Growth . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3.2 Application: Spheroid Tumor Growth . . . . . . . . . . . . 71
4.4 Growth with Elastic Deformation . . . . . . . . . . . . . . . . . . . . . . . 76
4.4.1 Growth of a Rod Between Two Plates . . . . . . . . . . . 76
4.4.2 Three Different Configurations. . . . . . . . . . . . . . . . . . 77
4.4.3 Homeostatic Growth . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.4.4 Application: The Growth of Neurons. . . . . . . . . . . . . 80
4.4.5 Is This Just Plasticity? . . . . . . . . . . . . . . . . . . . . . . . . 85
4.5 Application: The Growth of Plant Cells . . . . . . . . . . . . . . . . . . 88
4.5.1 Lockhart Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.5.2 Extending Lockhart’s Model . . . . . . . . . . . . . . . . . . . 92
5 Elastic Rods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.1 The Kinematics of Curves and Rods. . . . . . . . . . . . . . . . . . . . . 98
5.1.1 Curves and Frenet Frames . . . . . . . . . . . . . . . . . . . . . 98
5.1.2 Rods and General Frames . . . . . . . . . . . . . . . . . . . . . 99
5.1.3 Inextensible, Unshearable Rods . . . . . . . . . . . . . . . . . 102
5.2 The Mechanics of Elastic Rods . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.2.1 Balance of Linear Momentum . . . . . . . . . . . . . . . . . . 104
5.2.2 Balance of Angular Momentum . . . . . . . . . . . . . . . . . 106
5.2.3 Local Mechanics of Rods . . . . . . . . . . . . . . . . . . . . . 107
5.3 Constitutive Laws for Elastic Rods . . . . . . . . . . . . . . . . . . . . . . 108
5.3.1 Extensible and Shearable Elastic Rods. . . . . . . . . . . . 108
5.3.2 Inextensible and Unshearable Rods . . . . . . . . . . . . . . 108
5.3.3 Isotropic, Extensible, but Unshearable Rods . . . . . . . 110
5.4 Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.5 Bending and Torsional Stiffnesses . . . . . . . . . . . . . . . . . . . . . . 111
5.6 The Kirchhoff Elastic Rod Model: A Summary . . . . . . . . . . . . 113
5.7 Example: Helical Rods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.7.1 Geometry of Helices . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.7.2 Helical Equilibria . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.7.3 Overwinding or Underwinding Helices . . . . . . . . . . . 119
Contents xvii
5.8 The Planar Elastica: Bernoulli–Euler Equations . . . . . . . . . . . . 122

5.8.1 Static Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6 Morphoelastic Rods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.1 Kinematics of a Growing Rod . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.2 Mechanics of a Growing Rod . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.3 Evolution Laws for Growing Rods . . . . . . . . . . . . . . . . . . . . . . 129
6.4 Example: The Remodeling of Stems . . . . . . . . . . . . . . . . . . . . . 130
6.5 A Buckling Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
6.5.1 Example: Michell’s Instability . . . . . . . . . . . . . . . . . . 133
6.5.2 A General Perturbation Expansion . . . . . . . . . . . . . . . 136
6.5.3 Bifurcation Criterion for Elastic Buckling . . . . . . . . . 138
6.5.4 Example: The Growing Ring . . . . . . . . . . . . . . . . . . . 138
6.5.5 A Growing Ring with Remodeling . . . . . . . . . . . . . . 140
6.6 Growing Rods on a Rigid Foundation . . . . . . . . . . . . . . . . . . . 142
6.6.1 Example: A Growing Ring on a Foundation . . . . . . . 143
6.6.2 Example: A Straight Rod Growing
on a Foundation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
6.7 Application: Growing Vines . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
6.7.1 Perversion in Tendrils . . . . . . . . . . . . . . . . . . . . . . . . 149
6.7.2 Twining Vines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
6.7.3 Application: The Growth of Bacillus subtilis . . . . . . . 168
Part III Surface Growth: A Two-Dimensional Theory

7 Accretive Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
7.1 Intrinsic Accretive Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
7.1.1 Simple Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
7.1.2 Shape Planarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
7.1.3 Shape Invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
7.1.4 Self-Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
7.2 Application: The Growth of Seashells. . . . . . . . . . . . . . . . . . . . 187
7.2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
7.2.2 Geometric Description . . . . . . . . . . . . . . . . . . . . . . . . 189
7.2.3 Accretive Growth of Seashells . . . . . . . . . . . . . . . . . . 193
7.2.4 Other Accreted Structures . . . . . . . . . . . . . . . . . . . . . 194
7.2.5 The Role of Mechanics in Morphological
Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 196
8 Membranes and Shells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
8.1 Elastic Membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
8.1.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
8.1.2 Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
xviii Contents
8.1.3 Constitutive Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

8.1.4 A Complete Set of Equations . . . . . . . . . . . . . . . . . . 214
8.1.5 Application: The Shape of Sea Urchins . . . . . . . . . . . 216
8.2 Nonlinearly Elastic Shells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
8.2.1 Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
8.2.2 Scalings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
8.2.3 Application: The Rice Blast Fungus . . . . . . . . . . . . . 224
9 Growing Membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
9.1 Morphoelastic Membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
9.2 Application: Microbial Tip Growth . . . . . . . . . . . . . . . . . . . . . . 234
9.2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
9.2.2 Bacterial Filaments: Actinomycetes . . . . . . . . . . . . . . 234
9.2.3 Fungi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
9.2.4 Root Hairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
9.2.5 Modeling of Tip Growth . . . . . . . . . . . . . . . . . . . . . . 239
9.2.6 A Model for Hyphal Growth . . . . . . . . . . . . . . . . . . . 240
9.2.7 Tip Shapes for Filamentous Bacteria . . . . . . . . . . . . . 241
9.2.8 Lysis and Beading . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
9.2.9 Shear Stress and the Normal Growth Hypothesis. . . . 247
10 Morphoelastic Plates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
10.1 Elastic Plates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
10.1.1 Mean and Gaussian Curvatures . . . . . . . . . . . . . . . . . 252
10.1.2 Growing Elastic Plates . . . . . . . . . . . . . . . . . . . . . . . . 254
Part IV Volumetric Growth: A Three-Dimensional Theory

11 Nonlinear Elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
11.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
11.1.1 Scalars, Vectors, and Tensors . . . . . . . . . . . . . . . . . . 264
11.1.2 Spatial Derivatives of Tensors . . . . . . . . . . . . . . . . . . 266
11.1.3 Derivatives in Curvilinear Coordinates . . . . . . . . . . . 268
11.1.4 Derivatives of Scalar Functions of Tensors . . . . . . . . 270
11.1.5 The Deformation Gradient . . . . . . . . . . . . . . . . . . . . . 272
11.1.6 Volume, Surface, and Line Elements . . . . . . . . . . . . . 274
11.1.7 Polar Decomposition Theorem . . . . . . . . . . . . . . . . . . 276
11.1.8 Velocity, Acceleration, and Velocity Gradient . . . . . . 277
11.2 Balance Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
11.2.1 Balance of Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
11.2.2 Balance of Linear Momentum . . . . . . . . . . . . . . . . . . 280
11.2.3 Balance of Angular Momentum . . . . . . . . . . . . . . . . . 282
11.2.4 Many Stress Tensors . . . . . . . . . . . . . . . . . . . . . . . . . 283
11.2.5 Balance of Energy for Elastic Materials. . . . . . . . . . . 284
Contents xix
11.3 Constitutive Equations for Hyperelastic Materials . . . . . . . . . . . 285

11.3.1 Internal Material Constraints . . . . . . . . . . . . . . . . . . . 286
11.4 Summary of Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
11.5 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
11.6 Objectivity and Material Symmetry . . . . . . . . . . . . . . . . . . . . . 288
11.7 Isotropic Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
11.7.1 Adscititious Inequalities . . . . . . . . . . . . . . . . . . . . . . . 291
11.7.2 Choice of Strain-Energy Functions . . . . . . . . . . . . . . 293
11.8 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
11.8.1 A Simple Homogeneous Deformation . . . . . . . . . . . . 296
11.8.2 The Half-Plane in Compression . . . . . . . . . . . . . . . . . 298
11.8.3 The Inflation–Extension of a Tube. . . . . . . . . . . . . . . 299
11.9 Universal Deformations for Isotropic Materials . . . . . . . . . . . . . 305
11.10 Bifurcation, Buckling, and Instability . . . . . . . . . . . . . . . . . . . . 310
11.10.1 Example: Bifurcation of the Half-Plane . . . . . . . . . . . 314
11.11 Anisotropic Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320
11.11.1 One Fiber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
11.11.2 Two Fibers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
11.11.3 Example: The Fiber-Reinforced Cuboid . . . . . . . . . . . 324
11.11.4 Example: The Fiber-Reinforced Cylinder. . . . . . . . . . 328
11.11.5 Application: The Hydrostatic Skeleton. . . . . . . . . . . . 336
11.11.6 Fiber Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
12 The Kinematics of Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
12.1 A Thought Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
12.2 Relieving Stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
12.3 The Conceptual Hypothesis of Morphoelasticity . . . . . . . . . . . . 353
12.4 Example: The Growing Ring . . . . . . . . . . . . . . . . . . . . . . . . . . 355
12.5 The Problem of Incompatibility . . . . . . . . . . . . . . . . . . . . . . . . 358
12.5.1 A Differential Geometry Perspective . . . . . . . . . . . . . 359
12.5.2 An Analytic Perspective . . . . . . . . . . . . . . . . . . . . . . 371
13 Balance Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
13.1 The Slow-Growth Assumption . . . . . . . . . . . . . . . . . . . . . . . . . 376
13.2 Balance of Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
13.3 Balance of Linear and Angular Momenta . . . . . . . . . . . . . . . . . 380
13.4 Energy Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
13.5 Imbalance of Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
13.6 Elastic Constitutive Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
13.7 Summary of Volumetric Morphoelasticity . . . . . . . . . . . . . . . . 384
13.8 Simple Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
13.8.1 A Growing Cuboid . . . . . . . . . . . . . . . . . . . . . . . . . . 385
13.8.2 Two Growing Cuboids . . . . . . . . . . . . . . . . . . . . . . . 386
13.8.3 A Growing Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . 388
xx Contents
13.9 Mixture Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392

13.9.1 Classical Mixture Theory . . . . . . . . . . . . . . . . . . . . . . 393
13.9.2 Constrained Mixture Theory . . . . . . . . . . . . . . . . . . . 396
14 Evolution Laws and Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
14.1 Symmetry of the Growth Tensor . . . . . . . . . . . . . . . . . . . . . . . 400
14.2 Isotropic Growth and Gel Swelling . . . . . . . . . . . . . . . . . . . . . . 403
14.3 Discrete Growth Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404
14.4 The Thermodynamics Perspective . . . . . . . . . . . . . . . . . . . . . . . 405
14.5 Phenomenological Laws and Homeostatic Stress . . . . . . . . . . . 408
14.6 Dynamics of Homogeneous Deformations . . . . . . . . . . . . . . . . 410
14.6.1 Diagonal Deformations . . . . . . . . . . . . . . . . . . . . . . . 411
14.6.2 A Two-Dimensional Example . . . . . . . . . . . . . . . . . . 415
14.7 Remodeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
14.7.1 Fiber Remodeling of a Cuboid in Tension . . . . . . . . . 421
14.8 Growth Induced Instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
15 Growing Spheres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
15.1 The Growing Shell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
15.1.1 Kinematics of Growing Spheres . . . . . . . . . . . . . . . . 432
15.1.2 Stresses in a Growing Sphere . . . . . . . . . . . . . . . . . . 435
15.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437
15.2.1 Anisotropic Growth . . . . . . . . . . . . . . . . . . . . . . . . . . 437
15.2.2 Differential Growth . . . . . . . . . . . . . . . . . . . . . . . . . . 438
15.3 Limit-point Instability and Inflation Jump . . . . . . . . . . . . . . . . . 442
15.3.1 The Effect of Growth on the
Limit-point Instability . . . . . . . . . . . . . . . . . . . . . . . . 448
15.4 Singularities in Growing Solid Spheres . . . . . . . . . . . . . . . . . . . 449
15.5 Cavitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 450
15.5.1 Cavitation Induced by Tensile Loading . . . . . . . . . . . 452
15.5.2 Cavitation Induced by Growth . . . . . . . . . . . . . . . . . . 452
15.6 Instability Due to Anisotropic Growth . . . . . . . . . . . . . . . . . . . 453
15.6.1 A Numerical Scheme . . . . . . . . . . . . . . . . . . . . . . . . . 460
15.6.2 Thin-Shell Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462
15.6.3 Thick-Shell Limit. . . . . . . . . . . . . . . . . . . . . . . . . . . . 465
15.6.4 Bifurcation of the Growing Shell . . . . . . . . . . . . . . . . 467
15.7 Instability Due to Differential Growth . . . . . . . . . . . . . . . . . . . . 470
15.7.1 Instability in a Shrinking Shell . . . . . . . . . . . . . . . . . 471
15.7.2 Instability of a Shell Growing Inside a Medium . . . . 473
16 Growing Cylinders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475
16.1 Kinematics of the Growing Cylinder . . . . . . . . . . . . . . . . . . . . 476
16.2 Application: Cavitation in Plants. . . . . . . . . . . . . . . . . . . . . . . . 480
16.2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
16.2.2 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483
Contents xxi
16.2.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484

16.2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484
16.3 Bifurcation of Growing Cylinders . . . . . . . . . . . . . . . . . . . . . . . 487
16.3.1 Buckling Versus Barreling . . . . . . . . . . . . . . . . . . . . . 487
16.3.2 Bifurcation and Buckling in Growing Cylinders . . . . 493
16.3.3 The Effective Rigidity of a Growing Cylinder . . . . . . 496
16.3.4 An Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 498
16.4 Application: Tissue Tension in Plants . . . . . . . . . . . . . . . . . . . . 499
16.4.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
16.4.2 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501
16.4.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
16.4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 507
16.5 Application: The Buckling of Arteries . . . . . . . . . . . . . . . . . . . 509
16.5.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
16.5.2 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
16.5.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 518
16.5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
16.6 Circumferential Buckling and Mucosal Folding . . . . . . . . . . . . 524
16.6.1 Example: Circumferential Buckling
in a Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
16.6.2 Example: A Two-Layer Cylinder . . . . . . . . . . . . . . . . 528
16.7 Application: Asthma and Airway Remodeling . . . . . . . . . . . . . 530
16.7.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 530
16.7.2 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 532
16.7.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534
16.7.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 538
16.8 Residual Stress Through Fiber Contraction . . . . . . . . . . . . . . . . 539
16.8.1 Rotation of a Pressurized Anisotropic Cylinder . . . . . 540
16.9 Application: The Spiral Growth of Phycomyces . . . . . . . . . . . . 542
16.9.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542
16.9.2 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 544
16.9.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 546
16.9.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 551
Part V Conclusion: Where It Does Not End

17 Ten Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555
17.1 The Rheology of Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555
17.2 The Regulation of Growth and Growth Size . . . . . . . . . . . . . . . 557
17.3 The Elusive Growth Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 559
17.4 Multiscale: From Discrete to Continuous and Back . . . . . . . . . 561
17.5 Growth Versus Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564
17.6 Multiphysics: Coupling Growth with Other Fields . . . . . . . . . . 567
17.7 A Theory of Accretive Growth . . . . . . . . . . . . . . . . . . . . . . . . . 570
xxii Contents
17.8 Dynamics and Post-bifurcation Behavior . . . . . . . . . . . . . . . . . 572

17.9 Active Forces, Actives Stresses, and Active Strains . . . . . . . . . 574
17.10 The Mathematical Foundations . . . . . . . . . . . . . . . . . . . . . . . . . 578
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 637
Part I
Introduction: Where It All Starts
At first the infant,

Mewling and puking in the nurse’s arms.
Then, the whining school-boy with his satchel
And shining morning face, creeping like snail
Unwillingly to school...
Chapter 1
Basic Aspects of Growth
Growth is a generic term that describes processes in which the mass of a body changes
over time. In biology, the problem of growth is fundamental to all aspects of life, with
realizations as diverse as cell division, morphogenesis, development, maintenance,
cancer, and aging. All life forms experience growth to some appreciable degree
and one of the ultimate challenges of modern biology is to understand the role of
the genetic code in transforming cells into fully mature organisms and explaining
how these organisms manage to regulate shape and function through growth and
remodeling.
Growth processes also appear in some physical processes where new material is
added to an evolving system. For instance, in epitaxial growth, a thin layer of crystal
is produced by depositing raw material on an existing crystalline substrate. Growth
is also associated with phase transition phenomena, where an interface evolves in
time to produce new structures such as crystals [81, 687, 765]. These free boundary
problems are controlled by diffusion and, unlike biological systems, the interface is
a line of discontinuity with no particular material property. The swelling of gels is
arguably the closest nonbiological process that mimics growth as it is non-diffusive
and occurs in the bulk of the material. It can be used as a physical analog to gain
insight into the role of mechanics in biological pattern formation [116, 663, 904].
© Springer Science+Business Media LLC 2017 3
A. Goriely, The Mathematics and Mechanics of Biological Growth,
Interdisciplinary Applied Mathematics 45, DOI 10.1007/978-0-387-87710-5_1
4 1 Basic Aspects of Growth
Our study of growth will mostly be limited to non-diffusive processes relevant to

biological systems.
1.1 Classification
Aspects of growth and remodeling occur during the entire life of an organism. There-
fore, growth fulfills many purposes and, accordingly, is associated with qualitatively
different processes. Traditionally, a first classification is obtained by considering the
way growth alters a body, either by changing its volume, its material properties, or by
rearranging the relative position of material points. The three main growth processes
are:
• Growth. The term growth by itself refers to a change in mass. It is colloquially
understood as an increase in mass, but the concept extends naturally to describe
mass reduction or shrinking. Change in mass can be created either by addition of
mass at constant density, as found in the development of soft tissues, a change in
density at constant volume, as in the case of bone densification, or both as found
in a developing bone. Mathematically, a theory of growth must allow for changes
in mass, volume, and density and must be flexible to account for mass permeating
through the boundary of the body, accumulating at the boundary, or occurring
within the body itself.
• Remodeling. It is well known that in the process of aging, tissues may become
stiffer or softer. The term remodeling refers to an evolution of material proper-
ties in a system without change of mass such as stiffness, fiber orientation, fiber
strength, and so forth. These remodeling processes are due to a change in the
microstructure that determines the overall behavior of the tissue. For instance, the
typical composition of soft tissues in many animals is a mixture of collagen fibers
within an elastin matrix. Whereas elastin content remains mostly unchanged over
many years, there is a continuous turnover of collagen that depends on the local
biochemical and mechanical stimuli acting on the cells. The relative content of
different types of collagen fibers and elastin determines the overall response of the
tissue [635]. This process can occur without a change of mass, but it is crucial to
understand the response of a tissue under mechanical loads. From a mathematical
perspective, the variation of material properties can either be modeled by consid-
ering a separate evolution of the material parameters of a system or, at a lower
scale, by taking into account the evolution of separate tissue components.
• Morphogenesis. Early in embryonic life, new tissues and organs are formed. In
this process, major reorganization and differentiation of cells take place after cell
division, and, importantly, there is a restructuring of material elements. This reor-
ganization process can only happen if the adhesion between different components
is weak enough so that they can separate and reattach. This simple observation
has important consequences for modeling as tissues undergoing morphogenesis
exhibit rapid elastic stress relaxation and plastic-like flow. Mathematically, this
1.1 Classification 5
Figure 1.1 Growing root as

depicted by Duhamel in
1758 [898].
evolution is often described by modeling tissues as fluid or viscoelastic rather than

elastic, even though these two points of views are equivalent, as we will show.
Growth can be further classified by the location of material addition. It can occur
at the tip, on the surface, or in the bulk. We consider these processes in more detail
next.
1.1.1 Tip Growth
In his book, “The Physics of Trees” published in 1758, Duhamel du Monceau

describes the following discovery: “The observations that I made in plants that grow
in water helped me discover another peculiar phenomenon; that is, roots only grow at
the tip.” [898, p. 83] (see Figure 1.1). Tip or apical growth describes growth processes
that take place in a small region at the tip of a filamentary structure. As shown in
Figure 1.2, it is the main growth mechanism used by many microbial organisms and
plant systems such as fungi, filamentary bacteria, pollen tubes, and root hair [431,
457, 479, 480, 482, 541, 625, 714, 715, 899, 1073]. In these organisms, there is
a small active growing zone located near the tip where the outer shell is constantly
remodeled and new material is consistently added.
A. B. C. D.
E. F.
Figure 1.2 Tip growth. A. microbacterial filament, (bar is 1 μm) (Kitasatospora setae source:
Society for Actinomycetes Japan, photograph by Y. Takahashi and Y. Iwai); B. microbacterial
filament (Streptomyces coelicolor A3(2)), (bar is 1 μm); C. Allomyces; D. lily root hair (typical
diameter 15–20 μm) (images courtesy of Jacques Dumais). E. Branching in streptomyces A3(2). F.
First theoretical description of tip growth by Reinhardt in 1892 for the growth of fungus [1073].
For sufficiently small organisms, such as filamentary bacteria, the newly added
material is transported to the tip by diffusion processes whereas for larger organisms,
such as fungi and pollen tubes, active transport processes which require a complex
internal structure are required for propagation.
Typically growth occurs in a region that maintains a constant size. Therefore, the
typical scaling of mass with time is linear, as shown in Section 4.3. However, many
filamentary structures also undergo repeated branching as shown in Figure 1.2E.
These branches allow for one-dimensional structures to explore three-dimensional
volume for nutrients. This branching process transforms the scaling of mass from
being linear to exponential with respect to time since each new branch adds mass
linearly and can branch again, leading to an intricate, fractal-like, geometry.
Tip growth is sometimes referred to as primary growth in plants. Once a stem
or root is large enough, it may also undergo secondary growth, sometimes referred
to as growth in the cambium, by the addition of external layers to thicken itself.
Cambial growth in trees takes place between the old wood and the bark of the tree,
and seasonal variations lead to the formation of tree rings.
Mechanically, the main issue is to understand the interaction of the tip with its
environment and to determine its shape, internal stresses, and how a filamentary
structure evolves based on different laws for material addition.
1.1.2 Accretive Growth
Accretive, surface, or appositional growth describe mechanisms, such as deposition

in hard tissues, where new material is added to the boundary of an existing body. It
is the typical mechanism responsible for the formation of teeth, seashells, horns, and
corals [4, 914, 915, 1173]. At the microscopic level, surface growth is also found
in bones where changes in bone density occur by deposition or resorption of new
material on the surface of trabeculae, or the walls of canals excavated by osteoclasts
[1230].
Mathematically, many problems in accretive growth can be modeled geometrically
by studying the evolution of the boundary as a function of accreted mass. For instance,
the shell of the ammonite in Figure 1.3 is the result of material being accreted
with different velocities at different points of the opening. This velocity gradient
is necessary for coiling since the outer ridge requires more material than the inner
ridge. In general, if the boundary is known at time t, an accretion vector, giving the
local velocity of material accretion, can be defined at each point on the boundary as
shown in Figure 1.4D. The new boundary is then obtained after a small incremental
time t.
Figure 1.3 The fossil shell

of an ammonite of the genus
Promicroceras [916]. Arrows
denote the accretion vectors.
A. B. C.
D. 0
E.
Figure 1.4 Accretive growth. ABC. Growth of horns and seashells as described by D’Arcy Thomp-
son [1230]. D. The theoretical basis of accretive growth. For a given body at time t, a growth vector
is defined at the active growth boundary. At time t + t a new surface is created. E. Different body
forms obtained by different choices for the growth vector (adapted from [914]).
The key question from a modeling point of view is to derive an accretion law
based on the interaction between the soft part of the animal and the new hard accreted
part. Mathematically, the main problem is to understand and classify the shapes and
patterns that emerge through this process [202].
1.1.3 Volumetric Growth
Bulk, volumetric or interstitial growth all refer to processes in which local volume
elements in the body change over time rather than on its boundary as in accretive or
tip growth. Bulk growth is typical of many developmental, physiological, and patho-
logical processes and has been particularly well documented in arteries, muscles,
solid tumors, and the heart [242, 636, 1206].
Bulk growth encompasses hyperplasia, the increase of volume due to cell pro-
liferation which is typical of many developmental systems; hypertrophy, the change
of volume due to enlargement of its constituents, typical of many physiological
processes; and neoplasia, the abnormal and often unregulated growth or division of
cells found in cancer.
Mathematically, bulk growth offers many outstanding challenges. First, a local
volume element may not remain isotropic during growth which implies that a ten-
sorial description of the deformation is needed. Consider an infinitesimal spherical
volume element extracted from a tissue, as shown in Figure 1.5. In volumetric growth,
this sphere is transformed into an ellipsoid. This deformation is fully captured by the
transformation of three orthogonal vectors into three new vectors along the princi-
pal directions of the ellipsoid. Mathematically, this local mapping is described by a
tensor. Accordingly, the process of volumetric growth assigns a tensor, the growth
tensor G, at each point of the body, describing the change of a local volume element.
A theory of growth is therefore intrinsically rooted in the theory of tensors.
A second challenge in bulk growth is that it is relevant for soft tissues that are
elastic and can also be deformed by the application of loads. When observing the
deformation of a body from an initial state, we must determine whether it is due
to the growth or elastic response of the material, or even a combination of the two
processes. This problem, which will be discussed at length in further chapters, was
Figure 1.5 Volumetric growth. During growth, a local element inside the tissue can expand or
shrink anisotropically. The description of such a process requires the definition of a growth tensor
G, locally transforming an infinitesimal sphere to an infinitesimal ellipsoid.
first formulated by Hsu in 1968, arguably the first work to address the fundamental
problem of mechanical growth modeling: “If the form to which a body grows under
no applied loads is known, what will be the form of the body if some mechanical
loads are applied during its growth?” [627].
From a modeling perspective, bulk growth can also be used to describe the growth
of filamentary objects, such as neurons or plant stems that have distributed growth,
or surfaces, such as epidermal layers of cells that do not change their thickness.
The classification of growth as tip, accretive, or bulk is descriptive but not formal.
It depends on the scale at which the problem is being studied. Indeed, both tip and
accretive growth can be modeled as bulk growth processes where a thin soft layer,
close to the boundary, expands and stiffens in time. Essentially, tip growth is a bulk
process that is localized at the apical part of the filament. A detailed analysis of
this process requires an understanding of areal growth due to the insertion of new
material. Similarly, bone growth and wound healing can be modeled either as an
accretive process or a localized bulk process where density evolves.
1.2 The Scaling of Growth
The first question about growth is one that is shared by most parents: How does
the size of an organism change in time? The first historical records of growth were
on the height of human beings. In 1759, Count Philibert Gueneau de Montbeillard
started to record the height of his first son on his birth and he continued to do so
every six months until his son reached the age of 18. This record, shown in Figure
1.6, was published in the fourth volume of the supplement to Buffon’s “Histoire
Height (cm)
180
170
160
150
de Montbeillard’ son
140
130
120
110
100
90
Zébulon Goriely
80
70
60 Age (years)
50
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Figure 1.6 Record of longitudinal growth done by Count Montbeillard from 1759 (data from
[1123]), compared to the height of Zébulon, the author’s first son.
1.2 The Scaling of Growth 11
Figure 1.7 Growth curves obtained from statistical averages (from Stratz’s book published in 1904
[1200]).
Naturelle” [1123]. A quarter of a millennium later, it still captures the essence of

human development from birth to adulthood. Its gross features are similar to the
growth curve of the author’s first son as well as the curves found in modern medical
growth charts, shown in Figure 1.7. These curves are the result of statistical averages
over large populations and show variations over different geographic regions.
The first complete statistical study of the evolution of weight and height in human
beings was conducted by the Belgian polymath, Adolphe Quetelet, and published
in his “Treatise on Man” in 1835 [1053]. In this remarkable book, Quetelet pooled
and analyzed data from the Belgian population. He considered both general trends
and deviations around the average using the Gaussian distribution. This work is
considered to be the first application of the Gaussian distribution in a statistical
study. Quetelet also suggested a law of growth by fitting the data of height H (t)
against time t
b+t
H (t) = at + . (1.1)
1 + 43 t
If both mass M(t) and height H (t) are known, one can speculate on how these
two quantities might be related to each other. It was again Quetelet who proposed
that, independent of time, weight is related to height by a scaling law of the form
M = cH α , where α = 5/2 during childhood and α = 2 for adults. For α = 2, the
Quetelet coefficient c becomes the infamous Body Mass Index (BMI), which remains
a critical estimate of fitness and obesity despite its obvious shortcomings and many
criticisms [283]. Historically, Quetelet’s proposal that height scales with respect to
weight is the first case of an allometric law, that is, a power law between a given
physical quantity and the total mass of the organism [649, 816, 1126, 1147] (see
Section 1.3).
Quetelet’s book prompted similar studies and, by the end of the nineteenth cen-
tury, there was a flurry of activity dedicated to understanding growth in humans,
animals, and plants [292]. The idea emerged that growth in different biological organ-
isms could follow some universal laws that would determine sizes and forms. This
research program was presented in the seminal book “On Growth and Form” by
D’Arcy Thompson, first published in 1917 [1230]. This monumental book remains
a reference and a source of inspiration for many researchers to this day.
Bogin reports that, by 1972, no less than 200 different models or fitting functions
had been proposed to describe the evolution of human growth [110]. The modeling
of growth as a slow continuous process is only valid on the long-time scale of months
or years. A detailed daily recording and analysis of height in infants between the age
of one and two shows that growth mostly occurs through discontinuous aperiodic
saltatory spurts [759, 760] of about 0.5 to 2.5 centimeters separated by long intervals
typically lasting between 2 to 63 days, where no growth takes place as shown in Figure
1.8. Jumps and oscillations during growth are also encountered in other biological
systems including fungi [799], pollen tubes [188, 360, 590], plants [354, 728], and
invertebrates [78].
The first attempt to model the evolution of the total mass from physical principles
was based on an idea by Pütter, first published in 1920 [1048]. Pütter proposed that
animal growth can be seen as a balance between addition and removal of building
materials in the body. Growth proceeds as long as new material is added faster than
it is removed, and stops when both processes are balanced.
Typically, in such models, the rate of material removal is assumed to be propor-
tional to the mass M(t) itself, as in a standard exponential decay problem, whereas
the rate of addition of new material is proportional to a power of the mass. Combining
these two effects leads to [1285]
Height (cm)
74
72
70
68
66
64
62
60
Age (days)
58
80 100 120 140 160 180 200 220
Figure 1.8 Daily measurements of the height in a boy between the age of 90 to 218 days show
discontinuous growth spurts separated by periods of no growth (data from [760]).
Ṁ = M(a M − p − b), (1.2)
where Ṁ ≡ dM/dt denotes the time derivative of M(t) and ab > 0. The solution
of this equation for p = 0 describes a simple exponential process
M(t) = M0 e(a−b)t , (1.3)
where M0 = M(t = 0) is the mass at birth.

For p = 0, Equation (1.2) is a Bernoulli equation [460, p. 232] whose general
solution is given by
p
M(t) M0 p −bpt
=1− 1− e , (1.4)
M∞ M∞
where M∞ = (a/b)1/ p is the asymptotic mass. Figure 1.9 shows examples of mass
evolution as a function of time for various values of the parameters.
Equation (1.2) also contains the classic logistic model [1255] obtained for p = −1
with a and b negative, and the Richards model [1078], defined by p < −1, used to
model the growth of plants.
While this general approach is appealing, the choice for the exponent p is prob-
lematic. A typical argument is to assume that the addition of new material is limited
by energy input and metabolic rates. If the energy intake and growth rate are directly
proportional to the weight itself, we have Ṁ = αM for a certain α = 0. This behavior
is recovered with a > b > 0, p = 0, and exponential unlimited growth is observed.
M(t)
M
=1
3, b
=1
/
p=1
-1
,b
b=
1
p=
,
-1
p=
M0
0
0 t
Figure 1.9 Examples of growth curves given by Equation (1.4) for various values of the exponent p.
Figure 1.10 Example of

data fitting with p = 1/4 by
von Bertalanffy [1285]
(Abramis brama is the
common bream).
Time in days
This behavior, observed in the early developmental stages of insects, is only valid
until a new phase of growth takes place.
If we assume that the metabolic rate follows a scaling law based on geometry by
postulating that energy intake is proportional to surface area, we have Ṁ = αM 2/3 ,
which corresponds to the choice p = 1/3. This power law was originally proposed in
1839 by Sarrus and Rameaux [1118]. Based on the same general ideas, Ludwig von
Bertalanffy [1285] suggested that most biological systems grow in an intermediate
regime with exponent 0 < p < 1/3 and suggested p = 1/4 as suitable to fit different
data sets, as shown in Figure 1.10.
Figure 1.11 Kleiber’s law: metabolic rate as a function of mass [704]. The experimental data for
the variation shows a 3/4 power law. It is compared with a scaling with respect to surface area (2/3)
and weight (1).
More recently, it was proposed [1316] that the production of new material should
be related to Kleiber’s law [704], which states that the metabolic rate of an organism,
defined as the energy expended by an organism per unit time, scales as the three-
fourth power of the total mass as shown in Figures 1.11 and 1.12. This assumption
also leads to von Bertalanffy’s choice of p = 1/4.
It has been argued that this power law for the metabolic rate is widely applicable
to biological systems from bacteria to whales and even plants and forests [957],
spanning 27 orders of magnitude in mass. However, some authors have criticized
the validity of this statistical analysis and have argued for exponents closer to two-
third [572, 1126]. The problem is that the arguments used to derive such laws are
inevitably vague and based on a number of simplifying assumptions that neglect many
well-established physiological processes. Unfortunately, the absence of a theory for
growth linking mass to size leaves room for endless discussions, controversies, and
grandiose teleological theories with very little scientific content and no mathematical
foundation. As a result, the general field of scaling laws for growth remains in its
infancy [287, 1315].
For any choice of exponent p ∈ (0, 1], the growth curves show a sigmoidal
behavior observed in many organisms as shown in Figure 1.9. A slow initial phase is
followed by quick maturation, that slows down after the reproductive age, eventually
reaching an asymptotic limit.
Humans have evolved and adapted in such a way that they cannot be simply
described by many of the power laws applicable to other species. However, these
animal models can still be used to described human growth if pre- and post-pubertal
periods are considered independently [106].
Dimensionless mass ratio
Dimensionless time
Figure 1.12 Scaling and data analysis based on Kleiber’s law [1316]. Here M/M∞ is plotted
against the dimensionless time τ = bpt − ln(1 − (M0 /M∞ ) p ) as described by Equation (1.4) with
p = 1/4.
The key feature of this simple power-law model is that it identifies the specific
growth rate, defined as the growth rate per unit mass, Ṁ/M, as a central quantity.
The basic idea in all models showing a sigmoidal behavior is that a proportion of new
tissue generated by growth is capable of growing itself, but as time goes by, this ability
is reduced. These two typical trends were first presented as fundamental principles
for growth by Peter Medawar in 1941 [856] when he wrote: “What results from
biological growth is itself, capable, of growing” and “Under the actual conditions of
development, the specific acceleration of growth is always negative.”
Another model that has been used in the context of growth is the so-called Gom-
pertz’ law, first introduced in 1825 [455] for the evolution of human populations.
It was subsequently used in actuarial sciences as a modified compound law, then
rediscovered as a suitable growth law for organisms [244, 1329]. The Gompertz law
is obtained as the solution of a Malthusian equation of growth, Ṁ = K M, with
a growth rate K decaying exponentially in time, following Medawar’s principles.
Choosing K = a exp(−kt) with a, k > 0, we have

M∞
Ṁ = ae−kt M, a = k ln , (1.5)
M0
Figure 1.13 Gompertz’ law. Weight (gm)

Growth curves for male 800
(triangles) and female
(circles) rats (Rattus 600
norvegicus) fitted by a
Gompertz law (adapted from
[1195]).
400
200
Age (days)
0
0 50 100 150 200
which leads to −e−kt

M∞
M(t) = M∞ . (1.6)
M0
This equation, whose solution contains the exponential of an exponential and two
characteristic time scales (1/a and 1/k), has been shown to be a particularly good
fit for bacterial and tumor growth, among others [37, 965, 1090, 1341, 1369], and
remains widely used for the purpose of fitting experimental growth data as illustrated
in Figure 1.13.
1.3 Relative Growth
Extensive psychological studies have demonstrated that children and puppies are
cute and lovable [20, 727]. It has been argued that this cuteness factor, a propen-
sity or desire to cuddle or defend a person, is due in part to our perception of the
relative size of body and facial features in children versus adults [110, 803]. As
shown in Figure 1.14, the skull of babies is proportionally larger than those of adults
[159]. Further, the relative dimensions of the skull, such as height versus diameter
or roundness, are also different. Our ability to distinguish almost instantaneously
children from adults independently of their size implies that no adult organism is
simply a pure dilation of itself at birth.
The pure dilation of an organ or organism from birth to adult life, is referred
to as isometric growth, as shown in Figure 1.15, whereas the relative growth of an
organ with respect to the total weight of the organism is known as allometric growth.
Therefore, the second fundamental problem of growth is to understand how organs,
limbs, or tissues grow with respect to the total body mass. This is the general topic
of the theory of relative growth [648, 649, 857].
The subject of relative growth was first discussed by Galileo in 1638 when he
considered the relative size of bones in animals [414]. He writes: “Nature cannot
Figure 1.14 Relative proportion in the growth of man (adapted from Stratz [1200]).
Figure 1.15 Isometric

growth. Hecht found in 1916
that the dogfish Anchovia
brownii grows isometrically.
Its adult form (B) is a
uniform dilation of its infant
form (A) [557].
produce a horse as large as twenty ordinary horses or a giant ten times taller than an
ordinary man unless by miracle or by greatly altering the proportions of his limbs and
especially of his bones.” That is, the same bones in larger animals are comparatively
thicker than those in smaller animals, as illustrated in Figure 1.16B. The discussion
on giants by Galileo is based on the common belief at the time that large fossil bones
of dinosaurs or large mammals belonged to giants of the past. The biggest giant in
Figure 1.16A, next to an “ordinary man”, is based on bones discovered in Sicily.
Through simple scaling arguments, Galileo correctly concluded that if giants did
exist they would not be a simple dilation of an ordinary man.
1.3 Relative Growth 19
A.
Ordinary
man
B. C.
Figure 1.16 A. Improbable giants from “Mundus subterraneus” by Athanasius Kircher (1678)
[697]. BC. In his “Discoursi”, published in 1638, Galileo noted that bones are subject to allometric
growth and, therefore, giants could not have the same bones or proportions as ordinary humans.
Galileo used the physical analogy that to support increasing loads, a column must grow thicker
[414].
A systematic study of relative growth began at the end of the nineteenth century
with the work of Dubois and Lapicque, who compared brain sizes within and between
species [308, 309, 770]. Their work follows an early observation from Cuvier that
bigger mammals have relatively smaller brains [127, 141, 489]. Based on extensive
research, D’Arcy Thompson reflected on the inevitability of differential growth in this
passage [1230, p. 205]: “An organism is so complex a thing, and growth so complex
a phenomenon, that for growth to be so uniform and constant in all the parts as
to keep the whole shape unchanged would indeed be an unlikely and an unusual
circumstance. Rates vary, proportions change, and the whole configuration alters
accordingly.” The theory was further expanded on and applied to many different
biological systems by Julian Huxley. In his book, “Problem of Relative Growth”, he
coined the word allometry to describe relative growth that follows a power law with
respect to mass [647].
The central idea in Huxley’s work is that the mass m or length l of an organ scales
as a power of the total mass M of the organism, that is
m = k M α. (1.7)
By expressing m as a function of M, the explicit dependence on time is removed

enabling the comparison of the relative sizes of m and M at given points in devel-
opment for different animals. The case α = 1 corresponds to isometric growth, and
any other value of α characterizes relative or allometric growth. If a fixed exponent
α holds during growth (that is, when k and α are time-independent), allometry is
equivalent to the statement that the specific growth rates of an organ and an organism
are proportional:
ṁ Ṁ
m = k M α ⇐⇒ =α . (1.8)
m M
Starting in the 1920s, this simple, but fundamental, power law has played the role
of an ordering principle in comparative biology and evolution [488–490, 1321]. The
possibility for applications appears endless, as the size of any organ in a given species
can be compared to any other species at any point in its development as illustrated
in Figure 1.17. Generations after generations, students and researchers, ecologists
and biologists, have combed the beaches, swept the forests, and fished the seas to
gather data on size and weight in the hope of recovering existing laws or uncovering
new ones. Depending on the data and the scientific question, one compares either the
relative size of different organs in one species (intraspecific allometry), the relative
sizes of one organ within different species (interspecific allometry), or the relative
size of one organ to different organs, at different points in development for a given
species (dynamic allometry).
A typical allometric study proceeds with the following steps [849, 958, 1037]:
• Gather experimental or bibliographic data on a given physical quantity q and the
corresponding weight M of the animal. For instance, q can be the size or weight
1.3 Relative Growth 21
A. B.
C. claw weight (mg)
body weight (mg)
Figure 1.17 Isometric and allometric scaling. A. Six species of salamanders in the genus Desmog-
nathus retain their proportions despite being of different sizes (reproduced from [681]). B. Different
males in the beetles Dynastes showing the relative increase of horns as a function of size (1924
illustrations from Champy [182], reproduced from [427]). C. Allometric scaling for the claw of the
fiddler crab in the genus Uca (adapted from [1192]).
of an organ, the metabolic rate of an organism, the growth rate, or the number of
cells;
• Plot log(q) against log(M), so that a law of the form q = k M β would be repre-
sented as a line of slope β on this plot;
• From the plot, find the best linear fit and extract the slope β;
• Find a rational number, α, preferably with a small denominator, close to β;
• Justify, using arguments from geometry, physics, mechanics, or thermodynamics,
that this exponent α can be obtained from first principles;
• Discuss the relevance of this law in the context of physiology, pathology, ecology,
evolution, or optimal design;
• Discuss why some species or organs may not follow the expected law. This last
step provides a falsidical justification of the law through its exceptions.
This approach is illustrated in Figure 1.18 which demonstrates allometric growth in
plants. These studies, despite their obvious shortcomings from an epistemological
point of view, have been extremely successful, as power laws have been shown to
hold across diverse organisms [957, 1303]. As such, allometry remains a favorite
tool and provides a methodological framework for comparative biology [136, 446,
Log GT
Log MT
Figure 1.18 Allometry in plants. An interesting example of a possible allometric law for growth
arises in plants when the growth rate G T is plotted against the weight MT in a log–log plot (adapted
from [957]).
956, 1320]. These studies clearly demonstrated that aspects of sizes, growth rates, or
metabolic rates tend to cluster in a linear fashion against total weight when expressed
in log–log coordinates. Therefore, the mere existence of observed scaling laws sug-
gests that critical aspects of size and growth rate are governed or constrained by
geometry, mechanics, and physics. This fundamental observation was already put
forward by Medawar in 1941 [856]. However, Medawar believed that growth is
dominated and limited by diffusion, which turns out to be relevant only at the micro-
scopic scale.
At the organ or body level, relative growth can be tracked by simple external
measurements. However, different regions of an organ may experience different
growth rates. The skull of mammals elongates after birth, brains in humans fold
onto themselves during development [466], and roots tend to lengthen by limiting
their expansion in a small region at the tip. Typically, growth is neither isometric
nor homogeneous. Therefore, different points on the growing body may expand at
different rates creating, in the words of Huxley, growth gradients and growth sources
[647], or in the terminology preferred by nineteenth century plant physiologists,
differential growth [1107]. Differential growth is central to all aspects of growth
mechanics; it is responsible not only for shaping an organism, but also for creating
stresses through geometric incompatibility. However, to capture its key features, new
experimental and mathematical methods are needed.
1.4 The Kinematics of Growth 23
1.4 The Kinematics of Growth
The third fundamental problem in the modeling of growth is to track growth processes
both spatially and temporally. It implies that the expansion and relative deformation
of different points in an organ must be followed during growth.
This idea can be illustrated in one dimension. Initially at time t = 0, we paint
a number of marker points, say A0 and B0 , on a filament as shown in Figure 1.19.
During growth, we record the position of these marks as a and b, respectively. The
problem is then to extract information from the dynamics of these material points,
such as the growth rate or the region where expansion takes place. For instance, the
stretch of the initial segment from A0 and B0 is the ratio of the lengths x/X 0 . In
general, we want to define the stretch at a single point by taking the limit X 0 → 0.
To do so, we denote by S0 the initial position of a material point from one end, labeled
S0 = 0 at time t = 0. The position of this point at time t is denoted s = s(S0 , t). Then,
assuming that there is no other stretching process, the local stretch at the material
point S0 and time t due to growth is given by
∂s
γ(S0 , t) = . (1.9)
∂ S0
In particular, if growth is taking place independently of the position, we would have

simply γ = l/L 0 , the ratio of the two lengths.
Borrowing basic concepts from continuum mechanics, we define the Lagrangian
growth velocity V (S0 , t) as the velocity of a material point and the Eulerian growth
velocity, v(s, t) as the velocity at a spatial point s as
∂s
V (S0 , t) = (S0 , t), v(s, t) = V (S0 (s, t), t). (1.10)
∂t
Then, the Lagrangian growth rate γ̇ defines a local growth rate in terms of the initial
material coordinate S0 , which is given by
∂V ∂2s
γ̇ ≡ (S0 , t) = (S0 , t). (1.11)
∂ S0 ∂t∂ S0
Figure 1.19 Kinematics of growth in one dimension. Two arbitrary points A0 and B0 are marked
initially and tracked through growth. Of particular interest is the evolution of the relative distance
between these points (the increments X 0 ).
A. B.
Figure 1.20 Tracking growth. Initial marker points are placed on the stem (A) and root (B) and
are followed during growth (from Müller’s textbook [924]).
In terms of Eulerian properties, one can define the Eulerian growth rate lg as the
spatial gradient of the Eulerian velocity, that is
∂
lg (s, t) = v(s, t). (1.12)
∂s
This quantity expresses the rate of change of length of an infinitesimal element
located at position s. By the chain rule, the Lagrangian and Eulerian growth rates are
related by
γ̇ = lg γ. (1.13)
For problems where growth is localized in space, such as tip growth, an Eulerian
description may be advantageous since the function lg (s, t) describes the growth
experienced by a material point located in the growth region. If growth depends on the
cells, and not on their spatial position, a Lagrangian description may be more appro-
priate. This description is particularly convenient in certain developmental stages
when cell division dominates the growth process.
1.4 The Kinematics of Growth 25
This approach was already used by plant physiologists of the nineteenth century
to quantify the growth of stems [1159, 1182] and roots [187] and identify regions of
active growth, as shown in Figure 1.20. The function γ is an important quantifier; it
identifies spatial locations where growth takes place and is referred to as the relative
elemental growth rate [193, 236, 444, 1158, 1181], a term coined by Erickson and
Sax [344, 345]. To obtain this elemental growth rate in an experiment, the position
of initial markers is followed over time and interpolated to obtain an estimate for
both a discrete version of the Eulerian velocity and for its gradient [1157, 1302].
The analysis of growth kinematics in higher dimensions [1127, 1334] follows
the same principle. In the case where growth is considered as the only possible
deformation, we consider a body in an initial configuration B0 ⊂ R3 with coordinates
X0 denoting the position of material points in the body. The same body at a later
time t is in the current configuration, denoted Bt . We will assume that the current
configuration is related to the initial configuration by a smooth invertible map χ :
B0 → Bt such that a material point located at X0 in B0 is mapped to a point x(t) =
χ(X0 , t) in Bt . In this case, the Lagrangian and Eulerian growth vectors are defined
as
∂χ(X0 , t)
Vg (X0 , t) = , vg (x, t) = Vg (X0 (x, t), t). (1.14)
∂t
The vector field defined by these growth vectors creates growth lines that indicate
the direction in the body where expansion takes place, as shown in an experiment on
rabbit ears in Figure 1.21.
Assuming that the density remains constant, the evolution of the mass density
ρ(X, t) is governed by
∂t ρ + ρ div vg (x, t) = ργ, (1.15)
where ργ is the local growth rate (see Section 13.2 for a derivation). Therefore,
once the growth vectors are known, the computation of the divergence provides
information on the local growth rate through (1.15), as shown in Figure 1.22C.
t=15 days t=28 days t=49 days t=87 days
Figure 1.21 Tracking growth on rabbit ears. Streamlines and equipotential lines for the growth of
a New Zealand white rabbit ear from day 15 to day 87 [249].
A. B.
C.
(a) (b) (c)
(d) (e) (f)
Figure 1.22 Tracking growth on leaves. A. Initial marker points are placed on the leaves and
tracked through time (from Müller’s textbook [924]). B. Tobacco leaves from Avery [48]. C. Digital
analysis of growth. Starting from an original time lapse sequence (a), the displacement vectors are
calculated (b), a mask of nonmoving points is evaluated (c), and used to correct the displacement
vector (d). A smooth interpolation of the displacements (e) provides a way to compute the divergence
of the displacement giving the local rates of growth (f) (reproduced from Schmundt et al. [1127]).
The local mass density ρ is only related to the volumetric change of a volume
element, without giving any information on the local direction of the deformation;
that is, the anisotropy of growth. This information is contained in the gradient of the
deformation and will be discussed in Chapter 12.
Chapter 2
Mechanics and Growth
2.1 Growth is Influenced by Stress
From a biologist’s perspective, growth is mediated by gene activation and regulation.

In a simplified view, different genes will trigger different growth responses. For
instance, in the case of the leaf of the model plant Arabidopsis thaliana shown in
Figure 2.1, a number of genes and microRNAs have been identified that regulate
and control the shape of the plant’s leaves during development. Exquisite details
are known on the initial stage of leaf formation, including features such as adaxial–
abaxial polarity, symmetry, and flat morphology, as well as the precise control of
cell division, proliferation, and expansion. This genetic understanding of growth is
fundamental as it identifies, at the smallest scale, the essential features that control
cell division and expansion. It also provides information at the local level on the
change in shape and volume of the growing components of a body. However, a leaf
will only acquire its shape by developing physical forces between cells or within
a cell. There are important physical, geometric, and mechanical constraints in the
development of a tissue or an organ that need to be integrated with genetic and
biochemical signals to obtain a full picture of growth.
The idea that mechanics shapes organisms and should play a fundamental role in
the description of growth processes was first put forward by Wilhelm His in 1888,

28 2 Mechanics and Growth
Figure 2.1 A genetic view of growth. The shape and size of a leaf is determined during development
by a number of regulatory genes that have been identified by systematic genetic studies [693].
as shown in Figure 2.2 [578, 612]. In his treatise “On the Principles of Animal Mor-
phology”, he writes:“Embryology and morphology cannot proceed independently of
all reference to the general laws of matter, to the laws of physics and of mechanics.
This proposition would, perhaps, seem indisputable to every natural philosopher;
but, in morphological schools, there are very few who are disposed to adopt it with
all its consequences.”
His’ comment was echoed by Le Gros Clark in 1945, in his quest to understand
brain morphology [215]: “Yet, whatever may be the primary cause which initiates a
succession of growth processes leading ultimately to the differentiation of a specific
form or pattern, the morphologist cannot but accept the thesis that the proximate
factors, which immediately operate at the time when the form or pattern becomes
manifest in embryological development, are capable of expression mathematically
in terms of purely mechanical forces.”
The fact that mechanical loading has an effect on growth is not surprising, given
that there is much anecdotal evidence from our everyday lives. We know, for instance,
that muscles will grow when repeatedly strained during exercise; the space program
taught us that astronauts lose bone mass in space flights due to the effect of reduced
gravity [1262]; and our obsession with competitive sports has led to the observation
that tennis players have denser bones in their playing arms [1222]. We also know that
trees will grow shorter and stubbier in windier areas [237], that the earlobe will stretch
and grow under the action of heavy earrings (see Figure 2.3), and that the dreadful
custom of foot-binding results in smaller foot size and permanent deformations [87].
2.1 Growth is Influenced by Stress 29
A. B.
Figure 2.2 Wilhelm His’ mechanical analogy between A. the folding of a rubber tube and B. the
folding of a gut tube during morphogenesis (adapted from [445, 578]).
Figure 2.3 Ear gauging

stretches the ear lobe and
produces permanent
deformations.
A change in mass or in material properties typically requires a cellular response.

How is the information about forces and loadings carried through the cell? Mechan-
otransduction is the process by which mechanical signals, applied in the bulk or at
the boundary of the body, are mediated down to the cellular level and then transferred
to the nucleus. It is a complex process that is not yet fully understood. However, the
effect of the mechanical environment on the regulation of growth is undeniable and
it can be modeled mathematically, even in the absence of well-understood transduc-
tion mechanisms. Indeed, the precise influence of mechanical loading on growth and
physiological regulation is well documented in many systems, some of which we
briefly review next.
2.1.1 The Growth of Stems
Detailed experiments quantifying the relationship between growth and applied stress
were performed by plant physiologists in the nineteenth century. The idea depicted
in Figure 2.4 is simple and elegant: A string is attached to the tip of a growing stem
and tied around a pulley. Weights can be placed at the other end of the string to
apply a constant force to the growing stem. The length of the stem is then contin-
uously recorded by an ingenious device, the auxanometer. It is therefore possible
to record the length of the stem in different mechanical conditions by applying dif-
ferent weights. Once the length as a function time is known, the growth rate can be
extracted. Figure 2.4B presents the results of such an experiment on maize leaves.
The evaluation of the growth rate shows that there is a positive correlation between
growth rate and applied load. This type of experiments establishes the existence of
a relationship between external loading and growth.
2.1.2 The Growth of Axons
Another interesting example of stress-mediated growth in a biological system is found

in the development and branching of neurons. Most neurons have a very distinctive
morphology with a large cell body and long protoplasmic protrusions extending
from the cell body. These protrusions, called neurites, develop into either axons
or dendrites that connect to those of other cells to create a connected network; the
nervous system. The initiation, development, and growth of axons have been shown
to depend on applied mechanical forces [126, 559]. Initiation refers to the early
development of a neurite out of the cell body. It has been shown that the initiation
of axons of chick sensory neurons can be manipulated experimentally by the proper
application of tension on the surface of the cell body [1366]. Following initiation,
the elongation of an axon can be induced by gluing a glass needle to the neuron and
then towing it with constant force. These experiments clearly show that the growth
rate depends crucially on the applied load as shown in Figure 2.5.
A. B.
Figure 2.4 A mechanical view of growth. A. An auxanometer: The length of a growing stem is
continuously recorded while the plant is being pulled by different weights (from Sachs’ “Text-book
of Botany” [1107]). B. Results of similar modern experiments performed on maize leaves. Different
weights are attached, leading to varied growth velocities and growth rates as a function of the applied
force [251].
A. B. D.
Force ( µ dynes)
C.
Length changes (µm)
Figure 2.5 Experimental elongation of a neurite in a stage 2 hippocampal neuron induced by

applied tension. A. Neuron immediately before needle application. The arrowhead marks the posi-
tion where the needle was attached. B. Same neuron 30 min later during early stage of neurite
towing. C. 5 hours and 40 min after panel B, at the end of towing. Bar, 20 μm [757]. D. Axonal
elongation rate for chick sensory neurites as a function of experimentally applied tension (data and
graphs adapted from [560]).
Figure 2.6 Arteries’ thickness, diameter, and length are regulated by mechanical stimuli, among
other processes.
2.1.3 Thoma’s Law for Arteries
Arteries are the blood vessels that carry blood from the heart to organs and body
tissues. Mechanically, they are cylindrical structures with a triple layer of viscoelastic
material that can grow and remodel as shown in Figure 2.6. The arterial system is a
highly complicated and regulated mechanical system sustaining varying pressures,
axial loads, and shear stresses. It has been known since the early work of R. Thoma in
1893 ([1227], see also [91, 1207]) that the magnitude of blood flow in chick embryo
blood vessels regulates the vessels’ diameters and, furthermore, that this effect is
mediated through shear stress on the inner arterial wall. Thoma also showed that
the magnitude of pressure regulates the vessels’ thicknesses. These observations on
mechanical regulation of artery thickness and diameter have been reproduced and
refined experimentally over the years [779, 1301]. Arteries also respond to changes
in axial loading. A sustained increase in axial loading tends to lengthen arteries in
culture [264] and in vivo [659], suggesting that axial stress (or strain) is also regulated
toward a homeostatic value (that is, a level of stress considered to be the normal
physiological regime) [448].
2.1.4 Woods’ Law for the Heart
The heart is a complicated organ that pumps blood to the body by active muscular
contractions. The size and thickness of the heart are regulated during homeostasis to
maintain proper function. In 1892, R. H. Woods proposed that wall stress is a key
factor in this regulation process [153, 1339, 1359]. The original idea of Woods is to
look at the heart as an elastic membrane and use Laplace’s law to obtain information
on wall stresses. Laplace’s law (sometimes called Young’s law [1360] or Laplace–
Young’s law [771, 1223]) relates the geometry of an elastic membrane of thickness h
to the forces applied to it. In the simplest case of a sphere of radius R under pressure
P, the law states that the force per unit area, σ , experienced by the material is simply
given by
h
P=σ . (2.1)
R
The larger the pressure, the larger the force in the membrane. In general, the geometry
of a membrane is locally characterized by its principal curvatures κ1 and κ2 , as
defined in Chapter 10. Then, Laplace’s law states that for an elastic membrane, the
difference of pressure P across the membrane is related to the principal curvatures
and the membrane thickness h, and the corresponding principal stresses σ1 , σ2 , by
P = h (σ1 κ1 + σ2 κ2 ) . (2.2)
If we assume that the stresses are equal, σ1 = σ2 = σ , we obtain P = hσ (κ1 + κ2 ).

Therefore, we can obtain the stresses in the material from the pressure and the
geometry of a membrane. This is precisely what Woods achieved. In his method,
shown in Figure 2.7, Woods used a compass to measure the heart’s curvatures. By
evaluating the wall stresses at different points on the heart, he showed through direct
geometric measurements that the Woods’ number, defined as the ratio
Figure 2.7 Principal radii of curvature on the heart surface. Woods used necropsy hearts. He picked
several points on the surface and used a pin (N) to represent the normal direction to the surface.
Assuming that the heart is close to a surface of revolution, the two directions of principal curvature
are close to the horizontal and vertical directions (the arcs APB and CPD). Then. the problem is to
find the center of curvature (O and O’) such that a compass would trace the arcs. The opening of
the compass is then related to the radius of curvature (from Burton [153]).
Normal heart Ventricular dilation Wall thickening due

due to volume overload to pressure overload
Figure 2.8 Growth and remodeling in the heart. Pathology of the heart leading to an increase in
size. Left: normal heart, middle: ventricular dilation due to volume overload, right: increase in wall
thickness due to pressure overload (adapted from [738]).
P
CWoods = = h(κ1 + κ2 ), (2.3)
σ
is nearly constant across the heart.
More strikingly, Woods’ number varies only by about 22% across different mam-
mal and bird species [667, 834, 1206]. This observation suggests that wall stress
could be the prime regulator of heart size and thickness. In simple terms, Woods’ law
states that the heart grows and remodels to keep the wall stress constant. This law is
consistent with various pathologies and regulatory mechanisms [1206]. For instance,
during long periods of high blood pressure P due to pathological conditions, the heart
ventricles thicken by the addition of myofibrils which lower the wall stress over long
periods, as shown in Figure 2.8 [450, 1063]. Similarly, athletic exercise leads to an
increase in volume causing a decrease in the curvature κ. Subsequently, this change
in curvature is compensated for by an increase in muscle fiber length which further
causes an increase in the heart’s thickness h [449].
2.1.5 Wolff’s Law for Bones
The connection between mechanics and bone structure was noticed as early as 1832
when Bourgery [117, 124] first commented on the striking analogy between trabec-
ular bone orientation and mechanical loading as shown Figure 2.9.
This observation was further developed by Culmann [254, 949] through the anal-
ogy between the tension and compression lines in the human proximal femur and
those of the Fairbairn crane, which exhibits a similar shape. The analogy is based
on the observation that the internal organization of the trabeculae is similar to the
lines of principal stress obtained in an isotropic homogeneous material with the same
geometry under mechanical loading. Wolff further suggested that this architecture is
the result of the bone structure due to active processes, and that bones evolve accord-
A. B.
C. D. E.
Figure 2.9 Wolff’s law for bones. A. Drawing of a Femur from Bourgery 1832 [117]. B. Drawing
from Ward’s “Outlines of Human Osteology” [1298, p. 370]. C. A spiral model of bone due to
Dixon [683]. D. Carl Culmann’s illustration of a Fairbairn crane [254, p. 622]. E. Wolff’s structural
study of a bone [1335].
ing to some “mathematical laws” that were not given explicitly [1335]. Nowadays,
Wolff’s law for bone remodeling is not understood as a scientific law sensu stricto, but
is used instead to describe processes by which bones adapt by actively remodeling in
order to best resist the applied load [1096]. This functional adaptation principle was
first proposed by Wilhelm Roux in 1881 [1100]. Roux proposed that bone adaptation
is the result of a “quantitative self-regulating mechanism” at the cellular level [633].
The mineralized cartilage that forms the bone is the result of a complicated process
involving the interaction between many cell and tissue types. Early bone structure is
built by the action of osteoblasts, bone forming cells, and osteoclasts, bone resorbing
Figure 2.10 Bone growth. Normal (left) and osteoporic (right) bone architecture (3rd lunar verte-
brae). Soft tissues and cells have been removed. The damage in the osteoporic bone is caused by
osteoclasts (picture courtesy of Tim Arnett).
cells. Osteoblasts continue to produce extracellular matrix that develop into mature
bone cells called osteocytes. After ossification, osteoclasts and osteoblasts continue
to build the typical microstructures of cortical or cancellous bone tissue. After mat-
uration, the bone structure depends on the proper balance between new osteoclastic
resorption and subsequent osteoblastic formation. For instance, the case of osteo-
porosis shown in Figure 2.10 is the result of a failure of this adaptive response to
maintain the structure needed to withstand daily loading [39].
While the role of mechanics in bone remodeling is well established, the transla-
tion of different external forces into a cellular signal that controls the recruitment and
activity of bone resorbing or producing cells is more delicate [441, 529]. Mechan-
otransduction provides the fundamental bridge between the mechanical and biologi-
cal world. In the case of skeletal cells, the forces arising from mechanical loading can
generate a number of mechanical stimuli, including strain in the bone tissue, stress
in the central cavity and the outer cortex of bone shafts, microdamage or fracture,
transient pressure waves, shear forces through the microscopic canals between the
lacunae of ossified bone, and electric fields due to the motion of bone crystals [1206].
It is also known that remodeling is sensitive to changes in loading magnitude, the
number of loading cycles, the distribution of loading, and the rate of strain among
others [1102].
2.1.6 Davis’ Law for Soft Tissues
In addition to Wolff’s, Thoma’s, and Woods’ laws, the end of the nineteenth century
provided us with yet another conceptual law for growth and remodeling. Around
that time, the American surgeon Henry Gassett Davis developed innovative traction
methods in orthopedic medicine. In his book, “Conservative surgery” [263], Davis
studied the response of ligaments and tendons to the mechanical stresses imposed by
B.
A.
Figure 2.11 Davis and Nutt developed innovative techniques based on the idea that stresses influ-
ence growth. A. A traction shoe invented by Nutt to help elongate the gastrocnemius muscle [968].
B. Traction methods by Davis [263].
various mechanisms for straightening broken bones, as shown in Figure 2.11. Davis
writes “Ligaments, or any soft tissue, when put under even a moderate degree of
tension, if that tension is unremitting, will elongate by the addition of new material;
on the contrary, when ligaments, or rather soft tissues, remain uninterruptedly in a
loose or lax state, they will gradually shorten, as the effete material is removed, until
they come to maintain the same relation to the bony structures with which they are
united that they did before their shortening.”
These ideas were later summed up by John Joseph Nutt in his 1913 book “Diseases
and Deformities of the Foot.” When discussing the treatment of congenital club foot,
Nutt refers to the influence of stress on soft tissues as Davis’ law and the effect of
stress on bones as Wolff’s law [968, p. 157]. He writes “These physiological laws
are simply expressions of the influence of work, rest, and fatigue on metabolism and
the resultant hypertrophy and atrophy.”
Modern studies of tendons in simulated microgravity have confirmed that, in the
absence of stress, tendons tend to loose their stiffness similarly to known skeletal
muscle atrophy [1069].
Diameter (µm)
B.
A.
Time (days)
Figure 2.12 Multicellular tumor spheroids. A. Three-dimensional structure of a spheroid (bar is

250 mm) [920]. B. Growth kinetics of spheroids grown in agarose gels of different concentrations,
showing that, as the gel becomes stiffer, the asymptotic size of the spheroid decreases. (adapted
from [566])
2.1.7 Tumor Spheroid Growth
Cancer cells can be cultured in vitro to grow on a gel or in a pool of nutrients to form
a sphere-like shape, called a spheroid [391, 1203], as shown in Figure 2.12. These
spheroids are of great interest as they can be used in a variety of settings to understand
the response of cancer cells under different physical, chemical, and genetic treatments
[1116]. Mathematically, the delightful spherical symmetry of these aggregates sim-
plifies the problem tremendously as the shape is fully characterized by its radius. It
allows for a direct comparison to experiments and the analysis of the problem can
be reduced by considering the evolution of the spheroid radius [5, 158, 384, 393,
1098].
The relationship between mechanical stimuli and tumor growth was demonstrated
by Helmlinger et al. who showed that the final size of the spheroid depends on the
concentration of the agarose gels in which it is cultured [566]. This study suggests that
solid stress, among other factors such as nutrients and acidity, has a direct effect on
growth, as high stress may limit the ability of cells to divide due to contact inhibition
[530, 566].
2.2 Stress is Influenced by Growth
Wapas, Walapas, or Walabas are the generic names for trees belonging to the genus
Eperua. They are found through the French and British Guyanas where they are the
most abundant tree species. These trees, shown in Figure 2.13, grow rapidly and can
be cut every 30 years, representing an important economic resource [238]. However,
2.2 Stress is Influenced by Growth 39
A. B.
C.
Figure 2.13 A. Eperua venosa. Wapas trees burst open when cut due to the residual stress caused
by differential growth (picture by Jean Thiel, courtesy of Pierre Détienne). B. When they are cut,
they crack open [282]. C. Residual stress in plants can also be relieved by slicing them [412].
despite their abundance and rapid growth, Wapas are not widely exploited because of
the dangers associated with harvesting them. Locally, wapas are known as “les arbres
tueurs” (the killer trees) as they often burst open when they are cut, presenting a real
threat for workers [282]. This tendency to burst is a perfect and extreme example of the
release of residual stress in plants called tissue tension by plant physiologists [1031].
It is believed to be of prime importance in shaping plants [59]. In many biological
tissues, due to a combination of cellular, chemical, and mechanical factors, different
parts of the tissue experience distinct growth rates. The net result of this differential
growth is that the tissue may be under stress even when unloaded. Residual stress is
the stress field that may exist in a body when unloaded. The generation of residual
stress through differential growth is a critical feature of any mechanical theory of
growing bodies [1175]. Essentially, as growth takes place, parts of the body need to
be stretched or compressed to ensure integrity (no cavitation) and compatibility (no
overlap). In turn, the strains associated with these residual stresses are referred to as
residual strains [584, 1171].
A. B.
Epidermis
Pith
Figure 2.14 Residual stress in rhubarb. A. A rhubarb will naturally curve backward when cut. B.
Cross section of a rhubarb showing clearly the two different tissues, the pith inside and the epidermis
outside.
2.2.1 Tissue Tension in Plants
A smaller and simpler demonstration of the existence of residual stress in plants can
be achieved in the kitchen by slicing a stalk of rhubarb (Rheum rhabarbarum), as
shown in Figure 2.14A, and observing that it curls back to relieve some of its internal
stresses [45, p. 47]. As shown in Figure 2.14B, a stalk of rhubarb is composed of
an outer layer consisting of the epidermal tissue and the collenchyma layers, and
an inner layer, the pith, consisting chiefly of parenchyma cells. This observation
suggests that, in its natural state, the pith is in compression and the epidermis is in
tension.
We can further quantify these two different states by peeling the epidermis with a
kitchen peeler [1274]. If we carefully peel a strip of the stalk’s outer layer and place
next to its original position, we notice that the strip has shrunk in length by about 2%.
If we remove completely the epidermis, as shown in Figure 2.15, we observe that the
Figure 2.15 Residual stress in rhubarb. The middle segment of a long stalk of rhubarb was cut.
This segment, of initial length 20 cm, was then peeled. The peeled strips are now shorter, by about
2–4%, and the pith is longer, by about 6%. The mutual tensions between inner and outer layers have
been relieved.
Figure 2.16 Residual stress

is measured by placing a
peeled pith in a cylindrical
tube and adding weight on it.
The tube prevents the pith
from buckling. Extra weight
is added until the pith
recovers its initial length.
This weight is a direct
measure of the overall
residual stress developed in
the stem before the
eidpermis is removed [924].
pith has increased in length by about 6%. This elementary experiment confirms that,
before slicing, the outer wall is in a state of axial tension while the pith is in a state
of axial compression. The actual force developed in the inner layer can be found by
the ingenious experiment shown in Figure 2.16. The pith is placed in a cylinder to
avoid buckling and a weight is added to the top so that it recovers its initial length.
This weight is precisely the force exerted by the epidermis on the pith in its original
state.
The possible mechanical role of this combination of tissues and their associated
stresses can be appreciated by realizing that the peeled rhubarb has lost most of its
rigidity; so much so that it now buckles under its own weight. Coming back to our
first observation, if the rhubarb is cut along its axis, it will tend to bend outwards,
as part of the elastic stress is relieved when the pith elongates and the outer tissues
shorten by curving as shown in Figure 2.14A.
A. B.
a c b
Figure 2.17 Curvature pea test to determine the role of plant growth hormones. A. Varying the
concentrations of auxin induces different growth rates in the pith (solid line) and epidermis (dashed)
[1031]. B. If the pith grows faster than the epidermis, the pea (Pisum sativum) hypocotyl will curve
outwards (ac) or inwards (b) when cut (adapted from Thiman and Schneider [1226]).
The mutual tension between outer and inner tissues in rhubarb, and its possible
role in plant mechanics, was mentioned as early as 1848 by Brücke, described by
Sachs in 1857, and explored in detail by Hoffmeister in 1867, who noticed that the
outer and inner tissues of common grape vine (Vitis vinifera) contract and extend
elastically upon separation, respectively. The possible role of tissue tension in plant
mechanics was well described by Sachs, who writes: “We have here the case of an
elastic stiff body consisting of two parts, each in a high degree flexible and by no
means stiff; only in their natural connection do the epidermal tissue and internal
tissues together form an elastic rigid body” [1107, p. 216].
Following these early works, tissue tension became a central topic of interest
in plant physiology: It was observed in many plants, including wheat roots, fennel
leaves, rhubarb stalks, sweetgum trees, and hypocotyls of cucumber, sunflower, can-
taloupe, and squash [134, 135, 152, 562–564, 739]. Mechanically, these stresses
are attributed to the differential extension of the cell walls in outer and inner layers,
creating an irreversible change in the resting lengths of both tissues [562, 739, 741].
Tissue tension also played a central role in the discovery of auxin as a growth
hormone, through the so-called curvature pea test shown in Figure 2.17 [113, 837,
1030, 1128, 1226]. As auxin has a distinct effect on different tissues, the respective
growth of the epidermis and pith of pea hypocotyl can be controlled by varying
the auxin concentration, and explicitly tested, by slicing the pea along its axis and
measuring the resulting stem curvature.
Despite the early success of mechanics in plant physiology, the advent of genetics
and biochemistry changed the emphasis of plant research from physical to cellular
mechanisms. However, by the end of the twentieth century, the study of tissue tension
(sometimes called growth stresses in trees [55, 286, 379]) regained interest [365] and
the respective role of tissue tension, biochemical pathways, and genetic signaling in
growth regulation became a controversial topic once again [742, 1031].
2.2.2 Residual Stress in Physiology
It is through the work of both Fung and coworkers, and the joint work of Vaishnav
and Vossoughi in the 1980’s [406, 407, 1269, 1270], that the importance of residual
stress in physiology came to be fully appreciated. The classic experiment of Fung
[407] exploits the fact that slicing a disk of artery will release residual stresses by
opening as shown in Figure 2.18.
If we view the artery as a three-layer cylinder, the opening of the disk reveals that,
in the closed position, the inner layer, shown in Figure 2.6, is in a state of compression
whereas the outer layer is in tension. As shown in Figure 2.18B, residual strain can
be quantified by the opening angle between the two open ends and the furthest point
on the disk [408]. This angle has become a standard measure of residual strains in
the study of arteries. The stresses associated with the opening angle are known to
play a fundamental role in regulating the transmural stress gradient and lowering
circumferential stress at the inner walls [184, 601, 635, 636, 1054]. Experimental
observations [605, 607] also indicate that, not unlike the rhubarb, arteries exhibit
axial residual stress due to the relative axial growth of the different layers, as first
B.
A.
Figure 2.18 Residual stress in arteries. A. Slicing a disk of artery will result in the artery opening
to relieve some of its residual stress (picture courtesy of Gerhard Holzapfel and Ray Ogden). B.
Fung’s classic experiments (a) and the definition of the opening angle (b) [408].
A. B.
C.
D. E.
A (a1)
(a2)
Figure 2.19 Evidence of residual stress in physiological systems. A. The trachea [534]. Notice
how, after being cut, the disk relieves stress by opening. B. Bones [442]. Here, an incision creates
an opening due to tension. C. The esophagus [510]. Different tissue layers have distinct opening
angles due to varying residual stresses. D. The heart [986]. A slice of heart also opens when cut. E.
Stress relief in the brain [1345].
noted by Bergel in the 1960s [89, 168, 426]. However, the role of axial residual stress
in regular homeostasis still remains to be elucidated.
Once the importance of residual stress in arteries was established, it was not long
before residual stress was observed in a number of other physiological systems, as
shown in Figure 2.19. These systems include physiological tissues and organs such
as the esophagus [510, 511, 513, 1365], the aorta [49], the heart [986], the trachea
[534], the brain [140, 568, 1345], bones [443, 1214, 1347], cartilages [1138], and
the developing embryo [77, 79]. It was also found in non-physiological systems
such as solid tumors [26], sea urchins [902] and even in the drying of fruits [724]
and in the aging of meat [784]. Residual stresses are a hallmark of living tissues
and organs, but also play an important role in many classic elastic systems, as first
noted by Ciley in 1901 [212], and in man-made structures [1256–1258]. The reviews
[940, 941, 1330–1332] provide interesting discussions of the techniques used in
engineering and bioengineering to measure residual stresses and their various effects
on the properties and behaviors of different structures.
2.3 The Theory of Morphoelasticity
From a geometric point of view, residual stresses arise due to the incompatibility
between a specification of the local change in a volume element and the continuity and
integrity of the body. This incompatibility is easy to visualize in a one-dimensional
thought experiment analog to the rhubarb experiment. Consider a sandwich of three
identical elastic rods perfectly glued to each other. They are constrained so that they
only deform along their lengths as shown in Figure 2.20. Now, we let the middle
rod grow uniformly in length. To preserve the integrity of the composite body and
prevent it from breaking apart, the middle rod must be in a state of compression and,
accordingly, the outer two rods must be in a state of tension. As a result, this structure
has developed residual stress. This stress could be relieved by simply removing the
glue between the layers, hence allowing the three rods to return to their unstressed
length after growth. It could also be partially relieved if the upper layer is removed
and the two remaining layers are allowed to curve. Interestingly, the shape of the
grown composite body depends not only on the geometry, but also on the elasticity
of the three rods. If the outer rods are very stiff, negligible deformation will take
place, but large stresses will be generated.
In tension
Growing region In compression
In tension
t=0 t>0
Figure 2.20 A simple system consisting of three elastic rods glued together. The middle rod grows
in compression and places the outer rods in a state of tension, creating residual stress through the
structure.
This elementary example outlines the interplay between the elasticity of the mate-
rial, the local change in volume due to growth, and the overall geometry of the body.
Furthermore, stresses also influence growth through evolution laws, creating further
dynamical feedback in the development of stresses and geometry. Understanding this
coupling, for given materials and growth evolution laws in a given geometry, is the
central problem of the theory of morphoelasticity [464, 468]. The goal of such an
approach is not only to obtain the new grown shape and the residual stress developed
in the structure, but also to understand how such a new body evolves dynamically,
how it responds to loading, how it reorganizes itself by remodeling, and how it fulfills
key structural and biological functions. For example, the grown rhubarb is stiffer as
a structure than its different tissues separately. What is the respective role of growth
and elasticity in producing this final structure? Similarly, it is believed that arteries
develop residual stress to counteract hoop stress gradients that may cause tissue sep-
aration. How is this structure obtained dynamically? How does its material response
evolve during loading?
Yet another way that growth, geometry, and elasticity combine in biological mate-
rials is through the formation of new morphological patterns born as mechanical
instabilities [80]. It is well known that, under external loads, elastic materials can
develop instabilities such as buckling [486], wrinkling, or creasing [103] and a nat-
ural question is whether growth itself can generate sufficient stress so as to destabilize
the body. It is not clear that such an instability can exist since the geometric effect
of growth is to change the different length scales associated with the body, such
as thickness, height, and width. It is also possible that growth may destabilize the
body and trigger the development of new shapes. Examples of these growth-induced
patterns can be found in bulk growth [80, 463] as well as in the growth of plates,
shells, membranes [275, 277, 851, 852], and rods [917]. Growth may instead have a
stabilizing effect as, typically, stubby bodies are more stable than slender ones under
compression, and in the process support the function of the biological structure [485,
1274].
The general questions of interest in developing a mechanical theory of growth are
in understanding the origins of the coupling and the interplay between growth and
mechanics, as well as in determining possible consequences for biological functions
and shapes:
• What is the role of mechanical cues in growing tissues?
• How does growth modify the structural properties of a tissue?
• What is the combined role of mechanics, geometry, and growth in morphological
pattern formation?
2.4 A Short History of Growth Modeling
A brief summary of key discoveries connecting growth to mechanics and mathematics

is given in Table 2.1.
2.4 A Short History of Growth Modeling 47
On the modeling side, there have been many attempts to include the effects of
growth and remodeling in biological materials [21]. The plant physiologists from the
nineteenth century were interested in the coupling of mechanics and growth in plants.
Among others, the textbooks of Sachs and Müller include growth data, elastic mea-
surements such as Young’s modulus, and the yield stress, as well as discussions on the
role of physical stimuli in phenomena such as gravitropism, thigmotropism, circum-
nutation, aerenchyma, heliotropism, and tissue tension [45, 261, 412, 924, 1107].
However, no mathematical modeling was proposed, and the physical approach to
plant physiology was replaced in the twentieth century by genetics and biomolecular
considerations, as found in modern textbooks [1215].
Table 2.1 Early history of growth studies.

Year Scientist Study
1638 Galileo First discussion of scaling
1758 du Monceau First description of tip growth
1759 de Montbeillard Record of height from birth to 18 years
1825 Gompertz Population growth model
1832 Bourgery Connection between bone and mechanics
1835 Quetelet Population study and power laws
1838 Moseley Mathematical description of seashells
1839 Sarrus & Rameaux Allometric law for energy with respect to area
1848 Brucke First description of residual stress in plants
1867 Davis Law for soft tissue remodeling
1870 Wolff Mathematical law for bone design
1875 Sachs “Textbook of Botany”
1875 Cohn Tip growth in actinomycetes
1880 Roy Nonlinear response, pre-stretch in arteries
1881 Roux Functional adaptation principle applied to bone
1888 His “On the Principles of Animal Morphology”
1892 Woods Mechanical role of wall stress in heart
1892 Reinhardt Description of tip growth in microorganisms
1893 Thoma Remodeling of arteries
1897 Dubois & Lapicque Brain mass with respect to total mass
1913 Nutt “Diseases and Deformities of the Foot”
1917 D’Arcy Thompson “On Growth and Form"
1920 Pütter Growth viewed as an energy balance
1926 Cannon Concept of homeostasis
1932 Huxley “Problems of Relative Growth"
1941 Medawar Principles of growth
1947 Kleiber Metabolic rate scaling with mass
1957 von Bertalanffy Growth model
One of the earliest attempts to develop a mathematical model taking into account
the combined action of mechanics and growth, is found in the work of Lockhart on
plant cell wall growth [798] and in the theoretical work of Hsu [627]. But the pioneer
work in shaping a theory of growth was developed by Cowin and Hegedus, with a
theory of adaptive elasticity for bones [239, 246, 247, 855], and, in the same period,
by Murray, Maini, and coworkers in their models of wound healing and mechanical
pattern formation [926, 929, 1243].
The general idea that growth can be modeled by considering the effect of both
changes in mass and elastic deformations in a body can be traced to several theo-
retical works [307, 335, 627, 1190]. However, it was through the seminal work of
Skalak, Hoger, and coworkers that a full theory of growth based on the foundations
of continuum mechanics took shape [1171–1173, 1175, 1242]. They showed that
by considering growth as a mapping from an initial configuration to a new reference
configuration, the theoretical concepts of kinematics, such as velocity, velocity gra-
dient, deformation gradient, and so forth, can be used to quantify both the evolution
of a body and its response to mechanical loads. These lines of thought culminated in
a general statement of growth [1091] in terms of the so-called multiplicative decom-
position of the deformation gradient, a formulation that was previously confined to
dislocation theory and elasto-plasticity [176, 179, 730, 777, 810, 933]. The funda-
mental idea of Rodriguez et al. [1091], which can also be found in [225], is that the
deformation of a body can be decomposed into a growth step that describes the local
addition of a material, and an elastic response, characterizing the reorganization of
the body needed to ensure compatibility and integrity. Key to this approach is the fact
that stresses generated through growth depend solely on the elastic step and can be
computed by the theory of nonlinear elasticity. This early work on growth has since
been amply discussed [242, 306, 636] and there have been a number of attempts to
formalize it [284, 341, 811, 1350]. This kinematic theory is now the starting point for
a large body of scientific literature and it will be explored in detail in this monograph
[240, 422, 709].
Despite progress in this field, it has been commented by many authors
[26, 1208] that one of the key remaining challenges is to develop a constitutive
theory for the relationship between growth and mechanical variables such as stresses
and strains. The criticism is that the kinematic approach of Rodriguez et al. is mostly
a geometric modeling of growth that does not address the microscopic process by
which growth occurs. Humphrey, Rajagopal, and coworkers [447, 635, 637, 639–
641] have advocated that growth should be modeled within a mixture theory by
following mass production, resorption, and remodeling of individual constituents. In
this approach, growth is then represented by the evolution of separate natural con-
figurations and the differences between these configurations lead to residual stress.
2.5 A Review of Reviews 49
2.5 A Review of Reviews
There are a number of excellent reviews and introductory courses on the modeling
of biological growth that are worth mentioning as a guide to the literature.
The first review of growth modeling is due to Larry Taber [1206]. This seminal
review covers a great number of biological and physiological systems with detailed
descriptions of the phenomena observed as well as a description of various models.
Despite the fact that the field has expanded considerably since its initial publication,
it remains a must-read for anybody interested in the subject.
Stephen Cowin has authored a number of interesting reviews focused mostly on
aspects of adaptive elasticity, with a special emphasis on the response of cartilages
and growth in bones, and the scaling of growth. It also includes fascinating historical
perspectives not found in other reviews [241–244].
Krishna Garikipati gives a nice review of different aspects of the kinematics
of growth, including both a thorough discussion of accretive growth and a criti-
cal discussion on the assumption of behind the multiplicative decomposition [421].
Garikipati is also the lead author of a multi-author review discussing and contrasting
various aspects of a theory of growth [21].
Gareth Jones and Jon Chapman have coauthored an excellent review on growth
containing comprehensive discussions on cell division, growth models, a comparison
of different discrete computational models, and new ideas on how to combine these
discrete models with their continuous counterparts [672].
Ellen Kuhl and Andreas Menzel, give in a number of reviews, a complete descrip-
tion of the theory of multiplicative decomposition and fiber anisotropy from a theoret-
ical and computational perspective including many beautiful applications to bones,
arteries, the heart, the brain, muscles, tendons, lungs, and skin based on their own
modeling and computational work [349, 733, 873].
Jay Humphrey and colleagues, have discussed, in a series of reviews, the need
for mixture models for growth and remodeling, with special emphasis and details on
arteries and aneurysms [44, 635, 636, 638].
Finally, Marcelo Epstein authored an excellent chapter on the problem of growth
in his book [339]. It includes an exhaustive discussion of the kinematics of multiplica-
tive decomposition, widely used to describe growth and residual stress, and shows
how such a decomposition naturally follows from basic kinematic assumptions and
material symmetry.
Chapter 3
Discrete Computational Models
Since cells are responsible for most of growth in biological systems, it is natural, as
a first step, to develop computational models of growth focused on the interaction
of discrete entities, either representing the cell itself, or parameterizing the cells
through its boundaries. These discrete computational models have been particularly
helpful for biologists interested in developmental processes and pathology [895].
In particular, they have helped validate the differential adhesion hypothesis in cell
sorting that states that two populations of cells will arrange themselves in separate
phases based on their respective binding energy [128–130, 376, 1191]. We provide
here a brief overview of different discrete approaches to model growth.
3.1 On-Lattice Models
3.1.1 Cellular Automata
The first attempts to develop computational models of growth were based on cellular
automata. These models first introduced by von Neumann and Burks [945], and
popularized by Conway’s Game of Life [224], found their way in statistical mechanics
[1336] and biology [347]. The basic idea of cellular automata for biological systems

52 3 Discrete Computational Models
A.
B.
Figure 3.1 Cellular automata model of growth. A. In two dimensions, each square represents either
a normal or a tumor cell. Cells change type following rules dependent on their neighbors. B. In
three dimensions some of the lattice positions can be used to represent a capillary [310].
is that the domain of interest is divided into a number of lattice sites occupied by
cells. At each discrete step in time, the lattice is updated based on rules of interaction
between different cells.
For instance, the growth of cancer cells shown in Figure 3.1 is modeled by con-
sidering two types of cells: normal and tumor cells. Each cells occupies a position
in a two-dimensional square or three-dimensional cubic lattice. Depending on the
availability of free space, the nutrient, and other internal variables, each cell may
change its internal state at each time step. Based on appropriate rules, cell growth
and death are modeled by adding cells into a free space, removing a necrotic cell, or
changing a normal cell to a tumor cell.
There are well-known problems with cellular automata: The symmetry emerging
from these interaction rules typically follow the underlying symmetry of the lattice,
and mechanical and diffusion effects cannot be easily taken into account. Attempts
to overcome these problems include using irregular lattices [832], using density
as a surrogate for pressure [1050], and computing underlying concentration fields
separately [14].
3.1.2 Cellular Potts Models
An alternative method to use lattices to model cells is to assume that a cell occupies
different lattice sites. In these cellular Potts models, a cell location is defined by
several adjacent lattice sites. As the cell moves, grows, or deforms, the set of lattice
coordinates that defines the cell evolves. This evolution is determined by a mini-
mization process of a scalar function analogous to an effective energy in physical
models. This pseudo-energy is a sum over the entire system and contains terms to
3.1 On-Lattice Models 53
A.
B.
Figure 3.2 Cellular Potts models. A. Different types of cell patterning can be obtained by changing
either cell–cell binding energy or cell-matrix binding energy. (a) Initial conditions, (b) cell sorting,
(c) mosaic cell ordering, (d) engulfment. B. Application to vertebrate segmentation and somite
formation. O: Confocal image of one half of the presomitic mesoderm in a live chick embryo (scale
bar is 50 µm). P: corresponding cellular Potts simulation. The simulated morphology reproduces
the basic features of the morphogenetic process (reproduced from [571, 876]).
penalize binding, volume changes, and boundary changes. The role of internal cell
pressure is introduced by assigning a target volume to each cell. The volume term in
the effective energy is minimized when the cell reaches its target volume.
The actual process by which cell dynamics is computed is performed by a stochas-
tic modified Metropolis algorithm. It consists in randomly selecting two sites and
accepting probabilistically a change of cell type based on the difference in energy
that would result in this change. The change is automatically accepted if the energy
after change is lower than a given energy threshold and accepted with a decaying
exponential probability if it is higher.
Growth is implemented by changing the target volume, so that, over time, the
energy is decreased if the volume is increased. Examples of cell dynamics modeled
by such processes are given in Figure 3.2.
The introduction of cellular Potts models has resolved many issues of cellular
automata and helped scientists to either validate basic hypotheses in cellular biology
or explore the interaction of biological systems with multiple cell types [1134, 1205].
Yet, these methods require the introduction of many interaction rules and since the
pseudo-energy of the system is not a true energy, the exact form of different effects
is somewhat arbitrary and cannot be derived from first principles. In particular, the
correct application of loads and the inclusion of forces and stresses in these systems
remain problematic.
3.2 Off-Lattice Models
3.2.1 Center Dynamics Models
An alternative description for the interactions between N discrete cells is to assume

that cells are either spherical in three dimensions, or circular in two dimensions.
Restricting our discussion to the planar case, the positions of the cell centers, {ri ∈
R2 , i = 1, . . . , N }, fully characterize the entire system. The cell centers can either
be seen as the center of overlapping circles or, equivalently, as the centers of Voronoi
polygons in a tessellation of the system as shown in Figure 3.3. A two-dimensional
Voronoi tessellation is a partitioning of a plane into polygonal regions based on
distance to points in a given set. Each edge of a polygon is defined as the set of points
closer to a particular point than to any other [608].
The dynamics of this model is produced by the interaction between cells following
the equations of motion

N
m i r̈i = Fi − νi ṙi − νi j (ṙi − ṙ j ), i = 1, . . . , N , (3.1)
j=1,i= j
where m i is the mass of the i-th cell, νi the drag coefficient between cell and medium,
νi j the friction coefficients between cell i and cell j, and Fi the net force acting on
the i-th cell (excluding drag). Here ṙi and r̈i represent the first and second derivatives
of the position with respect to time, respectively. Typically, biological systems are
overdamped, in which case, the left-hand side terms can be neglected.
The key difficulty in modeling a system of interacting cells is to make the proper
choice for the friction coefficients νi j and the force Fi acting on each cell. Many
Figure 3.3 Configuration of A. B.

cells represented either as A.
overlapping spheres or B. a
Voronoi tessellation (adapted
from [672]).
rj rj
ri ri
3.2 Off-Lattice Models 55
different functional forms of forces have been proposed to capture the basic effects
of repulsion when the cells are overlapping, adhesion at short range, and vanishing
interaction at long range [299, 672, 1006]. Once the force is postulated, the dynamics
is obtained by integrating forward the system of differential equations (3.1). If the
forces derive from a potential, stochastic effects can be easily included using a Monte
Carlo approach [298, 301].
Growth is implemented by assuming that certain cells undergo mitosis [107]. As
the cell volume is increased, it will reach a critical size at which time it will be
replaced by two smaller cells with the same volume. At this point in time, a new cell
is added to the system [300, 1058]. This process can take place while keeping the
spherical symmetry, or in small increments transforming a sphere into a dumb-bell
shape as shown in Figure 3.4.
Center dynamics model have been used to model a number of multicellular sys-
tems such as intestinal epithelia [859], growing population of tumor cells [156], the
regeneration of liver tissues [589], slime mold cells migration and response to loads
A.
B.
Figure 3.4 Growth in center dynamics models. A. The radius is first increased to mimic the
doubling of the cell volume and a daughter cell is added to the system. B. During cell division, a
cell deforms from a perfect sphere into a dumb-bell in small steps. When the dumb-bell reaches a
certain size, it divides into two cells (adapted from [300]).
[258, 1005]. These methods remain an important part of cellular modeling and are
continually refined and generalized.
One of the main drawbacks of these methods is the difficulty in choosing the
correct interaction potential between cells. While many choices seem plausible at a
microscopic scale, the overall macroscopic behavior can lead to unphysical effects.
For instance, it was shown that if the attraction from next-nearest neighbors is too
large, the repulsion of nearest neighbors may not be sufficiently large to overcome
an overall attraction of the cells leading to massive cell overlap [1020]. In this last
publication, the authors further demonstrate that the response under loads may show
undesirable effects, especially under shear where the tissue may rupture. Neverthe-
less, these models are useful for exploring the behavior of cell population at the
discrete level.
3.2.2 Vertex Dynamics Models
In center dynamics model, the cell wall or cell membrane is characterized by either
a fixed distance from the center or by its Voronoi tessellation as shown in Figure 3.3.
Since both the shape of a cell and its interactions with other cells depend on its
interface, an alternative way to model packed cells is to follow their interfaces. In
these vertex dynamics models, cells in two dimensions are defined as polygons and
the positions of all vertices are the main dynamical variables in the model. These
models assume that cells are tightly packed so that each vertex is either the boundary
between two cells or part of the domain boundary. They are particularly suited to
model the various forces acting on cell membranes.
Consider a system of N polygonal cells with a total of M vertices. Each cell has
an area Aα and perimeter L α where α ∈ {1, . . . , N }. As shown in Figure 3.5, an
edge joining vertex i to vertex j has length li j and line tension ti j . We model the
mechanical response of this system by introducing an energy function of the form

N
μα
M−1
M
N
να
W = (Aα − A0α )2 + ti j li j + L 2α , (3.2)
α=1
2 i=1 j=i+1 α=1
2
where A0α is a target area and μα plays the role of an elastic area stiffness as it
describes the work stored in a cell when the area is modified away from its preferred
target area. The last term describes the boundary energy with a coefficient να that
describes the contractility of the cell boundaries. A change in adhesion between cells
can be modeled by varying the tension ti j . From the energy W , the forces acting at
each nodes can be derived as
∂W
fi = − . (3.3)
∂ri
A. B.
C.
Figure 3.5 Vertex dynamics model. The cell is represented by its edges. On each vertex a force
vector is defined and used for the global force balance. A. Original model of Weliky and Oster
[1310]. B. Model developed by Chen and Brodland [192]. C. Epithelia are sheets of columnar cells
connected via cell–cell adhesion [356]. Left. Example of a stationary network patterns generated
by repeated cell division. Bottom: rules for cell division. A cell increases its volume up to a critical
size, at which point, a new cell is created.
At equilibrium, these forces must vanish, which can be accomplished numerically

by integrating the Newtonian equations for each vertices in the presence of strong
damping. The dynamics of a set of cells can be simulated by either deterministic
integration of Newton’s equations or by stochastic Monte Carlo methods.
Growth in vertex dynamics models is implemented by two different processes.
First, the target areas of individual cells can be changed to simulate the enlargement of
cells. Second, when a cell is sufficiently large, cell mitosis is modeled by introducing
new vertices and edges in order to create new cells divide [131, 609]. Special rules
about how cells divide and where division occurs need to be implemented as shown
in Figure 3.5C. After division, each new cell is assigned a new target area.
3.2.3 Advantages and Drawbacks
Discrete, cell-level, models have the advantage to use directly information about cells,
their behavior and their interaction with other cells. Depending on the type of cell
assemblies, different methods should be considered. In a tight packing assembly with
well-defined boundaries, vertex dynamics models are suitable, whereas for dilute
collection of cells, center dynamics models are preferred. In two dimensions, these
models can be easily implemented leading to interesting emerging global behaviors
from simple local interactions. This computational framework constitutes an ideal
platform to test biological hypotheses, to explore the importance of different local
effects, and to suggest new behaviors in different parameter ranges.
However, discrete models, are still in their infancy and suffer from a number of
problems. While two-dimensional simulations can be readily implemented, as soon as
the geometry of cells is not spherical, three-dimensional simulations represent a major
computing challenge [896]. Keeping track of the correct geometry, the neighbors,
and their interaction requires dedicated algorithms that robs the interested scientist
of the original appeal of this approach. However, a number of authors have taken on
the challenge and produced convenient software packages to help with the process
[194, 580, 646, 1153].
Another issue is the observed macroscopic behavior. It is well understood that
starting from a continuum formulation of a problem we can obtain systematically
different discrete versions of the problem. However, starting from discrete systems,
there is no systematic way to obtain a continuum version of the problem. It could be
argued that biological systems do not have a continuum equivalent. Yet, assembly of
cells and tissues can be tested mechanically as a continuum using classical means.
We should expect that the characterization of tissues at the macro-scales could be
informed by the discrete interactions of their components. This connection is still
lacking and remains one of the main stumbling block in linking cellular properties
to tissue properties.
The specification of interaction rules can also become an issue. While it is rela-
tively easy to formulate plausible rules based on educated guesses, it is not clear that
these rules can be easily validated and quantified at the microscopic level. Discrete
models rely not only on basic principles dictated by a pseudo-energy but they also rely
on many evolution, geometric, and interaction rules. Each rule is typically encoded
by several parameters and the total number of parameters quickly increases with the
complexity and dimensionality of the problem. In many instances, it is tempting for
the researcher to add new rules to obtain expected behaviors. The problem is then
to decide whether discrete models can be reliably used as a predictive tool rather
than a way to mimmic known observations. Strict guidelines on how to set working
hypotheses for these models do not exist as the final product is obtained iteratively
by adding new components, new rules, new species, and new physical effects.
Nevertheless, since the main ingredients in discrete modeling can be directly
related to biological experiments and the output is mostly visual, these computer
models are a perfect discovery vehicle for multidisciplinary collaborations with biol-
ogists. As such they constitute an important toolbox to explore the dynamics of cells
and tissues [895]. For these reasons, discrete models have been and will continue to
be extremely successful, especially in developmental biology.
Part II
Filament Growth: A One-Dimensional
Theory
And then the lover,

Sighing like furnace, with a woeful ballad
Made to his mistress’ eyebrow. Then a soldier,
Full of strange oaths, and bearded like the pard,
Jealous in honour, sudden, and quick in quarrel,
Seeking the bubble reputation
Even in the cannon’s mouth. . .
Chapter 4
Growing on a Line
Filamentary structures can be observed in nature at all scales, from the microscopic
chains of molecules, to the gigantic braided magnetic flux tubes in solar flares [83,
1185]. They are three-dimensional objects that are characterized by one length scale
much larger than the other two scales. Due to their geometric similarities and despite
their widely different length scales and microscopic structures, filaments of all sizes
grow, move, and change shape and topology according to universal laws [682]. For
example, a sufficiently twisted rope will coil on itself [1362]. The same change
of configuration is observed to occur in bacterial fibers, DNA molecules [1228]
and telephone cables [58, 478]. Understanding the dynamics and growth in these
fundamental structures is not only of intrinsic theoretical interest, it also lies at the
heart of a host of important processes in biology [84, 826], physics [1163], and
engineering [561].
Here, we consider the modeling of growth as it applies to various biofilaments
such as bacterial fibers [473], bacterial filaments [1139], fungi [90, 1088], root hairs
[1095], stems [59, 910], roots [921, 980], tendrils [477, 684], neurons [758], umbil-
ical cords [461, 890], tendons [1059], arteries [18], and the spine [335].
Filamentary growth can occur essentially in three different ways. First, filaments
can extend axially as shown in Figure 4.1; that is by increasing the length of the

64 4 Growing on a Line
Figure 4.1 Axial growth of

a root. Growth acts by
extending the length of the
filament. Adapted from
[1035].
filament’s centerline. This change in length can be localized at the tip as discussed
in Section 1.1.1 or distributed along the length, especially when the filamentary
structure does not grow against a dense medium [756, 866, 1139].
A second type of growth occurs if the thickness or the density of the filament
increases. In this case, the filament adds mass without changing its length, leading
for instance to the formation of growth rings in trees [378–380]. This process, known
as secondary growth, is typically found in plants where it takes place after an initial
period of axial extension.
A third type of growth occurs when points in a cross section of a filament grow
axially at different rates. Then, differential growth takes place and the structure
may develop curvature and torsion [67]. A simple physical analog of growth-driven
curvature is found in the thermal buckling of a bimetallic beam first studied by
Timoshenko [1235] as shown in Figure 4.2. In that process, two straight metallic
Figure 4.2 Bilayer metallic strips curve due to the difference in thermal expansion of the two
materials.
4 Growing on a Line 65
Figure 4.3 Gravitropism. A plant turned on its side will change growth direction by developing
curvature (adapted from [1035]).
beams with different thermal expansion coefficients are fastened together. Under a
change of temperature, one beam expands faster than the other and induces curvature
for the entire structure [149, 150]. In biological systems, differential growth acts
mechanically in a similar fashion. For instance, in plants, variation of growth rates
in the section of a stem, branch, or root is the main mechanism responsible for the
gravitropic response; that is, the ability of plants to redirect their overall growth
direction vertically [68, 69, 415, 562, 911]. If a stem of Impatiens glandulifera
(policeman’s helmet) is turned on its side as shown in Figure 4.3, it will react by
growing faster on the lower side than the upper side until it is aligned with the
direction of gravity [740].
In a three-dimensional setting, curvature is closely coupled to torsion. If differen-
tial growth does not occur symmetrically in a cross section, it will induce torsion and
twist as well as curvature, resulting in helical growth. This phenomenon is observed
in twining vines, [1156, 1161] but also in bacteria [866] as shown in Figure 4.4,
fungi [483], and in plants [75, 155, 470].
At the mathematical level, the growth of biological filaments involves interest-
ing aspects of curve dynamics [914]. A natural starting point for the modeling of
Figure 4.4 Helical growth. The aerial growth of Streptomyces violascens shows clear helical
structure, Bar is 1.0 μm [1].
biofilaments is to consider them as slender elastic rods subjected to external con-

straints. The basic idea is to perform an average over a cross section of the forces
acting on the filament and then define a resultant force acting on the centerline of
the rod. This procedure leads to the Kirchhoff equations [33, 221, 285, 478, 1282],
relating averaged forces and moments to the curve’s strains characterized by the cur-
vatures, shears, and axial extension. These equations provide the starting point for
much theoretical analysis and numerical modeling and will be discussed at length in
Chapter 5 and 6.
The theory of growing three-dimensional structures requires great care. Growth,
through incompatibility, naturally changes the underlying geometry and induces
local residual stress. A full mechanical and geometric theory is needed to capture
these fundamental aspects of volumetric growth. This general theory will only be
developed in Part IV.
If a one-dimensional structure is allowed to grow both axially and in girth, no
residual stress can be created and the effect of growth can easily be captured by a
slight generalization of the classic theory of elastic rods, taking into account variations
in arc length and changes in material properties. Therefore a natural starting point is
to develop a general framework for growing material in one dimension.
4.1 Example: A Growing Rod in One Dimension
Rods are elastic structures that can support bending, twisting, stretching, and shear-
ing. However, when restricted to deformations in one dimension, rods support only
stretching. When allowed to grow this stretch can be due to either elastic deformation
or extension due to an increase of mass.
Figure 4.5 shows a rod where a material point is parameterized by S0 , denoting its
arc length in an initially unstressed configuration from one end at time t = 0. This
rod can change its length in time and a point initially at a position S0 is located in
the current configuration at time t at a position s = s(S0 , t). Accordingly, the total
=0
=
=0
)=
Figure 4.5 Extension of a filament constrained in one dimension. Initially of size L 0 , the filament
changes its length due to growth, stretching, or a combination of the two to reach a new size l.
4.1 Example: A Growing Rod in One Dimension 67
length at time t is l = s(L 0 , t), where L 0 is the initial length. Assuming that the
change in length at all points is smooth enough, the local change of length is given
by the stretch
∂s
λ(S0 , t) = . (4.1)
∂ S0
If all segments of the rod experience the same stretch so that the deformation is
uniform, λ is constant, in which case λ = l/L 0 . In general, wherever λ > 1, there
is a local increase in length with respect to the initial configuration, whereas λ < 1
represents a local decrease in length.
In the context of growing elastic structures, a change in length can be due to either
growth or elastic stretching due to tension or compression. These processes are very
different and we first consider them separately.
4.2 Purely Elastic Deformations
In a purely elastic deformation, the elastic energy stored by stretching or compressing

the rod can be recovered when the loads are removed. In such a case, we denote
this elastic stretch by α and since it is the only type of deformations admitted, we
have λ = α. The value of α for a given load and a given material is specified by a
constitutive law that relates stresses to strains. In general, a stress is a force per unit
area. In our one-dimensional setting, the only relevant stress, σ, is the force acting
on a cross section, along the axis, divided by the cross-section area. Here, we further
assume that this area does not change during the deformation. A strain is a measure
of the deformation. A simple choice for the strain is = α − 1 so that in the absence
of deformations an unstretched rod, i.e., α = 1, has no strain, i.e = 0.
For illustrative purposes, we start by assuming a spring-like Hookean behavior,
that is, we assume that the stress is proportional to the displacement:
σ = E = E(α − 1). (4.2)
The parameter E is the Young’s modulus. In the general theory of elasticity, the
Young’s modulus is a measure of the stiffness of a material. It is defined as the ratio
between a uniaxial stress and the corresponding uniaxial strain in small deformations.
Assuming that E is constant along the rod, this relationship completely specifies the
deformation as a function of the applied stress σ, that is, we have
σ
∂s
σ=E −1 =⇒ s= + 1 S0 , (4.3)
∂ S0 E
and the current length is l = (σ/E + 1)L 0 .

α
0.5 1.0 1.5 2.0 2.5 3.0
-5
-10
Figure 4.6 Stress–stretch curves. Comparison between the linear Hookean (dash) and nonlinear
(solid) response of a filament for μ = 1, E = 3 in arbitrary pressure units. Note how the nonlinear
response captures the expected behavior of a solid in high compression for which increasingly large
compressive stresses are needed to further compress the material.
It is important to realize that the Hookean law (4.2) is only valid for small defor-
mations. In larger deformations, it predicts that a finite compressive stress of σ = −E
would shrink any line element to a point (λ = 0), a clearly unphysical behavior.
This law can be corrected for large deformations by using the theory of three-
dimensional elasticity. We will show in Section 11.8.1 that the uniaxial extension of
an incompressible rectangular neo-Hookean bar leads to the following nonlinear law
for the extension of a filament:
σ = μ(α2 − α−1 ), (4.4)
where μ is the shear modulus. Close to α = 1, we recover a linear behavior that

justifies the use of the Hookean model. In that regime, we can also identify the
relationship, E = 3μ, between the shear modulus and Young’s modulus as shown
in Figure 4.6. In general, materials that show strain stiffening, that is, an increase in
stiffness for large deformations, or strain softening, a decrease in stiffness, can be
modeled by various functions of the stretch, in which case the constitutive response
in one dimension will be given by the general relationship
σ = f (α), (4.5)
with the requirement that f (1) = 0. For such systems, the Young’s modulus is defined
as E = f (1) and represents the linearized behavior for small deformations around
the stress-free state.
4.3 Growth Without Elastic Deformations 69
4.3 Growth Without Elastic Deformations
If a rod grows in the absence of applied loads and elastic response, it will increase in
length. This change in length can be described by a local growth stretch analogous
to the elastic stretch, that is, λ = γ. In a growth process that takes place in time, the
function γ evolves with time, so that γ = γ(t). It is often easier to describe such
processes as an evolution law describing the addition of mass occurring in a small
increment in time. Therefore, we postulate the existence of an evolution law of the
following form:
∂γ
= G(γ, s, S0 ). (4.6)
∂t
For example, uniform linear growth is achieved by taking G = 1, in which case
s(S0 , t) = t S0 . If the rod is made of cells that reproduce at a constant rate and each
new cell can reproduce again, the rate of growth is proportional to the local growth
stretch:
∂γ
= kγ, (4.7)
∂t
which results in exponential growth of the rod
s(S0 , t) = S0 ekt , l(t) = L 0 ekt , (4.8)
and we identify k as the relative growth rate introduced in Section 1.3.
4.3.1 Example: Tip Growth
Tip growth is a generic mechanism for many growing biofilaments as described in

Section 1.1.1. A detailed description of the mechanisms underlying tip growth is
given in Section 9.2.1. Here, we consider the kinematic description of an arbitrary
growth process localized at the end of a one-dimensional structure. The universal
feature of tip growth is that addition of mass occurs only at the end, which in the
simplest case can be described by the function

γk 0 < s ≤ a,
G(γ, s) = γk(1 − H(s − a)) = (4.9)
0 a < s < l,
where H(·) is the Heaviside function and a > 0 is the length of the growing zone.
In many respects, this simple model contains some of the difficulties that we will
encounter later. The main problem is that the position s of a material point S0 changes
as a function of time, so that s = s(S0 , t) and the equation ∂t γ = G = gγ cannot
be directly integrated. At s = a, we have a moving interface between a growing
λ(S0)
16
s(S0) 10 2.6
10
2.4 9
14
12 0 2.2 8
10 2.0 7
1.8 6
8
5
6 1.6
4
4 1.4 3
2
2 1.2
1
S0 S0
0
0 1.0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Figure 4.7 A simple kinetic model of tip growth obtained from the integration of Equation (4.13)
with t = 0.1, a = 4, k = 1, and initial length L 0 = 12. Here, 10 successive iterations are shown.
The grid on the filament corresponds to equidistant material points in the initial configuration.
and nongrowing material. An efficient way to solve the problem is to consider the
coupled system
∂γ
= gγ, (4.10)
∂t
∂s
= γ, (4.11)
∂ S0
with initial conditions γ(0) = 1, s(0) = S0 .

Before deriving the exact solution, we can solve this system numerically in time
by forward Euler integration
γ(t + t) = γ(t) (1 + tg(s(t))) , (4.12)

S0
s(t + t) = γ(t + t)dS0 . (4.13)
0
An example of a position profile is shown in Figure 4.7. At each iteration, the growing
rod extends its total length by a(ekt − 1) which, after a time t, amounts to a total
increase in length of a(ekt − 1)t/t. In the limit t→0, this process leads to
a linear law for the overall growth l(t) = L 0 + akt. Interestingly, despite the fact
that the tip is growing exponentially, the overall length increases linearly since the
expanding zone remains constant.
Now, consider an initial point at material position S0 = P. If P > a, the point
is not in the growing region and changing L 0 to P in the previous argument gives
s(P, t) = P + akt. If P ≤ a, this point is in the growing region and will move
exponentially away from the origin until it reaches the position a, at a time tc , so
Figure 4.8 The solution to S0

the growth profile, viewed in
s(S 0 )-l(t)
0 1 2 3 4 5 6 7 8 9 10 11 12
a reference frame placed at
0
the growing tip, shows
convergence to a self-similar
solution (a = 4, k = 1, and
L 0 = 12). Here, 16 −5
successive iterations are

shown from t = 1/4 to
t = 4. −10
t=1/4
−15
−20
t=4
−25
that s(P, tc ) = a = P exp(ktc ), or equivalently, tc = k −1 ln(a/P). For t > tc , this

point will not be in the growing region and its distance from the origin will increase
linearly according to the previous computation. We combine the different position
and time intervals together to obtain the general solution:
s(P, t) = (P + k at) H(P − a) +

kt
Pe [1 −H(t − tc )] + [a + k a (t − tc )]H(t − tc ) [1−H(P − a)] .
(4.14)
With respect to a reference frame placed at the growing tip, we observe a rapid
convergence to a self-similar profile as a function of the initial arc length as seen in
Figure 4.8. This profile is self-similar in the sense that for large time the function
s(S0 ) − l(t) viewed as a function of S0 becomes independent of time.
4.3.2 Application: Spheroid Tumor Growth
4.3.2.1 Background
As an application of the idea of growth in a localized region, we consider a simple

model for tumor growth where a spherically shaped tumor expands depending on the
level of nutrients diffusing through its boundary. This model is an example of a large
A. B. C.
b
Figure 4.9 Multicellular tumor spheroids. A. Three-dimensional structure. B. Cross section. In a

typical spheroid structure, the inner necrotic core is surrounded by a layer of quiescent cells. This
layer is itself surrounded by a layer of active cells at the outermost boundary (reproduced from
[920], bar is 250 μm). C. Simplified geometry of the problem, a core of inactive cells is surrounded
by a single spherical shell of active growing material.
class of experimental and theoretical models used in oncology, known as spheroid

models [385].
As shown in Figure 4.9, cancer cells can be cultured in vitro to grow on a gel or
in a pool of nutrients to form a sphere-like shape, the spheroid [391, 1203]. These
spheroids are of great interest as they can be used in a variety of settings to understand
the response of cancer cells under different physical, chemical, and genetic treatments
[566, 1116]. Mathematically, the spherical symmetry of these aggregates allows for
a simplification of the problem by considering the evolution of the radius; in essence,
reducing the problem to one dimension [5, 158, 384, 393, 1098].
4.3.2.2 A Model
We model a tumor spheroid as a time-dependent perfect sphere of radius b(t). Ini-

tially, the radial position of a material point inside the tumor is R0 with radius B0 and
we are interested in the evolution of each point as a function of time, r = r (R0 , t).
In particular, we look for the outer radius of the tumor as a function of time, that
is, b(t) = r (B0 , t). Since we assume radial symmetry, we can focus our attention
on any line from the origin to the boundary of the sphere. We assume that growth
is isotropic, that is, there is no preferred direction of growth; a volume element dV0
initially located at a point R0 will grow to a new volume element dv at the point
r = r (R0 , t) with a volumetric growth coefficient η = η(r, R0 ) dependent either on
the type of cells (dependence on the original R0 ) or its position r in the sphere:
dv = η(r, R0 )dV0 . (4.15)

Expressing both volume elements in spherical coordinates (so that dv = r 2 sin ϕdr dθ
dϕ) and assuming that the sphere remains spherical, the angles in the initial and
current configurations are identical. Integrating over these two angles leads to
∂r
= η R02 r −2 . (4.16)
∂ R0
By definition of the growth stretch, we have ∂r /∂ R0 = γ, so that
γ = η R02 r −2 . (4.17)
In the absence of an elastic deformation, this problem is equivalent to the growth of

a rod. We assume that at each point, growth is exponential and proportional to the
nutrient concentration u(r, t), so that
∂η
= kηu(r, t), (4.18)
∂t
∂r
= γ, (4.19)
∂ R0
where u(r, t) satisfies the diffusion equation

∂u D ∂ r 2 ∂u
= 2 − Q. (4.20)
∂t r ∂r ∂r
Here, D is the diffusion constant and Q is a constant nutrient uptake. The tumor
is assumed to sit in a well-mixed bath of nutrients so that the concentration at the
outer surface is u(b, t) = ub . Since this partial differential equation is second order
in space, we need a second boundary value. The second condition is ur (0, t) =
0 if u(0, t) > 0 or u(r, t) ≥ 0 ∀r ∈ [0, b] if u(0, t) = 0. This condition expresses
the constraint of no-flux if the concentration at the origin does not vanish and is
everywhere positive if the concentration at the origin vanishes.
4.3.2.3 Analysis
Starting with a small size tumor, we expect the nutrients to penetrate all the way to
the origin. However, as the tumor grows, it will reach a critical size bcr at time tcr so
that the concentration at the origin vanishes u(0, tcr ) = 0. Since the concentration
remains positive for all time, for t > tcr the concentration will vanish at a point a
with 0 < a < b as illustrated in Figure 4.10.
Assuming that the typical diffusion time of nutrients is much faster than growth,
it is reasonable to assume that the nutrient concentration has reached equilibrium so
that the steady state of Equation (4.20) can be solved exactly. If b ≤ bcr , we have:
Figure 4.10 Nutrient u(r) b1 bcr b3

concentration as a function ub
of the radius. For t < tcr , the
nutrient penetrates
throughout the entire t1< tcr
spheroid. For t = tcr , the t= tcr
concentration vanishes at the
origin and for t > tcr , the t3 >tcr
concentration vanishes in the
region r < a where a
necrotic core develops.
r
a3
Q 2
u(r ) = (r − b2 ) + ub . (4.21)
6D
The critical radius bcr is then defined as the smallest radius for which u(0) = 0. This
condition leads to
6Dub
bcr = . (4.22)
Q
For b ≥ bcr , we impose the condition u(a) = 0 for a value of a ∈ [0, b), and the
solution is now

0 if r ≤ a,
u(r ) = Qr 2 Q(b3 −a 3 )−6Dbub abQ(a 2 −b2 )+6abDub (4.23)
6D
+ 6(a−b)D
+ 6(a−b)Dr
if r ≥ a.
The value of the inner radius a is determined by a no-flux condition at r = a, ∂u |

∂r r =a
=
0. This condition implies that a is given as the first positive root of the polynomial
P(a, b) = Q(2a 3 − 3a 2 b + b3 ) − 6bDub . (4.24)
In Figure 4.10, the nutrient profile is shown for different values of the external radius.
The inner shell of radius a can be associated with the so-called necrotic core, a region
inside the spheroid where the cells die due to lack of nutrients. As the radius increases,
the proliferating zone (b − a) tends rapidly to a constant determined by taking the
limit of the solution of (4.24) for large b, that is,

2Dub
= lim (b − a) = . (4.25)
b→∞ Q
Once the nutrient profile is known, the growth of the tumor can be computed by
integrating (4.18). The position of a material point, labeled R0 in the original con-
figuration, is given by
R0
r =3
3
η(r )R02 dR0 . (4.26)
0
Taking the derivative with respect to time and using (4.18), Equation (4.26) simplifies
to R0
∂(r 3 )
=3 u(r )kη R02 dR0 , (4.27)
∂t 0
and by changing variables in the integral using (4.16), we have

∂r (R0 , t) r
= kr −2 u(ρ)ρ2 dρ, (4.28)
∂t 0
where u(ρ, t) = 0 for ρ ≤ a. In particular, the radius of the sphere is given by

∂b(t) b
= kb−2 u(ρ, t)ρ2 dρ. (4.29)
∂t 0
Initially, the entire spheroid is proliferating and exponential growth is observed as

shown in Figure 4.11. As the necrotic core forms, growth is still exponential but in a
region of constant width. Therefore, following the argument in the previous section
on tip growth, we conclude that, for large b, the radius increases linearly in time as
shown in Figure 4.11.
4.3.2.4 Discussion
We comment that this model is quite simplistic, since we expect a quiescent zone to
form where the cells do not proliferate but remain alive and are in a state of dormancy
[509, 694]. Nevertheless, this toy model captures the essence of the problem and gives
Figure 4.11 Evolution of b(t)

the spheroid radius in time.
Linear
For t < tcr , the tumor grows
exponentially. However,
growth
when t > tcr , a necrotic core
develops and only a spherical
shell proliferates. For large
time, linear growth is
observed in much the same
manner as systems
exhibiting tip growth. Exponential
bcr
growth
t
tcr
a basic understanding for the concepts of penetration length, proliferation zone, and
necrotic core frequently used in the study of tumor growth [815].
Apart from the effect of nutrients, it is believed that mechanical stress also plays
an important role as high stress may limit the ability of cells to divide through contact
inhibition [530, 566]. However, in order to investigate the effect of mechanics on
growth, an analysis of the full stress tensor at any point in the spheroid needs to be
performed [25, 29, 1098].
4.4 Growth with Elastic Deformation
So far, we have considered growth and elastic deformations as two independent

processes. In general, growth may induce stresses and stresses can depend on growth.
For instance, if we let a filament grow between two rigid plates, the ends cannot
move and resist extension, thereby generating an axial compressive stress in the rod.
Therefore, growth and elasticity are coupled and we must integrate them within a
single framework.
In the one-dimensional context, the fundamental assumption of morphoelasticity
is that a change in length can be produced by both processes, Mathematically, it
can be expressed by postulating that the full axial stretch is the product of a growth
stretch γ, and an elastic stretch, α:
λ = αγ. (4.30)
This apparently simple multiplicative decomposition is rather subtle and will be fur-
ther discussed and justified in the general context of nonlinear elasticity discussed
in Chapter 12. Before discussing this problem in more detail, we explore the conse-
quences of this decomposition by considering several examples.
4.4.1 Growth of a Rod Between Two Plates
As an example of the coupling between growth and stress, we first consider the
simple problem of a rod growing between two immovable plates. Initially, the rod
has a length L 0 equal to the distance between the plate. Since, the rod is constrained,
the total length, l(t) = L 0 , of the rod remains constant for all time. Therefore, we
simply have λ = l/L 0 = 1 for all time.
We assume that the rod grows uniformly and linearly in time, so that
γ = 1 + t. (4.31)
This growth creates a spatially homogeneous compressive stress. Indeed, since λ = 1,

we have α = 1/γ = 1/(t + 1).
4.4 Growth with Elastic Deformation 77
If we first assume a constitutive relationship of the form (4.2), the stress increases
with time as
Et
σ=− . (4.32)
t +1
Paradoxically, as t → ∞, the stress saturates asymptotically to σ∞ = −E. Indeed,

for a Hookean law (4.2), it takes a finite stress, namely σ∞ , to compress a rod to zero
length.
Clearly, real materials will not follow this behavior as a Hookean law is only valid
for small displacements. Corrections to the Hookean law will enter the constitutive
relation when compression or tension becomes sufficiently large. For instance, using
the neo-Hookean law (4.4), we obtain

1
σ=μ −t −1 , (4.33)
(1 + t)2
and an infinite compressive force (σ → −∞) is generated in the limit t → ∞.
4.4.2 Three Different Configurations
A slightly different way to understand the coupling between growth and elasticity
in the previous problem is to separate the response of the material into two distinct
processes as shown in Figure 4.12. First, a growth deformation is applied; we let
the rod grow without external constraints on the ends, so that at any given time
t, it is in an unstressed state. In this configuration, the material point S0 is at a
position S(S0 , t) = S0 (1 + t) with total length L(t) = L 0 (1 + t). Since this state
Initial configuration Current configuration
The rod has length L 0 The rod has length l=L 0
Growth deformation γ Elastic deformation α
Virtual configuration
The rod grows to a length L(t)
Figure 4.12 The growth of a constrained rod between two plates can be thought of as the composi-
tion of two distinct processes: A growth deformation which creates a stress-free virtual configuration,
made of a rod of length L(t), followed by an elastic deformation necessary to enforce the boundary
conditions: a rod of length l(t) = L 0 .
is not realized in the experiment, we refer to it as a virtual configuration; a useful

mathematical construct needed to find the solution of the problem. Second, the grown
rod is placed back in between the plates, and we find the elastic stretch necessary to
satisfy this constraint to be α = 1/(t + 1).
This example also provides a simple illustration of one of the fundamental concept
related to the modeling of growth; that is, the evolution of the reference configuration.
In order to evaluate elastic deformations and stresses in a growing material, the
strain must be defined. Strain is a measure of the deformation with respect to a
reference configuration where the material is unstressed. As growth takes place,
the reference configuration of the material evolves. In our example, the reference
configuration is not the initial configuration but the virtual configuration obtained by
removing stresses. The corresponding length of the growing filament in this virtual
configuration is L = L 0 (t + 1).
We use the following notation throughout this monograph: Whenever possible
most quantities in the current configuration are denoted by lowercase variables.
The corresponding quantities in the initial stress-free configuration are written in
uppercase with subscript 0. The same quantities in the virtual configuration are des-
ignated by the same variable in uppercase but without subscript, or with a subscript t.
For instance, the length in the initial, virtual, and current configuration is denoted as
L 0 , L, and l, respectively.
The standard picture of elasticity is to consider an initial reference configura-
tion without stress and a current configuration where loads are applied and stresses
can be computed from the mechanical balance and constitutive equations. We see
in this simple example that an understanding of the stresses in a growing elastic
medium requires another configuration, namely, the virtual configuration, where the
material is grown in the absence of stress. In the case of one-dimensional exten-
sions, the removal of stresses by unloading the material can always be carried out
in an experiment. However, we will see that in many situations this configuration is
not physically realizable, but can only be defined locally, hence the names virtual
configuration or virtual reference con figuration; a mathematically useful but often
physically unattainable construct.
4.4.3 Homeostatic Growth
Next, we consider a situation where growth depends on stress. We assume that the
filament has a natural homeostatic stress σ ∗ ; a stress that the filament, through the
active process of growth, tends to recover when perturbed from its natural state.
Here, for illustrative purposes, we assume that this process can be modeled by the
evolution law
∂γ
= γ(σ − σ ∗ ). (4.34)
∂t
This law implies that exponential growth or resorption will take place until the stress
σ reaches homeostatic stress σ ∗ [1209, 1276]. Assuming again a Hookean stress–
strain relationship σ = E(α − 1), we see that there is a unique homeostatic strain
α∗ = 1 + σ ∗ /E.
Next, we consider the following problem: We assume that at time t = 0 the fil-
ament is resting in a state of homeostatic stress with a length linit . The filament is
instantaneously extended to a constant length l. How will the filament evolve to
reestablish its homeostatic stress?
The elastic stretch at any time t is α(t) = λ/γ(t) = l/(L 0 γ(t)), and the initial
length L 0 is
linit linit linit E
L0 = = ∗ = ∗ , (4.35)
αinit α σ +E
so that
l(σ ∗ + E)
σ = E(α − 1) = − E, (4.36)
linit γ
which, upon substitution into Equation (4.34), leads to a single equation for γ which
can be solved with the initial condition γ(0) = 1 to find

l l ∗
γ(t) = + 1− e−(E+σ )t . (4.37)
linit linit
Examples of the dynamics of growth for stretching and compression are given in
Figure 4.13. When the filament is instantaneously extended, it is in a state of tension.
The system reestablishes its homeostatic stress by adding mass via growth which
increases the internal compressive stress. Similarly, if the filament is compressed
instantaneously, it is in a state of compression and the homeostatic stress will be
recovered by shrinking the material. The notion of homeostatic stress is central in
understanding normal maintenance and function of many tissues [1209] and will be
considered again in Chapter 14.
Figure 4.13 Growth γ(t )

recovery of a filament with
homeostatic stress σ ∗ of l/l =2
length linit either extended to 0
a filament of length l = 2linit
or compressed to a filament
of length l = linit /2 (σ ∗ = 1
l/l =1
in both case). 0
l/l =1/2
0
t
4.4.4 Application: The Growth of Neurons
4.4.4.1 Background
An interesting example of stress-mediated growth in a biological system is found

in the development and branching of neurons. Most neurons have a very distinctive
morphology with a large cell body from which long protoplasmic protrusions, the
neurites, extend. Neurites develop into either axons or dendrites that are intertwined
to create a network, the nervous system. The axon carries the electrical signals to
other neurons. Individual axons are about a micron in diameter but may extend
across the entire body. The initial formation, development, and growth of axons
have been shown to depend on applied mechanical forces [126, 559]. For instance,
it has been shown that the initiation of axons in a chick’s sensory neurons, chick’s
forebrain neurons, and rat’s PC12 cells can be experimentally triggered by the proper
application of tension on the surface of the cell body [1366]. Following initiation, the
elongation of an axon can occur either spontaneously in vivo or it can be manipulated
by external forces.
In the first case, tension is produced either by the motion of the growth cone,
the specialized tip of the axon [755] or by the overall growth of the tissue in which
it is embedded [1036]. The growth cone generates tension in the axon through an
actin-based motor–clutch system [57].
In the second case, shown in Figure 4.14, a calibrated glass needle is glued to a
neurite with the proper adhesive proteins and is towed [560], thereby controlling the
elongation process. The applied tension at the end of the axon can be imposed by the
deflection of the glass needle. This setup can be used to control either the velocity
of the growing tip or to prescribe a pulling force with high accuracy [560, 1036].
Figure 4.14 Elongation of a

neurite in a hippocampal
neuron induced by pulling.
A. Before needle application.
Arrowhead marks
attachment position. B. Same
neuron towed 30 min.
C. Same neuron 5:40 h/min
after panel B (Bar is 20 μm,
reproduced from [757]).
Figure 4.15 Axonal Growth rate (μm/hmr)

elongation rate for chick
sensory neurites as a function
of experimentally applied
tension. Each line reflects the
data of a single towed neurite
(adapted from [560]).
Tension (μdynes)
The typical range of applied forces is 1-8 nN (nano-Newtons) and the position of the
axon tip can be measured by standard microscopy.
Remarkably, various analyses of the relationship between applied tension and tip
velocity, as illustrated in Figure 4.15, reveal that, after a critical tension, below which
no growth occurs, the growth rate is mostly linear with the applied tension [126, 181,
560, 755, 1366]. Note that both the slope and critical tension for these relationships
differ from cell to cell [757]. This change in length results from stretching the neurons
over long periods of time.
If the axon is suddenly plucked, its response will depend on its material property
rather than on its growth response to external stimuli. On these small time scales,
the axon shows a typical linear, spring-like, response past a critical rest tension that
corresponds to the tension in axons in the absence of external loads. Experimental
results are shown in Figure 4.16.
Most studies on axonal growth assume that growth occurs at the growth cone where
most of the activity and microtubule assembly take place [807]. However, detailed
observations of markers such as axonal branch points and docked mitochondria on
the axon itself reveal that growth is neither restricted to the distal end (close to the
growth cone) nor is it homogeneous along the axon. Instead, the growth rate varies
Figure 4.16 Axial tension Force (μdynes)

of chick sensory neurites as a
function of experimental
stretching (obtained by
“plucking” at their midpoint
by a calibrated glass needle).
Observe the Hookean
behavior past a critical
tension. This rest tension is
the tension present in the
neurite in the absence of
external loads (adapted from
Length changes (μm)
[560]).
from the proximal (close to the body) to the distal site, accelerating closer to the
growth cone and resulting in assembly and deposition of new material along the
length of the axon [756].
4.4.4.2 A Model
From these basic observations, we model a growing axon as a one-dimensional elastic

structure subject to growth as described above. The filament is of length l = l(t) and
is anchored at the point s = 0. Following the analysis and the ideas presented in [273,
1002], we assume that the tension in the axon, σ, is due to both the applied tension
at the end, σ L , and an adhesion force, f , between the substrate and the axon. At any
given time, we assume that the axon is elastic and operates in small deformations,
so that it is characterized by a Hookean relationship
σ = E(α − 1), (4.38)
where α = ∂s/∂ S, and s(S, t) is the current arc length of a material point S(t), where
S(t) is the reference arc length corresponding to the distance from s = 0 at time t
in the unstressed virtual configuration. Since there is an adhesion force f acting on
the filament due to its interaction with the substrate, the tension along the filament
is given by
∂σ
+ f = 0. (4.39)
∂S
This relation can be obtained from the general Equations (5.79) for a growing Kirch-
hoff rod described in Section 6.2 by restricting the motion to a line and taking
n = (0, 0, σ) and a body force f = (0, 0, f ) in the tangential direction.
We further assume that the adhesion force can be modeled by a simple Hookean
law
f = K (s(S) − S). (4.40)
Taking an extra derivative of (4.39) and (4.40), and using the Hookean relation-
ship (4.38), we obtain
∂2σ σ
+ 2 = 0, (4.41)
∂ S2 a
√
where a = E/K is a characteristic length for the problem. Since the proximal
end of the filament (s = 0) remains fixed, s(0, t) = 0, which implies f (0, t) = 0
and ∂σ/∂ S = 0 at S = 0 for all time t. The other end is pulled with a tension
σ(S = L , t) = σ L .
To model the growth process, we need a law for the evolution of S(t). At time
t = 0, the filament is parameterized by its initial arc length S0 and the change in
reference configuration is described by γ = ∂ S/∂ S0 and a growth law of the form
∂γ
= γg(σ), (4.42)
∂t
where growth takes place when the tension is larger than a critical homeostatic tension
σ ∗ so that
g(σ) = k(σ − σ ∗ )H(σ − σ ∗ ), (4.43)
where H(·) is the Heaviside function and k is a constant.
4.4.4.3 Analysis
The solution of (4.41) with the given boundary conditions is
cosh(S(t)/a)
σ(S(t)) = σ L . (4.44)
cosh(L(t)/a)
A large value of a corresponds to either a small adhesion force or a very stiff filament.
In this case, the tension in the filament becomes uniform since we have
lim σ(S) = σ L . (4.45)

a→∞
For large adhesion force or a very compliant filament, a is small and the tension is
localized at the pulling side with an exponential decay of characteristic length a.
In an experiment where the end is pulled with a constant tension σ L , the behav-
ior of the filament would still appear Hookean since it will extend to a length
l = L + (a/E) tanh(L/a)σ L , corresponding to an extension proportional to the
applied tension.
Next, we consider the growth process for long times. If γ is known, the evolution
of a material point, initially at position S0 , is given by
S0
S(S0 , t) = γ dS0 . (4.46)
0
Therefore, the velocity of this point is

S0 S0
∂t S(S0 , t) = (∂t γ) dS0 = γg(σ) dS0
0 0
S
= g(σ(S)) dS. (4.47)
0
Using both the explicit form (4.44) for the tension and the growth law (4.43), the
equation for S(t) is
Figure 4.17 Velocity of V(S)

points along a growing axon 20
(L = 200, σ L = 1, σ ∗ =
5
a=6
1/10, k = 1).
15
0
a=4
10
a=10
5
S
50 100 150 200
S
cosh(S/a)
∂t S = kH(S − S ∗ ) σL − σ ∗ dS
∗ cosh(L/a)
S
sinh(S/a) − sinh(S ∗ /a)
= kH(S − S ∗ ) σ ∗ (S ∗ − S) + aσ L , (4.48)
cosh(L/a)
where
σ∗ L
S ∗ = a arccosh( cosh ). (4.49)
σL a
This last equation gives, at any length L, the velocity of a material point S at time
t, that is,

sinh(S/a) − sinh(S ∗ /a)
V (S, t) = kH(S − S ∗ ) σ ∗ (S ∗ − S) + aσ L . (4.50)
cosh(L/a)
A typical velocity profile is shown in Figure 4.17.

Note that Equation (4.48) for S(t) cannot be solved directly as both S and L
depend on t. Therefore, to solve this free-boundary problem, we first consider this
equation for the particular point S(t) = L(t):
sinh(L/a) − sinh(S ∗ /a)

∂t L = kσ ∗ (S ∗ − L) + kaσ L , (4.51)
cosh(L/a)
with the initial condition L(0) = L 0 . No useful closed-form solution exists for
this equation. Instead, we consider the limit where S ∗ → 0, which leads to ∂t L =
kaσ L tanh(L/a), that is,

L(t) = a arcsinh ekσL t sinh(L 0 /a) . (4.52)
Once the function L(t) is known, it can be substituted into Equation (4.48) and an
equation for the evolution of S(t) can be found for all initial points of the form
Figure 4.18 Length as a L(t)

function of time in the limit 65
a=
of S ∗ → 0 (L = 200, σ L =
150
1, σ ∗ = 1/10, k = 1).
40
a=
100
50 a=10
t
1 2 3 4
S(t = 0) = S0 . After an initial exponential phase where the entire filament experi-
ences growth, extension becomes limited to a finite zone close to the tip of char-
acteristic size a which results in the length increasing linearly with time, as shown
in Figure 4.18. This behavior is similar to the problem of growing spheroids seen
previously.
4.4.4.4 Discussion
This simple model has focused on macroscopic aspects of axonal growth and its
relationship to applied tension. Other mathematical models have focused on the
tubulin production in the cell body, transport processes to the growth cone, and the
assembly necessary to create new microtubules to extend the axon [494, 688, 847,
848, 988, 1272]. Furthermore, the extent and pattern formations of axons in three
dimensions during development, rather than on the line, raise many fundamental
questions about pattern formation and pathfinding [12, 494, 592, 1167, 1225].
4.4.5 Is This Just Plasticity?
For the reader familiar with the theory of plasticity, the discussion on the decompo-
sition of deformations in elastic and growth processes may seem familiar. In a plastic
deformation, the material flows irreversibly after a critical stress and we could argue
that, superficially, the problem of growth is equivalent to the problem of plastic defor-
mation. Many authors, especially in the plant literature, have argued along these lines
and have adopted the view that elastic growth should be based on the theory of plas-
ticity [514]. There are, however, many distinctions between the two theories and great
care should be exercised before modeling growth processes as plastic processes. At
the kinematics level, there are many similarities such as the decomposition between
plastic and elastic deformations and the evolution of a reference configuration.
However, plasticity has been developed chiefly to describe the properties of metals.
The evolution laws for plastic strains are a direct consequence of the microstructure
of metals, with special emphasis on slip lines and shear stresses. There is no reason
to believe that biological systems follow the same principles.
It is therefore important to be aware and to take advantage of the formal concepts
and mathematical methods that have been developed for plasticity, but it can be
misleading to restrict the deformations, due to growth, to plastic processes. Further,
many of the relationships between growth and stresses observed in tissues and plants
are different from the evolution rules given by the theory of metals.
The two theories, elasto-plasticity and morphoelasticity, are two instances of the
theory of anelasticity: The general theory of continua that considers both conservative
elastic and nonconservative anelastic deformations regardless of their origin. In our
case, the anelastic deformations are due to growth, but other possibilities of physical
theories involving anelastic deformations include elasto-plasticity, thermoelasticity,
viscoelasticity, and dislocation theory [321, 843].
The deformations of a one-dimensional structure offer a good starting point for a
discussion of the similarities and differences between growth and plasticity. A purely
elastic material is one in which the response to applied stresses is instantaneous and
reversible. Other materials which exhibit irreversible deformations can be modeled
by the proper choice of constitutive relationships and multiplicative decomposition.
Here, we postulate that a material can exhibit both an elastic deformations character-
ized by strain, el = l/L − 1, and an anelastic irreversible deformation characterized
by the strain, an = L/L 0 − 1. Using the multiplicative decomposition, we have
λ = λel λan = 1 + el + an + el an . (4.53)
If we further assume that these deformations are small, so that the product el an can
be neglected, we obtain that the total strain is decomposed additively into
= λ − 1 = el + an . (4.54)
Since we have two processes responsible for the deformation, we need to supply two
separate constitutive relationships that characterize the material behavior.
• An ideal fluid: For an ideal fluid, we have no elastic response el = 0 and the rate
of anelastic strain is proportional to the applied stress:
∂an ∂ 1
= = σ, (4.55)
∂t ∂t η
where η is a viscosity. In this case, the material exhibits only irreversible flow.
• A Maxwell fluid: A Maxwell fluid combines both an elastic response of a Hookean

material with an anelastic fluid-like response. That is:
1 ∂an 1
el = σ, = σ. (4.56)
E ∂t η
Combining these two equations leads to a single relationship for the total strain
∂ 1 1 ∂σ
= σ+ . (4.57)
∂t η E ∂t
Note that if the strain is turned off, the stress relaxes as σ ∼ e−t/τr , where τr = η/E.
• A Kelvin solid: A Kelvin solid (or Kelvin–Voigt solid) is an elastic material in

which the response to the stress occurs over a finite time determined by the viscous
nature of the material. Again, we assume a Hookean response for the elastic part,
but the anelastic component decays both as a function of the stress and the anelastic
strain itself:
1 ∂an E 1 ∂σ
el = σ, = − an − . (4.58)
E ∂t η E ∂t
These two relations can be written as a single equation
∂
σ = E + η . (4.59)
∂t
If η = 0, the material relaxes with a characteristic viscoelastic time τr = η/E as
∼ e−t/τr describing the evolution of strains when the stress is turned off. This
decay should be contrasted with the equivalent stress relaxation of a Maxwell fluid.
• A Bingham material: In a simple fluid there is irreversible flow in response to

applied stress, however small. For non-Newtonian fluids such as paint, flow does
not begin until a critical yield stress, σ ∗ , has been exceeded. This behavior is
captured by the Bingham model which, in its simplest form, is expressed as
∂an 1
el = 0, = (σ − σ ∗ )H(σ − σ ∗ ), (4.60)
∂t ν
where H(·) is the Heaviside function and ν/σ ∗ is a typical characteristic relax-
ation time. This behavior models an irreversible flow once a critical stress has
been exceeded. It has been used as a paradigm for most plant growth models, as
discussed in Section 4.5.
• A Maxwell–Bingham material: The Bingham model can be generalized to a

Maxwell-Bingham type system by assuming an elastic response
1 ∂an 1
el = σ, = (σ − σ ∗ )H(σ − σ ∗ ), (4.61)
E ∂t ν
which leads to
∂ 1 1 ∂σ
= (σ − σ ∗ )H(σ − σ ∗ ) + . (4.62)
∂t ν E ∂t
where ν/E is a characteristic relaxation time. Regarding terminology, the conven-
tion is to call a Maxwell fluid viscoelastic, reflecting the combination of irreversible
Kelvin solid Maxwell fluid

1.0
σ
2.0
0.8 ε
ε
1.5
0.6
1.0
σ
0.4
0.2 0.5
0.0
t t
0.0
0 1 2 3 4 0.0 0.5 1.0 1.5
Bingham fluid Maxwell-Bingham system

2.0
1.0 σ
ε
0.8 1.5
ε
0.6
1.0
σ
0.4
0.5
0.2
t t
0.0 0.0
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5
Figure 4.19 Comparison of strains produced by a ramping of the stress in time for four different
materials, ν = 1/4, σ ∗ = 1/2, E = σ0 = η = 1.
flow generated by viscous stresses with an elastic component, and to term a Bing-
ham fluid plastic to mean an irreversible deformation beyond a critical yield stress.
A comparison of the strains produced by the different models, where stresses are
ramped up linearly with respect to time, is given in Figure 4.19.
4.5 Application: The Growth of Plant Cells
The one-dimensional theory of elastic growth described in the previous section is

elementary. Nevertheless, these ideas can be applied to simple models for the elonga-
tion of plant cells. The earliest quantitative model of plant cell growth is by Lockhart
[798] and has been applied successfully to gain some insight into the role of plant
growth hormones such as auxin.
4.5 Application: The Growth of Plant Cells 89
4.5.1 Lockhart Model
4.5.1.1 Background
The axial organs of plants such as stems or coleoptiles are long filamentary struc-
tures that are generated mostly by cell expansion. The protoplasts of plant cells,
consisting of the cytoplasm and nucleus, are surrounded by elastic cell walls, a com-
posite structure formed by the complex interactions of a gel-like structure made of
pectins, with a network of stiff fibers composed of cellulose microfibrils tethered by
hemicelluloses, and glycoproteins [172, 741, 1016]. Due to the strongly anisotropic
properties of the cell walls, axial organs grow almost exclusively in length. These cell
walls are infiltrated by water as seen in Figure 4.20. Whereas the protoplast has large
concentrations of solutes, the water in the cell walls has negligible amounts. This
difference in solute concentrations creates an osmotic pressure Posm on the order
of 0.3 to 1 MPa [231, 1065, 1367]. This osmotic pressure is counteracted by the
hydrostatic pressure inside the cell usually referred to as the turgor pressure Ptur . By
analogy with hydraulic systems, the difference between these two pressures is called
the water potential
P = Posm − Ptur , (4.63)
and, the cell can be thought of as a membrane osmometer. When the cell is fully
turgid, that is, fully swollen, the osmotic pressure is completely balanced by the
hydrostatic pressure and P = 0. In the absence of a growth mechanism, induced
by cell wall loosening, the water potential drives reversible changes of cell wall
volume which can be described by the basic laws of hydraulics.
4.5.1.2 A Model
We model the plant cell as an axisymmetric cylinder of constant radius, but varying
length l(t) with initial length L 0 and unstressed length L(t), which corresponds to
removing all pressure. Then, assuming that there is no change in the cross-section
area, the relative rate of volume change is equal to the relative rate of length change.
At equilibrium, the force acting on the cylinder cap is balanced by a constant elastic
stretching α = l/L so that dα/dt = 0 and the change in extension is proportional to
the water potential. Hence, we have
1 dλ
= K P, (4.64)
λ dt
where λ = l/L 0 and K is the water conductance. Therefore, changes in osmotic

concentration in the extracellular space drive reversible plant motion by changing
the relative osmotic concentration. This process is, for instance, the mechanism by
which guard cells on the surface of leaves regulate photosynthetic activity by opening
Vacuole (v)
Uptake of
Nucleus (k)
water
Cell wall (cy)
Axial elongation
of the cell
Figure 4.20 Elongation of plant cells. The vacuole increases in size due to uptake of water causing
an axial elongation of the cell (adapted from [924]).
and closing pores due to varying physical conditions [1215]. Since the law for the
change of volume due to change in water potential is of the form of a flow, the
deformation is reversible but inelastic, in the sense that the work required to expand
the cell during deformation cannot be recovered.
Plant cells also have the ability to increase their volume in an irreversible manner.
This process of irreversible cell enlargement is referred to as the growth of the plant
cell. The change in volume from the initial meristematic size to maturity can be up
to a 1,000 fold for hair cells on the surface of young cotton seeds and 30,000 fold for
xylem vessels [232]. After the initial cell division in the meristem region, growth is
mostly achieved by wall loosening. This stress relaxation produces a decrease in the
turgor pressure that drives water uptake. In order for this process to take place over
long periods of time and to allow for steady growth, osmoregulation is necessary to
balance the dilution of solutes with the addition of water. Furthermore, walls need to
be constantly rebuilt to prevent thinning and turgor pressure needs to be maintained
to keep the cell walls in a state of tension necessary for elongation [1066]. The cell
wall extension created by loosening is usually referred to as a creep or plastic process
and has been modeled as a Bingham-like response.
Let γ(t) = L(t)/L 0 be the growth stretch associated with this growth process.
Assuming that the material response of cylindrical walls is isotropic and the cell
can only be deformed in the axial direction, the axial stress is given by the pressure
exerted on the caps of the cylinder. Then, the creep-response can be written directly
in terms of the turgor pressure rather than the axial stress, that is,
1 dγ
= (Ptur − P ∗ )H(Ptur − P ∗ ), (4.65)
γ dt
where P ∗ is the yield pressure threshold above which irreversible extension takes
place with extensibility coefficient , the amount of extension above threshold.
Since both water uptake and enlargement take place at the same time, we model the
extension as a morphoelastic process where λ = αγ, which implies
1 dλ 1 dα 1 dγ 1 dγ
= + = , (4.66)
λ dt α dt γ dt γ dt
where we have used dα/dt = 0. Therefore, the relative rate of volume changes given
by (4.64) and (4.65) must be equal and, for Ptur > P ∗ , we can eliminate Ptur . After
simplification, we recover Lockhart’s equation,
1 dλ K
= (Posm − P ∗ )H(Posm − P ∗ ). (4.67)
λ dt +K
4.5.1.3 Analysis
We first note that we can recover, from Lockhart’s equation, both Equations (4.65)
and (4.64) as limit cases when either K or K , respectively. In the first case,
growth is limited by the microscopic processes that allow for cell wall extensibility.
While in the second case, growth is limited by water conductance.
How do we use such an equation? Note that the left-hand side of this equation
is the relative elemental growth rate g = (1/λ) (dλ/dt) used by plant physiologists
to describe the growth of stems as defined in Section 1.4. It can be obtained exper-
imentally for steady growth by tracking the evolution of material markers in time.
Similarly, the water conductance can be measured independently and the osmotic
pressure can be controlled experimentally. Therefore, Lockhart’s equation provides
a way to obtain the cell extensibility parameter and the yield threshold P ∗ by
measuring the sensitivity of growth with respect to changes in the turgor pressure
[231].
Different techniques to control the turgor pressure have been used such as:
• the incubation of excised tissues in solutions of varying concentrations of
osmoticum [505, 506, 586];
• the water-stressed treatment obtained by letting the plants dry down, resulting in
lower water potential [319];
• the control of pressure by enclosing the growing plant in a pressure controlled

chamber [230];
• the injection of silicone oils into the vacuoles of cells [998];
• the use of a pressure probe to inject or remove cell solution [1367].
According to Lockhart’s model, the growth rate g varies linearly with the turgor
pressure, up to a critical pressure P ∗ , where the growth rate vanishes. So by plotting
g as a function of Posm or Pturg , the parameters and P ∗ can be obtained from the
slope and the horizontal intercept, respectively.
4.5.1.4 Discussion
As is often the case in biology, this simple idealized linear model is relevant for many
different systems such as the growth of soybean leaf [147], but many other plants
have relative growth curves which are not linear with changes of turgor pressure. For
instance, the growth rate in Nitella measured in [505] is best fitted by a nonlinear
function whereas Chara corallina exhibits linear rather than exponential growth. In
this case, the relative growth rate is best modeled by taking g to be constant [1045].
The study of the variations of the coefficient , under growth induced by various
concentrations of the plant hormone auxin, has revealed that cell wall extensibility
is greatly affected by auxin while the turgor pressure remains essentially unchanged
[587]. This basic result together with the measure of through stress relaxation
experiments [231] and the inhibition of growth by abscisic acid have demonstrated
the importance of cell wall loosening or stiffening as factors controlling plant cell
growth [1131].
Despite the wide use of the word “plasticity” to describe the irreversible deforma-
tion of plant cell walls, the actual rheological properties of cell walls greatly differ
from those of a plastic material. In particular, growth stops after a few minutes fol-
lowing inhibition of respiration, despite possible changes in turgor pressure [1131].
Therefore, irreversible wall elongation is an active process mediated by biochemical
regulators that depend on cellular mechanisms. It is not a purely physical property
of the wall. Furthermore, experiments on segments of maize coleoptiles that have
been frozen then thawed [588] have shown that these cells did not exhibit plastic
flow under a loading–unloading cycle, but only reversible viscoelastic properties, as
demonstrated in Figure 4.21.
4.5.2 Extending Lockhart’s Model
Lockhart’s model assumes both a state of constant elastic equilibrium, as a conse-

quence of the assumed constancy of the stresses, and the absence of external loads
applied to the tissue. If either the turgor pressure is changed rapidly or extra tensile
loads are applied to the growing tissues, the growing walls simply follow the con-
Figure 4.21 Load-extension

curve for a frozen-thawed Load (g)
segment of maize (Zea mays 48
L.) coleoptile obtained by a
sequence of loading steps
followed by a sequence of 40
unloading steps in a time
frame of 90 minutes. Note
that despite the fact that the 35
loading and unloading
curves are different, the
segment returns completely 24
Loading
to its original length after
unloading. This closed
16
hysteresis loop indicates a
viscoelastic response of the
walls rather than an 8 Unloading
irreversible plastic
deformation (data from
[588]). 0 Extension (μm)
-1000 -800 -600 -400 -200 0 200
stitutive relation given by (4.67). If external tensile loads are applied to a growing
tissue, there is a temporary increase in the growth rate which is later modulated by
an auto-regulatory mechanism. This response of growth to stresses was observed
by Heyn [573, 574] who observed elastic and irreversible deformations in 1936 by
loading and unloading a growing stem with weights. As illustrated in Figure 4.22,
later experiments to distinguish between elastic and irreversible responses of grow-
ing tissues are based on the same idea; that is, a temporary application of an external
loading force on the tissue [743, 744] or a sudden increase of turgor pressure [1045].
The influence of an elastic response on the rate of growth can be modeled by
assuming that the total deformation is decomposed into an elastic and growth com-
ponent [229, 994] so that we now have
1 dλ 1 dα 1 dγ
= + . (4.68)
λ dt α dt γ dt
Assuming that the water conductance is much greater than the extensibility of the
cell wall, we can restrict our attention to the irreversible component of the extension
rate given by (4.65). The elastic component is taken to be a simple Hookean law
σ = E(α − 1) which leads to
1 dα 1 dP
= . (4.69)
α dt E dt
A.
P (MPa) C.
Length
B.
Time (min)
D.
L (μm)
Extension
time (min) Time (min)
Figure 4.22 Elastic and growth deformations in plants. AB. Change in length of Chara corallina
internode cells under a rapid change of pressure P (adapted from [1045]). C. Change in length
following a change of applied forces in soybean hypocotyls in a well-watered (1× Elongating),
water-stressed (1/8 × Elongating), and mature state (adapted from [964]). D. Length of a coleoptile
segment of Zea mays during the measurement of in vivo extensibility in the extensometer (adapted
from [744]). An intact segment was extended by applying a load of 20 g, (+F). After 15 min, the
force was removed (-F) and reapplied after further 5 min (+F). Etot, Eel, Epl denote total, elastic,
and plastic extensibilities.
The two effects combined together give
1 dλ 1 dP
= (Ptur − P ∗ )H(Ptur − P ∗ ) + . (4.70)
λ dt E dt
Note that this equation assumes that under fixed turgor, the steady state of growth
increases exponentially in time. For plants whose length increases linearly in time
[1045], this equation can be replaced by
dλ L 0 dP
= (Ptur − P ∗ )H(Ptur − P ∗ ) + , (4.71)
dt E dt
where L 0 is the initial cell length. Interestingly, this form of growth response is
formally equivalent to a Maxwell–Bingham material (see Equation (4.61)) and it is
therefore tempting to associate the plant behavior with the behavior of an elasto-
plastic material.
In order to isolate the growth contribution from the elastic contribution, one can
subject the growing stem to changes in pressure. As long as the pressure remains
constant, there is no contribution from the elastic part since the last term in Equa-
tion (4.70) vanishes identically. The elasticity of plant tissues can be tested by tra-
ditional methods in mature tissues ( = 0), but mature cells tend to be more rigid
than growing cells. Therefore, in order to probe the elasticity of a growing tissue,
a variation of the pressure must be applied. In the thought experiment shown in
Figure 4.23, we apply a constant pressure P1 from t = 0 to t = t1 , then suddenly
increase the pressure to P2 for a time t = t2 − t1 . The change in length after that
time L = λ(t2− ) − λ(t2+ ) depends both on A and as follows
Figure 4.23 Theoretical P(t)

change in length of a P2
growing morphoelastic
filament under rapid changes
of the pressure P.
P1
P*
t
λ(t)
ΔL
ΔL gr
L0
t
0 t1 t2

L0
L = (P2 − P1 ) (t2 − t1 ) + , (4.72)
E
from which the Young’s modulus can be found. In particular, if the duration of the
pressure pulse is small enough, we have
L0
E= (P2 − P1 ). (4.73)
L
The Lockhart model has been used for the study of the impact of various chemical
effects such as salinity, acidity, nitrogen supply [845]; biochemical factors such as
growth hormones [121, 744]; and physical quantities, such as water potential, [840,
1317], and light action [233]. It has also been used as a basis for further model-
ing including elasticity [994], biochemical pathways [187], and spatial variation of
Young’s modulus [874]. It has become a starting point for many plant physiological
studies and has been obtained as a continuum limit of microscopic processes by
modeling the cell wall as a viscous fluid with reinforced fibers [67, 317, 318].
In later chapters, we will further study the mechanics and growth of fiber-
reinforced plant cells viewed as three-dimensional structures.
Chapter 5
Elastic Rods
We now consider the general problem of growing elastic filaments in space. A natural
starting point is to model growing filaments as thin elastic rods subject to external
constraints. The basic idea behind the theory of elastic rods is to replace the stresses
acting inside a three-dimensional tubular structure by a resultant force and moment
vectors attached to a curve. This procedure leads to the Kirchhoff equations [33,
221, 285, 478, 1282], relating averaged forces and moments to the curve’s strains,
which are characterized by its curvatures, shears, and extension. These equations
provide the starting point for much theoretical analysis and numerical modeling and
have been successfully used to model one-dimensional structures as varied as DNA,
proteins, neurons, springs, tree branches, and underwater telecommunication cables.
The extension from elastic rods to morphoelastic rods is performed by taking into
account possible axial extension as well as changes in material properties such as
density, cross-sectional area, and intrinsic curvature, due to growth and remodeling.
Various aspects of a theory of growing elastic Kirchhoff rods were discussed in
[451, 470, 473, 477, 478] mostly in biological contexts, and by O’Reilly and co-
workers who developed a theory of planar growing plants with special emphasis
on tip growth, branching, and constitutive laws [357, 991, 992] (see also [521] for

98 5 Elastic Rods
a three-dimensional extension of these ideas). Other related works include elastic

growth in a fluid environment [1075, 1337] and thermo-elastic rods [213, 499, 500].
In this chapter, we first summarize the theory of Kirchhoff elastic rods by deriving
balance laws for relevant mechanical quantities attached to a space curve. In the
next chapter, using various examples, we develop a theory of growing elastic rods
following [786, 917]. We show that a growing filament can develop stresses due either
to its global geometry or boundary conditions. In turn, these stresses can be sufficient
to cause buckling and, in the case of biological systems, initiate morphogenesis.
5.1 The Kinematics of Curves and Rods
5.1.1 Curves and Frenet Frames
We define a dynamical space curve r(S, T ) as a function of a material parameter S

and time T , i.e., r : [0, L] × [T0 , T1 ] → R3 . We assume that the curve is of class
C 3 in S and C 2 in T . If the first two S-derivatives of the position vector are linearly
independent at each S ∈ [0, L] and for all time T , the curve is said to be regular.
At any time t, the arc length s of a point S, measuring the Euclidean distance
along the curve, is defined as

S ∂r(σ, T )
s(S, t) = dσ , (5.1)
0 ∂σ
√
where |r| = r · r is the usual Euclidean norm.
The arc length is a convenient parameterization of the curve to extract basic geo-
metric properties and we view, interchangeably, the space curve as either a function
of the material parameter S or the arc length s unless otherwise specified. The unit
tangent vector τ to the space curve, r, is
∂r
τ = ≡ r . (5.2)
∂s
We use the tangent vector and its derivatives to construct the standard Frenet–
Serret frame referred to here simply as the Frenet frame. The Frenet frame, {τ , ν, β}
is a set of three orthonormal vectors consisting of the tangent, τ , normal, ν, and
binormal, β, vectors at each point of the curve r. The normal vector is the unit vector
along the derivative of the tangent vector with respect to the arc length
∂τ
= κ ν, (5.3)
∂s
where the curvature
∂τ
κ = = r , (5.4)
∂s
5.1 The Kinematics of Curves and Rods 99
measures the turning rate of the tangent along the curve and is geometrically given
by the inverse of the radius of the best fitting circle at a given point.
The binormal vector is obtained as the vector product of the tangent and normal
vectors: that is,
β = τ × ν. (5.5)
Since β = τ × ν + τ × ν = τ × ν , β is parallel to ν and we can write
β = −τ ν, (5.6)
for a certain function τ , the torsion, that measures the rotation of the Frenet frame
around the tangent τ as a function of arc length. The torsion is a measure of the curve
nonplanarity. For a regular curve, we have
∂β (r × r ) · r

τ = −ν · = . (5.7)
∂s |r × r |2
Along the curve, the Frenet frame moves as a function of arc length according to
a closed system of first-order differential equations, the Frenet equations:
∂τ
= κ ν, (5.8)
∂s
∂ν
= τ β − κ τ, (5.9)
∂s
∂β
= −τ ν. (5.10)
∂s
Given a regular curve r = r(s), a unique curvature κ(s) and torsion τ (s) can be
obtained through (5.4) and (5.7). The converse statement is the fundamental theorem
of curves in space: Given differentiable functions κ(s) > 0 and τ (s), s ∈ [0, L],
there exists a unique curve with τ = 0, such that s is its arc length, κ(s) is its
curvature, and τ (s) its torsion, up to a global rotation, translation, and orientation of
the curve [169]. This result is a direct consequence of the existence and uniqueness of
solutions for the initial value problem consisting of the system of 12 linear ordinary
differential equations (5.2) and (5.8)–(5.10) with initial conditions on the position
r(0) and frame {τ (0), ν(0), β(0)}.
5.1.2 Rods and General Frames
In order to study the motion of elastic filamentary structures, we extend the notion
of space curves to rods. Geometrically, a rod is a space curve r(S, T ), known
as the centerline, equipped with two additional unit orthonormal vector fields
(d1 (S, T ), d2 (S, T )) of class C 2 , neither of which are along the tangent. These two
100 5 Elastic Rods
d2 r(s,t)
d1
d3
Figure 5.1 A rod consists of a centerline and a pair of unit orthonormal vectors. For a given point
on the centerline, the vectors d1 and d2 are chosen to lie in the principal directions of the cross
section. The vector d3 is a unit vector normal to the cross section.
unit vectors represent the orientation of a material cross section at S as illustrated in

Figure 5.1. As before, T ∈ [T0 , T1 ] is time and S ∈ [0, L] is a material parameter. A
configuration is the shape of a rod. The current configuration is the shape of the rod
at a given time t, whereas the reference configuration is a distinguished configura-
tion used to evaluate strains. Here, we choose the material parameter S to be the arc
length in a distinguished reference configuration, r, of the space curve. Hence, in this
configuration, ∂ Sr is a unit vector. We will later identify this reference configuration
with a stress-free configuration [205]. In the current configuration, we have
∂r
= v, (5.11)
∂S
where v is the stretch vector, with |v| > 0 not necessarily equal to one.
The general frame, sketched in Figure 5.2, consists of the three orthonormal
vectors {d1 , d2 , d3 } obtained by defining d3 (S, T ) = d1 (S, T ) × d2 (S, T ). We note
that, by construction, {d1 , d2 , d3 } forms a right-handed orthonormal basis.
A complete kinematic description of the frame is given by
∂di
= u × di , i = 1, 2, 3, (5.12)
∂S
∂di
= w × di i = 1, 2, 3, (5.13)
∂T
where u is the Darboux vector and w is the spin vector. A rod is therefore character-
ized by six kinematic quantities. The Frenet frame is a particular case of the general
frame obtained by choosing the vector d3 as the tangent and either d1 or d2 along
the normal.
d3(s’)
d1(s’)
d2(s’)
d3(s)
d1(s)
e3 d2(s)
e2
e1
Figure 5.2 The general frame represents the evolution of a local basis along the rod. At two
different material points s and s , the general frame has different orientations. A vector attached on
the centerline of the space curve can be expanded locally in these bases.
Since the general frame is a local basis, any vector a attached to the centerline
can be expressed as a linear combination of the basis vector. Explicitly, we write
a = a1 d1 + a2 d2 + a3 d3 , (5.14)
where the components in the local basis are denoted by a = (a1 , a2 , a3 ). We note
that |a| = |a|. Here, we follow Antman’s notation [33] and use sans-serif fonts to
denote the components of a vector in the local basis. The first two components v1 , v2
of the stretch vector v represent transverse shearing of the cross sections while v3 = 0
is associated with stretching and compression. Since the vectors di are normalized,
the norm of v gives the stretch of the rod during deformation away from the reference
configuration: α = |v| = |v|. Equivalently, we can write α = ∂s/∂ S, where s is the
current arc length.
The first two components (u1 , u2 ) of the Darboux vector are associated with bend-
ing while u3 represents twisting, that is, the rotation of the basis, but not the curve,
around the d3 vector. It contains information on both the torsion of the centerline and
on the rotation of the cross section for increasing values of S.
To understand the mathematical structure of the system, it is convenient to intro-
duce an orthonormal matrix D describing the basis:

D = d1 d2 d3 . (5.15)
This matrix can be used to extract the local components of a vector:

102 5 Elastic Rods
a = Da. (5.16)
An alternative representation of the kinematics is obtained in terms of matrices by

introducing the Darboux matrix U(s, t) and the spin matrix W(s, t) as follows:

∂D ∂d1 ∂d2 ∂d3
≡ = DU, (5.17)
∂S ∂S ∂S ∂S

∂D ∂d1 ∂d2 ∂d3
≡ = DW. (5.18)
∂T ∂T ∂T ∂T
These two new matrices are the skew-symmetric matrices associated with the axial
vectors u and w, respectively,
⎡ ⎤ ⎡ ⎤
0 −u3 u2 0 −w3 w2
U = ⎣ u3 0 −u1 ⎦ , W = ⎣ w3 0 −w1 ⎦ . (5.19)
−u2 u1 0 −w2 w1 0
The entries of U and W are not independent. By differentiating (5.17) with respect to
time and (5.18) with respect to arc length and then equating their cross-derivatives,
we obtain a compatibility relation for U and W:
∂U ∂W
− = UW − WU. (5.20)
∂T ∂S
We note that the two systems of linear partial differential equations (5.17) and (5.18)
can be interpreted as a Lax pair for the nonlinear partial differential equations for the
elements of the Darboux and spin vectors generated by the compatibility condition
(5.20). Indeed, the natural kinematic structures of evolving rods have led to a large
and beautiful body of work connecting the motion of curves to integrable systems
[289, 453, 750, 752, 935]. However, this construction is purely geometric and has
little to do with the evolution of elastic rods as governed by mechanical principles.
5.1.3 Inextensible, Unshearable Rods
A particularly important class of rods is obtained by choosing v1 = v2 = 0, and

v3 = 1. In this case, α = 1, and since we have s = S, there is no stretch and we
can choose S to be the arc length in any configuration. The vector d3 is the tangent
vector and there is no shear: The possible deformations of the rod are restricted so
that the vectors spanning the cross sections remain perpendicular to the axis.
This geometric constraint is used to model rods which are both unshearable and
inextensible (for simplicity, we refer to rods that are both inextensible and unshearable
as inextensible rods). Geometrically, the vectors d1 and d2 lie in a plane normal to
the axis and are related to the normal and binormal vectors by a rotation through the
register angle ϕ,
d1 = ν cos ϕ + β sin ϕ, (5.21)

d2 = −ν sin ϕ + β cos ϕ. (5.22)
This rotation implies that
∂ϕ
u = (κ sin ϕ, κ cos ϕ, τ + ), (5.23)
∂S
where κ and τ are the usual Frenet curvature and torsion. These relations can also
be inverted to yield ϕ, κ, and τ as functions of the Darboux vector components:
u2
cot ϕ = , (5.24)
u
1
κ = u21 + u22 , (5.25)
u2 u1 − u1 u2
τ = u3 + . (5.26)
u21 + u22
The quantities τ , ∂ϕ/∂ S, and u3 play related but distinct roles. The torsion τ is a
property of the centerline alone and is a measure of its nonplanarity. Hence, a curve
with null torsion is a plane curve. The quantity ∂ϕ/∂ S, the excess twist, is a property
of the rod only. It represents the rotation of the local basis with respect to the Frenet
frame as the arc length increases. An untwisted rod, characterized by ∂ϕ/∂ S = 0
is called a Frenet rod. In a Frenet rod, the angle ϕ between the binormal b and the
vector field d2 is constant. Hence, the binormal is representative of the orientation
of the local basis {d1 , d2 , d3 }. The twist density, u3 , is a property of both the space
curve and the rod: It measures the rotation of the local basis around the space curve,
defined by the Frenet equations, as the arc length increases. A rod with a straight
centerline, τ = 0, but with a twisted frame, ∂ϕ/∂ S = 0, may have the same twist
density as a Frenet rod with helical centerline for which τ = 0 and ∂ϕ/∂ S = 0.
Note that in the general case of an extensible and shearable rod, the components
of the Darboux vector cannot be described purely by geometric parameters such as
curvature, twist, and torsion since the material parameter S is not the arc length.
5.2 The Mechanics of Elastic Rods
Before considering the case of a growing rod, we review the fundamental equations
for the dynamics of Kirchhoff rods. The dynamics of elastic rods can be derived
from Euler’s laws that are based on fundamental physical principles, namely the
conservation of linear and angular momenta. Euler’s first law states that the rate
of change of linear momentum is equal to the total force acting on a body. Euler’s
second law postulates that the rate of change of angular momentum with respect to
a given point is equal to the total couple acting on the body with respect to the same
point.
104 5 Elastic Rods
p(S-Δ S,x1,x2) p(S, x1,x2)
d1
Front
Back side
side
S(S-ΔS) r (S)
S(S)
d2
S1 S-ΔS S S2
Figure 5.3 Forces and stresses acting on a given cross section with arc-length coordinate S due to
neighboring sections at S < S − S and at S > S.
5.2.1 Balance of Linear Momentum
The balance of linear momentum applied to a rod, as depicted in Figure 5.3, states
that the rate of change of linear momentum of any segment of the rod is equal to
the total force acting on the segment. We assume that the density is constant in time
and only varies along the arc length. The shape of a cross section is also assumed
invariant with respect to time.
We consider a segment of arbitrary length S lying between the cross section
S(S) and the cross section S(S − S). The linear momentum of that segment is

S
∂
dS dx1 dx2 ρ (r + x1 d1 + x2 d2 ) , (5.27)
S−S S ∂T
where ρ is the mass density defined as mass per unit reference volume, and x1 and x2
are the coordinates of a material point in the section along the directions d1 and d2 ,
respectively. By the definition of the centerline, the point x1 = 0, x2 = 0 in (5.27) is
the centroid of the cross section (see Figure 5.4). Hence, the terms linear in x1 and
x2 vanish identically and the rate of linear momentum is
S
∂2r
(ρA) dS, (5.28)
S−S ∂T 2
where A is the cross-sectional area and (ρA) defines a lineal density, that is, the mass
per unit reference length.
The total force acting on the same segment is the sum of the internal contact forces
and the external body forces. We define p(S, x1 , x2 ) as the contact force per unit area
5.2 The Mechanics of Elastic Rods 105
Figure 5.4 Cross section of

a rod and local coordinates.
d1 (x1,x2)
d3
d2
exerted at the point (x1 , x2 ) on the back side, in the sense of increasing S of S(S),
by the sections S(S ) with S > S. The contact force n(S) exerted on the back side
of S(S) by the previous sections is called the resultant force at S, and is given by

n(S, T ) = dx1 dx2 p(S, x1 , x2 , T ). (5.29)
S
The component n3 of the force n along d3 , in the sense of increasing S, is negative

if the rod is subject to compressive forces, and positive in the case of tensile forces.
Note that by the action–reaction principle, the contact force per unit area exerted
at the point (x1 , x2 ) on the front side, in the sense of decreasing S, of S(S) by the
segment S(S − S) is just −p(S − S, x1 , x2 , T ). As a consequence, the resultant
force exerted on the segment, including the contributions from both sides,
S
∂n(S, T )
n(S, T ) − n(S − S, T ) = dS. (5.30)
S−S ∂S
Therefore, the total forces acting on the segment are the sum of the above contact
force and the total body force acting on the segment, that is,

S
∂n(S, T )
S
n(S, T ) − n(S − S, T ) + dS f(S, T ) = dS f(S, T ) + ,
S−S S−S ∂S
(5.31)
where f(S) is the body force per unit reference length applied on the cross section at
S. The balance of linear momentum dictates that
S S
∂2r ∂n(S, T )
(ρA) 2 dS = dS f(S, T ) + , (5.32)
S−S ∂T S−S ∂S
106 5 Elastic Rods
that we rewrite as

S
∂n(S, T ) ∂2r
dS f(S, T ) + − (ρA) 2 = 0. (5.33)
S−S ∂S ∂T
Assuming that the integrand is continuous in S and given the fact that this balance
law is valid for arbitrary S, the integrand must vanish identically and we obtain
∂n ∂2r
+ f = ρA 2 . (5.34)
∂S ∂T
The process of obtaining a partial differential equation from an integral relationship
expressed over an arbitrary domain such as (5.33) is a central tool of continuum
mechanics called a localization procedure.
5.2.2 Balance of Angular Momentum
For the balance of angular momentum, we compute the total couple with respect to
the origin and the total angular momentum about the same point. The total angular
momentum is given by

S
∂
dS dx1 dx2 ρ (r + x1 d1 + x2 d2 ) × (r + x1 d1 + x2 d2 ). (5.35)
S−S S ∂T
We choose the vectors d1 and d2 so that they lie along the principal directions of
inertia of the cross section. Therefore, the product of inertia S dx1 dx2 x1 x2 vanishes
identically and the only nonvanishing terms in (5.35) are

S
∂r ∂d1 ∂d2
dS (ρA)r × + ρ I2 d1 × + I1 d2 × , (5.36)
S−S ∂T ∂T ∂T
where the second moments of area I1 and I2 are defined as

I1 = x2 dx1 dx2 , I2 =
2
x12 dx1 dx2 . (5.37)
S S
We define m(S, T ), the resultant couple, as the couple acting on a section at S

due to the adjacent segment S(S ) with S > S. However, the total couple exerted on
the reference segment is not simply m(S, T ) − m(S − , T ), because the moments
m(S, T ) and m(S − S, T ) are not taken with respect to the same point. Integrating
over the section, the resultant couple produced by the contact force p is
5.2 The Mechanics of Elastic Rods 107
m(S, T ) − m(S − S, T )

+r(S, T ) × n(S, T ) − r(S − S, T ) × n(S − S, T )
S
+ dS [r(S, T ) × f(S, T ) + l(S, T )]
S−S

S
∂m ∂
= dS + (r × n) + r × f + l , (5.38)
S−S ∂S ∂S
where we have taken into account both the couple created by the body force and an
external body couple l(S). The conservation of angular momentum now reads

S
∂m ∂
dS + (r × n) + r × f + l
S−S ∂S ∂S
S
∂ ∂r ∂d1 ∂d2
= dS (ρA)r × + ρ I2 d1 × + I1 d2 × .
∂T S−S ∂T ∂T ∂T
(5.39)
Assuming again continuity of the integrands and using Equation (5.34) to simplify
derivatives of n, we can localize this relation to obtain

∂m ∂r ∂ 2 d1 ∂ 2 d2
+ × n + l = ρ I2 d1 × + I 1 d2 × . (5.40)
∂S ∂S ∂T 2 ∂T 2
In problems with adhesion or contact, the integrands appearing in (5.33) and (5.39)
may not be continuous. In this case, the possibility of jumps has to be taken into
account [990]. However, this case will not be considered here.
5.2.3 Local Mechanics of Rods
To summarize, the stress acting on the cross section at r(S) from the part of the rod
with S > S gives rise to a resultant force n(S, T ) and resultant moment m(S, T )
attached to the central curve. The balance of linear and angular momenta yields the
two fundamental equation of mechanical balance in elastic rods [33]:
∂n ∂2r
+ f = ρA 2 , (5.41)
∂S ∂T

∂m ∂r ∂ 2 d1 ∂ 2 d2
+ × n + l = ρ I2 d1 × + I d
1 2 × , (5.42)
∂S ∂S ∂T 2 ∂T 2
where f(S, T ) and l(S, T ) are the body force and body couple per unit reference
length applied on the cross section at S. These body forces and couples can be
used to model different effects such as short- and long-range interactions between
108 5 Elastic Rods
different parts of the rod, or can be the result of active stress, self-contact, or contact
with another body.
The quantity A(S) denotes the cross-sectional area, in the reference configuration
and ρ(S) is the mass density, defined as a mass per unit reference volume. Here I1 (S)
and I2 (S) are the second moments of area of the cross section corresponding to the
directions d1 and d2 , respectively, at a material point S. They are given explicitly
in (5.37) and several examples are given in Section 5.5.
To close the system, we must specify a constitutive relationship between the
resultant stresses and strains, that is, in the case of rods, between the resultant force
and couple and the curvature and stretch vectors.
5.3 Constitutive Laws for Elastic Rods
5.3.1 Extensible and Shearable Elastic Rods
First, we consider the case where the rod is both extensible and shearable. We further
assume that the rod is hyperelastic: We assume that there exists a strain-energy density
function (also referred to simply as a strain-energy function) W = W (u− û, v− v̂, S)
such that the constitutive relations for the resultant moment and force in the local
basis are given by
m = f(u − û, v − v̂, S) = ∂y W (u − û, v − v̂, S), (5.43)

n = g(u − û, v − v̂, S) = ∂z W (u − û, v − v̂, S), (5.44)
where y = u − û and z = v − v̂.

Here, v̂, û are the strains in the unstressed reference configuration, that is, m =
n = 0 when u = û, and v = v̂. Without loss of generality, one can choose the
general frame so that v̂1 = v̂2 = 0. Furthermore, if S is the arc length of the
unstressed configuration then v̂3 = 1.
Typically, W is assumed to be continuously differentiable, convex, and to satisfy
the growth condition at infinity:
W (y, z)
→ ∞, as |y|2 + |z|2 → ∞. (5.45)
|y|2 + |z|2
The rod is uniform if its material properties do not change along its length (i.e., W
has no explicit dependence on S) and the stress-free strains v̂, û are independent of
S as well.
5.3.2 Inextensible and Unshearable Rods
In the second case, we assume that the rod is inextensible and unshearable. There-
fore, we take v = d3 and the material parameter becomes the arc length in all
5.3 Constitutive Laws for Elastic Rods 109
configurations, that is, S = s. In this case, there is no constitutive relationship for

the resultant force and the strain-energy function is a function of only (u − û):
m = ∂y W (u − û) = f(u − û). (5.46)
For a quadratic strain-energy W = yT Ky, the constitutive relations for the local
basis components m are
⎡ ⎤
K 1 K 12 K 13
m = K(u − û), K = ⎣ K 12 K 2 K 23 ⎦ , K 1 ≥ K 2 . (5.47)
K 13 K 23 K 3
The condition K 1 ≥ K 2 corresponds to the choice of labeling of the vectors d1

and d2 . If, in a basis K 1 < K 2 then a proper relabeling of these vectors leads to
K 1 ≥ K 2 . Also, note that, in general, we can choose the vector d1 so that either
K 12 = K 21 = 0 or û 1 = 0 or û 2 = 0.
In the simplest, and most widely used case, the energy is further simplified by
taking the matrix K to be diagonal, in which case
W1 = K 1 (u1 − û1 )2 + K 2 (u2 − û2 )2 + K 3 (u3 − û3 )2 , (5.48)
where û is the intrinsic Darboux vector that defines the shape of the rod when
unloaded. The coefficients K 1 and K 2 are the principal bending stiffnesses of the
rod, and K 3 is the torsional stiffness.
In order to obtain explicit relationships between the geometric properties of the
cross sections and the material properties of the elastic material that make up the rod,
the classic approach is to compute the response of an elastic cylinder under bending
and torsion within the framework of linear elasticity. We outline these results in
Section 5.5. Explicitly, the resultant moment and coefficients {K 1 , K 2 , K 3 } are
m = E I1 (u1 − û1 )d1 + E I2 (u2 − û2 )d2 + μJ (u3 − û3 )d3 , (5.49)
where E is the Young’s modulus, μ is the shear modulus, J is a parameter that

depends on the cross-sectional shape and I1 and I2 are the second moments of area
given by (5.37). By identification with (5.48), we have E I1 ≡ K 1 and E I2 ≡ K 2 , as
well as μJ ≡ K 3 . Explicit forms for J and I1 , I2 are given in Section 5.5.
From (5.49), we can obtain two important dimensionless parameters
I2 μJ J
a= , b= = , (5.50)
I1 E I1 2I1 (1 + σ)
where we have used the relations μ = 2E(1 + σ). Here, σ denotes the Poisson’s
ratio, a measure of a material compressibility defined for an isotropic linear elastic
material as the ratio of transverse contraction strain to longitudinal extension strain
110 5 Elastic Rods
in the direction of a stretching force. For most materials, it varies between 0 and 1/2,
the latter value corresponding to incompressible materials.
The constant a measures the asymmetry of the cross section. Our convention is
to orient the vector fields d1 and d2 such that I1 and I2 are the larger and smaller
second moments of area, respectively. In this case, we have
0 < a ≤ 1, (5.51)
the value 1 being reached in the dynamically symmetric case in which the second
moments of area are identical.
The constant b is the scaled torsional stiffness. It involves the constant 1/(1 + σ),
which ranges from 2/3 attained for incompressible materials to 1, corresponding to
materials for which there is no restriction under stretching.
5.3.3 Isotropic, Extensible, but Unshearable Rods
For many applications, rods can be modeled as inextensible and unshearable. How-
ever, for the modeling of a growing rod, it will be interesting to consider the case
where the rod is unshearable and extensible with a quadratic energy. In this case, in
addition to (5.49), we have an equation relating the elastic stretch α ≡ v3 ≡ ∂s/∂ S
to the tension:
m = E I1 (u1 − û1 )d1 + E I2 (u2 − û2 )d2 + μJ (u3 − û3 )d3 , (5.52)
n3 = E A(α − 1). (5.53)
For a straight rod under uniaxial tension, this last constitutive equation is simply
Hooke’s law. Note that since S is not the arc length, the Darboux vector u is not
given by (5.23), but it is scaled by an elastic stretch factor α so that
∂ϕ
u = α(κ sin ϕ, κ cos ϕ, τ + ). (5.54)
∂s
However, the intrinsic Darboux vector û is given by the intrinsic geometric curvatures
in the reference configuration; it is not scaled by α since it is a material property of
the rod in the reference configuration.
5.4 Scaling
To further simplify the system in the quadratic inextensible case with equal second
moments of area I = I1 = I2 , we scale the variables by choosing a combination of
the length [L], time [T ], and mass [M] units in the following way:
5.4 Scaling 111
√ [M] [L]3
[M] = ρ AI , = EI. (5.55)
[T ]2
Explicitly, this scaling reads

t → t I ρ/AE s → s I /A,
√
n → AEn m → mE AI ,

u → u A/I w → w AE/I ρ. (5.56)
Under this rescaling, the properties of a rod with a circular cross section can be conve-
niently described by a single parameter: the ratio of bending to torsional stiffnesses,
given by
μJ 1 2
≡ = ∈ ,1 , (5.57)
E I1 1+σ 3
where σ denotes the Poisson’s ratio.
5.5 Bending and Torsional Stiffnesses
Within the framework of linear elasticity theory [761, 806, 1179] it is known that
the parameter J , introduced in (5.49), is determined by

∂ϕ ∂ϕ
J = I1 + I2 + dx1 dx2 x1 − x2 . (5.58)
S ∂x2 ∂x1
The warping function ϕ(x1 , x2 ) provides a measure of the displacement, δs, that each
material point in a given cross section experiences when the rod is twisted uniformly.
In the linear theory, this displacement is determined by
δs(x1 , x2 ) = u3 ϕ(x1 , x2 ). (5.59)
The warping function is the solution of a Neumann problem for the Laplace equation,
namely
∂2ϕ ∂2ϕ
+ = 0 inside S, (5.60)
∂x12 ∂x22
∂ϕ ∂ϕ
N1 + N2 = x2 N1 − x1 N2 on ∂S, (5.61)
∂x1 ∂x2
where N1 and N2 are the components of the exterior unit vector normal to the bound-
ary ∂S of the cross section in the d1 and d2 directions.
In the case of simply connected cross sections, the expression (5.58) for J can
be formulated in terms of the stress function ψ, defined as the harmonic function
conjugate to ϕ:
112 5 Elastic Rods
∂ϕ ∂ψ ∂ϕ ∂ψ
= , =− . (5.62)
∂x1 ∂x2 ∂x2 ∂x1
The stress function ψ is the solution of the following Dirichlet problem for the
Laplace equation:
∂2ψ ∂2ψ
+ = 0 inside S, (5.63)
∂x12 ∂x22
x12 + x22
ψ= on ∂S . (5.64)
2
Substituting (5.62) into (5.58), we obtain

∂ψ ∂ψ
J = I1 + I2 − x1 + x2 dx1 dx2 , (5.65)
S ∂x1 ∂x2
which, using the divergence theorem and (5.64), can be simplified to

J =2 ψ dx1 dx2 − I1 − I2 . (5.66)
S
The explicit computations of I1 , I2 , and J for various shapes can be found in classic
textbooks [1179]. Here, we give the results for elliptic, circular, and rectangular
sections:
• Elliptic cross sections: If we take A and B (A < B) to be the semi-axes of an
ellipse in the d1 and d2 directions, respectively, we have
π AB 3 π A3 B
I1 = , I2 = , (5.67)
4 4
π A3 B 3 4I1 I2 A2 1 2a
J= = , a= , b= . (5.68)
A +B
2 2 I1 + I2 B2 1+σ1+a
√
Applying the scaling (5.56), the scaled semi-axes become, respectively, 2 a and 2.
• Circular cross sections: In the particular case of rods with uniform circular cross
section of radius R, the specialization of the previous case with A = B = R gives
J π R4
I1 = I2 = = , (5.69)
2 4
which implies a = 1 and b = .
• Rectangular cross sections: If A and B denote the lengths of the rectangle in the
d1 and d2 directions, respectively, where A < B, we have
5.5 Bending and Torsional Stiffnesses 113
1
b
B
A
0.8
0.6
0.4
0.2
a
0 0.2 0.4 0.6 0.8 1
Figure 5.5 The domains covered in the (a, b) plane by various cross section shapes are enclosed
in black lines (solid = ellipses, dash = rectangles, dot = right triangles, dash-dot = limaçons).
AB 3 A3 B
I1 = , I2 = , (5.70)
12 12
A3 B 64 A4 πB
J= − tanh ± 0.502%, (5.71)
3 π5 2A
3

A2 1 384a 2 πB
a= , b= 2a − tanh ± 0.502%. (5.72)
B2 1+σ π5 2A
√ √
Under the scaling (5.56), the scaled lengths of the rectangle are 2 3 and 2 3a.
In the case of a square cross section (a = 1), the expression for b reduces to
1
b = 0.859 ± 0.502%. (5.73)
1+σ
Figure 5.5 shows the domain covered in the (a, b) plane for different particular
shapes. More data can be found, for example, in [1119]. Rods with cross sections
made up of assembled flat pieces, like X or H shapes, have low torsional stiffnesses
(small b).
5.6 The Kirchhoff Elastic Rod Model: A Summary
The most commonly used model of elastic rods is obtained by assuming that the rod is
unshearable and inextensible as detailed in Section 5.3.2 with a diagonal constitutive
relationship. In this case, we have the following set of equations for the position
114 5 Elastic Rods
vector r, basis D = [d1 , d2 , d3 ], Darboux and spin matrices U, W, and resultant

force and couple n and m
1. Kinematics
∂r
= d3 , (5.74)
∂s
∂D
= DU, (5.75)
∂s
∂D
= DW, (5.76)
∂t ⎡ ⎤
0 −u3 u2
U = ⎣ u3 0 −u1 ⎦ , (5.77)
−u2 u1 0
⎡ ⎤
0 −w3 w2
W = ⎣ w3 0 −w1 ⎦ . (5.78)
−w2 w1 0
2. Mechanics
∂n ∂2r
+ f = ρA 2 , (5.79)
∂s ∂t

∂m ∂r ∂ 2 d1 ∂ 2 d2
+ × n + l = ρ I2 d1 × + I d
1 2 × . (5.80)
∂s ∂s ∂t 2 ∂t 2
3. Constitutive law
m = E I1 (u1 − û1 )d1 + E I2 (u2 − û2 )d2 + μJ (u3 − û3 )d3 . (5.81)
A few comments are in order:

1. The material is described by three stiffnesses, E I2 , E I2 , and μJ , its density ρ
and by its geometry defined by the cross-sectional area A, the length L, and the
intrinsic Darboux curvatures û.
2. The system is characterized by 15 variables: three coordinates r, three curva-
tures u, three spins w, three resultant forces n, and three resultant moments m.
The last three sets of equations (5.79–5.81) provide nine equations. The first
equation (5.74) provides another three equations. The second and third equa-
tions (5.75–5.76) are a set of nine equations, but only three of these nine equa-
tions are independent since the matrix D is orthonormal, and these two sets are
connected to each other by the compatibility condition (5.20).
3. The body force f and body couple l are given based on the physics of the problem.
They can represent gravity, magnetic forces, electrical forces, self-contact, etc. In
most problems of interest, body couples are not generated and can therefore be
neglected.
5.6 The Kirchhoff Elastic Rod Model: A Summary 115
4. The rotary inertia terms, given by the right-hand side of (5.80), are often ignored
based on an analysis of typical time scales [46]. They can, nevertheless, be of
importance in some dynamic phenomena. For instance, they are required to cap-
ture the vibration frequencies of a loaded rod close to its buckling point [942].
5. A rod is said to be initially straight if the intrinsic curvatures vanish identically
û = 0.
6. A remarkable feature of the model, known as the Kirchhoff kinetic analogy, is that
the equations governing the equilibrium solutions are formally equivalent to the
Euler equations describing the motion of a top, that is, a rigid body with a fixed
point under an external force field. This equivalence is achieved by identifying
the axis of the top with the tangent vector of the rod, and the time with the rod’s
arc length. For each equilibrium configuration of a rod, the tangent vector as
a function of the arc length defines a curve on a unit sphere as time evolves.
Similarly, in the motion of a top, the axis draws a curve on the unit sphere. The
Kirchhoff analogy establishes that to each solution of the Euler equations for tops
in time, there exists an equilibrium solution of the Kirchhoff equations for rods
in space. In the case of planar motion (where the rigid body reduces to a simple
pendulum) this analogy was already known by Bernoulli and Euler [685, art. 620].
The statics of rods is thus intimately connected to the dynamics of spinning tops.
For instance, if û2 vanishes identically, the equations for the rods are equivalent
to the Lagrange case of the Euler equation (a symmetric top) which are known to
be completely integrable [471, 959]. An example of a solution of the top, together
with the corresponding elastic rod, is shown in Figure 5.6.
Figure 5.6 The Kirchhoff analogy. The evolution in time of the axis of the Lagrange top is equiv-
alent to the evolution in space of the tangent vector of an inextensible transversely isotropic rod
with quadratic elastic energy. Left. The tip of the top axis draws a periodic curve on the unit sphere.
Right. The equivalent rod forms a torus knot in space.
116 5 Elastic Rods
5.7 Example: Helical Rods
An important class of solutions consists of elastic rods with constant Frenet cur-
vature and torsion. They can be either straight rods if both curvature and torsion
vanish, circular rods if only torsion vanishes, or helical if neither vanishes. Helices
arise in nanosprings [846], carbon nanotubes, α-helices [831], proteins [943, 944],
DNA [92, 826] and macromolecules [675, 1261], lipid bilayers [565, 1129], bacte-
rial flagella in Salmonella and Escherichia coli [679, 817], aerial hyphae in actino-
mycetes [814], bacterial shapes in spirochetes [454], horns, tendrils, vines [470],
screws [133], springs, and helical staircases [226, 417]. In many situations it is
beneficial to model such structures as elastic rods.
We first show that there is a large family of helical structures that are solutions
of the static inextensible Kirchhoff equations in the absence of distributed body
forces and couples, i.e., f = l = 0 in (5.79–5.80). For simplicity, we consider
here the transversely isotropic quadratic case where W has the form (5.48) with
K 2 = K 1 . The existence of helical equilibria for arbitrary strain-energy functions W
was considered in [205].
5.7.1 Geometry of Helices
We consider a helical space curve, parameterized by arc length, s, turning around a

cylinder of radius R whose central axis points along the fixed e1 -axis:
r = Pδse1 + R cos(δs)e2 + R sin(δs)e3 , (5.82)

√
where δ = ±1/ P 2 + R 2 . The choice + (P > 0) defines a right-handed helix. We
use Maxwell’s convention [844] and define right-handedness as the handedness of
a common screw. The height (along the e1 -axis) per turn of the helix, known as the
pitch, is h = 2π|P| and the length of the curve per turn is l = 2π/δ. For a helix of N
turns, the total height is H = 2π|P|N and the total filament length is L = 2π N /|δ|.
A sketch of a helical rod with an applied twist Lγ is shown in Figure 5.7. The Frenet
frame for this helix is
τ (s) = Pδe1 − Rδ sin(δs)e2 + Rδ cos(δs)e3 , (5.83)

ν(s) = − cos(δs)e2 − sin(δs)e3 , (5.84)
β(s) = −Rδe1 + Pδ sin(δs)e2 − Pδ cos(δs)e3 , (5.85)
with Frenet curvature, κ, and torsion, τ given by
R P
κ = Rδ 2 = , τ = Pδ 2 = . (5.86)
P 2 + R2 P 2 + R2
5.7 Example: Helical Rods 117
Lγ
2R e1
2π P
Figure 5.7 A right-handed helical rod characterized by an applied twist Lγ, a radius R, and a pitch
2π P.
A twisted helical rod is obtained by a rotation of the Frenet frame in the plane normal
to τ by an angle γ:
d1 = ν cos(γs) + β sin(γs), (5.87)

d2 = −ν sin(γs) + β cos(γs), (5.88)
d3 = τ . (5.89)
We only consider solutions with constant excess twist ( ∂ϕ∂S

in (5.23) constant) since
it was shown that helical rod solutions with nonconstant excess twist do not exist
for our choice of strain-energy function. Such solutions only exist for very particular
choices of W [205]. The Darboux vector for this rod is
u = (κ sin(γs), κ cos(γs), τ + γ) . (5.90)
5.7.2 Helical Equilibria
We restrict our attention to inextensible rods with constitutive law (5.48) and K 2 =
K 1 . We further assume that the unstressed state of the rod is also a helix defined by

u = (0, κ̂, τ̂ ). (5.91)
For an inextensible rod in the absence of body loads, the equilibrium equations are
n = 0, (5.92)
m + d3 × n = 0, (5.93)
118 5 Elastic Rods
where (·) = ∂s (·). We write the force and moment vectors in the director basis
n = n1 d1 + n2 d2 + n3 d3 , (5.94)
m = m1 d1 + m2 d2 + m3 d3 , (5.95)
and introduce n = (n1 , n2 , n3 ) and m = (m1 , m2 , m3 ). The equilibrium equations

can then be written as
n +u × n = 0, (5.96)

m +u × m + v × n = 0, v = (0, 0, 1). (5.97)
There are three limiting and exceptional families of helices with either û = 0,
u = 0, or n = 0.
• Naturally straight rods: If û = 0, i.e., κ̂ = τ̂ = 0, a helical solution exists with
n = [K 3 ω + (K 3 − K 1 )τ ] (κ sin(ωs), κ cos(ωs), τ ), (5.98)

m = (K 1 κ sin(ωs), K 1 κ cos(ωs), K 3 (τ + ω)), (5.99)
where ω is arbitrary. If we further assume that τ = 0 and L = 2π/κ, the rods are
twisted circular rings. These twisted helical and ring solutions are illustrated in
Figure 5.8ABC.
• Straight rods: When u = 0, the general frame steadily translates, that is, the
centerline is a straight line, as illustrated in Figure 5.8D. The moment m is given
by the constitutive relations, and the force n is parallel (or anti-parallel) to the
helix centerline with arbitrary magnitude.
• Zero-force helices: The third family of singular equilibria are solutions with n =
0, which implies ω = 0. In this case, m is parallel to u. Therefore, the vector m
is constant and parallel to the axis of the helical centerline.
All other helical equilibria have κ = 0 and are twistless, i.e., ω = 0. These
solutions are maintained by a resultant force and moment given by

n = K 3 κ(τ − τ̂ ) − K 1 τ (κ − κ̂) (0, 1, τ /κ), (5.100)
m = (0, K 1 (κ − κ̂), K 3 (τ − τ̂ )). (5.101)
In terms of a fixed basis, the moment and force necessary to maintain a helical shape
can be expressed as a wrench, that is, a combination of a force N and moment M
along the helical axis e1 . Explicitly, we have

M ≡ n · e1 = δ −1 K 1 κ(κ − κ̂) + K 3 τ (τ − τ̂ ) , (5.102)

N ≡ m · e1 = δκ−1 K 3 κ(τ − τ̂ ) − K 1 τ (κ − κ̂) . (5.103)
A. B. C. D.
Figure 5.8 Particular twisted solutions. ABC. Twisted helical rods and twisted rings require κ̂ = 0.
D. If κ̂ = 0, then κ = 0 and only straight-twisted solutions exist. (A. κ = 3, τ = 1, ω = 5, B.
κ = 3, τ = −1, ω = 5, C. κ = 1, τ = 1, ω = 3, D. κ = 0, τ = 0, ω = 3).
These relationships for the wrench in terms of the material parameters and geometry
of a helix are the classic helical spring formulas first studied by Thomson and Tait
in 1867 [1229, Sect. 605] (see also Love [806, Sect. 271]).
5.7.3 Overwinding or Underwinding Helices
An interesting application of the helical spring formula is found when we restrict our
attention to helices maintained solely by axial tension, that is, M = 0. This condition
can be easily realized by pulling a spring.
Physically, the spring is extended but free to twist around its axis. The same
experimental conditions have been reproduced in single-molecule experiments [459,
792]. As sketched in Figure 5.9, in these experiments, molecules of DNA are pulled by
optical and magnetic tweezers while the angular rotation of a bead attached along the
DNA is tracked. The key question is to understand the rotation of the helical structure
as the axial load increases. Experiments show an unexpected behavior where the DNA
systematically overwinds before unwinding. In Figure 5.10, overwinding means that
as N increases, θ also increases. The authors write “simple intuition suggests that
DNA should unwind under tension” [459]. A simple question is to understand the
rotation of a helical spring as it extends: Will it unwind or overwind?
The helical spring formulas (5.102–5.103) imply that for constitutive relations
appropriate for metal springs with circular cross section; as tension N > 0 is
120 5 Elastic Rods
Figure 5.9 In a
single-molecule experiment,
a molecule of DNA is
attached to a substrate and
pulled or twisted at the other
end where a paramagnetic
bead is attached. The
rotation can be observed by
tracking the motion of an
attached fluorescent bead
increased monotonically from zero with M = 0, the rotation first increases (over-
winding the helical axis in the same direction as the reference handedness) and
then decreases, returning to its initial value and then unwinding as illustrated in
Figure 5.10C.
We define the coiling angle per unit arc length to be ρ = |u|, so that in the
unstressed state ρ̂ = |û|. Close to the unstressed state, we can then express the axial
stretch z = τ /ρ − τ̂ /ρ̂ and rotation angle θ = ρ − ρ̂ as a function of N :
N (K 1 κ̂4 + K 3 κ̂2 τ̂ 2 )
z= + O(N 2 ), (5.104)
K 1 K 3 (κ̂2 + τ̂ 2 )3
N (K 1 − K 3 )κ̂2 τ̂
θ= √ + O(N 2 ). (5.105)
K 1 K 3 κ̂2 + τ̂ 2
The second equation gives the linear approximation at the origin to the graph in
Figure 5.10C, which demonstrates that a helix will initially overwind when pulled
from its minimum energy state if and only if = K 3 /K 1 < 1. We remark that if the
filament is formed from a three-dimensional homogeneous linearly elastic material
with Poisson’s ratio in [0,1/2], e.g., standard metals, then ∈ [2/3, 1], so that initial
overwinding in response to positive tension always arises for simple helical springs; a
slightly surprising yet completely classic effect. Equation (5.104) defines an effective
Hooke constant for small extension.
Geometrically, the condition M = 0 implies, from (5.102), that all helical equi-
libria lie on an ellipse in the curvature–torsion plane, given by
κ(κ − κ̂) + τ (τ − τ̂ ) = 0. (5.106)

A. B. C.
τ θ b
nt
b
ta
c
ns
co
Γ<1
ρ
a ^ ,^
(κ τ)
c N
a
θ constan
e3
Γ<1,M=0
θ κ
t
e2
0
Figure 5.10 A. A pure axial tension applied to a helical spring gives rise to another helical equilib-
rium (adapted from Hooke’s 1678 “De Potentia Restitutiva” [611]). B. The solutions with M = 0
zero axial torque helical equilibria lie on an ellipse in the curvature–torsion plane determined by
the value of . Close to the unstressed helix with N = 0 (point a), N > 0 equilibria correspond to
|τ | > |τ̂ | (case τ > 0 shown). Initial overwinding in extension occurs for < 1. C. The relative
coiling angle as a function of the axial force. For < 1 and N > 0 the helix first overwinds (θ > 0),
and then unwinds (θ < 0).
For given material parameters , κ̂, τ̂ , all M = 0 helical solutions lie on this ellipse
and the helix always extends when pulled (|τ | > |τ̂ | for N > 0). A point on the
ellipse corresponds to overwound configurations with respect to (κ̂, τ̂ ) when it lies
outside the circle centered at the origin and of radius ρ̂. It follows that for N > 0 a
helix will initially overwind when pulled from rest if and only if < 1. Furthermore,
the maximum overwinding, corresponding to point b in Figure 5.10B, is obtained
when the ellipse (5.106) intersects the hyperbola
2(1 − )κτ + τ̂ κ − κ̂τ = 0, (5.107)
which comes from (5.103), and the coiling angle returns to its original value when
the ellipse re-intersects the circle ρ2 = ρ̂2 (corresponding to the point c).
We conclude that the seemingly counterintuitive behavior found in DNA experi-
ments is in fact generic for most macroscopic helical elastic filaments. This behavior
is found in the classic Kirchhoff equations and has been confirmed experimentally
as early as 1902 [891, 1093].
Confusion over the sign of rotation in helical structures also arises in the problem
of growth of stems and roots where cell wall anisotropy is responsible for the overall
handedness that is observed. A typical argument is that the cell wall is a cylindrical
structure with reinforced, say, right-handed helical microfibrils that rotate clockwise
during growth (viewed from the top) [155, 483]. Therefore, unwinding is expected
because “helices unwind when they are stretched” [797]. This explanation of hand-
edness is again falsidical and will be further discussed in Section 16.9.
122 5 Elastic Rods
5.8 The Planar Elastica: Bernoulli–Euler Equations
We now consider a reduction of the three-dimensional basic rod above. We assume

that the rod is planar unshearable, inextensible, has a circular cross section, is nat-
urally straight, and that there are no body forces or couples acting on it. Therefore,
it has no torsion and we further assume that it has no twist. The Darboux vector is
then simply given by
u = (0, κ, 0). (5.108)
A convenient representation of the rod is obtained by assuming that it lies in the

e1 − e2 plane and by introducing the angle θ between the tangent vector and the
e1 -axis. That is, we have
τ = d3 = cos θ e1 + sin θ e2 , (5.109)
which implies
κ = θ , (5.110)
and d2 = e3 . By writing n = Fe1 + Ge2 , f = f e1 + ge2 , r = xe1 + ye2 , we can

simplify the equations of motion (5.79–5.80) to
F + f = ρA ẍ, (5.111)

G + g = ρA ÿ, (5.112)
E I θ + G cos θ − F sin θ = ρI θ̈. (5.113)
These equations form a system of three equations for the three unknowns (F, G, θ).
5.8.1 Static Solutions
We consider the static case of (5.111–5.113) in the absence of body forces. Therefore,
ẍ = ÿ = 0 and we conclude that F and G are constant meaning that n is a constant
vector. Without loss of generality, we choose e1 to be along this constant force, so
that G = 0 and
E I θ − F sin θ = 0, (5.114)
or
θ + α sin θ = 0, (5.115)
where α = −F/E I is positive for compressive forces, i.e., F < 0. We recognize

in (5.115) the equation for a hanging pendulum under gravitation. For the pendulum
θ is the angle with respect to the vertical, and α is the ratio of the gravity acceleration
by the length of the pendulum. Therefore, any solution of the pendulum is a solution
5.8 The Planar Elastica: Bernoulli–Euler Equations 123
Figure 5.11 Euler’s drawings of spatial equilibria of the elastica [350] together with the solutions
of (5.115) (adapted from [959]).
of the elastica. This equivalence is the two-dimensional version of the Kirchhoff

analogy between Lagrange tops and rods at equilibrium and was already known by
Euler [350]. Since all the solutions of the pendulum are known in terms of elliptic
functions, all the solutions of the planar elastica can be obtained systematically [959]
as illustrated in Figure 5.11.
Chapter 6
Morphoelastic Rods
We can now adapt the theory of elastic rods presented in the previous chapter to
include the contribution of growth and remodeling in filamentary structures. Growth
takes place on time scales much longer than any time scales associated with elastic
propagation or viscous relaxation. Indeed, elastic wave propagation takes place at
the speed of sound, whereas growth time scales are of the order of minutes, hours,
and days. Growth is therefore slow enough so that it is uncoupled with inertial terms
and it can therefore be considered as a quasi-static process. Different attempts have
been made to model growing rods. Essentially, one can distinguish three different
approaches.
In the first approach, which we refer to as parameter variation, we consider fam-
ilies of rod solutions parameterized by one of the material parameters. For instance,
in the growth of a tree, one may consider the length and width as two parameters
that evolve in time. At each growth time, the value of such parameters is increased
and a new static solution is recomputed [470].
The second approach is remodeling. The idea is to consider a separate evolution
law for the material parameters that may depend on both time and on the history of
the material. For instance, through differential growth, an initially straight rod can
develop intrinsic curvature and the resulting equilibrium configurations for given
boundary conditions can be studied [451, 853]. This is fundamentally different from

126 6 Morphoelastic Rods
the previous approach since the material parameters are now coupled with the evo-
lution of the rod in space and time.
The third approach, morphoelasticity, combines all aspects of growth and remod-
eling by considering the evolution of the reference configuration separately from
the mechanical response of the rod. In the general theory of morphoelasticity, a
distinction between growth and remodeling is usually made. Growth is associated
with a change of volume of the unstressed elastic material, whereas remodeling
refers to a change of material properties. This distinction is clear in the general
three-dimensional theory of growing elastic bodies, presented in Part IV. However,
in the theory of rods, growth and remodeling are irremediably intertwined. Indeed,
rod theory is a reduction from three-dimensional elastic bodies to one-dimensional
structures that mixes geometry and material parameters. For instance, the bending
stiffness, K 1 = E I1 in Equation (5.48), is a product of a material coefficient, the
Young’s modulus, E and a geometric property, I1 , the second moment of area. As
the rod grows and remodels, the material can become stiffer or increase its radius,
both changing the stiffness K 1 .
Rods can also grow in length. Again, the effect of this type of growth is not obvious
as a uniform change of length in a rod can easily be rescaled without affecting the basic
equations. Therefore, the effect of a change of length in a rod can only be appreciated
through nonhomogeneous or boundary effects arising from the interactions of the
rod with its environment, either through boundary conditions at each end of the
rod or through body forces and couples coupling the rod with its surrounding. The
special case where a rod interacts with itself, nonlocally, through periodic boundary
conditions can also lead to stressed configurations even in the absence of external
forces or couples.
In order to describe interesting aspects of growing rods in space, we first define
the mechanics of a single growing rod through the evolution of its reference configu-
ration. Second, we consider the case of a growing rod attached to a given rigid curve
in space via an elastic force. For instance, one may consider the case of a growing
rod either attached to a substrate, growing inside a sphere, or lying on a cylinder.
Finally, we can consider the case of two rods growing at different rates while attached
together [786].
For the rest of this chapter, our basic assumptions are:
1. The growth dynamics is slow enough so that the unstressed rod is in elastic
equilibrium at all times. We use the variable t to denote this slow time evolution.
This assumption implies that the growth dynamics is decoupled from the elastic
dynamic of the rod. We assume the existence of a dissipative process that takes
the rod back to a stable equilibrium and we consider the effects of inertial terms
to assess the stability of such equilibria.
2. The rod is assumed to be hyperelastic and characterized by a quadratic strain-
energy function W . Due to remodeling and growth, the parameters defining the
intrinsic shape û and the coefficients of W are also functions of the growth time t.
6.1 Kinematics of a Growing Rod 127
6.1 Kinematics of a Growing Rod
Based on the general approach of growth in nonlinear elasticity through a multiplica-

tive decomposition [80, 1091], we consider three different configurations for the rod.
The initial configuration B0 is the unstressed configuration of the rod at time t = 0;
all quantities in this state are denoted by a subscript 0. The virtual configuration V
is the unstressed configuration at a given time t and the current configuration B is
the actual configuration of the rod at time t for given external loads, body loads, and
boundary conditions. Note that the unstressed configuration may not be realizable
in Euclidean space. For instance, given a ring of radius one and intrinsic curvature
one, an increase in its length would create a stress-free configuration that would be
self-penetrating, as shown in Figure 6.1. However, unlike the three-dimensional case
studied in Chapter 12, there is no problem with local compatibility or generation
of residual stress associated with the local definition of a growth and elastic tensor
[468]. Therefore, all local quantities are well defined and the virtual configuration is
appropriate for the computation of the current configuration.
At time t = 0, the rod is described by its intrinsic curvature û0 defined with respect
to its arc length S0 , its total length L 0 , density ρ0 (S0 ), cross-section area A0 (S0 ), and a
stiffness matrix K0 . This intrinsic curvature evolves so that at any given time t the rod
has unstressed curvature û = û(S0 , t), with arc length S, total length L(t), density
ρ(S0 , t), cross-section area A(S0 , t) and a stiffness matrix K(t). In this description,
S0 is a material parameter for all other configurations. Following the discussion of
Section 4.4, it is related to the arc length S of the virtual configuration V by the
growth stretch γ, i.e.
∂S
γ(S0 , t) = , (6.1)
∂ S0
B0 B
Figure 6.1 The three configurations of a morphoelastic rod: initial B0 , reference V , and current
B. The virtual configuration is stress free. It evolves in time and reflects the change due to growth.
Initially, the rod has intrinsic curvature. As growth proceeds the cross section and length change
inhomogeneously. The virtual reference configuration can also intersect itself as it does not represent
a possible physical realization of the rod. The current configuration is the actual configuration with
imposed boundary conditions and body loads.
so that γ characterizes the local increase of length of a material segment located at a

material point S0 at time t. This virtual configuration is required in order to compute
the current shape of the rod at a later time t > 0 for given loads and boundary
conditions. This decomposition is illustrated in Figure 6.1.
6.2 Mechanics of a Growing Rod
At time t, the balance of linear and angular momenta in the virtual configuration
yields
∂n
+ f = 0, (6.2)
∂S
∂m ∂r
+ × n + l = 0, (6.3)
∂S ∂S
where f and l are the body force and couple per unit reference length. The reference
variable S is the natural choice to express all quantities as the constitutive equations
are given in this configuration. This equation together with the appropriate boundary
conditions and with one of the two constitutive models proposed (either (5.47) for
inextensible rods or (5.52) for extensible rods), can be solved to obtain the current
configuration.
We define α to be the elastic stretch and λ the total stretch between the initial and
current configuration. We have
∂s ∂s ∂ S
λ = αγ ⇐⇒ = . (6.4)
∂ S0 ∂ S ∂ S0
A change of independent variable leads to an alternative formulation in the current

and initial configuration. Namely, in the current configuration we have
∂n
+ α−1 f = 0, (6.5)
∂s
∂m ∂r
+ × n + α−1 l = 0, (6.6)
∂s ∂s
where α−1 f and α−1 l are the body force and couple per unit current length. Finally,
in the initial configuration, we have
∂n
+ γf = 0, (6.7)
∂ S0
∂m ∂r
+ × n + γl = 0, (6.8)
∂ S0 ∂ S0
where γf and γl are the body force and couple per unit initial length.
6.3 Evolution Laws for Growing Rods 129
6.3 Evolution Laws for Growing Rods
Following the slow-growth assumption, the rod is characterized by one of the three
constitutive models described in Section 5.3. Since both reference arc length and
stretch can change at time t, great care must be exercised to ensure that these changes
in length are taken into account correctly. To avoid confusion, we use the simple prin-
ciple of first defining all material properties in the virtual reference configuration B
with respect to the arc length S. All other formulations of the rod equations follow
directly from a change of dependent variables as shown above. Note that the parame-
ters entering the constitutive equations, such as stiffnesses and intrinsic curvatures,
may also evolve with the slow time t.
We now consider a possible law for the evolution of intrinsic curvature. We first
consider the case of an inextensible rod. In this case, the only material and structural
parameters evolving in time are the radius, density, stiffness matrix K, and intrinsic
Darboux vector. Whereas, radii, density, and stiffnesses are quantities whose evolu-
tion must be prescribed from biological laws, the intrinsic curvatures û depend both
on intrinsic quantities and on the geometry of the deformation. To understand this
dependence, we adapt the reasoning of [991] to the present situation.
First, we consider an inner rod with uniform intrinsic curvatures ûin and stiffness
matrix Kin . Next, we apply a moment so that the new shape has uniform curvatures
ûout . In this new configuration, we add an external outer cylindrical layer with the
current curvature and with bending stiffness Kout . Ignoring the possibility of internal
residual stresses in the material and treating this new structure as an elastic rod, the
problem is to determine its new properties: its curvature û and stiffness matrix K.
For any given uniform shape given by u, the total moment of the composite rod is
K(u − û), which is also equal to the sum of both moments due to the inner and outer
rods. Hence, we have
Kin (u − ûin ) + Kout (u − ûout ) = K(u − û). (6.9)
This relation must be valid for all u so that we have both
K = Kin + Kout , (6.10)
and

û = K−1 Kin ûin + Kout ûout . (6.11)
It is now possible to generalize this procedure by continuously adding new layers.

Consider a deformed rod and assume that at time t it has curvatures u(t), and material
properties û(t) and K(t). At time t + t, a thin new layer is applied to the rod. The
balance of moment now reads
K(t)[u(t + t) − û(t) − k 2 (t)t] + K[u(t + t) − u(t) − k 1 (t)]

= K(t + t)[u(t + t) − û(t + t)], (6.12)
where K = K̇t is the bending stiffness of the new layer, k 1 (t) is the difference
between the curvatures of the new layer and the curvatures of the inner rod, and k 2 (t)
is a source term for the intrinsic curvature of the inner structure. By expanding all
quantities to order t, we obtain
K̇(u − û + k 1 ) = K(û˙ − k 2 ). (6.13)
If there is no increase of mass or change of stiffness in the section, we have K̇ = 0

and the evolution of the curvatures is simply governed by û˙ = k 2 and a choice for
k 2 must be made. For instance, in the plane, the only nonzero component of the
curvatures is the Frenet curvature κ = u2 . If we assume that the intrinsic curvature
relaxes to the current curvature with a relaxation time η −1 , we have
κ̂˙ = η(κ − κ̂). (6.14)
The implication of this law and its application to the remodeling of vines is explored
next.
6.4 Example: The Remodeling of Stems
In many growth processes, the evolution of a structure directly influences the evolu-
tion of the material parameters. For instance, the branch of a tree can be trained to
grow in a certain shape, which makes for beautiful alleys in French formal gardens
such as in the Jardin des Plantes in Paris. At first, the elastic structure is loaded and
stressed into a particular shape. As time passes, the structure remodels itself in such
a way as to relieve the stresses, and the structure permanently sets in, even in the
absence of loading.
An example of such a process is shown in Figure 6.2 where an initially straight
tendril coils and eventually adopts the coiled shape. The cut shape differs little from
the intact one, showing that the tendril actually passes slowly through a continuum
of quasi-equilibrium states to reach its current steady state. For a morphoelastic
filament, this property can be modeled by considering an evolution of the rod’s
material properties. In the simplest case, the intrinsic curvatures evolve slowly in
time to match the current configuration, hence releasing all stresses.
As a simple example, consider the case of an unshearable, inextensible planar rod
under end compression [451]. As the rod buckles it takes a new shape. At this point it
is assumed that the natural state of the rod will evolve toward this equilibrium shape,
that is, its intrinsic curvature κ̂ = û 2 evolves in time toward its actual curvature.
Following the previous discussion, we use a simple morphoelastic model for the
6.4 Example: The Remodeling of Stems 131
Figure 6.2 A tendril of the genus Passiflora (passion flowers). An intact tendril (a) with perversions
(i, ii) maintains its shape when its connection to the plant is cut (b, arrow). Scale bar is 1 cm
(reproduced from [451], picture courtesy of Ray Goldstein).
relaxation of curvature:
∂ κ̂
= η(κ − κ̂), (6.15)
∂t
where 1/η is a characteristic time for relaxation. It is important to notice here that,
after buckling, the curvature changes at all points and when the intrinsic curvature
relaxes it takes different values at all points. The process depends on the loading
and the parameters, but also on the history of the loading process which makes it
fundamentally different from previous modeling through parameter variation.
To emphasize this point, consider Figure 6.3 where an initially straight rod was
loaded with a given ramp and allowed to relax following (6.15). If the thrusting force
is gradually removed at some time following a ramp, then the filament shape will
soon adapt to the preferred curvature and it will then reach the minimum of the elastic
energy, with κ(s) = κ̂(s) everywhere.
Figure 6.3 shows five sequences of trained shapes which result from a protocol
of the force consisting of a ramp followed by exponential relaxation (with varying
decay rates), where the total thrust vanishes. As the force is removed ever more
slowly, the filament shape remains highly bent. Each final trained shape is typically
very close to that of an elastic filament subjected to some particular external force,
but of course the latter would not remain in place after the removal of that force.
A. B.
(v)
(iii)
(ii)
(i)
Figure 6.3 Filaments trained with time-dependent forces. A. For each of the force-time profiles
shown, the corresponding sequence of filament shapes is shown at intervals of t = 0.01 in (B).
The dashed line indicates the end of the linear ramp. As can be clearly seen in picture (i), the
filament has remodeled into its shape and retains it after the load is removed. With longer unloading
times ((ii) to (v)), the filament when unloaded, partially relaxes to its original shape (reproduced
from [451]).
6.5 A Buckling Criterion
An interesting aspect of growing rods is the possibility of growth-induced buckling

instabilities. We first derive a general formulation to obtain a bifurcation criterion for
a rod with arbitrary body force and couple. We assume here that the rod has constant
intrinsic curvature û. Our starting point is the full kinematic and mechanics descrip-
tion for an unshearable rod in the steady state, expressed in the virtual configuration
as
r = αd3 , (6.16)
d = u × di , i = 1, 2, 3, (6.17)
n + f = 0, (6.18)
∂r
m + × n + l = 0, (6.19)
∂S
m = K 1 (u1 − û1 )d1 + K 2 (u2 − û2 )d2 + K 3 (u3 − û3 )d3 , (6.20)
n3 = E A(α − 1), (6.21)
where (·) = ∂(·)/∂ S.

6.5 A Buckling Criterion 133
We assume the existence of a one-parameter family of solutions. This base solution

is defined by {α(0) , r(0) , n(0) , u(0) , d1(0) , d2(0) , d3(0) }. Viewed as a bifurcation problem,
the general idea is to consider a small perturbation to this solution and determine the
critical value of the control parameter at which a different solution to the perturbed
system exists. In the case of growing rods, a natural choice for the growth parameter
is γ. The question is then to find the critical value of the parameter γ = γ ∗ such that
the Kirchhoff equations admit a solution that is different from the base solution.
6.5.1 Example: Michell’s Instability
Before giving the general formulation for the bifurcation of a rod, we consider the
simple case of a twisted elastic ring as depicted in Figure 6.4. This problem was first
stated by Thomson and Tait in their classic, “Treatise on Natural Philosophy” [1229].
In paragraph 123, they discuss the problem of the respective stability of the circle
versus the eight-form shown in Figures 6.4 and 6.5 and state that, “the circular form,
which is always a figure of free equilibrium, may be stable or unstable, according
as the ratio of torsional to flexural rigidity is more or less than a certain value
depending on the actual degree of twist.” Motivated by this statement, John Henry
Michell solved the problem in 1889 by determining the exact value of the twist
necessary to trigger an instability in a ring [462, 885]. Michell writes: “If a wire of
isotropic section and naturally straight be twisted, and the ends joined so as to form
a continuous curve, the circle will be a stable form of equilibrium for less than a
certain amount of twist.”
Reformulated in our terminology, the problem is to assess the stability of an inex-
tensible, transversely isotropic, elastic ring with no intrinsic curvature. The system
can be easily realized in an experiment: Take a straight uniform elastic rod with
circular cross section, paint a straight line on it and shape the rod so that the rod’s
centerline is a circle. At the junction, the tangent from the two ends agrees, but the
cross sections can be rotated so that the line painted on the straight intrinsic shape
Figure 6.4 Untwisted (left) and twisted rings (right). The total twist is given by the total angular
rotation of an arrow pointing at any circle on the surface of the untwisted ring. Here the total twist
is 13π/2.
Figure 6.5 Buckling of an unstable ring when the critical twist is first reached. The twist is not
shown on the ring. The solution shown is obtained as an asymptotic solution close to the bifurcation
point [478].
twists around the central curve. The twist is the total angular rotation of the line with
respect to the central curve. The line will close on itself at the junction if the twist is
an integral multiple of 2π. After closure, the rod is glued and released.
For small values of the twist, the twisted ring is observed to be stable. For suffi-
ciently high twist, the elastic ring will become unstable and will start writhing out
of the plane. The phenomenon is quite striking as the instability is subcritical; the
ring suddenly buckles and loops back on itself by forming an eight-shape with self-
contact, as shown in Figure 6.5. We refer to the twisted elastic ring instability as
Michell’s instability and the problem is to identify the value of the critical twist at
which the instability sets in.
For an inextensible rod α = 1, (6.21) is neglected, and S is the arc length for all
deformations. In the absence of body force or couple f = l = 0 and for naturally
straight rods with transversely isotropic cross sections, (6.16)–(6.20) take the simple
form
n = 0, (6.22)
m + r × n = 0, (6.23)
m = K 1 κβ + K 3 ωτ , (6.24)
where, as defined in Section 5.1, {τ , ν, β} are, respectively, the tangent, normal, and
binormal vectors to r, and τ , κ are the Frenet torsion and curvature.
The twist ω describes the rotation of the material frame. By writing Equa-
tions (6.22–6.24) in the Frenet frame, the resultant force can be eliminated in terms
of κ, τ , ω to obtain
ω = 0, (6.25)
κτ − ωκ = 0, (6.26)
κ2 (ω − 2τ ) τ + κ (κ3 − κ ) + κκ = 0, (6.27)
where = K 3 /K 1 . The first equation implies that the twist density ω must be con-
stant. To identify bifurcation points, where the ring may lose stability with respect
to out-of-plane and torsional deformations, we consider small variations of the cur-
vature and torsion
κ = κ(0) + κ(1) , τ = τ (0) + τ (1) , (6.28)
and expand Equations (6.26)–(6.27) to first order in .

The base solution is simply κ(0) = 1/R, τ (0) = 0 and ω arbitrary, where R is the
radius of the ring. This twisted ring solution is a one-parameter family of solutions
parameterized by ω and is a particular case of the twisted helical equilibria discussed
in Section 5.7. After simplification, to first order in , a single equation for κ(1) = k(S)
is obtained
k + R −2 + 2 ω 2 k = 0. (6.29)
This equation supports periodic solutions of the form k = exp(inS/R) with n integer,
such that (6.29) becomes
2 ω 2 R 2 = n 2 − 1. (6.30)
The first nontrivial solution occurs for n = 2 and corresponds to

√
3
ωc = . (6.31)
R
The total twist T w is the integral of the twist density ω over the circumference. It
corresponds to the total angular rotation of the cross section along the rod; that is,
T w = 2π Rω. The twisted ring becomes unstable for values of the total twist larger
than √
2π 3
T wc = . (6.32)

For most materials, typical values for lie between 2/3, in the incompressible limit,
to 1 with most metals around 4/5. Other filaments such as DNA may present higher
values of twist to bending rigidity [1125]. Therefore, a ring becomes unstable when
it has been twisted by about two full turns (with limits 1.73 to 2.6 corresponding to
α between 2/3 and 1).
The problem of twisted elastic rings is important in the study of macromolecules

such as DNA molecules [83, 84, 826, 1284] but also plants [477, 1156] and microbial
filaments [452]. In particular, the analysis of mini-DNA rings made out of a few
hundred bases offers a unique perspective to characterize physical properties of
DNA [537, 1013, 1051]. Twisted elastic rings are the natural theory to understand
and extract these properties [818, 985].
6.5.2 A General Perturbation Expansion
We can now generalize the linearization process for the general Kirchhoff equations
assuming the existence of a one-parameter family of base solutions {r(0) , n(0) , u(0) ,
d1(0) ,d2(0) ,d3(0) } with elastic stretch α(0) . Let be a small parameter. Based on the
general ideas outlined in [472, 474, 475, 478], we begin by expanding the basis
itself,
di = di(0) + di(1) + O(2 ). (6.33)
If we express the correction terms di(1) in terms of the unperturbed basis, and require
orthonormality to O(), we find that the O() corrections can be expressed in terms
of a single vector
c = c1 d1(0) + c2 d2(0) + c3 d3(0) , (6.34)
such that
di = di(0) + c × di(0) + O(2 ). (6.35)
We now expand all quantities, such that the first-order perturbation of all vector
quantities is expressed in the frame {di(0) , i = 1, 2, 3}. For any scalar a and vector
v = v1 d1 + v2 d2 + v3 d3 , (6.36)
we consider the following expansion in :
a = a (0) + a (1) + O(2 ), (6.37)

3
v= (vi(0) + vi(1) )di(0) + O(2 ). (6.38)
i=1
Note the following relationship: If vi = vi(0) + vi(1) , we have vi(0) = vi(0) and vi(1) =

vi(1) − (c × v(0) )i , where v(0) = i vi(0) di(0) .
We expand r, n, and u per (6.38), and α = α(0) + α(1) + O(2 ). We assume that
the body force f and body couple l can be expanded similarly. By assumption, the
base solution solves the Kirchhoff equations. We compute the linearized equations
by expanding all equations and variables to order . Doing so, we obtain a linear sys-
tem for the independent variables {r1(1) , r2(1) , r3(1) , n (1) (1) (1) (1) (1) (1)
1 , n 2 , n 3 , u 1 , u 2 , u 3 , c1 ,
c2 , c3 , α(1) }.
The kinematic equation ∂r/∂ S = αd3 is
(r1(1) ) + u (0) (1) (0) (1) (0)

2 r 3 − u 3 r 2 = α c2 , (6.39)
(r2(1) ) + u (0)
3 r1
(1)
− u (0)
1 r3
(1)
= −α c1 , (0)
(6.40)
(r3(1) ) + u (0)
1 r2
(1)
− u (0)
2 r1
(1)
=α . (1)
(6.41)
The frame equations ∂di /∂ S = u × di is
c1 = u (1) (0) (0)

1 + c3 u 2 − c2 u 3 , (6.42)
c2 = u (1)
2 + c1 u (0)
3 − c3 u (0)
1 , (6.43)
(1)
c3 = u3 + c2 u (0)
1 − c1 u (0)
2 . (6.44)
The force balance n + f = 0 to first order is given by
(n (1) (0) (1) (0) (1) (1)

1 ) = u 3 n2 − u 2 n3 − f1 , (6.45)
(n (1)
2 )

= u (0)
1 n3
(1)
− u (0)
3 n1
(1)
− f 2(1) , (6.46)
(n (1)
3 )

= u (0)
2 n1
(1)
− u (0)
1 n2
(1)
− f 3(1) , (6.47)
and, similarly, the moment balance reads
(m (1) (0) (1) (0) (1) (1) (1) (0) (0) (0) (1)
1 ) = u 3 m 2 − u 2 m 3 − l 1 − α n 1 + α (c2 n 3 + n 2 ),
(6.48)
(m (1)
2 )

= u (0)
1 m3
(1)
− u (0)
3 m1
(1)
− l2(1) + α(1) n (0)
2 +α (0)
(c1 n (0)
3 − n (1)
1 ),
(6.49)
(m (1)
3 )

= u (0)
2 m1
(1)
− u (0)
1 m2
(1)
− l3(1) −α (0)
(c1 n (0)
1 + c2 n (0)
2 ), (6.50)
where the first-order moment is given by the constitutive relationship also expanded
to first order:
⎛ ⎞
3

m i(1) = K i u i(1) + ⎝c × K j (u (0) (0)
j − û j ) − u j d(0)
j
⎠ , i = 1, 2, 3.
j=1
i
(6.51)
In the case of an inextensible rod, α(0) = 1, α(1) = 0 and there is no constitutive law
for n3 . For an extensible rod, the constitutive equation (6.21) gives
n (1) (0) (0) (1)

3 = c2 n 1 − c1 n 2 + E Aα . (6.52)
6.5.3 Bifurcation Criterion for Elastic Buckling
To obtain a bifurcation criterion, we define the nine-dimensional vector
μ = {r1(1) , r2(1) , r3(1) , n (1) (1) (1)

1 , n 2 , n 3 , c1 , c2 , c3 }. (6.53)
Note that we can eliminate u i(1) immediately using (6.42–6.44), and α(1) using (6.52).
The system can be written in compact matrix form
M2 μ + M1 μ + M0 μ = 0, (6.54)
where each Mi is a 9 × 9 matrix whose coefficients are functions of the base solution.
If the family of equilibria can be written in a basis where it is constant in S,
these matrices are also constant. In that case, we can look for a solution of the
form μ = ξeinS + ξ̄e−inS . Inserting this form into (6.54), the system is satisfied if
det(M) = 0, where
M = −n 2 M2 + inM1 + M0 . (6.55)
For problems related to uniform growth, it is convenient to consider γ as a control

parameter. Assuming that the base solution exists for a range of values of γ, the
problem is to determine the critical growth γ ∗ at which a bifurcation takes place,
that is, the values of γ at which the determinant vanishes. The corresponding value
of n determines the mode or wavelength of the bifurcation mode, depending on the
particular geometry.
6.5.4 Example: The Growing Ring
A classic problem in the theory of elastic rods is the buckling of a ring to a noncircular
shape as discussed in Section 6.5.1 for the twisted ring. Similarly, the problem of the
stability of a ring with intrinsic curvature under a variety of loadings and material
parameters has been considered by many authors [462, 527, 825, 1150, 1237, 1238].
The case of multicovered rings, that is, rings that wrap around multiple times on
themselves has also been studied in detail [474, 825, 827].
Here, we use this simple configuration to demonstrate the effect of growth on
the buckling of rods. We further consider the role of remodeling and of an elastic
foundation in the next sections. We assume that the ring is a constitutively isotropic,
inextensible rod of initial and unstressed radius 1, with cross-section radius a. We
assume that it grows linearly in time so that, without loss of generality, γ = 1 + t,
while all other material properties are unchanged. In this case, we have û = (0, 1, 0),
L = 2πγ.
First, we study the buckling of the growing ring in the absence of external forces.
In the absence of remodeling, the buckling is strictly determined by the critical
parameter γ. For this system there exist, for all γ ≥ 1, circular solutions character-
ized by a curvature u2 = 1/γ and vanishing force n = 0. These solutions are main-
tained by a bending moment m = E I (u2 − 1) generated by the periodic boundary
conditions. The ring is initially situated in the x–y plane with general frame
d1(0) = − cos(S/γ)ex − sin(S/γ)e y , (6.56)

d2(0) = ez , (6.57)
d3(0) = − sin(S/γ)ex + cos(S/γ)e y . (6.58)
Then, the only nonvanishing components of the base solution are
r1(0) = −γ, u (0)

2 = 1/γ. (6.59)
Inserting this into (6.39–6.50), the determinant condition can be solved explicitly for
γ, and is found to be

1−+ 1 + (4n 2 − 2) + 2
γ(n) = , (6.60)
2
where is given by (5.57) and the buckling mode n must be an integer to satisfy the
periodic boundary conditions of the closed ring. The critical growth is the smallest
possible value of γ > 1, and it occurs for n = 2.
The first-order corrections for the centerline are shown in Figure 6.6. Equation
(6.60) provides the critical point at which the ring buckles. After buckling, a full
numerical solution for the post-buckled ring can be obtained by solving the full
system via a shooting method [942] and parameter continuation. This solution is
shown in Figure 6.7. Note that self-contact is not taken into account. Nevertheless,
the ring follows an interesting deformation path. By γ ≈ 2.19, the closed rod has
collapsed back into a planar ring, tripling on itself. As γ continues to increase, this
tripled ring grows, so that at γ = 3, it forms a tripled ring of radius 1. At this point,
the curvature everywhere matches the intrinsic curvature; the ring is back to a stable
stress-free state.
The tripled covered ring will become unstable if we further increase γ. The critical
value of γ where it becomes unstable can be found by realizing that the buckling
mode n in (6.60) is now an integer multiple of 1/3, since the rod has tripled its length.
Figure 6.6 The shape of a growing ring after buckling, computed from a linear stability analysis.
Figure 6.7 The shape of a growing ring after buckling, computed from a numerical analysis
( = 0.7) [917]. The parameter γ denotes the growth stretch.
For = 0.7, the second bifurcation occurs at γ ≈ 3.83, with mode 4/3. This solution
is shown in Figure 6.8 where the out-of-plane amplitude is plotted as a function of
the growth parameter γ.
6.5.5 A Growing Ring with Remodeling
We look again at a growing ring of initial unstressed radius 1. We assume that it

grows linearly in time so that γ = 1 + t, where t is the growth time. During growth,
the intrinsic curvature also evolves to the current curvature following the law (6.14)
in such a way that if growth is interrupted, the ring evolves to an intrinsic state with
0.8
analytical
numerical
0.6
amplitude of
0.4
0.2
1.0 1.5 2.0 2.5 3.0 3.5 4.0
Figure 6.8 Comparison between the out-of-plane amplitude of the ring obtained from the numerical
simulation (dashed) for = 0.7 and the asymptotic value predicted by an asymptotic model (solid)
given in [917].
a characteristic time 1/η. As long as the central axis of the rod remains circular, the
curvature is, as before, u2 = 1/(1 + t), which can be used directly in (6.14), giving

∂ û2 1
=η − û2 . (6.61)
∂t 1+t
The solution of this last equation is

û2 (t) = e−η(1+t) eη + ηEi(η(1 + t)) − ηEi(η) , (6.62)
where Ei(·) denotes the exponential integral function. We now have a competition
between growth and the remodeling of the intrinsic curvature and the ring may
become unstable as shown in Figure 6.9. A bifurcation happens when
√
u2 −1+ 2 + 14 + 1
≤ c2 ≡ , (6.63)
û2 8
at which time the solution does not remain planar and the expression for the intrinsic
curvature (6.62) ceases to be valid. In this situation where we have two dynamical
processes, the history of the deformation has to be taken into account. A similar situ-
ation for the Euler buckling of a straight rod and the twisted ring has been discussed
in [451, 1337].
u2 / ^
1.0
u2 η>ηc
0.8
η=ηc
0.6
0.4
η<ηc
0.2
0.0
t1 t
0 1 2 3 4 5
Figure 6.9 Stability of a growing ring with evolving intrinsic curvature ( = 3/4). The condition
for buckling is c2 ≈ 0.54, for η < ηc ≈ 0.57, whereby the instability occurs at time t1 . For η > ηc ,
the intrinsic curvature evolves sufficiently fast so as to remove stresses in the ring and to prevent an
instability developing (upper curve: η = 2, lower curve η = 0.1).
6.6 Growing Rods on a Rigid Foundation
Next, we consider a growing rod attached to a rigid foundation via an elastic force, and
explore the effect of the body force f and body couple l generated by the attachment.
We assume that the foundation is a rigid space curve to which the rod is attached
via a restoring body force depending on the distance from the rod to the curve. This
attachment induces a body force and body couple acting on the rod. The position
of the rigid curve rA (s) is specified with respect to the rod by a pair of functions
φ : [0, l] → [0, 2π[ and A : [0, l] → [0, l]ˆ where l and lˆ are the length of the rod
and the rigid curve, respectively. The situation is depicted in Figure 6.10.
Figure 6.10 A rod is

attached on a rigid
foundation ρ by an elastic
force.
6.6 Growing Rods on a Rigid Foundation 143
The function φ(s) is the angle from the d1 axis at which the attachment occurs.
Defining a(s) to be the perpendicular distance from the centerline to the attachment
point on the surface of the rod, the point rA (s) of attachment is
rA = r + a [cos φ d1 + sin φ d2 ] . (6.64)
The function A(s) defines the location on the foundation to which a point with
material coordinate s is attached, i.e., a map between the centerline of the rod and the
foundation curve ρ(ŝ). The attachment map A(s) is a one-to-one map from the rod
to the foundations such that ŝ = A(s), A(0) = 0. That is, the point rA (s) is attached
to the point ρ(A(s)).
The two functions φ and A, as well as the foundation curve ρ, completely define
the geometry of attachment. To describe the mechanical force connecting the curve
to the rod, we introduce
E = ρ(A(s)) − rA (s), (6.65)
and assume that the body force generated by the rigid foundation is along that direc-
tion, and only depends on the distance relative to a rest length q(s),
E
f(s) = f (|E| − q(s)) , (6.66)
|E|
where q(s) is the rest length of the spring, so that f (0) = 0 when |E| = q. We further
assume that f derives from a convex potential for which the origin is a minimum.
The particular case where f is linear corresponds to the classic case of a Winkler
foundation with a Hookean spring [100, 326, 983].
Since the attachment force acts on the rod, it also generates a body couple given
by
l = (rA − r) × f. (6.67)
6.6.1 Example: A Growing Ring on a Foundation
We consider the case in which the inside edge of a growing rod of initial radius 1
is attached to a ring of radius (1 − a), located in the x–y plane and centered at the
origin. As the ring grows, the foundation exerts an increasing force directed toward
the center of the ring, as illustrated in Figure 6.11. To compute the circular planar
base solution, we consider a ring in the x–y plane with a frame given by (6.56–6.58).
The foundation is attached at the point rA = r + ad1 , where a is the cross-section
radius of the rod, and the foundation force is assumed to be linear:
foundation
Figure 6.11 Circular solutions for the growing ring with foundation on inside edge, viewed from
above. The intrinsic curvature is û2 = 1.
f = k(ρ − rA ), (6.68)
where k is the foundation stiffness.

Since ρ = −(1 − a)d1(0) , the circular base solution is given by:
r1(0) = −γ, r2(0) = r3(0) = 0, (6.69)

f 1(0) = k(γ − 1), f 2(0) = f 3(0) = 0, l (0)
= 0, (6.70)
u (0)
2 = 1/γ, (6.71)
(0)
n 3 = −kγ(γ − 1). (6.72)
To compute the critical value of γ, the body force (6.68) and couple (6.67) are
expanded to O(), yielding
f 1(1) = −kr1(1) , f 2(1) = −k(r2(1) + ac3 ), f 3(1) = −k(r3(1) − ac2 ), (6.73)

l1(1) = 0, l2(1) = −a( f 3(1) + c2 f 1(0) ), l3(1) = a( f 2(1) − c3 f 1(0) ). (6.74)
Following Section 6.5, a determinant condition is formed for the critical growth
γ ∗ . The result, and the influence of the foundation on buckling, is illustrated in
Figure 6.12, in which γ ∗ is plotted against the log of the foundation stiffness k for
modes n = 2, 3, 4.
For small k, the solution is similar to the case of no foundation: The ring buckles
at mode 2, and as k → 0, the buckling criterion (6.60) is recovered. As the stiffness
increases, the critical growth decreases for increasing buckling mode n and the ring
buckles earlier. As the foundation force strengthens, it becomes energetically favor-
Figure 6.12 Critical buckling growth against the log of the foundation stiffness k, for modes 2,
3, and 4. As the stiffness increases, buckling occurs after less growth, and at higher modes ( = 1
and a = 0.07).
able to buckle at higher modes. These solutions have higher bending energy, but they
reduce the foundation energy by remaining closer to the foundation.
6.6.2 Example: A Straight Rod Growing on a Foundation
Next, we consider the buckling of a growing planar rod on a foundation in the absence
of body couples, as shown in Figure 6.13. This scenario is similar to the classic
problem first studied by Winkler in 1867 [395, 983, 1287, 1328]. It is instructive
to see how buckling emerges within our framework. The rod is naturally straight,
initially planar, and constrained in the plane so that the Darboux vector is u =
(0, ακ, 0), where κ is the Frenet curvature and α = ∂ S s is the elastic stretch. It is
convenient to use the planar representation of a rod given in Section 5.8. The rod lies
in the x–y plane and is parameterized by the angle θ between the tangent vector and
the x-axis:
τ = d3 = cos θex + sin θe y , (6.75)
which implies
∂θ ∂θ
κ= = α−1 , (6.76)
∂s ∂S
Figure 6.13 Setup of the buckling of a straight planar rod on a foundation. A rod is attached to
a foundation and is allowed to deform only in the plane. An increase in length leads to a buckling
instability.
and d2 = ez . By writing n = Fex + Ge y , f = f ex + ge y , r = xex + ye y , we can

simplify the equations of motion (6.2–6.3) to a system of five equations:
∂x ∂y
= α cos θ, = α sin θ, (6.77)
∂S ∂S
∂F ∂G
+ f = 0, + g = 0, (6.78)
∂S ∂S
∂ θ
2
E I 2 + αG cos θ − αF sin θ = 0. (6.79)
∂S
These equations are supplemented by the constitutive law for the foundation, given
below, and a constitutive law for tension
F cos θ + G sin θ = E A(α − 1), (6.80)
where A is the cross-sectional area as before. We use this last relationship to express
α in terms of F, G, and θ in the equations above.
We consider the case of a clamped uniformly growing rod of initial length L 0 = 1
whose end positions are fixed for all time:
x(0) = 0, x(L) = 1, y(L) = y(0) = y0 , θ(0) = θ(L) = 0, (6.81)
where y0 is the distance between the rod and the rigid foundation, taken to be the
segment of the x-axis between 0 and 1.
Different assumptions about the nature of the attachment between the rod and
the foundation can be made. We consider the case where the rod is initially glued to
the axis: A point (S0 , 0) on the x-axis is attached to a point (S0 , y0 ) on the rod. In
the current configuration, the two points are still connected elastically and are now
located at (S/γ, 0) and (x(S), y(S)). Therefore, the body force acting on the rod
from the foundation is
h()
f= (x − S/γ)ex + (y − y0 )e y , (6.82)
γ
where
we have assumed that the rest length of the foundation is y0 and =
(x − S/γ)2 + (y − y0 )2 is the distance between these two connected points. Note
the factor 1/γ which indicates that the attachment was made in the initial configu-
ration and no subsequent remodeling takes place. The function h() is chosen such
that h(0) = 0 and h (0) = −Ek < 0.
The base solution is a straight compressed rod given by
1−γ
x (0) = S/γ, y (0) = θ(0) = G (0) = 0, F (0) = E A . (6.83)
γ
To find the critical value of γ where a bifurcation first occurs, we expand our
variables as
x = x (0) + x (1) + O(2 ), y = y (0) + y (1) + O(2 ), (6.84)
and, as before, linearize the system around the trivial compressed state. To first order,
it is easy to show that x (1) = F (1) = 0, and the problem is reduced to a set of three
linear equations
dy (1) θ(1) dG (1) y (1)

= , = Ek , (6.85)
dS γ dS γ
d2 θ(1)
EI + E A(γ − 1)θ(1) + γG (1) = 0, (6.86)
dS 2
which can be further reduced to a single fourth-order differential equation for θ(1)
d4 θ(1) d2 θ(1)
+ 2a + b2 θ(1) = 0, (6.87)
dS 4 dS 2
where
A(γ − 1) k
a= , b= , (6.88)
2I γ 2 I γ3
and the boundary conditions, vanishing for all linearized variables, now read
dθ(1) d 3 θ(1)
θ(1) (0) = A(γ − 1) (0) + I γ 2 (0) = 0, (6.89)
dS dS 3
dθ(1) d 3 θ(1)
θ(1) (γ) = A(γ − 1) (γ) + I γ 2 (γ) = 0. (6.90)
dS dS 3
We now look for solutions of (6.87) of the form θ(1) ∼ eiωs , which leads to the four
roots
ω12 = a + a 2 − b2 , ω22 = a − a 2 − b2 . (6.91)
The condition a = b gives the first bifurcation condition for the existence of oscil-
latory modes on an infinite domain. Explicitly, it reads
A2 (γ − 1)2 − 4I kγ = 0, (6.92)
which leads, for a circular cross section of radius r , to the first condition
kr
(r + kr 2 + 4π),
γ1 = 1 + (6.93)
2π
√
associated with a typical mode number n = a; that is, a typical wavelength
1/4
I π 5/4 3/4
= √ γ1 r k −1/4
3/4
ζ = 2π/ω1 = 2πγ1 (6.94)
k 2
1/4
√
π π 1/4 1
=r √ πk −1/4 − k + k 3/4 + O(k 5/4 ). (6.95)
2 4 32
In a finite domain with clamped boundary conditions, there is a delay at the bifurcation
and the value of γ2 > γ1 for which the system is unstable can be found from the
boundary conditions. Assuming a > b, the nontrivial solution, shown in Figure 6.14,
reads
θ(1) = C1 [cos(ω1 S) − cos(ω2 s)] + C2 sin(ω1 S) + C3 sin(ω2 S), (6.96)
Figure 6.14 Buckling of a clamped growing rod on an elastic foundation. The rod is constrained
to lie in the unit interval and clamped at the boundary (with y0 = 1/2). Here k = 1, r = 0.02, and
γ2 = 1.19934. The amplitude is arbitrary and chosen to be C1 = −0.2. The centerline is indicated
by a dashed line.
where
C1 ω1 b(cos(γω1 ) − cos(γω2 ))
C2 = , (6.97)
(2aω1 − ω13 ) sin(γω2 ) + (ω23 − 2aω2 ) sin(γω1 )
C1 ω2 b(cos(γω2 ) − cos(γω1 ))
C3 = , (6.98)
(2aω1 − ω13 ) sin(γω2 ) + (ω23 − 2aω2 ) sin(γω1 )
with the condition
a sin(γω1 ) sin(γω2 ) + b cos(γω1 ) cos(γω2 ) − b = 0. (6.99)
The first positive root of (6.99) is γ1 , but it leads to the trivial solution θ(1) = 0.
Therefore, the critical value of γ is the second root γ2 > γ1 . However for r L 0 ,
the root γ1 provides an excellent first approximation for the critical value γ2 [720].
This linear analysis can be used as a starting point for a nonlinear analysis that
predicts the amplitude of the buckled solution as a function of the load [764].
6.7 Application: Growing Vines
Climbing plants have developed a fascinating array of mechanical strategies to

achieve vertical growth without being able to support themselves. Hookers, lean-
ers, weavers, rooters, stickers, clingers, tendril-bearers, or twiners are just a few
realizations of the thirty different ways vines manage to grow by taking advantage
of their surroundings [869]. The growth process of climbing plants raises many
interesting mechanical questions already noted by nineteenth century plant physiol-
ogists and further studied by Silk, Holbrook, and coworkers [838, 1124, 1155, 1160,
1162]. We now consider two problems in the growth of vines that can be modeled
as morphoelastic rods.
6.7.1 Perversion in Tendrils
6.7.1.1 Background
As a first example, we consider the evolution of tendrils in plants, following

[476, 853]. Among the many different mechanisms climbing plants use to climb
along supports, the so-called “tendril-bearers” constitute an important class, e.g., the
grapevine, the hop, the bean, the melon. A tendril is a modified leaf that can be found
at the extremities of some climbing plants, and these are used to achieve vertical
growth, as shown Figure 6.15. In the first stage of their development, tendrils are
tender, soft, curly, and flexible organs originating from the stem. As they grow, the
Figure 6.15 Tendrils are

modified leaves that are used
by some climbing plants to
attach themselves to a
support [161].
Tendrils
tendrils circumnutate [772, 812]: The tip of the tendril describes large loops in space
by completely rotating on itself until it touches a support, such as a trellis, pole, or
branch. If the circumnutation does not result in contact, the tendrils eventually dry
and fall off the stem, as illustrated in Figure 6.16. The tendrils which are in contact
with a support enter another phase of their development; their tissues develop in such
a way that they start to curl and tighten up, eventually becoming woody, robust, and
tough [111]. This curling provides the plant with an elastic spring-like connection to
the support, which enables it to resist high winds and loads.
Physically, a tendril is a filament under tension. Once a tendril has grasped a
support, it starts to develop curvature by differential growth until it bifurcates into a
shape made out of two or more helical structures with opposite handedness as shown
in Figure 6.17. This shape can also be observed in telephone cords. If one completely
untwists the helical structure of the telephone cord and pulls the ends, a straight cord
can be obtained. Now, if one slowly releases the ends, the filament suddenly changes
to a shape composed of two helical structures with opposite handedness and linked
by a small inversion, as seen in Figure 6.18. Mechanically, this phenomenon is due to
the fact that the original structure has no twist and neither end is allowed to rotate. The
total twist remains zero at all times and the elastic energy is minimized by creating
helices.
The German mathematician J. B. Listing [94, 793] refers to an inversion of chiral-
ity as a perversion. D’Arcy Thompson used the idea to characterize seashells: “the
one is a mirror-image of the other; and the passing from one to the other through
the plane of symmetry (which has no “handedness”) is an operation which Listing
called perversion” [1230, p. 820]. Maxwell also uses the word perversion, in his
“Treatise on Electromagnetism” published in 1881: “They are geometrically alike in
all respects, except that one is the perversion of the other, like its image in a looking
glass” [844]. The usage of the word perversum actually originated in the description
of rare left-handed specimens of seashells from a species of overwhelmingly right-
handed individuals. For instance, there are only six known left-handed specimens out
of the millions of known Cerion, a West Indian land snail [491, 1089]. Accordingly,
6.7 Application: Growing Vines 151
A.
Tendrils
B.
Tendrils
Figure 6.16 Growth and attachment of tendrils. A. In the first stage, the tendrils are circumnutating
until they find an attachment. B. In the second stage, the tendrils are attached and perversion sets
in. Adapted from Darwin [261].
following the terminology introduced in [477], we refer to any filamentary structure

supporting both left- and right-handed helices as a perversion.
The phenomenon of perversion in climbing plants has a long interesting scientific
history. Inversion of helicity in tendrils already appears in the illustration of Linnaei
in “Philosophia Botanica” in 1751 [791] shown in Figure 6.19. However, according to
De Candolle [161, 162], the first record of a scientific observation of perversion goes
back to a letter of the French scientist Ampère to the French Academy of Sciences.
Most major nineteenth century plant physiologists, such as Dutrochet in 1844 [316],
von Mohl in 1852 [897], and Léon in 1858 [782, 783] have described the perversion
found in tendrils.
Tendril
perversion
Right-handed
coils
Left-handed
coils
Figure 6.17 After attachment, the tendril of Bryonia dioica curls on itself to create a tendril
perversion connecting coils of opposite handedness. Drawing adapted from Darwin [261].
It is Charles Darwin, inspired by the related studies of his friend, the American
botanist, Asa Gray [495], who gave the first complete and truly scientific analysis
of the growth of climbing plants in his delightful little book, “The Movements and
Habits of Climbing Plants” [261], based on an essay presented at the Linnean Society
in 1865. In this book, he devotes an entire chapter to the problem of tendril growth
and gives the first qualitative explanation of the phenomenon: “When a tendril has
caught a support and is spirally contracted, there are always as many turns in one
direction as in the other; so that the twisting of the axis in the one direction is exactly
compensated by the twisting in the opposite direction... I cannot resist giving one
other illustration, though superfluous: When a haberdasher winds up ribbon for a
customer, he does not wind it into a single coil; for, if he did, the ribbon would twist
itself as many times as there were coils; but he winds it into a figure of eight on his
thumb and little finger, so that he alternately takes turns in opposite directions, and
thus the ribbon is not twisted. So it is with tendrils, with this sole difference, that
they take several consecutive turns in one direction and then the same number in an
opposite direction; but in both cases the self-twisting is avoided.” This situation is
illustrated in Figure 6.17.
The onset of twisting is due to the development of intrinsic curvature in the
filament. In physical systems, there are various mechanisms for creating intrinsic
curvature in an initially straight unstressed filament. In the case of tendrils, differential
growth takes place between the ventral and dorsal side of the sections as already
explained by de Vries in 1877 and illustrated in Figure 6.20.
Another mechanism is heat setting, as used, for instance, in the textile industry,
where it is known as the false-twist technique [556, 882]. In this process, an initially
Figure 6.18 Example of a telephone cord perversion.

Figure 6.19 Tendril

perversion illustrated by
Linnaeus in “Philosophia
Botanica” [791].
straight filament is heated and shaped as a ring, or helix, and cooled down. This
method is used to roughen yarns and is based on the creation of inversion along the
fibers. Due to the thermoplasticity of many materials, the material conserves its new
shape at lower temperatures. Finally, in a common experiment happening daily in
hair salons, chemical unbonding and rebonding is performed on hair to artificially
create curls. This process is based on the destruction of certain chemical bonds while
setting the filament in the desired shape where the bonds are then recreated by both
chemical and heat setting.
Another striking occurrence of perversion can be found in the structure of human
umbilical cords. As seen in Figure 6.21, umbilical cords are generally made out of two
Figure 6.20 Transverse

section through a tendril
from “Physiological Plant
Anatomy”, Haberlandt, 1884
and based on de Vries
studies (1877). The ventral
side is on top and the dorsal
side at the bottom.
Figure 6.21 The umbilical cord of Zéphyr, the author’s third son. An umbilical cord is composed
of two arteries and a vein winding to form a left-handed triple helical structure.
arteries and one vein embedded in a compact gel-like structure known as Wharton’s
jelly [823]. The arteries are longer than the vein, which is itself longer than the jelly,
and therefore the cord forms a triple helix. The handedness of this helix was already
described in 1521 by Berengarius [322], fascinating medical doctors and scientists
ever since [186].
The umbilical cord is, on average, 50–60 cm long with up to 40 helical turns,
and helicity can be observed as early as 42 days’ gestation [334]. Umbilical cords
can be either left-handed, right-handed, straight or with mixed helicity. Strangely
enough, the ratio of left-handed to right-handed cords is about 7 to 1, the usually
accepted ratio of right-handed to left-handed adults. However, there is no statisti-
cal correlation between cord handedness and hand preference [747]. According to
different statistical studies, inversion from left-handed to right-handed structures in
umbilical cords, as shown in Figure 6.22, varies from 2% to 26% of all cases [186,
461, 1120]. This discrepancy comes from the fact that most studies have not focused
on perverted umbilical cords and have classified some of them as predominantly
right- or left-handed. To date, there is no model for the growth of umbilical cords
that would explain the difference of handedness, its inversion or even the occurrence
of helicity.
Both genetic and mechanical factors seem important as indicated by the correlation
between umbilical cord helicity in monozygotic twins [322] and experiments on the
effect of tension on fetal activity in laboratory rats [61, 894]. In light of the generic
development of perversions in growing helical filaments, it is reasonable to assume
Figure 6.22 Example of umbilical perversion [85].
that the onset of umbilical perversion is the result of biased differential growth of
the cord with external constraints, such as tension and the blockage of twist.
Inversion of helicity also appears in the microscopic world. For instance, cotton
[1300] and cellulose fibrils [163] exhibit perversions, as seen in Figure 6.23. The
flagella of some bacteria, such as Salmonella or Escherichia coli, have helical shapes
[622, 679]. These flagella, which are about 4 μm long, are composed of a few
protofilaments made out of a single protein. Depending on the respective positions
of the different protofilaments, the flagella can be, in their unstressed configuration,
either right-handed or left-handed [174, 177, 817, 1348]. Flagella are attached to the
cell body through a molecular rotary motor [88] and are observed to flip from a left-
to a right-handed helix when the running motion is interrupted by a tumble motion
where the flagella bundle disassembles [680, 1260]. This inversion of helicity plays
a crucial role in the ability of the bacterium to change its direction of motion and it is
performed by propagating a right-handed helix onto a left-handed helix, henceforth
creating a perversion [452].
Figure 6.23 Example of cellulose perversions obtained by electrospinning from liquid crystalline
solution [163] (Bar 100 μm).
A. B.
Figure 6.24 Perversion in bacteria. A. Flagella of Salmonella, with coexisting left- and right-
handed helices (arrows) (picture courtesy of H. Hotani). B. Solenoidal form of Bacillus subtilis
fiber, with coexisting helices (picture courtesy of M. Tilby). Scale bars: 5 and 2 μm, respectively.
Perversion can also be observed in a variety of other microscopic biological sys-

tems as shown in Figure 6.24: It is found in the shape of certain bacteria such as
spirochetes [454], in some mutant forms of Bacillus subtilis [1232], in Salmonella
flagella, and in the shape of miniature (1 to 2 mm long) seashells [1278]. Another
example, orders of magnitude smaller, involves two distinct conformations of DNA.
Experiments indicate stretches of left-handed Z-DNA spontaneously forming from
right-handed B-DNA due to the stresses generated by twisting during the process of
transcription [570].
6.7.1.2 The Model
The creation of these helical structures from a straight filament is the result of dif-
ferential growth that, in the first instance, can be understood within our framework
in terms of parameter variation. The tendril is modeled as an initially straight fila-
ment under tension. Due to faster growth on the dorsal side of the tendril, intrinsic
curvature increases slowly so that the filament is in static equilibrium at all times.
The problem reduces to exploring the possibility of a bifurcation from a straight
solution to a solution asymptotically connecting two helical structures of opposite
handedness.
We assume that the tendril is an inextensible, unshearable homogeneous rod with
circular cross section and intrinsic curvature κ̂, with no body load. Assuming that
only static solutions are relevant for our problem, (5.79–5.80) read
n = 0, (6.100)
m + r × n = 0, (6.101)
with
m = (u1 − κ̂)d1 + u2 d2 + u3 d3 , (6.102)
where we have rescaled the equations according to the scaling given in Section 5.4
and = K 3 /K 1 , as given by (5.57).
This system of equations can be written as a six-dimensional autonomous dynam-
ical system for the variable X = (n1 , n2 , n3 , u1 , u2 , u3 ) where
X = F(X; κ̂). (6.103)
The boundary conditions are given by a wrench, a pair of couple and force, (M, N )
applied at both ends.
6.7.1.3 Analysis
The question of finding a perversion solution can now be cast into three parts. First,
we identify possible asymptotic helical states with different handedness. Second, we
find the critical bifurcation value for the parameter κ̂ at which the straight filament
becomes unstable. Third, we build a solution connecting the two asymptotic regimes
after the bifurcation.
Following Section 5.7.2, the helical solutions of interest are twistless and have
the form
u = (0, κ, τ ), (6.104)
where κ and τ are constant. This restriction leaves, as possible asymptotic states,
a two-parameter family of solutions given by curvature and torsion. If we further
restrict these solutions to those with no total twist, the set of possible asymptotic
helices is given by (5.106) with τ̂ = 0; that is,

κ̂ 2 κ̂2
κ− + τ 2 = . (6.105)
2 4
In the curvature–torsion plane, this set is an ellipse, as shown in Figure 6.25. For the
dynamical system (6.103), a perversion is a heteroclinic orbit, an orbit asymptotically
connecting two fixed points. The fixed points are all located on this ellipse and an
inversion of handedness is obtained when a heteroclinic orbit connects fixed points
with opposite torsion. We first consider an infinitely long filament maintained by a
force N = n.e3 > 0 applied in opposite directions at s = ±∞ for which we have
N = γτ δ, (6.106)
where
κ̂
γ=( − 1 + ), δ= κ2 + τ 2 . (6.107)
κ
K/ 2 K
Figure 6.25 Each point in the curvature–torsion plane represents a helix. The set of helices that
can connect asymptotically to another helix with opposite torsion forms an ellipse in the curvature–
torsion plane.
For a given N , there is a single pair of asymptotic curvature and torsion (κ, ±τ ).
Therefore, the moment and force at infinity are given by
m = (κ − κ̂)d2 ± τ d3 , (6.108)
n = γτ u. (6.109)
Physically, when a rod is held straight and the tension is slowly released, a moment
must be applied at the ends in order to keep them from rotating. The condition for a
perversion solution is that the total twist is zero, which is equivalent to the condition
that the ends do not rotate as the tension is released.
These infinite heteroclinic orbits are not physical solutions, as there is no infi-
nite filament. However, the heteroclinic solution approaches the asymptotic helices
exponentially fast (in arc length) so that a finite piece of a heteroclinic orbit sym-
metric with respect to torsion, is close to the limiting helical solutions such that
the boundary conditions would be close to those of a finite rod. The normal form
analysis performed in [853] confirms the existence of a family of periodic orbits with
arbitrary period converging to the heteroclinic orbit. These solutions can be matched
to specific symmetric boundary conditions for finite length perversion solutions.
The study of finite length solutions also reveals more complicated behavior with
the possibility of multiple inversions on a finite domain [291] and a complex dynamics
creating them past the bifurcation point [794].
Third, the linearized system for Equation (6.103) around the fixed point X∗ =
(0, 0, N , 0, 0, 0) representing a straight, untwisted rod under end force N , has linear
eigenvalues

√ κ̂2
0, ± N , and ± N− . (6.110)

At the critical tension, Nc = κ̂2 / , two eigenvalues coalesce. This coalescence

is known in dynamical systems as a Hamiltonian pitchfork bifurcation [828]. For
0 < N < Nc , a perversion solution exists. A dynamical analysis of the Kirchhoff
equations (5.79–5.80) was performed in [477], where it was revealed that for N < Nc ,
the straight filament with intrinsic curvature is unstable. We can therefore study the
phenomenon of perversion near this critical value of the tension as a bifurcation
problem. This is done by combining center manifold reduction with normal form
expansion to obtain a reduced system that is valid close to the bifurcation point.
This system can be solved explicitly and solutions to the asymptotic problem can
be obtained as shown in [853]. These solutions together with numerical solutions
obtained by direct simulation of the full system are shown in Figure 6.26.
6.7.1.4 A Twistless Spring
A tendril perversion is made out of two helices with opposite handedness connected
by a small inversion. This structure has interesting properties compared to a regular
spring. In tension, the total twist remains zero. Following [853], we call such a
structure a twistless spring, i.e., a spring with zero total twist. To understand the
properties of such a spring, we compare it to a helical spring and an ideal spring.
The ideal spring is characterized by a Hookean law N = k z, where N is the axial
force, z is the displacement along the axis of the spring from its rest length, and k is
its Hooke’s constant.
The helical spring is made out of an inextensible helical rod with circular cross
sections. The rod is subjected to a wrench as described in Section 5.7.2, consisting
of an axial force N and axial moment M needed to prevent rotation of the ends. The
response of a helical rod away from its intrinsic curvature vector û = κ̂ d1 + τ̂ d3 is
fully specified by the spring formula (5.102–5.103). As tension is applied, the helix
deforms to a new helix with curvature κ and torsion τ .
Preventing the ends from rotating imposes the constraint that the total twist
remains constant:

T w = u3 ds = τ̂ ds = τ ds. (6.111)
This condition constrains the torsion τ to remain constant; that is, τ = τ̂ . We calcu-
late the length the spring is stretched by a tension T along the axis of the asymptotic
helices. Consider two points on successive rings P1 and P2 as shown in Figure 6.27.
When tension is applied the points separate. The twistless condition imposes a
moment at the end and we can use the helix spring formulas (5.102–5.103) to compute
the resultant force
4 5
3
0.6 2
1
0.4
0.2
0.2 0.4 0.6 0.8 1

-0.2
-0.4
-0.6
Figure 6.26 Family of heteroclinic orbits in the curvature–torsion plane together with the corre-
sponding filament shapes (κ̂ = 1, = 3/4) (reproduced from [853]).
Figure 6.27 Axial

separation of two
consecutive points on a helix
provides a simple
comparison with an ideal
spring.
P2
P2 d
d
P1 P1

κ̂ τ τ2
T = K1 √ − . (6.112)
1 − d2 d
We define the distance d between the z−coordinates of P1 and P2 . In the unstressed

state, this distance is d = d̂ = τ̂ /δ̂. We see that, as d varies from d̂ to 1, the force N
varies from 0 to ∞. Writing d = d̂ + z, we have

τ̂ 2
T = k z + O(z 2 ), k = E I1 4 π 2 1 + 4 π 2 2 . (6.113)
κ̂
Next, we compute the response of a twistless spring made of two connected

helices of opposite torsion in which the ends are held so that they do not rotate as the
axial force is changed. As tension is applied the helix deforms to a new helix with
curvature κ and torsion τ constrained to lie on the ellipse (6.105). The distance d is
now d = 2πτ /δ κ̂ and the axial force is
d
T = k 2 , (6.114)
1− d 2 (1 − )
where k = 4π 2 E I1 . To third order in z, (6.114) reads
T = h z + 2 h z 3 (1 − ) + O(z 5 ). (6.115)
Since 0 ≤ (1 − ) ≤ 1/3 and 0 ≤ z ≤ 1, the deviation of this relation from Hooke’s

law is small for small z. However, as the rod is stretched to its full length, d → 1,
the cubic term dominates. The tension T varies from 0 to h/ 2 as d varies from 0
to 1, in contrast to the ideal spring in which the tension varies from 0 to h. Note that
as → 1, the elastic properties of a perversion spring approach those of an ideal
spring.
Helix
Perversion
2
k /Γ
k Hooke's law
d
1-d 1
Figure 6.28 Comparison of different springs made of perversions and helices. The dash–dot curve
is the plot of the tension T versus the distance d for a perversion. The solid curve is the plot for
a helical spring. The dashed curve is the plot of the tension versus the displacement for an ideal
spring given by Hooke’s law.
In Figure 6.28, we compare the responses of the three springs. We see that the
twistless spring is much closer to an ideal spring. Physically, as tension is applied to
the ends of the twistless spring, it unwinds, relieving some of the torsional energy.
Another advantage of the twistless spring is that it does not apply moments at the
ends since moments are balanced in the middle of the spring.
Finally, we note another peculiar behavior of the twistless spring. We saw in
Section 5.7.3 that springs under pure axial force tend to overwind first then unwind.
Since a perversion is essentially made out of two helical springs under pure axial
force, we expect, when pulling on such a structure, to see the inversion connecting
the two asymptotic helices to first turn one way before changing its rotation. This
behavior is readily observed by manipulating vine tendrils.
6.7.2 Twining Vines
6.7.2.1 Background
Next, we turn our attention to another class of climbing plants, the twiners as shown
in Figure 6.29A. Twiners, such as garden peas, climbing jasmines, and morning
glories, are perhaps the most studied of all vines [1049]. The growing tip waves
around in a circular motion known as circumnutation until it finds an appropriate
upright support. It then starts wrapping around it to extend itself upwards as seen in
Figure 6.29B. The tip of the vine keeps nutating and the vine continues to climb by
forming a spiral around the support.
Can a given twining plant climb around supports of different sizes? This question
was first raised by Charles Darwin in his book “The Movements and Habits of
A. B.
b
Figure 6.29 A. Purple morning glory (Ipomoea purpurea). B. Chocolate vine (Akebia quinata)
Climbing Plants” [261]. He comments: “Most twining plants are adapted to ascend
supports of moderate though of different thicknesses. Our English twiners, as far
as I have seen, never twine round trees... By what means certain twining plants are
adapted to ascend only thin stems, whilst others can twine round thicker ones, I do
not know.”
As an example, Darwin noted that Solanum dulcamara can twine around supports
of 3 mm but not on supports of 5 or 6 mm (see many other examples in [1103]). The
natural question is to determine the critical cylinder radius above which a plant is no
longer able to twine.
In the process of establishing themselves on a pole, twining plants rely on friction
[1049, 1160]. As noted by Julius von Sachs [1107], the vine may slide off “...if the
surface of their vertical support is too smooth to furnish a strong mutual friction.”
What is the effect of friction in the vine’s ability to grasp the pole? Similarly, what is
the pressure generated by a plant on the pole? How does it change with its intrinsic
properties and shape?
The stems of most plants, such as trees or flowers, are in compression. However,
a peculiar feature of twining plants is that their stem is in tension [1162]. How is
this tension generated? There is no applied load pulling the tip of the growing plant,
gravity only increases compression and although the stem can build compressive
and tensile domains through differential growth, the net effect, once averaged over
the cross section, vanishes. Therefore, a vine in continuous contact with a pole
cannot generate tension. As we will see, tension is actually produced when the plant
establishes discrete points of contact, known as anchorage points (see [470] for
details).
Before lignification, the vine is a long, thin, flexible filament subject to twisting and
bending. Due to the small lineal density of the vine and the large stresses developed
through self-contact, the gravitational load on the vine has been experimentally found
to be negligible by comparison to other forces involved in the problem. For instance
the linear weight in Pharbitis nil (ivy morning glory) is about 0.4 mN cm−1 , but it
can exert a contact force of 300 mN cm−1 [1162]. Therefore, the gravitational load
will be omitted in the analysis of the grasping problem.
During circumnutation, the vine at the apex develops intrinsic curvature and tor-
sion. We assume that these curvatures are constant and uniform. This assumption
is also consistent with the vertical and lateral oscillations of the vine tip observed
during growth. In Figure 6.29A, the shoot apex, the growing flexible part of the vine,
spans from a to b. Below the point a, the vine is set and woody by lignification.
6.7.2.2 A Model
Viewed as a growth problem, we can study the possible equilibria of a rod with intrin-
sic curvature, torsion and increasing length, in contact with a cylinder. Therefore,
the problem reduces to finding suitable solutions with the increasing length taken as
our control parameter.
We model the growing vine as an inextensible, unshearable elastic Kirchhoff
rod with intrinsic curvature and torsion. A further simplification of the problem is
obtained by assuming that the rod is confined to the plane. In this case, the support
is a rigid disk of radius R and the problem is to find equilibrium configurations of
an elastic rod with intrinsic curvature 1/Rc in contact with the disk as illustrated in
Figure 6.30. The filament is assumed to be strongly anchored at its base while its tip,
located at s = L, is simply supported. These conditions imply that at the origin, we
have r(0) = (R, 0), r (0) = (0, 1) whereas at s = L, the rod lies on the disk and its
curvature is equal to the intrinsic curvature; that is, κ = 1/Rc .
The Kirchhoff equations, together with the boundary conditions, completely spec-
ify the problem for any given values of L , R, Rc , and numerical continuation of
equilibrium solutions of this boundary-value problem can be performed. In order to
gain insight into the possible solutions, two series of equilibrium configurations for
filaments with different intrinsic curvatures are shown in Figure 6.31. In both cases,
the disc radius is kept constant while the length is increased. For the first series,
presented in Figure 6.31A, a stable configuration develops and a continuous contact
with the support is established.
Conversely, in the second series, shown in Figure 6.31B, no continuous contact
configurations exist. Instead, past a certain length, the filament starts to wind outside
the disc. Therefore, the limiting ratio R/Rc between climbing and non-climbing
Figure 6.30 A rod with simple y nL

intrinsic curvature is support
clamped on one end of a
rigid disk and simply
supported at the other end. er α
eθ s =L
R s=0 x
clamping
states is found to be in the interval 3 < R/Rc < 3.5, as illustrated in Figure 6.32. A
detailed bifurcation analysis [470] reveals that the critical ratio is R/Rc ≈ 3.31.
The analysis so far has been restricted to the frictionless case. However, it has
been emphasized by different authors [1049, 1160] that friction plays an important
role in the ability of a climbing plant to support its weight. Here, friction is modeled
by a tangential component of the disk reaction force on the tip of the anchor. To
understand its effect, we first consider an equilibrium configuration obtained in the
absence of friction. We then change the length of the filament without moving the
tip. This is made possible by introducing a tangential component in the reaction force
along eθ at s = L due to friction. As L increases, the tangential force component
reaches a critical value equal to μ p where μ < 0 is the friction coefficient and p is
the normal forced exerted on the disk. Similarly, if L decreases, the tangential forces
reach μ p with μ < 0. Past this value, the tip slips incrementally on the disk to find
a new nearby equilibrium configuration where the tangential component equals the
critical tangential value. Therefore, to identify the equilibrium configurations with
friction, we increase the length and find configurations where the tangential force
equals μ p. For each friction coefficient μ we compute, as detailed above, the critical
value Rc = Rc (μ) defining the bifurcation between filaments that can grow around
the disk and filaments that leave the disk. The results in Figure 6.33 show that friction
can have a crucial effect on the ability of a plant to remain on a large disk. Note that
in the case of a twining plant, once the tip is in contact with the pole, friction prevents
the plant from sliding back, therefore corresponding to positive values of μ.
A.
B.
Figure 6.31 Equilibrium configurations for a naturally curved filament around a disk with A.
R/Rc = 3 and B. R/Rc = 3.5. A. Configurations with continuous contact eventually develop. The
vine is successfully attached to the pole. B. No configuration with continuous contact exists. The
vine fails to grasp the pole. (Figure courtesy of Sébastien Neukirch).
5 L
α
0 0.5 1 1.5 2 2.5 3 3.5 4
Figure 6.32 Bifurcation diagram corresponding to the equilibrium configurations of Figure 6.31B.
Intuitively, one can understand Darwin’s problem as follows: For thin supports,
the vine tip and the tangent of the support are almost parallel and the vine follows
the support without curling back on itself. For thicker supports, in order to maintain
its grip, the tip needs to touch the support surface at a larger angle α. If the support
becomes too large, so too will the angle α, and the vine will curl back on itself. For
increasing friction the contact angle needed for this to happen will be higher and,
consequently, the plant can wind on thicker supports.
We now turn our attention to the three-dimensional case. The main properties
found in the two-dimensional problem still hold; namely, discrete points of contact
exist between the plant and support, tension is generated by these contact points,
and attachment is possible only by the difference between intrinsic curvature and
ρc
8
μ
-1 -0.75 -0.5 -0.25 0.25 0.5
Figure 6.33 Change of critical radii ratio ρc = R/Rc as a function of friction. Positive friction
coefficients act when the tangential force prevents the filament from sliding back whereas negative
friction coefficients correspond to situations where friction prevents the tip from slipping forward.
Range of μ values are consistent with experimental data [1160]. The prediction of an upper bound
of ρc around 8 is consistent with the observation of Bell [75] (reproduced from [470]).
Figure 6.34 A sequence of three-dimensional solutions to the attachment problem. Note the con-
tinuous, almost helical, solution, followed by the anchor that provides tension in the filament (repro-
duced from [470]).
geometry of the support. An example of such solutions is given in Figure 6.34.

Numerical studies suggest that the value of Rc obtained in the planar case provides
an upper bound for the three-dimensional case.
The analysis of the growing vine shows that a model for a growing elastic rod in
contact with a pole can provide a simple explanation for the limitation of twining
plants to wind around thick supports, and for the role of friction to boost the plant’s
ability to achieve vertical growth. The model also explains how tension is generated
and shows that both continuous and discrete contact points between stem and pole
exist, a prediction consistent with experiments [655].
6.7.3 Application: The Growth of Bacillus subtilis
As an example of the effects of a local growth law, we consider the evolution of

length and twist in a model for growing Bacillus subtilis.
6.7.3.1 Background
The individual cells of the bacterial strain Bacillus subtilis are rod-shaped and typi-
cally of length 3−4 μm and diameter 0.8 μm. Under certain circumstances, they are
found to grow into filaments consisting of the cells linked in tandem due to the failure
of daughter cells, produced by growth and division, to separate. As they elongate,
these filaments, which are immersed in a liquid environment whose temperature and
viscosity can be controlled, are observed to twist at a uniform rate. The degree of
twist and handedness can be controlled experimentally and a wide range of states,
from left-handed to right-handed forms, can be produced. The actual twist state of
the cells seems to be related to properties of the polymers which are inserted into the
cell wall during growth.
Eventually, the filaments are observed to writhe and deform into double-helical
structures. These structures continue to grow and periodic repetition of this process
results in macroscopic fibers termed macrofibers, with a specific twist state and
handedness as shown in Figure 6.35. Two different schematic looping mechanisms
of Bacillus subtillis are given in Figure 6.36.
A striking feature of this iterated process is that at every stage of the self-assembly,
the handedness of the new helical structures that are created is the same, e.g., a
right-handed double helix gives rise to a right-handed four-strand helix and so on.
The nature of the environment does, however, influence certain aspects of the self-
assembly. In a viscous environment, the basic writhing instability leads to an insta-
bility at the middle of the filament with the formation of a tight central loop; this is
followed by a helical windup which starts at the base of this loop (Figure 6.36B).
Figure 6.35 Macrofibers of Bacillus subtilis (picture courtesy of Neil Mendelson).

L R
A. L R
B.
L
R
L
R
L
R
L R
L
R
L
R
L R R L
L R
Figure 6.36 The two basic looping mechanisms of Bacillus subtilis. A. In nonviscous medium. B.
In viscous medium (picture courtesy of Neil Mendelson).
By contrast, in a nonviscous medium, the instability causes the filament to fold over
into a large loop closed by contact between the ends of the filament. This closure
is then followed by a helical windup starting at the instability point (Figure 6.36A).
In both cases, this self-assembly conserves handedness and usually continues over
long periods until macrofibers, several millimeters long, are formed.
The dynamics of the self-assembly and the mechanical properties of the bacterial
threads have been studied in great detail, by Mendelson and coworkers [866–868].
In addition to the fascinating questions of growth and form raised by this process,
the macrofibers themselves offer the prospect of unusual biomaterials that can be
mineralized and packed in ways that are of practical biomedical and biotechnical
use.
6.7.3.2 A Model
First, we propose a simple qualitative explanation of the observation that, in Bacillus

subtilis, the twisting and supercoiling are of the same handedness. The key to this
transition is the buildup in time of an intrinsic twist, described in the Kirchhoff
model by a nonvanishing parameter û3 . We assume that in Bacillus subtilis each
bacterial cell possesses an intrinsic twist. Reproductive growth of all the cells in
the bacterial filament results in an exponential growth of its length accompanied
th
tial grow
Exponen
al waves
ion Torsion
ed tens
Grow th induc
al waves
Torsion
Deficit
m Twist
Maximu ta bility
=> Ins
Figure 6.37 Exponential growth and linear torsional wave competition. As the filament grows
exponentially, the linear twist wave propagating from the end does not reach the middle of the
filament where the twist deficit is maximal.
by a reduction of twist density. Consider a file of such cells making a filament with
right-handed intrinsic twist, represented by drawing marker lines with a right-handed
helical pitch, as shown in Figure 6.37. One end is twisted in the opposite direction
to the marker lines until they are approximately straight. The twist density of the
filament now appears to vanish whereas its natural state is one of nonzero twist
density—as indicated by the original helical marker lines. Thus we now have a twist
deficit, given that û3 = 0. To return to its natural state, the filament must make up
for this deficit by restoring twist. This can be achieved in two different ways. If one
end is free, the tube winds sending a twist wave down the rod. Alternatively, if the
ends are held but brought toward each other, the tube will relax by supercoiling with
the same handedness as the intrinsic twist. This is the behavior observed in Bacillus
subtilis.
To obtain a full picture of the dynamics of Bacillus subtilis, we can use the
Kirchhoff equations with uniform growth and model the viscous environment as a
local drag force, as done in a series of computer simulations by Isaac Klapper [700,
701, 1337]. Among other things, the model gives quantitative predictions about the
self-assembly geometry, such as the way the loop size scales with environmental
conditions. The computer simulation of growing rods with intrinsic twist predicts
the formation of looping in filaments remarkably similar to the ones found in the
experiments as seen in Figure 6.38. A great deal of experimentation has shown that
the twist state and helix handedness of Bacillus subtillis macrofibers stem from the
individual cell from which the fiber is derived. The information required to control
macrofiber morphogenesis appears to reside in the growth plan of this cell and all its
t= 0 t= 2 8
t= 4 0 t= 4 7
t= 5 2 t= 5 5
t= 5 8 t= 6 1
Figure 6.38 Writhing dynamics of Bacillus subtillis. Left: Experiments (courtesy of N. Mendel-
son). Right: Simulation of Kirchhoff rods with growth (courtesy of I. Klapper [700, 701]).
descendants. Intrinsic twist, a key feature of the dynamic model described here, is a
logical candidate for the mechanical information in the growth plan that dictates all
subsequent growth and form.
Part III
Surface Growth: A Two-Dimensional
Theory
And then the justice,

In fair round belly, with a good capon lin’d,
With eyes severe, and beard of formal cut,
Full of wise saws, and modern instances,
And so he plays his part. . .
Chapter 7
Accretive Growth
We now consider the growth of biological structures that can be modeled as two-
dimensional objects; that is, three-dimensional bodies that have a transverse length
scale much smaller than the other two typical length scales. In developing useful
models for the growth of surfaces, it is important to distinguish between hard and
soft surfaces.
Seashells, horns, antlers, and teeth can be modeled as hard surfaces that cannot
be easily deformed. In these structures, accretive growth takes place by adding mass
locally on the boundary of the structure since there is no possibility of deforming the
material in the bulk in order to add mass. From a continuum mechanics perspective,
surface growth poses a particularly interesting challenge. Since new material points
are continually deposited on the surface of the body, the domain on which all physical
quantities are defined also evolves in time, so there is no fixed reference configura-
tion. Therefore, incorporating accretive growth within classic continuum mechanics
requires particular care [421]. There is significant interest in this topic from a funda-
mental mechanical standpoint as well, and several theories and continuum balance
laws have been proposed to address the issue [338, 579].
Other biological structures can be modeled as two-dimensional soft surfaces, such
as leaves, pollen tubes, and flowers. In these systems, there is a fair amount of elastic
deformation. In contrast to the previous case, growth is typically distributed over

176 7 Accretive Growth
the entire area. The shape is the result of a competition between the local addition
of mass and the bending and stretching properties of the surface. By analogy with
the general three-dimensional case, we refer to this type of surface growth as area
growth.
We start with a geometric description of accretive growth in simple hard struc-
tures for which elastic deformations are negligible. We assume that as the object
grows it remains mostly two-dimensional. Historically, there have been two differ-
ent approaches to the growth of hard surfaces. The first approach is to consider the
problem from a mechanistic rather than a descriptive point of view. Here accretion
takes place at the boundary of the surface and is characterized by a certain depo-
sition rate in a given spatial direction that is fixed by the intrinsic geometry of the
surface [914, 915, 1077, 1173]. The second approach is mostly descriptive, that is,
one tries to characterize the object by a parameterized surface with a few parameters
that can be obtained from direct measurements. Along these lines, the description
of seashells is the most developed theory [861], and due to its historical importance
and intrinsic beauty, we use seashells as a paradigm for accretive growth. We start
with the mechanistic description and, later, compare this approach with the classic
description of seashells by surface parameterization.
7.1 Intrinsic Accretive Growth
We adopt a local constructive view for the general problem of growth of hard bodies
through surface accretion following the theory of Moulton and coworkers [914, 915].
We consider the growth of a surface starting with a generating curve. At time t = 0,
the curve is assumed to be a twice continuously differentiable curve R(S) = r(S, 0) ∈
R3 defined by its arc length S ∈ [0, L]. Accretion is modeled by the evolution of
this initial curve by defining at each point S on this curve a growth vector q(S, 0)
representing the local evolution of material points during growth as illustrated in
Figure 7.1. This vector field is the initial growth velocity of the material point located
at S. At any time t, we specify a growth field q(S, t) defining the growth velocity of
a material point S at time t so that the problem is to integrate this velocity field in
time to obtain a curve r(S, t).
Since the velocity vector describes a process of accretive growth, this curve can
be thought of as a cell track, initially at position R(S) on the surface r(S, t), i.e, the
track that would be left by the cell depositing the new material over time. Since the
direction and amplitude of the accretion process determined by the velocity vector
is a local process, it is reasonable to assume that the velocity vector is a function of
the local coordinate system, as well as any physical; chemical; or biological field,
such as rate of accretion, morphogen gradient, temperature, pH, etc.
Following our discussion of curves in Chapter 6, we attach a general frame to the
curve r(S, t):
D = d1 d2 d3 , (7.1)
7.1 Intrinsic Accretive Growth 177
A. S0
B.
qS
S
C.
Figure 7.1 Creation of a surface from an initial generating curve. A. At time t an initial curve is
given. This curve evolves in time by the specification of a local velocity field attached to the outer
edge of the surface. B. In the incremental process of growth, at any time t, a curve is given together
with a velocity field q. C. The new curve at t + t is obtained by moving the original curve by t q.
The curve created by the accretive process is the new generating curve for subsequent accretive
processes.
where d3 is the unit tangent vector to r(S, t) and is given by
r = v = v3 d3 = λd3 , (7.2)
where (·) ≡ ∂ S (·) and λ = v3 is a coefficient that describes the stretch of the curve
with respect to the initial arc length S.
The general frame {d1 , d2 , d3 } forms an orthonormal basis for all S and t, that is,
D is an orthonormal matrix:
DT D = DDT = 1, (7.3)
which implies
D ≡ ∂ S D = DU, (7.4)
Ḋ ≡ ∂t D = DW, (7.5)
where U is the Darboux matrix, a skew-symmetric matrix describing the rotation

of the local basis at a given time t along the curve and W is the spin matrix, a
skew-symmetric matrix describing the rotation of the local basis at a point S as time
evolves. As discussed in Section 5.1.2, these matrices read

⎡ ⎤ ⎡ ⎤
0 −u3 u2 0 −w3 w2
U = ⎣ u3 0 −u1 ⎦ , W = ⎣ w3 0 −w1 ⎦ , (7.6)
−u2 u1 0 −w2 w1 0
with the associated axial vectors u = (u1 , u2 , u3 ), and w = (w1 , w2 , w3 ) in the

general frame. Following the convention of Chapter 5, a vector a in a fixed basis
{e1 , e2 , e3 }:
a = a1 e1 + a2 e2 + a3 e3 , (7.7)
can be written in the general frame as a = Da = a1 d1 + a2 d2 + a3 d3 . The general

basis D of a curve of class C 3 is determined up to an arbitrary choice of angle between
the vector normal to the curve and the vector d1 as explained in Section 5.1.3, and
the elements of U, which are related to the usual notion of curvature and torsion.
Recalling that the time t represents the evolution of growth, at any point on a
curve r(S, t) we attach a growth velocity vector q(S, t) so that
ṙ(S, t) = q(S, t) = Dq, (7.8)
˙ ≡ ∂t (·). Note again that q = (q1 , q2 , q3 ) are the components of the velocity
where (·)
vector q expressed in the local frame D as opposed to (q1 , q2 , q3 ), the components
of the velocity vector q in the fixed reference frame.
We assume that growth is controlled locally, that is the information about growth
is contained in the local components of the growth vector. Therefore, we restrict our
attention on the vector of components q.
Assuming that the curve r is at least twice differentiable in t and S, the symmetry
of second derivatives provides two sets of compatibility conditions

∂ ∂ ∂ ∂
r = r , (7.9)
∂ S ∂t ∂t ∂ S

∂ ∂ ∂ ∂
D = D . (7.10)
∂ S ∂t ∂t ∂ S
Using again the orthonormality constraint DDT = 1, we obtain, after simplification
q − v̇ = Wv − Uq, (7.11)
W − U̇ = UW − WU, (7.12)
where v is the stretch vector defined in (7.2). In components, these two equations
read
q1 + u2 q3 − u3 q2 = λw2 , (7.13)

q2 + u3 q1 − u1 q3 = −λw1 , (7.14)
q3 + u1 q2 − u2 q1 = λ̇, (7.15)

u̇1 − w1 = u2 w3 − u3 w2 , (7.16)
u̇2 − w2 = u3 w1 − u1 w3 , (7.17)
u̇3 − w3 = u1 w2 − u2 w1 . (7.18)
For a given growth vector q, these equations form a set of six nonlinear first order
partial differential equations for the seven dependent variables {u, w, λ}. The extra
degree of freedom is given by the extra degree of freedom in the choice of the general
basis. If we assume that the curve is at least three times differentiable, then we can use
the Frenet frame, and {d1 , d2 , d3 } = {ν, β, τ } are the normal, binormal, and tangent
vectors, respectively. Then, the Darboux matrix U takes the simple form
⎡ ⎤
0 −τ κ
U = λ⎣ τ 0 0⎦, (7.19)
−κ 0 0
where κ ≥ 0 and τ are the usual geometric curvature and torsion as given by (5.4) and
(5.7). This choice implies u1 = 0 thus reducing the system to a set of six equations
for six unknowns which can be solved for an initial profile and growth field. Once
u(S, t) and λ are known, the surface built through the accretive process is obtained
by integrating the system of twelve nonlinear ODEs given by
∂ ∂
D = DU, and r = λd3 . (7.20)
∂S ∂S
To understand the computational steps involved in solving this system of equations,
we consider a few simple solutions.
7.1.1 Simple Examples
7.1.1.1 Cylinders, Cones, and Tori
We first consider the case of a growing cylinder. We work in the Frenet frame and
take the generating curve C0 to be a circle of radius 1, r(S, 0) = (cos(S), sin(S), 0).
We choose a constant growth velocity field along the binormal, that is, q = (0, 1, 0).
In this simple case, the compatibility equations and frame evolution are trivial, and
the surface is found by integrating (7.8):

t
r(S, t) = q dt + r(S, 0) = (cos(S), sin(S), t), (7.21)
0
producing, as shown in Figure 7.2A, a parameterization of a cylinder.

A. B. C.
c1=0, c2=1, c3=0 c1=0, c2=1, c3=1 c1=-1/2, c2=1, c3=1/2

Figure 7.2 A. A straight cylinder. B. A twisted cylinder. C. A twisted cone.
If we add a constant velocity component in the tangential and normal directions

to the generating curve by assuming that the growth velocity field is q = (0, c2 , c3 )
with c2 c3 = 0, we obtain
r(S, t) = (cos(S + c3 t), sin(S + c3 t), c2 t) , (7.22)
which is a parameterization of a twisted cylinder, as shown in Figure 7.2B, and

different from (7.21). The difference between the two cylinders can be understood
in terms of cell track motions. In an accretive growth process, a cell in Figure 7.2A
deposits new material directly in the binormal direction to the ring whereas a cell in
Figure 7.2B has an additional tangential component.
Next, we add a normal velocity component, so that q = (c1 , c2 , c3 ) with c1 c2 c3 =
0. In this case, the frame evolution is the same as the twisted cylinder case, but since
λ̇ = −c1 from (7.15), the surface parameterization is
r(S, t) = ((1 − c1 t) cos(S + c3 t), (1 − c1 t) sin(S + c3 t), c2 t) , (7.23)
and the resulting surface is a cone with twisted cell tracks, as depicted in Figure 7.2C.
line of rotation
d2
d1 d3
growth axis
Figure 7.3 Growth of a torus. On the left is a schematic illustrating the linear growth along the
growth axis. On the right is a growing torus along with growth velocity field on the generating curve
(b1 = 2, b2 = 1).
More interesting shapes are obtained if the local growth vector is not constant.
Taking again a unit circle as an initial curve, coiling is achieved by taking q2 to be a
linear function along a growth axis in the plane of the generating curve
q2 = b1 + b2 cos(S), (7.24)
where b1 and b2 are constants. We show in Section 7.1.2 that the generating curve
remains circular for all time. We can obtain a torus from (7.24) by choosing q1 =
q3 = 0 and integrating the governing equations (7.13–7.18):

1 q
r(S, t) = cos(b2 t)q2 − b1 , sin(S), 2 sin(b2 t) , (7.25)
b2 b2
an example of which is given in Figure 7.3.
7.1.1.2 Planar Horns and Shells
We extend this construction to generate more complex surfaces by adding a constant

normal component to the growth vector defined by (7.24). There are three different
velocity profiles that can produce normal growth in this context:
q1 = −c, q2 = b1 + b2 cos(S), q3 = 0, (7.26)

q1 = −c, q2 = λ(b1 + b2 cos(S)), q3 = 0, (7.27)
q1 = −c, q2 = b1 + b2 λ cos(S), q3 = 0. (7.28)
A. B. C.
Figure 7.4 Three primary forms of growth with a linearly expanding circular generating curve.
For each case, we show from top to bottom: the local velocity, a schematic illustrating the binormal
growth gradient as the radius changes, a sample structure, and the curve traced by the center of the
generating curve [914].
These three different choices lead to surfaces with expanding generating curves as
7.1.1.3 Three-Dimensional Structures
We note that in the examples given so far, each point of the generating curve follows
a path in time that is either linear or planar. For more complex structures such as
seashells, we need to consider nonplanar growth. For simplicity, we keep a circular
generating curve. We consider two different methods of growth that produce similar
surfaces, yet have distinct and important differences from a biological viewpoint.
• Rotating growth axis: To create out-of-plane growth, we can rotate the growth
axis in time. For the binormal velocity q2 = b1 + b2 cos(S), the growth axis is
always on a line contained in the x–z plane, and thus the rotation remains in
this plane. If this growth axis locally rotates during the evolution, the direction
A. B. C. Cell tracks on Turitella
Figure 7.5 A. and B. Comparison of two primary forms of local velocity for out-of-plane growth
giving rise to the same shape. The plotted surfaces are identical, but with different parameterizations,
reflecting the distinct difference in cell tracks. The parameters are c = 0.12, b1 = 1.6, b2 = 1,
b3 = 0.55 for each surface. C. The cell tracks of a Turitella shell can be seen by tracing the small
antimarginal ribs. A particular track has been highlighted. The tracks stay in the same relative
position on the shell aperture during growth (reproduced from [915]).
of rotation will also vary and growth will be nonplanar. To illustrate this effect,
consider the growth velocity field
q1 = −c, q2 = b1 + b2 cos(S − b3 t), q3 = 0, b2 > 0. (7.29)
Maximal growth occurs at material point S = b3 t, thus the parameter b3 dictates

the rotation rate of the growth axis. An example is shown in Figure 7.5A.
• Tangential velocity: An alternative form to rotating the growth axis is to add a

tangential velocity component, so that
q1 = −c, q2 = b1 + b2 cos(S), q3 = b3 , b2 > 0. (7.30)
Maximal growth always occurs at a fixed material point,the point S = 0, and this
point moves in the tangential direction with time, hence rotating the growth axis
around the generating curve as illustrated in Figure 7.5B.
While these two forms can produce the exact same surface by using the same para-
meter values, there is a subtle but important biological difference between these two
models. In (7.29), there is no tangential motion and the material points remain in
the same location on the generating curve, but the growth rate for any given material
point on the generating curve varies with time. Hence the location where growth is
maximal also varies. This idea is illustrated in Figure 7.5, in which surfaces are plot-
ted using a rotating growth axis and a tangential velocity component. Note that the
forms are slightly different from Equations (7.29) and (7.30) as the scale factor λ has
been included in the velocities. Included on each surface are curves representing a
cell track for the particular value S = 0. In the case of the rotating growth axis, shown
in Figure 7.5A, the binormal velocity of the cell varies with time, and the cell track
oscillates between the inner and outer edge of the global structure. By comparison
in the tangential velocity case, shown in Figure 7.5B, S = 0 always corresponds to
the point of maximal growth, and the cell track follows the outer edge of the global
structure.
From a biological perspective, either form could be achieved a priori. They can
only be differentiated by following cell tracks. For example, in Turitella shown in
Figure 7.5C, one can trace the cell tracks by following the locations of the small
antimarginal ribs. We observe that the cell tracks take the form of Figure 7.5B,
suggesting that out-of-plane growth was produced by a growth velocity field of the
form (7.30).
Next, we derive conditions that the local growth velocity fields have to satisfy to
guarantee universal shape properties. In particular, we identify conditions so that the
generating curve remains either planar or similar, up to a global dilation.
7.1.2 Shape Planarity
Consider an arbitrary planar curve r(S, t). Suppose, without loss of generality, that
the curve is initially parameterized by its arc length, so that |r (S, 0)| = λ0 = 1. What
are the general constraints on the growth velocity field so that the curve remains planar
at all time? Planarity implies that at any given time the curve can be expressed as
r(S, t) = p(t) + x(S, t)X̂ + y(S, t)Ŷ, (7.31)
where p is the position vector of a point in the plane of the curve and the unit
vectors X̂, Ŷ form a basis for the plane of the curve. This construction is illustrated
in Figure 7.6. In this representation, the quantities λ and κ take the form
x y − x y
λ= x 2 + y 2 , κ= . (7.32)
(x 2 + y 2 )3/2
In order for the generating curve to remain planar during its evolution, its torsion
must vanish identically, i.e., u3 = u̇3 = 0. Using this condition in (7.13–7.18) leads
to the following equation for q2 :
growth axis
Figure 7.6 A planar curve is represented by a position vector p and the unit vectors X̂ and Ŷ, which
form a basis for the plane of the curve. The growth axis in this figure is given by Ŷ = −b2 /b3 X̂ .

2λ u2 2λ2 λ u λ
q
2 − q2 + + q2 u22 + 2 − + 2 = 0, (7.33)
λ u2 λ λ u2 λ
the general solution of which is
q2 = b1 (t) + b2 (t)x(S, t) + b3 (t)y(S, t). (7.34)
The form (7.34) implies a remarkable universal feature of growth. Namely, if the
generating curve remains planar, then there exists a distinguished axis along the
vector b2 X̂ + b3 Ŷ such that the binormal growth on the generating curve is a linear
function when projected onto this axis. This axis is the growth axis that we defined
in the case of a circular generating curve. In particular, for the initial curve r(S, 0) =
(cos(S), sin(S), 0), the form is q2 = b1 + b2 cos(S) + b3 sin(S). Taking b3 = 0 sets
the x–axis of the initial plane to be the growth axis, and we recover the growth law
for the torus given by (7.24) and shown in Figure 7.3.
7.1.3 Shape Invariance
We further restrict the growth velocity field so that the planar shape of the generating
curve remains similar, that is, identical to the original curve up to a global dilation.
This constraint implies that the curve can only undergo translation, rotation, and
dilation. Therefore, we look for conditions on q1 and q3 so that the shape of the
generating curve does not change as the curve evolves. Shape invariance implies that
at any time t, |r | = λ = λ(t), and that the curvature κ(S, t) is the initial curvature
scaled by the dilation factor λ, so that
κ(S, t) = λ−1 κ(S, 0). (7.35)
Working in the Frenet frame, the shape variables are then given by
u1 = u3 = 0, u2 = λκ = κ(S, 0). (7.36)
Therefore, we conclude that to ensure shape planarity, two conditions must be satis-
fied:
u̇2 = 0 (7.37)
q1 + u2 q3 + u22 q1
+ u2 λ̇ − λ w2 = 0. (7.38)
Motivated by the notion that the allowable transformation of the cross section should
consist of dilation, rotation, and translation, we find that in global coordinates, the
solution is of the form
q = c1 (t)r + c2 (t)r⊥ + a(t), (7.39)
where r⊥ is a vector perpendicular to the position vector r with the same magnitude.
The first term describes a growth velocity in the radial direction and accounts for
dilation of the curve. The second term gives a growth velocity in the circumferen-
tial direction and accounts for rotation. The arbitrary vector, a(t), in the third term
characterizes a rigid translation. To see that the form (7.39) satisfies (7.38) and to
determine the explicit form of c1 and c2 , we express q in the local coordinates. This
is achieved by writing
r = β1 d1 + β3 d3 , r⊥ = β3 d1 − β1 d3 , a = α1 d1 + α3 d3 . (7.40)
The local velocities are then
q1 = c1 β1 + c2 β3 + α1 , q3 = c1 β3 − c2 β1 + α3 . (7.41)
Recalling that r = λd3 and using the Frenet equations d1 = −u2 d3 , d3 = u2 d1 , we
have
β1 = −u2 β3 , β3 = λ + u2 β1 , (7.42)

α1 = −u2 α3 , α3 = u2 α1 . (7.43)
Then, by inserting (7.41) into (7.38), we find that this latter relation is satisfied if
λ̇
c1 (t) = , c2 (t) = w2 (t), (7.44)
λ
which highlights the relationship between the dilation factors c1 and λ̇ as well as the
equivalence between the rotation factor c2 and the frame rotation component w2 .
7.1.4 Self-Similarity
A remarkable feature of many seashells is the property of scale invariance also

known as self-similarity. Self-similarity is a global property of an object. Globally,
we may define self-similarity in space as the requirement that the ratio of the distance
between successive landmark points is constant. For instance, the ratio of the distance
between successive whorls in a logarithmic spiral shell is constant [226], as shown in
the next section. The question is now to define self-similarity at a local level: Which
local growth velocity fields ensure global self-similarity? The key property for scale
invariance is that, for each material point, the local direction of the growth vector
does not vary in time. In the case of q3 = 0, this is expressed by the requirement

q2
∂t arctan = 0. (7.45)
q1
Property (7.45) is required for a surface to be scale invariant. However, it does not
imply an invariance in growth rate, only in the growth direction. To enforce invariance
in time, we further require that at each point the ratio of the magnitude of the growth
vector to the size of the generating curve (i.e., radius of the circular cross section
under consideration) should not vary in time. Note that if the initial radius is 1, then
the radius at time t is λ(t), and the requirement is

|q|
∂t = 0, (7.46)
λ
√
where |q| = q · q is the magnitude of the growth vector. This condition motivates
the growth law:
q1 = −λ(t)c, q2 = λ(t)(b1 + b2 cos(S)), q3 = 0. (7.47)
The difference with (7.26) is that here each velocity component is scaled by the
stretch factor λ(t). The direction of the growth vector is still time independent, i.e.,
the ratio q2 /q1 is fixed for each S. This growth law produces a structure similar to
Figure 7.4A with the difference that the magnitude of its growth rate is also scale
invariant, i.e. (7.46) is also satisfied.
7.2 Application: The Growth of Seashells
7.2.1 Background
By their beauty and simplicity, seashells have captured the imagination of mankind
for centuries. They have been used as currencies, jewelry, ornaments, tools, and are
the favorites of many insatiable collectors. Most seashells are the hard mineralized
Figure 7.7 Organization of

the generative zone in most
modern mollusks. The
mantle is part of the living
animal and the periostracum
is a thin organic layer of the
shell that acts as a promoter
of deposition for the
mineralized shell [916].
shells of gastropod mollusks. More precisely, the shells of mollusks are microlam-
inate composite bioceramics made of minerals and biopolymers arranged in layers
of prismatic calcite crystals or brick wall aragonite crystals [824]. The regular and
complex arrangements of these layers, which account for up to 98% of the total mass
of the animal, provide the shell with a strength that exceeds that of the mineral itself
[38, 1176]. The shell of most modern mollusks is the result of a repeated secretion
process generated by the mantle, a thin soft tissue of the mollusk that extends out
of the shell to wrap around its edge to cover the periostracum, a thin outer organic
layer [1164] shown in Figure 7.7.
During growth, the mantle moves forward slightly beyond the calcified shell edge
whilst secreting the periostracum, which isolates the extrapallial fluid from which the
calcified shell is precipitated [1114]. The periostracum is secreted in the periostracal
groove, between the outer and middle mantle lobes, and is extruded between the two
mantle lobes, whereby a stiffening process of sclerotization takes place [1288]. It is
subsequently turned inside out around the outer mantle lobe and reaches its external
position where no further thickening occurs.
When calcification occurs, the periostracum becomes fixed on the outer shell
surface. The unfixed periostracum surrounds the outer mantle lobe and is attached at
both extremities along the calcified shell edge and inside the periostracal groove. The
shape of the generative zone, namely the stiff periostracum surrounding the softer
outer mantle lobe, is therefore incrementally set in the calcified shell during growth.
In turn, the calcified shell edge acts as a template for the new growth increment.
Strikingly, despite the fact that gastropod mollusks are a diverse group of meta-
zoans with over 100,000 extant and fossils known, their external shells seem to follow
a remarkably simple design based on an expanding spiral [533, 1077] that closely
matches the logarithmic spiral [539], identified as the key feature of seashells by
Moseley in 1838 [908].
This morphological regularity has long been recognized and appreciated, and has
inspired a large body of research. In particular, D’Arcy Thompson described the
elegance and simplicity of the spiral coils of the molluscan shell in mathematical
and geometric terms [1230], as shown in Figure 7.8. Following the work of Raup
7.2 Application: The Growth of Seashells 189
[1061, 1062], who triggered the field of research termed “theoretical morphology,”
many models and approaches have been devised [274, 1199]. While most of the
early works were descriptive, focusing on the form and shape of shells, research has
shifted in recent decades to questions of biology and evolution [1077], understanding
pigmentation patterns [108, 381, 862, 863], and the growth process underlying shell
form [532]. Nevertheless, even in the case of the molluscan shell, which has been
reproduced mathematically numerous times, a full understanding of how growth
occurs, e.g., the biomechanical basis for coiling, remains elusive.
7.2.2 Geometric Description
Mathematically, the ideal form of these seashells can be described with the general
framework of curves and rods of Chapter 5. The central axis of the seashell is a log-
arithmic spiral and the radius of the rod increases with arc length. A convenient way
to represent the surface of a seashell is to consider a centerline curve p = p(t) where
A.
B. C.
Figure 7.8 Example of shells in mollusks. A. Turritella terebra. B. Neptunea angulata (drawing
by P.H. Nyst, 1878 [970]). C. Architectonica perspectiva (clear sundial snails, drawing by L.C.
Kiener, 1880 [689]).
d2
Π
z ρ(S
) )
r(S, t d1
p(S )
y
x
Figure 7.9 The generating curve C defining the cross section of the seashell.
t is a parameter. As before, at each point on this curve we attach three orthonor-

mal vectors d1 (s), d2 (s), and d3 (s) such that d3 (s).p = 0. We define for each t, a
generating curve C in the plane spanned by the first two directors of a simple curve:
ρ(S, t) = r1 (S, t)d1 (t) + r2 (S, t)d2 (t), (7.48)
such that ρ(S, t) is a periodic and twice differentiable function in S and t. The surface
of the seashell is then given by
r(S, t) = p(t) + ρ(S, t). (7.49)
With appropriate guesswork and proper choice of functions p and ρ most seashells
can be represented within this framework.
Here, we focus on a particular set of surfaces whose central curve p is given by a
logarithmic spiral and whose generating curve C is a circle as shown in Figure 7.9.
These simplifying assumptions lead to a family of curves with three parameters
whose geometry can be easily explored.
The central curve is taken to be a logarithmic spiral in 3 dimensions, that is,
p(t) ≡ (x, y, z) = Aeαt (sin β cos t, − sin β sin t, cos β), (7.50)
where A, α, and β are real parameters. This curve, sketched in Figure 7.10, is con-
structed as follows: It is an equiangular (or logarithmic) spiral in the horizontal
x − y plane. An equiangular spiral is defined by the property that the angle between
a line from the center to any point on the curve and the tangent at the same point
makes a constant angle. If the parameter t is chosen to be the angle θ in the x − y
plane, then the radial line makes an angle γ = arccot(α) with the tangent vector at
A. z
y
x
y
B. ρ=A sin(β) e α θ
C. z
θ)
ρ( γ
γ θ
x
β
β x
γ
Figure 7.10 A. The logarithmic spiral. B. In the x–y plane, the logarithmic spiral is an equiangular
spiral. The angle between the radial line at a point and the tangent vector is constant and is equal to
γ = arccot(α) (three different points shown). C. In the x–z plane, the curve is tangent to two lines.
These lines have an opening angle 2β.
this point. In the x − z plane, and in any plane containing the z−axis, the projection
of the curve is a sine function modulated by a linear function, so that the curve in
space lies on a cone of opening 2β.
The logarithmic spirals are self-similar in the sense that given n points {pi =
p(t0 + 2iπ ), i = 1, ..., n}, the distance ratios between three successive points,
d(pi−1 , pi )
, (7.51)
d(pi , pi+1 )
is independent of the choice of i. The computation of this ratio also gives access to
the value of α. For instance, from the z-distance between successive coils, we find

1 z i − z i+1
α = log . (7.52)
2 z i−1 − z i
Conversely, we can test the self-similarity of a given seashell by measuring the ratio
of similar points and by checking that this ratio remains close to a constant. It is a
remarkable fact of nature that many seashells are actually extremely close to perfect
self-similar shapes [539].
We can now build the actual seashell surface by considering a generating curve C
around the centerline and scaling its size with the same exponential function. Here,
for simplicity, we take C to be a circle whose radius expands exponentially with t;
that is,
ρ(S, t) = Beαt (cos(S)d1 (t) + sin(S)d2 (t)) , S = 0, . . . , 2π. (7.53)
We still have to define the orientation of the plane with respect to the loga-
rithmic spiral. Historically, two different parameterizations of seashells have been
considered. The first one consists in choosing the plane so that it contains the
z-axis. This constraint is satisfied by taking d2 = ez and d1 to be perpendicular to
the curve. In this case, we have
r(S, t) = eαt (cos t (Asin β + B sin S), −sin t (A sin β + B sin S), −A cos β +cos S) .
(7.54)
This original description of seashells found in the work of Moseley in 1838 is still widely
used. Models of this type are almost exclusively descriptive, since a fixed reference
frame is an artificial construction disconnected from any actual growth process and
the existence of a fixed axis in space does not allow for any variation away from the
logarithmic spiral. However, this representation is particularly easy to implement and
provides an explicit form for the surface [643, 644, 813, 1061, 1062, 1122].
There are, however, some clear disadvantages to identify the plane with a vertical
plane: The plane is defined with respect to the axis and is not perpendicular to the
tangent vectors. Therefore, this representation is closely linked to the particular choice
of a logarithmic spiral as a central curve. For other choices of central curves, this
representation cannot be generalized. Thus, it is natural and mathematically appealing
to define to be the normal plane to the curve at every point. This choice was originally
proposed by Okamoto [981, 982] and subsequently adopted by multiple authors [381,
653, 813]. In this case, we identify the directors d1 and d2 with, respectively, the normal
and binormal vectors which, together with the unit tangent vector d3 define the Frenet
frame:
∂t p ∂t d3
d3 = , d1 = , d2 = d3 × d1 . (7.55)
|∂t p| |∂t d3 |
In the case of a logarithmic spiral, a straightforward computation leads to

cos t +α sin t sin t − α cos t
d1 = − , ,0 , (7.56)
δ1 δ1

α(sin t −α cos t) cos β α (cos t +α sin t) cos β δ1
d2 = − , , − sin β , (7.57)
δ1 δ2 δ1 δ2 δ2
Figure 7.11 An example of a seashell surface produced by Equations (7.59–7.61) with A = 1, ρ =

0.17, α = 0.1, β = 0.3. Note that the solid curves on the surface are either at a constant angle in the
normal plane or in the normal plane (circular cross section). The inclination of the normal planes
clearly indicates that the circular aperture is not in a plane containing the z-axis.

(− sin t + α cos t) sin β (cos t + α sin t) sin β α
d3 = ,− , − cos β , (7.58)
δ2 δ2 δ2
√
where δ1 = α 2 + 1 and δ2 = α 2 + sin2 β. The surface of the shell is then given by
x(S, t) = eαt [A sin β cos t + ρ (d1x sin S + d2x cos S)] , (7.59)

y(S, t) = eαt −A sin β sin t + ρ d1y sin S + d2y cos S , (7.60)

z(S, t) = eαt −A cos β + ρ (d1z sin S + d2z cos S) , (7.61)
where d1x = d1 .ex and so on. We define the variables S ∈ [0, 2π] and t ∈ [0, 2nπ]
where n is the number of coils. The above parameterization defines a 4-parameter
family of surfaces. However, we can set, without loss of generality, the distance of the
central curve at t = 0 to the z-axis to be A = 1 and measure all lengths with respect
to this distance. Doing so, we obtain a 3-parameter family of shapes parameterized by
α, β, ρ. The parameter ρ is the distance from the central curve to the opening at time
t = 0. An example of such a seashell surface is given in Figure 7.11.
7.2.3 Accretive Growth of Seashells
The previous representation of seashells is purely geometric as it describes shapes from

an external viewpoint. We now compare this approach to an intrinsic local view of
seashells by describing growth as a local accretion process. To do so, we first relate the
local rate of growth to the growth velocities and the evolution of the growth axis.
First, we impose shape invariance, shape planarity, and self-similarity in space as
defined in Section 7.1. These three conditions lead to two possible growth laws for
linear or planar growth given by (7.26) and (7.47); that is,
q1 = −c, q2 = b1 + b2 cos(S), q3 = 0, (7.62)

q1 = −λ(t)c, q2 = λ(t)(b1 + b2 cos(S)), q3 = 0. (7.63)
These two laws produce identical structures, and only the time evolution can distinguish
between the different growth processes as noted in [1077, 1267]. Note that (7.62) is not
time invariant and so the coiling rate, i.e. the time necessary to complete one full whorl
in the terminology of seashells, is not fixed. The time to coil increases exponentially
while the aperture size increases linearly.
For the process (7.63), the time to complete one coil is fixed, independent of size
or time, hence the material deposition rates must increase exponentially. This growth
process appears to be inconsistent with the observation that shell growth rates decrease
with increasing octogenetic age [104, 652, 1130]. We can directly relate the area growth
rate to the deposition velocity. We define the material rate ṁ as the rate of increase of
surface area, as follows

2 1/2
ṁ(t) = |q| ds = q1 + q22 + q23 ds. (7.64)
C C
If we assume that the material rate, ṁ is constant in time, we obtain the self-similar
form
c b1 + b2 cos(S)
q1 = − , q2 = , q3 = 0. (7.65)
λ λ
This form has a coiling rate logarithmic in time while the radius increases as t 1/2 .
Therefore, assuming a constant growth rate, the existence of a self-similar accreted
structure, generated by a circular opening, implies that the growth velocity must fol-
low (7.65). Further assumptions on the growth rate can match either microscopic models
of accretion or data on growth.
7.2.4 Other Accreted Structures
Once the key ingredients of the local accretion model are understood, it is fairly easy
to understand the balance between the different components of the growth vector to
reproduce different shapes observed in nature. For instance, several horns and seashells
are shown in Figure 7.12 with their associated growth vectors.
Another example with a noncircular generating curve is the bivalve shell shown
in Figure 7.13. This shape was obtained by using a simple initial curve and binormal
growth that is linear along a given direction with time varying coiling rate.
A.
q1=-(c+k sin(S )),
q2=b1+b2cos(S )
Turitella seashell
q3=k cos(S)
B.
q1=-(c+k(t) sin(S )),

q2=b1+λ b2 cos(S)
q3=k(t) cos(S),
with k =k1+k 2t Antelope horns
C.
q2=b1+b2(t) cos(S)
with b2 of form:
b2
b1
t1 t2 t
-b1
plus global velocity: Nipponite shell
{
k1ey : t<t1
qg= k2ex : t1<t<t2
-k3ey : t>t2
Figure 7.12 Examples of surfaces generated by accretive growth. A. A Turritella shell. B. Antelope
horns. C. A Nipponite shell. The form of the growth velocity field and a sample structure are shown.
In each case, an explicit solution for the curve evolution can be obtained. Parameters: for A: b1 = 1,
b2 = 10, c = −0.025, k = 0.32, and initial radius 0.1; for B: b1 = 0.75, b2 = 0.3, c = −0.04, k1 =
0.25, k2 = 0.02, and initial radius 0.1; for C: t1 = 12, t2 = 20.4, k1 = 0.8, k2 = 0.73, k3 = 1.2,
and radius 1 (adapted from [914]).
A.
growth axis
cross section
B.
Figure 7.13 Noncircular apertures. A. Growth of a bivalve shell using an initial curve r0 =
(R cos(S), R sin(S), 0) with R = 1 + 0.05 cos(10S) and velocity q1 = −c(r0 · d1 (S, 0)), q2 =
b1 + b2 (t)R cos(S), q3 = −c(r0 · d3 (S, 0)) (with c = 3, b1 = 0.1, b2 = 5 − 0.5t). B. Various
seashells obtained from different initial aperture profiles (explicit values of the parameters can
be found in [915]).
7.2.5 The Role of Mechanics in Morphological Patterns
The kinematic analysis of accreted growth in the previous section reveals that the loga-
rithmic spiral structure of seashells is controlled by simple local laws. It also reveals the
existence of a growth gradient along the aperture. How this growth gradient is estab-
lished and maintained through distant points so that the shell recognizes the growth
direction remains a mystery and to-date there is no model or even a qualitative expla-
nation for this basic property.
Nevertheless, we can use this basic growth structure as the underlying evolution of
the shell shape and focus our attention on the morphological patterns appearing on top
of the basic coiled shape. There are two basic morphological patterns superimposed on
the self-similar structures: commarginal or antimarginal ornamentations.
Commarginal ornamentation or transverse ribbing is a common feature of a number
of mollusk shells. It is characterized by an oscillatory component to the dilation in the
aperture as shown in Figure 7.14. The generating curve remains similar up to a global
dilation factor.
Figure 7.14 Commarginal ornamentation in Dictomosphinctes. (picture courtesy of Régis Chirat).
Antimarginal ornamentation are patterns that are orthogonal to the aperture. There-
fore, they are a modification of the shape of the generating curve different from a simple
dilation. This includes antimarginal ribs as well as more complex patterning such as
the spines shown in Figure 7.15 and fractal-like structures.
These morphological patterns are created through the interaction of the soft mantle
and the hard shell. At each accretion step, the mantle, a soft tissue growing with its
own timescale, may not take the exact shape of the aperture. This potential mismatch in
growth between accretion and soft tissue growth is responsible for the patterns observed
empirically by a number of authors [189–191, 1193] and has been proposed as a theo-
retical mechanism to explain pattern formation in [531, 905, 906]. Following this idea,
Moulton and coauthors developed a mathematical model based on the complex interac-
tion between accretion and soft mantle growth to explain various known morphological
patterns [202, 914–916].
7.2.5.1 Commarginal Ornamentation
Commarginal ornamentation is particularly apparent in some species of ammonites.

Ammonites are an iconic group of extinct cephalopods that are characterized by nearly
perfect logarithmic spiral shells with regular ribbing patterns. This pattern is almost
A.
B.
Figure 7.15 Antimarginal ornamentation: Spines in A. Murex pectem and B. Chicoreus (Triplex)
palmarosae. (pictures courtesy of Régis Chirat).
A.
attaches to mantle in compression,

grows shell edge shell expansion increases
mantle edge
- thin elastic ring
shell
edge mantle in tension,
shell expansion decreases
B.
Figure 7.16 Model of commarginal ornamentation. A. Model of growth showing the mantle and
the shell layer, and the attachment forces acting between them [915]. B. Commarginal ribs in
Peltoceras schroederi and primary variables for the model. The shell has radius s and orientation
angle φ with respect to the growth axis z [916].
universally present in ammonites and forms along the direction of the accreted shell,
appearing as an oscillation in the expansion of the shell cross section as illustrated in
Figure 7.14.
We model the mantle as a circular growing elastic ring with a given growth rate as
shown in Figure 7.16A. When the mantle adheres to the current shell edge, it is subject
to both a compression/stretching force and an attachment force. We assume that the
compressive force acting on the mantle is sufficiently small so that it remains circular
and that the equilibrium diameter is a balance between these two forces.
At equilibrium, the mantle is not in a stress-free state, and we assume that the
deposition of new layers is controlled by the level of stress through an adjustment of
the angle of deposition, as shown in Figure 7.16A. If the mantle is in compression,
it increases the angle of deposition as to increase the radius of the new aperture and
therefore relieve its stress. Conversely, if the mantle is in tension, the angle is decreased
and so is the rate of increase of the aperture radius. This simple law naturally creates
oscillation of the radial profile as a function of time due to a mismatch between accretion
and growth.
Let kg be the constant growth rate, so that in a time t, kg t is the length of material
added to the shell. The angle of material deposition is denoted by φ. Denoting rs as the
shell radius, it follows that
ṙs = kg cos θ sin φ, (7.66)
where θ is the fixed angle of the growth vector between the normal and tangent direction.
The key assumption is that the mantle can change the angle φ based on its stress level.
Let n 3 be the axial stress in the mantle edge; note that n 3 > 0 if the mantle is in tension
and n 3 < 0 if the mantle is in compression. This axial stress n 3 depends on the shell
radius rs as well as the diameter of the mantle edge in the unstressed state, determined
by the growth function γ . The feedback between the deposition angle and the tension
in the mantle is modeled by
kφ
φ̇ = 1 − e K n3 . (7.67)
rs
The system of ordinary differential equations (7.66–7.67) can be solved numerically

once the tension is computed from either rod theory or three-dimensional elasticity, as
shown in [915]. An example of a three-dimensional structure combining the kinematic
of an expanding logarithmic seashell with an oscillating radii provided by this model is
shown in Figure 7.17B where the typical ribbing pattern is obtained through this feed-
back process. A detailed analysis of ammonite ribbing leads to an excellent match with
known forms as shown in Figure 7.17. This model provides a mechanistic explanation
for most of the trends and patterns observed in fossil records [916].
7.2.5.2 Antimarginal Ornamentation
We model the formation of antimarginal ribs by assuming that the mantle is a circular
growing inextensible elastic ring with a given growth rate attached to the aperture. We
further assume that the aperture at the next accretion step depends on the shape of the
mantle since it creates a template for the next accretion cycle. Since the growing mantle
will exhibit an excess of length compared to the previous growth increment, it will be
subject to buckling upon adhering to the aperture, as shown in Figure 7.18. The new
deposited layer of shell material takes the deformed shape set by the next foundation
upon mineralization. Therefore, the mechanical deformation of the mantle dictates the
evolution of the shape of the aperture.
The theory of growing rings developed in Chapter 6 can be directly coupled to an
evolving foundation. The centerline of the mantle is the curve ρ with total length l, and
ˆ We assume that lˆ > l to account
the centerline of the aperture is r̂ with arc length l.
for growth of the mantle, and that the connection provided by the periostracum acts as
a linear foundation for the mantle. Note that in this configuration the mantle induces
both a body force and a body couple as sketched in Figure 7.19.
Following the notation of Section 5.1.3, we write the normal and binormal vectors
A. B.
C. D.
Figure 7.17 Commarginal ornamentation. A. Ammonite Promicroceras. B. Simulated ammonite.

Stress in the mantle governs the dilation of the shell edge and is combined with a fixed rule for
coiling (AB from [914]). C. Theoretical rib profile overlaid on a shell of Peltoceras schroederi.
D. Asymmetric ribbing profiles produced by a model with bimodal elastic response in tension and
compression. Left: Stephanoceras, Right: Xipheroceras (CD from [916]).
ν = − cos(S/γ )ex − sin(S/γ )e y , β = ez . (7.68)
The angle of rotation of the cross section due to the applied couple is a nonzero register
angle, ϕ, that is used to define the director basis for the unperturbed system:
(0)
d1 = ν cos ϕ + β sin ϕ, (7.69)
(0)
d2 = −ν sin ϕ + β cos ϕ. (7.70)
elastic rod (mantle)
attaches to mantle deforms
rigid foundation (aperture)
Figure 7.18 Mantle buckling model. The shape of the deformed mantle, described by the curve r̂,
is found as the minimizer of the sum of bending energy and spring energy due to attachment to the
foundation given by the curve ρ.
According to (5.23), the components of the Darboux vector are

(0) (0) (0)
u1 , u2 , u3 = γ −1 sin ϕ, γ −1 cos ϕ, 0 . (7.71)
The foundation is assumed to be a circle of radius 1 located in the x–y plane, i.e.,
ρ = −ν. Therefore, the foundation force is of the form
f = k(ρ − rA ), (7.72)
with rA = r − aβ and a is the radius of the rod. We are interested in computing the first
bifurcation to a noncircular solution. Before bifurcation, the rod’s centerline is a circle
of radius γ = L/2π, located in the plane z = h. Mechanical equilibrium is satisfied if
the plane of the ring satisfies h = a cos ϕ, and the register angle solves
E I γ −1 sin ϕ + ka cos ϕ(a sin ϕ − γ + 1) = 0, (7.73)
where E I is the bending stiffness of the mantle.

The critical length of the mantle at which the first bifurcation occurs is computed by
following the general method given in Section 6.5. In Figure 7.19, γ ∗ is plotted against
the foundation stiffness. We note that within this context, modes 2 and 3 do not exist
for all values of k, and mode 2 has a minimum value.
Next, we can apply this basic idea to compute the growth of a bivalve clam shell.
The initial foundation is taken to be a semicircle with a fixed linear increase in length
occurring at each increment of mantle growth. Fixed rates of coiling, dilation, and
translation are imposed with the shape change. Figure 7.20 shows the evolution of the
fixed plane curve as well as the full surface produced when mechanics is coupled with
kinematics. It can be noted that the folds of the aperture deepen over time through the
incremental mantle deformation process.
initial
grown
foundation
Figure 7.19 Left: Critical buckling growth in the case of the foundation underneath the initial
ring. The dashed line is the envelope of the critical growth—it is not monotonic, and suffers a
discontinuous jump in the transition from mode 3 to mode 4. Right: Growing ring with foundation
underneath.
7.2.5.3 Spine Formation
Spines are also antimarginal ornamentation since they change the shape of the gen-
erating curve. However, in distinction with the previous example, they emerge during
phases where there are large excesses of length in the mantle as shown in Figure 7.21.
The formation of spines can be understood in terms of the changing morphology of the
shell edge, driven by the growth and elastic response of the mantle. The basic idea is to
predict the form that is incrementally generated by the edge of the thin elastic mantle
with an excess of length relative to the previously formed and fixed configuration. The
mantle edge is modeled as an inextensible elastic rod and is assumed to grow at a fixed
rate. At each growth increment, the mantle adheres to the current shell edge through
the elastic periostracum. As the mantle has grown since the last secretion, it is longer
than the shell edge and deforms upon adhering. The shape of the deformed mantle is
the elastic equilibrium shape obtained as a balance of two components: the bending
moment of the mantle dependent upon the geometric curvature of the mantle edge, and
an attachment force between the mantle and the shell edge (the template that serves as
a mechanical rigid foundation).
The development of the shell proceeds by first incrementally increasing the length of
the mantle edge; second, determining the mantle shape through mechanical equilibrium,
and, lastly, by updating the shell edge as the shape of the deformed mantle. The emergent
form is governed by the length dimensions and elastic properties of the mantle edge,
the strength of adhesion, and the marginal growth rate, i.e., the excess of length as
A.
B.
Figure 7.20 Growth of a giant clam shell (Tridacna gigas). A. Mechanical deformation of the
mantle dictates the evolution of the aperture shape. B. Combined with an overall logarithmic growth
to simulate a three-dimensional shell [914].
shell edge
mantle edge
periostracum
...
Figure 7.21 Modeling spine formation. The elastic mantle edge, showing an excess of length
compared to the previous and fixed shell edge, deforms and then secretes new material in the
deformed shape [202].
variation,
Growth rate,
Figure 7.22 Phase diagram showing the effect of the growth rate and the degree of stiffness
variation on the evolving spine. Variations in spine morphology among species of the same fam-
ily, Muricidae, are shown for comparison: A. Hexaplex erythrostomus, B. Bolinus brandaris, C.
Pterynotus phyllopterus, D. Ceratostoma burnetti. Scale bar is 10 mm. Reproduced from [202].
compared to the previous growth increment. Examples of these processes are given in
Figure 7.22.
Chapter 8
Membranes and Shells
The two-dimensional structures described in the previous sections were grown from
hard materials whose kinematics was either prescribed or derived from the interaction
with a soft organism. We now turn our attention to two-dimensional biological struc-
tures that are soft and best described as either morphoelastic membranes, plates, or
shells. We start with a discussion of the elastic response of axisymmetric membranes
and shells.
Our main motivating example will be the description of tip growth in filamen-
tous cells such as fungi, actinomycetes, root hairs, and pollen tubes introduced in
Section 1.1.1. Although there are many fundamental differences in the structure, life
cycles, size, and function of these diverse organisms, the overall pattern of growth
appears to share certain fundamental features; most notably, in these systems the
hyphal growth is apical, i.e., the growth and incorporation of new cell wall material
is concentrated at the tip of the cell. The fact that filamentary cells grow with this
general morphology reflects an interplay between the anisotropy of the cell wall, the
mechanical stresses, and the growth process. We will explore this inhomogeneous
morphoelastic process in Section 9.2.
Since we are interested in the fundamental problem of tip growth we further
focus our attention on the case where the surface remains axisymmetric at all times.
This constraint limits the possible deformations to those that respect this particular
symmetry, but it allows for the derivation of a full mechanical theory based on simple
fundamental principles.
208 8 Membranes and Shells
Before proceeding, we comment on various terminology. Depending on the field

of study, the word “membrane” has different meanings. In biology, it is used to char-
acterize a thin layer separating two compartments with no precise physical properties.
It is sometimes equivalent to the term used in biophysics and sometimes equivalent
to the term used in mechanics. In mechanics, precise definitions are available: A
membrane is a two-dimensional elastic structure in a three-dimensional space that
can resist tension but not compression or bending. A plate is an initially flat structure
that resists bending, tension, and compression. It can be unshearable or shearable
depending on the theory. A shell is an initially curved surface that resists bending,
tension, and compression. It can be unshearable or shearable. In biophysics, a lipid
bilayer is also called a membrane, but it is modeled as a fluid membrane, that is,
a shearless structure that is fluid along the surface. Some biophysical membranes
can resist bending and stretching, whereas others support stretching but not bending.
Technically, fluid membranes are elastic shells or elastic membranes with a special
material symmetry group. They can be derived from standard rational mechanics
arguments within the theory of one-director Cosserat surfaces or a Kirchhoff–Love
shell with constraints [278, 279, 1188].
8.1 Elastic Membranes
We begin by considering an extensible axisymmetric elastic membrane filled with

an incompressible viscous fluid under pressure, whilst neglecting shear and bending
resistance. This type of formulation has been used successfully to describe the shape
of red blood cells and other biomembranes [1135, 1174] and we adapt it to include
the effects of pressure induced stretch, growth, bending, and geometry-dependent
elastic properties of both membranes and shells based on the derivations in [353,
479, 480, 1340]. First, we give a full derivation of the equations describing an elastic
membrane under loads from basic principles. As usual in the derivation of an elastic
theory, we proceed in three key steps by considering kinematics, mechanics, and
constitutive laws.
8.1.1 Kinematics
As shown in Figure 8.1, the axisymmetric membrane surface S is defined by revolving

a planar curve C around the z-axis. The planar curve C is parameterized by the material
coordinate S counted from the intersection O of the surface with the z-axis. At any
given time t, the current configuration is characterized by the curve C:
r(S, t) = r(S, t)er + z(S, t)ez , S ∈ [0, L], (8.1)

8.1 Elastic Membranes 209
A. B.
r y
ν
τ
τ ν
S ϕ z
s=s
x
S
θS rS
s = S =0
O z
Figure 8.1 Basic membrane and shell geometry. A material point S is measured by its arc length,
s(S) from the apex of the shell and its position r(S) on a curve C . Here, ν and τ denote the normal
and tangent vectors at that point. The angle θ(S) is the angle between the normal direction and the
z-axis. The surface is assumed to be axisymmetric with an azimuthal angle ϕ.
where the vectors er and ez are the canonical basis vectors in the r − z plane. It
is convenient to parameterize the curve C by the pair of functions (r(S, t), θ(S, t))
where θ(S, t) is the angle between the normal to C at S and the z-axis. The arc length,
measured from O, is s = s(S, t). Before an elastic deformation and keeping with
the general formalism used so far, the material parameter S is identified with the
arc length in a stress-free configuration, so that S = s(S, 0). The initial stress-free
shell configuration is referred to as the reference configuration and its shape is fully
specified by R(s) = r(S, 0).
The geometric variables satisfy the equations
∂r ∂z
= cos θ, = − sin θ, (8.2)
∂s ∂s
from which we define the local unit tangent, τ and normal vectors ν:
τ = cos θer − sin θez , (8.3)

ν = sin θer + cos θez . (8.4)
Two other important measures of the geometry of the surface are the principal cur-
vatures [169, 496] which are given by
∂θ sin θ
κs = , κϕ = . (8.5)
∂s r
Figure 8.2 Deformation of

an axisymmetric surface. r
The initial configuration
(dashed) parameterized by current
R(S) is mapped to the
current configuration (solid)
parameterized by r(S). A reference
material point in the
reference configuration with R(S)
material coordinate S is S
mapped to a new point with
arc length s(S). The radial r(S)
s(S)
stretch is the ratio αϕ = r/R.
z
Next, we consider axisymmetric deformations of the surface from the reference

to the current configuration as shown in Figure 8.2. These deformations are specified
by two variables, the radial stretch and the tangential stretch. The radial stretch
r(S, t)
αϕ (S, t) = , (8.6)
R(S)
at a given material point is the ratio of the new radius r to the original radius R at
that material point S. The tangential stretch, or just stretch for short,
∂s(S, t)
αs (S, t) = , (8.7)
∂S
is the amount of stretching of the body coordinates with respect to the reference arc
length. These two stretches (αϕ , αs ) completely define the deformation of the initial
axisymmetric surface.
A physical membrane also has thickness. During deformation, the thickness varies
from its initial value H to its current value h(S, t) and has an associated normal stretch,
αn = h(S, t)/H, measuring changes in the normal thickness of the shell. Depending
on the constitutive assumptions, the change of thickness during deformation can take
different forms. If we assume that the membrane is made out of an incompressible
elastic material, αn , is simply related to αs and αϕ by the incompressibility condition
αs αϕ αn = 1. If, however, we model a lipid bilayer, the material cannot be modeled as
a three-dimensional elastic material; a convenient choice to describe a fluid membrane
[403] is to choose both αn = 1 and αs αϕ = 1, denoting area incompressibility [545].
8.1.2 Mechanics
We now define the tensions acting on the membrane surface. Let ts be the line tension
on the surface along the tangent τ , in the direction of increasing arc length. The line
tension ts has the dimensions of a force per unit length. Similarly, we define tϕ as the
line tension along the binormal vector β = ν × τ , normal to τ in a plane tangent to
S, and in the hoop direction. That is, the direction of increasing azimuthal angle ϕ
in Figure 8.3. We consider an infinitesimal piece of this surface as seen in Figure 8.4
subject to an external body stress f = τc τ + Pν. The shear stress τc can be taken to
represent the drag forces exerted by the surrounding medium on the membrane and
the pressure P is the pressure difference across the membrane.
Next, we balance the total force acting over that surface element of area rdsdϕ,
that is,

(ts τ r)|s+ds − (ts τ r)|s dϕ + (tϕ β)|ϕ+dϕ − (tϕ β)|ϕ ds + frdsdϕ = 0. (8.8)
In the limit ds → 0 and dϕ → 0, we obtain

∂ ∂
(rts τ ) + (tϕ β) + rf = 0. (8.9)
∂s ∂ϕ
Using the fact that ∂s τ = −θ ν and ∂ϕ β = τ cos θ + ν sin θ, we find three scalar
equations
∂tϕ
= 0, (8.10)
∂ϕ
θ rts + tϕ sin θ − rP = 0, (8.11)
∂
(rts ) − tϕ cos θ + rτc = 0. (8.12)
∂s
x
Pressure difference P
ts τ
tϕβ
ϕ
z
r(S)
s=S=0
Figure 8.3 The surface S with the two principal tensions (ts , tϕ ) along the vectors τ and β.
ϕ ϕ
r ϕ
r ϕ
ϕ ϕ ϕ
Figure 8.4 An infinitesimal surface element is subject to an external body force f as well as contact
forces resulting from the rest of the surface.
The first equation expresses the fact that the force in the hoop direction remains
constant along that direction as to maintain the axisymmetry. Using definition (8.5),
the remaining two equations can be written as
P = κ s ts + κϕ tϕ , (8.13)
∂(rts ) ∂r
− tϕ + rτc = 0. (8.14)
∂s ∂s
Equation (8.13) is a generalized version of the celebrated Laplace’s equation, some-
times called Young’s equation or Young–Laplace’s equation in biophysics [1223]. It
expresses the fundamental balance between the pressure across a surface, the curva-
tures, and the line tensions acting on a surface.
In many problems involving surface tension in liquids, we have tϕ = ts = σ [266].
In this case, Laplace’s equations reduces simply to the traditional form
P = (κs + κϕ )σ. (8.15)
Furthermore, if the surface is locally spherical, κs = κϕ = 1/R, and (8.15) special-

izes to P = 2σ/R.
We further investigate Equations (8.13–8.14) in the absence of shear stress. In
the case of a constant pressure P we now show that there is a first integral, that is, a
function of the variables that remains constant along the curve C, given by [1038]

C = r 2 2ts κϕ − P . (8.16)
In particular, for all solutions (r(s), θ(s)) crossing the z-axis, we have C = 0.
8.1.2.1 Derivation of the Constant
In order to derive the constant (8.16) it is convenient to collect together the governing
equations in the form
P = ks ts + kϕ tϕ , (8.17)
∂(rts )
= tϕ cos θ, (8.18)
∂s
∂r
= cos θ, (8.19)
∂s
where ks , kϕ are given in (8.5). From (8.5) we easily see the purely geometric rela-
tionship
∂(rkϕ ) ∂r
= ks . (8.20)
∂s ∂s
On multiplying both sides of (8.17) by ∂r/∂s and using (8.20) it follows that
∂(rkϕ ) ∂r ∂r
ts + kϕ tϕ =P . (8.21)
∂s ∂s ∂s
From (8.18) and (8.19) we observe that
∂(rts ) ∂r
= tϕ cos θ = tϕ , (8.22)
∂s ∂s
and hence
∂r ∂(rts )
kϕ tϕ = kϕ . (8.23)
∂s ∂s
Equation (8.21) can now be written as
∂(rkϕ ) ∂(rts ) ∂r
ts + kϕ =P , (8.24)
∂s ∂s ∂s
and on multiplying both sides by r this last relation can be re-expressed in the form
∂(r 2 ts kϕ ) ∂r
= Pr . (8.25)
∂s ∂s
Integrating both sides with respect to s gives

C
r 2 ts kϕ = Prdr + , (8.26)
2
where C is an integration constant. Thus, for constant pressure P = P0 we have
C = r 2 (2ts kϕ − P0 ). (8.27)
As noted, this first integral is particularly useful in problems where the pressure is
constant and the axisymmetric surface intersects the z-axis.
8.1.3 Constitutive Laws
8.1.3.1 Linear Laws
We now relate stresses to strains by introducing a constitutive relationship that

expresses the elastic response of the membrane under load. To start, we consider a
simple law for isotropic incompressible membranes in small deformations and relate
the components of the material strain tensor to the extension ratios [353, 1174]
1 2
es = (α − 1), (8.28)
2 s
1
eϕ = (αϕ2 − 1). (8.29)
2
For small deformations, where both es and eϕ are small, a first order expansion of
the general nonlinear stress–strain relation for isotropic membrane (presented in the
next section) leads to the simplified law
ts = A(ess + 2eϕϕ ), (8.30)

tϕ = A(eϕϕ + 2ess ), (8.31)
where A = EH/3 characterizes the elastic property of the membrane, with E the
Young’s modulus and H the membrane thickness. Note that A can be a function of
the material parameter S. In terms of the extension ratios the stress–strain relation
can be written as:
ts = A(2αs2 + αϕ2 − 3), (8.32)

tϕ = A(αs2 + 2αϕ2 − 3). (8.33)
8.1.4 A Complete Set of Equations
Using the linear constitutive law and the definition for the curvatures (8.5), the equa-
tions for mechanical equilibrium, in the absence of shear stress, reduce to a system
for the membrane variables for a given initial profile R = R(S)
∂s
= αs , (8.34)
∂S
∂z
= −αs sin θ, (8.35)
∂S
∂r
= αs cos θ, (8.36)
∂S
∂θ αs sin(θ) r 2 + R2 (S)(2αs2 − 3) − prR2 (S)
= , (8.37)
∂S r[2r 2 − 3R(S)2 + αs2 R2 (S)]
where we have introduced p = P/A as the inverse of a characteristic length that

measures the relative effect of the pressure with respect to the stiffness modulus A.
This system is closed by expressing αs as a function of the other variables through
the first integral (8.16),

r2
C = 2Ar sin(θ) αs + 2 2
2
− 3 − r 2 P. (8.38)
R (S)
If we further assume that there is a point on the membrane such that r = 0, we have
C = 0, which leads to
r2 r
αs2 = 3 − 2 2 +p . (8.39)
R (S) 2 sin θ
8.1.4.1 Simple Examples
As a first example, it is straightforward to verify that the inflation of a sphere of radius

R0 into another sphere of radius R1 by a pressure p, is a solution of these equations.
We do this by choosing R = R0 sin(S/R0 ), θ = S/R0 , r = R1 sin(S/R0 ), which
leads to

R0
R1 = pR0 + p R0 + 144
2 2
12

pR0 p2 R02
= R0 1 + + + O(p R0 ) .
3 3
(8.40)
12 288
As a second example, we consider an infinite cylinder whose axis coincides with

the z–axis. In this case, we have θ = π/2, αs = 1, R = R0 , and r = R1 . In this
case, we have

1
R1 = p2 R0 4 + 4R0 2 + pR0 2 (8.41)
2

pR0 p2 R02
= R0 1 + + + O(p3 R03 ) . (8.42)
2 8
In tip growth, the shape of a tip can be approximated as a spherical cap connected
to a long cylindrical support. The two simple solutions for the sphere and cylinder
show that in order to have a membrane that connects a cylinder to a sphere, there
must be a steep change in the parameter p which, to a first approximation, needs
to decrease by a factor six from the sphere to the cylinder. Hence, there must be a
corresponding increase in the elastic parameter A. This simple analysis demonstrates
that in order to have a tip shape supported by an elastic membrane, there must be a
significant change of stiffness close to the tip.
8.1.4.2 Boundary Conditions
Depending on the specific problem, different sets of boundary conditions can be

imposed. If the pressure remains constant and the surface crosses the z–axis, (8.39)
can be used in the system (8.34–8.37). These equations form a system of four equa-
tions for the four unknowns (r, z, θ, s). Choosing S = 0 at one intersection with the
axis, we always have s(0) = 0. Without loss of generality, we can fix the translation
along the axis by setting z(L) = z0 where S = L denotes a reference material point
taken as the end point of the membrane, and z0 is arbitrary. Then, depending on the
nature of the attachment we can set either r(L) = RL (fixed radius at the boundary) or
θ(L) = θL (fixed angle at the boundary), or any combination of these two conditions.
There remains one condition to be fixed at σ = 0. The condition is simply r(0) = 0,
but since R(0) = 0, Equation (8.37) has a geometric removable singularity that leads
to computational difficulties. This problem is easily dealt with by considering the
boundary conditions at σ =
where
1 and expanding all variables in power
of
. The lowest nonzero contribution will dictate the choice of the last boundary
condition.
Finally, if we fix the volume to be a constant rather than the pressure, then the
pressure becomes a Lagrange multiplier that enforces the volume constraint. Com-
putationally, a solution with a given volume can be obtained starting from a guess
pressure and varying the pressure up to the desired volume.
8.1.5 Application: The Shape of Sea Urchins
It was D’Arcy Thompson who first realized the similarity between the shape of
urchins and the shape of sessile drops of liquid sitting on a table [1230, p. 944]:
“The full mathematical investigation of the forms of a drop is not easy, and we can
scarcely hope that the systematic study of the Echinodermata will ever be conducted
by methods based on Laplace’s differential equation; but we have little difficulty in
seeing that the various forms represented in a series of sea urchin shells are no other
than those which we may easily and perfectly imitate in drops.”
The problem of determining the shape of a sessile liquid drop is a classic problem
going back to Laplace [65, 771] and for which many details and properties have
Figure 8.5 A sessile water

drop on a substrate.
been established [363, 364, 1311]. The shape, shown in Figure 8.5 is the result of
a mechanical balance between a pressure gradient induced by gravity and surface
tension. Similarly, it has been argued that the shape of an urchin is the result of
a balance of forces generated by the coelomic pressure (the internal pressure that
invertebrates develop in their body cavity), the podia (the tubed feet), and self-weight.
The skeleton of sea urchins, also known as a test, is made of a number of calcite
plates that are loosely attached together at their sutures through collagen threads.
During growth, new plates are inserted at the apex and each plate expands by a
number of accretion, shrinkage, and fitting processes at the edge not dissimilar to
the process described in the previous chapter for seashells [328, 1361]. It has been
argued that despite their rigid shell-like structure, the great number of plates and their
loose attachment during growth provide a flexibility that allows them to deform and
relax through the action of competing forces of gravity and pressure, such that it is
best described by the action of forces on an elastic membrane [668].
Therefore, we follow here the original idea of D’Arcy Thompson as well as the
models of Barons [60] and Ellers [328, 329] to determine the basic shape of sea
urchins as an application of the axisymmetric elastic membrane model. The basic
assumption in this model is that growth takes place so that the line tension ts and tϕ
are equal at all points as to maximize strength. In that case, this problem is exactly the
problem first described by Timoshenko of determining the optimal shape of a water
tank [1233, p. 385]. Since the line tensions are assumed to be equal, it follows directly
from (8.37) that ts is constant at all points on the surface. Therefore, to compute such
a shape we assume that the urchin sits flat so that the pressure increases linearly with
gravity P = ρgz. We are then left with (8.37) and (8.2) which can be written
sin θ
θ = az − , (8.43)
r

r = cos θ, (8.44)

z = sin θ, (8.45)
where we have flipped the z-axis so that increasing z corresponds to increasing

pressure. The parameter a = ρg/ts is the inverse of the square of the capillary length,
the characteristic length at which both tension forces acting on the membrane and
gravitational forces are of the same magnitude.
The boundary condition at the apex are θ(0) = 0, r(0) = 0 and z(0) = 2b where
the length 2b sets the pressure at the apex. Without loss of generality, we set b = 1
and the urchin shapes are defined by a single parameter a up to a dilation.
To solve this system numerically, we consider a first order expansion of the bound-
ary conditions around s = 0 so that
θ(
) = a
, r(
) =
, z(
) = 2,
1. (8.46)
Equations (8.43–8.45) with conditions (8.46) reduces to a simple initial value prob-
lem that can be integrated up to the point S = L at which θ(L) = π, assuming that,
unlike a drop, the angle between the flat interface and the curve is exactly π.
As expected, for large values a, the shape tends to become spherical and, as a
decreases it becomes increasingly flat, since the extent to which gravity plays a
role decreases with respect to the size of the urchin. This computation leads to a
one-dimensional morphospace. A morphospace is a space of shapes parameterized
by several parameters that can be matched to actual shapes found in nature. In our
case, we have a single parameter a, and the morphospace reduces to a morpholine.
The match against various species is actually surprisingly good considering that we
have a one-parameter family and that the assumptions are only vaguely connected
to biological reality. An example of such a fit is shown in Figure 8.6.
The key assumption on the equality of the line tensions that gives rise to this drop-
like solution is particularly difficult to justify as these shapes cannot be obtained by
starting from a spherical elastic membrane filled with water, in contradiction to
D’Arcy Thompson’s claim. Indeed, if the tensions are equal, they are constant and
so are the stretches. In particular αϕ is constant which implies that r(S) = αs R(S).
Therefore, the shape obtained must be realized from a stress-free membrane with
A. B.
Figure 8.6 A. An urchin test together with a fit obtained for a = 3.98. B. The full shape obtained
by integration of the model.
A.
B.
Figure 8.7 A computational model for a sea-urchin taking into account the creation, accretion,
and fitting of plates. From Zachos [1361].
the same shape. Therefore, if the drop solution is indeed a valid representation of
the skeleton shape in urchins, there must be both a relaxation process that constantly
evolves the reference shape to the current shape, and a growth process that balances
the deposition of new material in order to equilibrate the line tensions. How line
tensions are being sensed and how the process is maintained and controlled during
the growth of urchins remains a mystery, but one that could be partially elucidated
by looking at plausible models of growth.
Finally, we mention an interesting computational model by Zachos [1361] who
looked at the creation and evolution of plates developing from the apex. Some of
the model results are shown in Figure 8.7. Note, however, that there is no mechanics
in this specific model and the drop-like shape is obtained by mapping a tessellated
spherical solution to the drop solution shown above.
8.2 Nonlinearly Elastic Shells
Next, we consider axisymmetric shells, that is, elastic objects that can be represented
by a surface, support bending, and have a non-flat unstressed shape. At the level of
kinematics, there is no difference in the geometric description of the deformations.
Therefore, we adopt the same notations and move directly to mechanics.
8.2.1 Mechanics
The equations for mechanical equilibrium, including bending moments can be

obtained by balancing linear and angular momenta on a small surface element. Long
and tedious force and couple balances that are better done in the privacy of one’s
office lead to a system of three equations [353, 695]
d(rqs )
= rqn − r κs ts + κϕ tϕ , (8.47)
ds
d(rts )
= tϕ cos θ + rκs qs − rτs , (8.48)
ds
d(rms )
= mϕ cos θ − rqs , (8.49)
ds
where ts and tϕ are, respectively, the meridional and azimuthal stresses; ms and mϕ
are the bending moments; and qs is the shear stress normal to the surface. In (8.47),
which represents the balance of normal stresses, qn represents the total normal stress
exerted on the shell. If the problem is pressure driven then qn = P, namely the
pressure difference across the shell; and if it is a combination of pressure and active
forces, represented by some function τn , then qn = P+τn . In (8.48), which represents
the balance of tangential stresses, τs is the external tangential shear stress acting on
the shell and will be used here to represent the friction between the deformed shell
and its environment.
8.2.1.1 Constitutive Relationship
In order to close the system of mechanical and geometric, equations constitutive

relations must be introduced. These are developed through the introduction of an
elastic strain-energy density function W. Assuming material isotropy, it can be shown
that such a strain-energy function, specified as energy per unit volume, can only
depend on the stretches (αs , αϕ , αn ). In Section 11.3, we will present a full theory
of constitutive relationships. For our purpose here, it is sufficient to state that a
convenient way of writing the dependence of the strain-energy density function is to
introduce three strain functions, the so-called strain invariants:
8.2 Nonlinearly Elastic Shells 221
I1 = αs2 + αϕ2 + αn2 , (8.50)

I2 = αs2 αϕ2 + αs2 αn2 + αϕ2 αn2 , (8.51)
I3 = αs2 αϕ2 αn2 . (8.52)
Then, the strain-energy density function is
W = W (I1 , I2 , I3 ), (8.53)
Assuming that the shell is incompressible, we have
1 ∂W
αn = , = 0, I3 = 1. (8.54)
αs αϕ ∂I3
Once a function W is given, the constitutive equations relating line tensions to

stretches can be obtained by differentiation of W with respect to its invariants. In our
case, the line tensions are given by the relations

∂W 2 ∂W
ts = 2Hαn (αs2
− αn2 )+ αϕ , (8.55)
∂I1 ∂I2

∂W ∂W
tϕ = 2Hαn (αϕ2 − αn2 ) + αs2 , (8.56)
∂I1 ∂I2
where H is the shell thickness, and the factor Hαn is the wall thickness in the current
configuration [1087]. Using these formulas we note that

∂W ∂W
ts − tϕ = 2Hαn (αs2 − αϕ2 ) + αn2 . (8.57)
∂I1 ∂I2
The choice of W depends on the problem at hand. A popular choice for elastomers
is the Mooney–Rivlin model,
Wmr = C1 (I1 − 3) + C2 (I2 − 3), (8.58)
where C1 and C2 are certain elastic moduli. When C2 = 0, (8.58) reduces to the
so-called neo-Hookean model Wnh = C1 (I1 − 3) and, for small deformations, C1 is
related to the Young’s modulus, E, by E = 6C1 . We note that despite its linearity
in the invariants, the Mooney–Rivlin potential is still capable of describing finite
deformations and is a highly nonlinear function of the stretches.
A variety of nonlinear material responses can be captured by considering other
functional forms of W (See Section 11.7.2), e.g., for soft tissue the Fung model [408],
C1 γ(I1 −3)
Wfu = e −1 , (8.59)
γ
is often used. The parameter γ is used to describe strain-stiffening, that is, the increase
of stiffness as the deformation increases. In the limit γ → 0, the Fung model reduces
to the Neo-Hookean model Wnh . Given any model, we write the constitutive rela-
tionships (8.55–8.56) in the form
ts = Afs (αs , αϕ ), tϕ = Afϕ (αs , αϕ ), (8.60)
where fs , fφ are dimensionless functions and A provides the dimensional factor appro-
priate for the scaling of the equilibrium equations. For the neo-Hookean and Mooney–
Rivlin models, A = 2C1 H = EH/3.
In the limit of small deformations, the neo-Hookean, Fung, and Mooney–Rivlin
models, all give the linear constitutive relations introduced in Section 8.1.3.1. Finally,
we need to specify constitutive relationships for the bending moments. The bending
moments are assumed to be isotropic and proportional to the change in the surface’s
mean curvature, so that we have
mϕ = ms = B(κs + κϕ − 2K0 ), (8.61)
where K0 is the initial mean curvature and B is the bending modulus [353, 712].
8.2.1.2 The Complete Shell Equations and the Membrane Limit
The geometric and mechanical equations can be combined to give a closed system.
It is convenient to express all the derivatives in terms of the material coordinate, S,
leading to
dz
= −αs sin(θ), (8.62)
dS
dr
= αs cos(θ), (8.63)
dS
dθ
= αs κs , (8.64)
dS

dκs cos θ sin θ qs
= αs − κs + , (8.65)
dS r r B

dts cos θ qs τs
= αs A (fϕ − fs ) + κs − , (8.66)
ds r A A

dqs qn sin θ qs cos θ
= αs A − κ s fs − fϕ − , (8.67)
dS A r A r
where (8.65) is obtained from (8.49) using the constitutive relation (8.61) and (8.5)
is used to express κϕ in terms of r and θ. In (8.66) and (8.67), ts and tϕ are expressed
in terms of αs and αϕ through the scaled constitutive relations (8.60–8.60), and
Equation (8.66) is converted into a differential equation for αs by eliminating αϕ
through the relation αϕ = r/R.
The six ordinary differential equations (8.62–8.67) together with the relationships
(8.60–8.61) and αϕ = r/R form a closed system for the variables (z, r, θ, κs , αs , qs )
that can be solved for a given initial profile R(S), elastic parameters A, B, prescribed
normal and tangential stresses qn and τn , and appropriate boundary conditions.
When bending moments are neglected, the shell no longer supports an out-of-
plane shear force, which implies qs = 0. Then, (8.67) reduces to

qn = A κs fs + κϕ fϕ = κs ts + κϕ tϕ , (8.68)
and we recover Laplace’s law as seen before. The system of shell equations then
simplifies to the general form of the membrane equation in large deformations,
namely
dz
= −αs sin θ, (8.69)
dS
dr
= αs cos θ, (8.70)
dS
dθ
= αs κs , (8.71)
dS

dκs cos θ sin θ
= αs − κs , (8.72)
dS r r

dts cos θ τs
= αs A (fϕ − fs ) − . (8.73)
ds r A
As before, using the constitutive relationship ts = Afs (αs , r/R) results in a closed
system of five equations for (z, r, θ, κs , αs ) with initial profile R(s), tangential shear
τs , and boundary values.
8.2.2 Scalings
It is useful to introduce the following dimensionless number for the problem:
Peff w
ξ= , (8.74)
3A
where w is a characteristic length, Peff is a measure of the normal force per unit area
acting on the walls and A characterizes the wall’s elastic properties. The number ξ
gives a measure of the magnitude of the deformation: Shells under different normal
loads and with different elastic properties will experience the same magnitude of
deformation from their initial unstressed state when measured in terms of their length
w. High values of ξ, representing either a soft material or a high pressure, result in
large elastic deformations, whereas low values of ξ, corresponding to either a rigid
material or low pressure, result in small elastic deformations. For a neo-Hookean
material we can relate A to the Young’s modulus E = 3A/H, where H is the thickness
of the shell or membrane, so that
Peff w
ξ= . (8.75)
HE
To set the scales, estimates of the elastic and bending moduli are required. An order of
magnitude estimate can be obtained from the Laplace’s law for a spherical membrane
under pressure. For a pressure difference P across the membrane wall of thickness
H, the sphere radius R, and membrane stresses T , we have
HT
P = 2 , (8.76)
R
and, to a first approximation, these stresses scale as T ∼ E, where E is the membrane’s
elastic modulus.
8.2.3 Application: The Rice Blast Fungus
8.2.3.1 Background
The fungus Magnaporthe grisea, commonly referred to as the rice blast fungus, is a
pathogen that attacks the rice leaves. The fungus attaches to the rice leaf and forms
a dome-shaped structure, the appressorium, in which enormous turgor pressures are
generated and used to blast a penetration peg through the rice cell walls and infect the
plant. A cartoon of the overall process is given in Figure 8.8. The process is as follows:
A conidium, i.e. a small spore, lands on the rice leaf surface. Once attached, a germ
tube grows out of the conidium, hooks into the surface of the rice leaf, and begins to
form an appressorium. The germ tube is eventually shut off from the appressorium
leaving the latter as a separate, independent unit. The mechanics of the free standing
appressorium and how it manages to develop sufficient pressures to blast through the
leaf is the subject of our analysis.
The appressorium, shown in Figure 8.9, adheres to the plant surface by means of a
highly potent adhesive in a ring around its base. The inner region of contact between
the appressorium and leaf surface lacks any obvious appressorial cell wall structure.
germ tube appressorium

conidium
pore ring
penetration peg
glue ring
Figure 8.8 Pictorial representation of the development of the rice blast fungus.
From the available electron micrographs the overall appressorial wall thickness is
of the order of 0.1 µm as illustrated in Figure 8.9. Once firmly attached, a melanin
layer about 100 nm thick develops within the appressorium. This dark melanin layer
allows only water molecules to permeate the wall, and on interaction with an internal
glycerol concentration of approximately 3.2 M [673], enormous internal pressures of
Figure 8.9 A typical wild-type appressorial structure prior to peg penetration [625]. The two
arrowheads at the base of the appressorium indicate the range of the appressorial pore. The dark
inner circle, indicated by the small arrows, is the melanin layer. In this image the appressorium is
sitting on a cellophane substrate.
the order of 3–5 atmospheres (0.3–0.5 MPa) can develop in the appressorium [900].
After melanization, at least a threefold increase in turgor pressure occurs which
may reach up to 80 atmospheres (8 MPa) [624]. Interestingly, despite the large
increase in pressure, there does not appear to be any significant change in the size
of the appressorium. After a period ranging from tens of hours to a few days, during
which time turgor pressure builds up, a penetration peg is blasted through the surface
of the rice leaf [718]. The forces involved are so great that, as demonstrated in
laboratory experiments [623], the peg can penetrate thin sheets of mylar. Estimates
for the forces involved can be found in [481]. Once invasion takes place, a network
of hyphae develop, then break through the leaf surface depositing new conidia and
the life cycle is perpetuated.
8.2.3.2 A Model
We model the appressorium as an axisymmetric elastic shell with basic geometry

shown in Figure 8.10. The characteristic length scale w ≈ 3 µm is chosen to be
that of the appressorial radius. We assume that the appressorium is glued in a small
gn
s
gt (s)
g r(s)
z
Figure 8.10 Basic geometry of the appressorium. The origin of the z-axis is set at the top of the
appressorium. The arc length, s, to a given material point is measured from the top of the shell,
r(s) is the radial distance from the vertical axis to that point, and ν denotes the normal direction.
The shell is taken to be axisymmetric where ϕ is the azimuthal angle. The glue is modeled as a
downward, Hookean force connecting the shell to the leaf surface. The glueing force g is assumed
proportional to the thickness of the glue at each point and is resolved into both normal, gn , and
tangential, gs , components.
region close to the surface through a linear foundation between the shell and the leaf
surface, and proportional to the glue layer thickness at each point. The glueing force
is expressed as
g = −µz ez , (8.77)
where, µ is the foundation stiffness, z is the vertical thickness of the glue layer at each
point, and ez is the vertical unit vector. The glueing force is resolved into normal, gn ,
and tangential, gs , components. The net effect of gn is to increase the shell pressure
in the glueing region. This effect is taken into account by identifying gn with τn ,
while gs plays the role of τs in Equation (8.66). The extent of the glueing region is
characterized by a parameter η representing this quantity in terms of a percentage of
the total arc length from the base of the appressorium to its top. The beginning of the
glue ring is taken to coincide with the initial point in the numerical boundary-value
problem (in effect corresponding to the edge of the appressorial pore). Therefore, if
L is the initial arc length of the profile and in the absence of other external stress, we
can identify
(gs , gn ) for 0 ≤ σ ≤ ηL,
(τs , τn ) = (8.78)
0 otherwise.
To estimate the parameters, we note that a typical radius of a microscopic spherical

shell is about 3 µm, and the recorded pressure is P ≈ 1−8 MPa with a wall thickness
of h ≈ 0.1 µm. The Young’s modulus E is of the order of magnitude
E ≈ 10 − 100 MPa,
which is consistent with that of many other biological materials shown in the elastic
moduli atlases of Ashby et al. [41]. We also note that this estimate is based on a
linear constitutive relationship, so if the radius stays approximately constant under
an increase in turgor pressure, the corresponding estimate of the elastic modulus
increases proportionately. Thus, if the turgor pressure increases fivefold and the
radius is, to a first approximation, unchanged, then the elastic modulus of the wall
could be considered to have increased fivefold. This change of apparent stiffness
can be explained by assuming that the wall material has a nonlinear stress-stiffening
response. For instance, this behavior is compatible with the Fung model with a large
parameter γ.
Elastic shell theory [761] also tells us that, within the framework of linear elasticity
theory, the bending modulus for a thin sheet is given by the relation
EH 3
B= , (8.79)
12(1 − ν 2 )
where ν is the Poisson’s ratio. Although not typically known for many cellular walls,
we can assume ν = 1/2, i.e., an incompressible material. For the same example, this
leads to the estimate
B ≈ 10 − 100 × 10−13 Nm.
Clearly, bending effects will only be significant if the cell wall exhibits regions of very
high curvature. These types of estimates for E and B have been used by Boudaoud
[114] to analyze morphological scaling laws for a large class of microorganisms.
They have been used to distinguish families that are either tension dominated, where
E is the significant parameter, or bending dominated, where B is the significant
parameter.
8.2.3.3 Analysis
A detailed analysis of the effect of the different parameters of the system was per-
formed in [1241] where it was shown that for regular pressure levels after melaniza-
tion, but well before the occurrence of peg penetration, a simple neo-Hookean model
captures correctly the shape of the appressorium, as shown in Figure 8.11.
Of particular interest is the persistence of the shape after melanization and large
pressure increase. Within a neo-Hookean model, a large increase in pressure implies
a large change in shape which is not observed experimentally. This observation
suggests that a Mooney–Rivlin material is not suitable to model the increased rigidity
under high stresses. However, if we use a stress-stiffening model such as the Fung
A. B.
Figure 8.11 Computed appressorial shape for a neo-Hookean model (heavy line). The dashed line
represents the glued portion of the appressorium. Parameters: ξ = 1, µ = 3, η = 0.2, B = 0.1.
The plot in B. is the 3-dimensional version of the numerical solution for the same profile shown in A.
A. Neo-Hookean membrane
B. Fung membrane
z z
8
ξ 0.4
0.2
6
0
4 -0.2
-0.4
2
ξ
-0.6
ξ -0.8
0
ξ -1
x x
-4 -2 0 2 4 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
Figure 8.12 Comparison of pressure increase. A. A neo-Hookean model shows a large increase in
radius when the pressure is raised from ξ = 1 to ξ = 10. B. For a Fung model, the same increase of
pressure ξ = 10 for all curves except the inner one leads to little change in the appressorial shape
when the Fung exponent γ is large enough. Parameters: ξ = 1, µ = 3, η = 0.25, B = 0.1, all
length units scaled to the original width before pressurization.
model (8.59), this effect is well captured as shown in Figure 8.12. This process
clearly establishes the importance of both nonlinearity and large deformations theory
to model significant changes in stress even at the scale of microscopic organisms.
Chapter 9
Growing Membranes
In the previous chapter we gave a theory of axisymmetric elastic shells and mem-
branes. Biological membranes have the additional ability to change their internal
configuration to grow and remodel. Since we are considering biological membranes,
we are mostly interested in systems in which the membrane is under pressure. At the
mechanical level when an elastic shell in its reference configuration is pressurized
it will expand until a new stress equilibrium is achieved in its current configuration.
During the course of this deformation the initially uniformly distributed material
points will become redistributed, typically separated under expansion, as a manifes-
tation of the new strained configuration. Furthermore, the shell expansion leads to a
reduction in the shell wall thickness—although this may not be uniform around the
shell. Once equilibrium has been achieved no further expansion will occur unless
there is a further increase in pressure, or the stresses in the shell are somehow relieved.
Through a number of mechanisms, such as addition of material, unfurling of mem-
branes, contractile phenomena, remodeling of fibers and so on, a biological shell
may further evolve. Due to this internal reconfiguration, if loads are removed, the
shell will retract to a shape that is different from the reference configuration. The
evolution of this new virtual reference configuration characterizes the growth and
remodeling of the biological shell.

232 9 Growing Membranes
9.1 Morphoelastic Membranes
At the mathematical level, the process of growth is achieved by considering an

evolving reference configuration parameterized by a slow time t. At any given time,
the current configuration is obtained by loading the reference configuration with the
pressure P and computing the resulting shape for the given boundary conditions.
The problem is then to describe the evolution of this reference configuration.
Following the discussion in the one-dimensional context of Sections 4.4 and 6.2,
the basic idea of shell morphoelasticity is that the geometric stretches defined between
the initial configuration and the current configurations are decomposed into growth
stretches and elastic stretches. We denote again with a subscript “0” all variables
relevant to the initial configuration, that is, the reference configuration at time t = 0.
We define the initial profile C0 by
R0 = R0 (S0 )er + Z 0 (S0 )ez , S0 ∈ [0, L 0 ], (9.1)
where S0 is, as before, the arc length of the contour with total length L 0 . The geometric
stretches between the initial and current configurations are defined as
∂s r
λs = , λϕ = . (9.2)
∂ S0 R0
We further define an evolving virtual reference configuration by the pair (S(S0 , t),
R(S0 , t)) with initial values S(S0 , 0) = S0 and R(S0 , 0) = R0 , or equivalently, by
the growth stretches
∂S R
γs = , γϕ = . (9.3)
∂ S0 R0
The current configuration at time t is defined by (s(S, t), r (S, t)) and is obtained by
solving the mechanical problem of the shell under load, that is, by computing the
elastic stretches from the reference configuration at time t to the current configuration
∂s r
αs = , αϕ = . (9.4)
∂S R
Therefore, the geometric stretches can be decomposed according to the morphoelastic
decomposition as follows:
∂s ∂s ∂ S
λs = αs γs ⇐⇒ = , (9.5)
∂ S0 ∂ S ∂ S0
r r R
λϕ = αϕ γϕ ⇐⇒ = . (9.6)
R0 R R0
9.1 Morphoelastic Membranes 233
This decomposition is a particular case of the general principle, described in

Chapter 12, whereby the deformation tensor is decomposed into the product of two
tensors. The growth tensor maps the initial configuration to an unstressed configu-
ration and the elastic deformation tensor maps the evolving reference configuration
to the current configuration.
The introduction of growth stretches requires new constitutive relationships. Since
they describe the evolution of the material in time, these relationships are usually
referred to as evolution laws and take the convenient form of differential equations
in the slow time t as they relate the reference and current configurations at time to a
new reference configuration at a later time. If we assume that the remodeling process
depends possibly on the stress and the material position, these evolution laws take
the general form
∂γs
= ηγs Gs (γs , ts , tϕ , S), (9.7)
∂t
∂γϕ
= ηγϕ Gϕ (γϕ , ts , tϕ , S), (9.8)
∂t
where η −1 is a characteristic relaxation time.

In the simple case where we assume that, at any given time, the reference config-
uration relaxes toward the current configuration, the evolution law takes the simpler
form
∂γs ∂γϕ
= η(λs − γs ), = η(λϕ − γϕ ). (9.9)
∂t ∂t
Taking advantage of the fact that the time scale for the growth process is much
larger than the elastic or viscoelastic relaxation time scales, minutes or hours com-
pared to milliseconds in most cases, we assume that the shell is in a quasi-static
equilibrium at all times. A new elastic equilibrium is obtained by considering the
evolution of the new reference configuration. In the long-time limit, this amounts to
re-parameterizing the distribution of material points around the shell back to a uni-
form distribution thereby making the current stressed configuration a new reference
configuration. We may also think of this process as equivalent to having rebuilt the
wall back to its original thickness while relieving all stresses.
Computationally, the growth process is discretized, that is, we start with an initial
shape as shown in Figure 9.1 defined by the function R0 (S0 , t), 0 ≤ S0 ≤ L 0 , and
compute the new shape r (S, t) by solving the mechanical equations subject to the
boundary condition r (L 0 ) = R0 , r (0) = 0. The new shape r (S, t) represents the new
mechanical equilibrium of the membrane. The new reference shape R(S, t + dt) and
parameterization S(S0 , t) is computed based on the current configuration. The shape
R(S, t + dt) is then used to compute a new mechanical equilibrium r (S, t + dt) and
so on.
r r
. . .. .
s(S0 ,t+dt)
s(S0 ,t)
S0 S0 S
R(S0 )
R0(S0) r(S0 ,t) R0(S0 )
r(S0 ,t+dt)
Figure 9.1 A sketch of membrane growth in the context of tip growth. In Step 1, the membrane
is inflated to an equilibrium shape. In Step 2, a new reference configuration is computed based on
the profile of Step 1. Since the reference configuration has evolved, a new current configuration is
obtained and the evolution of the membrane proceeds.
9.2 Application: Microbial Tip Growth
9.2.1 Background
At the microbial level, tip growth or apical growth is the dominant process for mass
addition. It is observed in different systems, such as bacterial filaments, fungi, root
hairs, pollen tubes. Despite appearing in vastly different organisms, it shares some
common notable features. For example, the actual shapes of the tip vary for different
cells, and for the same cell under different conditions. Also, there are periods in which
tip growth appears to be self-similar. In that case, the tip is just a translation of the tip
shape at the previous instant [480]. Another interesting feature is that growth is mostly
“normal,” that is, labeled points on the tip appear to move in a direction normal to the
tip surface. This possibility was discussed in the late nineteenth century by Reinhardt
[1073] and has been confirmed in experimental studies of fungal filaments [63] and
root hair [313, 1143], as shown in Figure 9.2. Despite these common features in the
overall growth dynamics there are major differences in the biological mechanisms
that drive the growth for different cell types [428, 430]. Therefore, before modeling
tip growth, we briefly review key characteristics of various microorganisms.
9.2.2 Bacterial Filaments: Actinomycetes
Actinomycetes have been a topic of considerable interest to microbiologists as a

model organism for filamentous bacterial growth since their discovery by Ferdinand
Cohn in 1875 [220, 775] (see Figure 9.3). They are broadly defined as a family of
gram-positive prokaryotic organisms with DNA high in cytosine and guanine con-
tent. Their cellular structure is composed of linear nuclear material in a cytoplasmic
9.2 Application: Microbial Tip Growth 235
Figure 9.2 Observation of normal growth. The analysis of trajectories of material points in the
fungus Rhizoctonia solani shows that markers (indicated by arrows on and circles) follow a trajectory
normal to the surface (Bar is, 5 μm. Adapted from [63]). The other lines correspond to possible
trajectories when a non-normal component of the velocity is included.
Figure 9.3 First description

of actinomycetes by Cohn in
1875 (adapted from [1]).
medium. Their cell wall structure is dominated by a characteristic network of stiff

chitin molecules embedded in a polymer complex [183, 1314]. The streptomycetes
are of particular interest since they are a natural source of antibiotics.
A.
B.
Figure 9.4 A. Life cycle of Streptomyces. B. Mycelial growth of S. coelicolor on solid agar
medium (microscopy performed by Michael Tabor and the author).
In a typical growth cycle, the streptomycete spores bud into long filamentary hyphae
which grow in and on a nutrient source. The growing hypha eventually undergo
branching leading to the formation of a dense mycelium as illustrated in Figure 9.4.
In this so-called vegetative phase, apical growth is energetically more efficient
than growth along the entire filament since the latter would result in more friction
with the nutrient medium. A typical filament of Streptomyces coelicolor in this

phase is less than 1 μm in diameter and can grow to lengths of 50–100 μm. The
vegetative phase creates an intricate network of filaments called mycelium. This
phase is usually followed by an aerial phase of growth in which the hyphae grow
out of the mycelium, sometimes deforming into helical structures, and ultimately
breaking up into spores—which then begin the growth cycle again [1]. We note that
in the aerial phase hyphal growth is no longer apical, but it is distributed along the
length of the filament.
The apical growth of streptomycetes in the vegetative phase has been demonstrated
in a variety of studies tracking the presence of tritiated GlcNAc, a precursor of
peptidoglycan. For example, Gray et al. [497] used this technique in their study of
Streptomyces coelicolor A3(2) to demonstrate that the bulk of the labeled material
was incorporated close to the hyphal apex.
The interplay between turgor pressure and cell wall rigidity has also been demon-
strated by the action of lysing agents which degrade the cell wall. In their studies of
Streptomyces coelicolor, Gray et al. [497] showed that b-lactam antibiotics caused
apical swelling of the hyphae leading, in some cases, to lysis. These and other exper-
imental studies of hyphal extension in Streptomyces coelicolor show that the main
factors determining growth in the vegetative phase are (i) turgor pressure, (ii) addition
of newly synthesized wall-building material in the apical extension zone, and (iii)
the rigidification of the distal cell wall. Thus the mechanism of growth, consistent
with these observations, is a continuous and complex process in which wall-building
materials are transported to, and incorporated into the tip, which is stretched by the
turgor pressure, and as the tip is continually stretched and rebuilt, the more remote
portions of the hyphal wall rigidify [1046].
9.2.3 Fungi
Filamentous fungi are typically at least an order of magnitude bigger than actino-
mycetes and have a much more complex internal cellular structure. Their highly
complex and multilayered cell wall typically consists of a network of chitin with
cross-linking b-glucan chains [1314]. Despite all of these differences, there are still
striking similarities with the overall growth dynamics of other filamentous cells—
especially those of apical hyphal growth. For example, studies of Neurospora crassa
showed that the incorporation of tritiated wall precursors decreased rapidly from the
hyphal tip [457]. For most fungi, turgor pressure also appears to be a factor in the
growth process, but there is some experimental evidence indicating that oomycetes
continue to grow even when they do not maintain turgor. Therefore, the precise role
of turgor has been the topic of some discussion [541, 542, 899]. However for all
fungi, it seems clear that the cytoskeleton plays a central role both in the growth
process and in transporting wall-building materials to the tip, although a complete
understanding of the different roles played by actin filaments and microtubules in the
tip growth process is still evolving. Current thinking suggests that the inhibition of
Figure 9.5 Oscillatory growth in Candida albicans (picture courtesy of Alexandra Brand [123]).
actin can stop hyphal elongation and results in the formation of bulbous tips, whereas
the inhibition of microtubules may result in distortion of the hyphal morphology [62].
It has also been proposed that a vesicle supply center, or Spitzenkörper, organizes
and/or directs the traffic of vesicles containing the wall-building materials to the
apical region [517]. Although the Spitzenkörper is not always visible under optical
microscopy, studies using electron microscopy indicate a dense region near the tip
consistent with the Sptizenkörper concept. Furthermore, laser tweezers can be used to
manipulate the Spitzenkörper trajectory and directly influence changes in the hyphal
growth direction [122, 1342].
A striking result is the observed normal cell wall expansion in filamentous fungi.
By tracking carbon particle markers on the tip of Rhizoctonia solani, Bartnicki-Garcia
et al. [63] were able to show that the markers moved in a direction perpendicular to
the surface of the hyphal apex, thereby confirming the old prediction of Reinhardt
concerning normal growth. Their results also support the view that turgor pressure
is a major driving force in the hyphal growth process.
Another interesting aspect of growth is that depending on the substrate, the hyphae
may present different morphologies. For instance, on hard surfaces, the hyphae of the
dimorphic fungus, Candida albicans, follow, typically, a linear trajectory. However,
on semisolid surfaces, the hyphae of Candida albicans grow in an oscillatory manner
[123], as shown in Figure 9.5.
However, for all these complexities and subtleties, the overall picture parallels
that of hyphal growth in filamentous bacteria, namely, apical growth is driven by a
complex process involving transport of wall-building materials and a tip-straining
mechanism [429].
9.2.4 Root Hairs
Root hairs have been one of the most enduring topics of study in microbiology with
scientific publications dating back almost two hundred and fifty years as discussed
in Section 1.1.1. As with filamentary fungi, it was established long ago that cell wall
expansion is concentrated at the tip [171]. More refined microscopic studies have
also shown that the expansion is both inhomogeneous, that is, the rate of expansion
is maximal around the tip, and anisotropic: The meridional expansion is greater than
the hoop expansion. The latter observation suggests a varying stiffness of the cell
wall, perhaps due to a corresponding variation in microfibril cross-linking. Further

experimental studies of growth in Medicago truncatula root hair [1143] also support
the notion of normal motion of material points on the growing tip.
9.2.5 Modeling of Tip Growth
The mathematical modeling of hyphal growth involves three equally important levels
of description: The biomechanical level aims to describe the growth and morphology
based on mechanical principles. The biophysical level aims to translate the structural
details of the cell wall into macroscopic material properties and must be incorporated
in the mechanical models. Finally, the biomolecular level aims to understand the
molecular basis for the detailed structure and expression of the cell.
The first attempts to model tip growth were mostly based on geometric arguments.
In these models the increase in wall area of an advancing tip is balanced with the
rate of incorporation of wall building material [64, 1082]. Other models consider the
consequences of having a given tip shape, such as a hemisphere or an ellipse [1121,
1249].
A geometric model following the idea that the local velocity vector attached at
a point on the surface is a function of the curvature was investigated in [467] (see
Figure 9.6). While this approach can sometimes yield formulas for idealized tip
shapes, it cannot capture the fundamental interplay between the mechanical stresses
and the material responses of the cell wall.
By contrast, mechanical models can capture the tip growth process through a
proper description of the physical stresses and strains experienced by the growing
hypha. Different mechanical modeling approaches have also been proposed. Early
models [715] that draw on concepts from shell theory and the Laplace law for mem-
branes suggest connections between experimentally observed tip shapes and their
Figure 9.6 Tip profile in a geometric model. The shape is obtained by assuming that areal growth
is proportional to the Gauss curvature [467].
material properties. Other models are based on the use of linear plasticity theory
to model irreversible extensions of the cell wall [314]. This approach follows the
idea behind the Lockhart model of plant cell growth described in Section 4.5. We
now turn to the modeling of tip growth as a morphoelastic process by following the
general method presented in this chapter.
9.2.6 A Model for Hyphal Growth
We model the tip of growing hyphae as a morphoelastic membrane or shell. Here,

we consider an idealized situation where we do not explicitly model the material
transport. Therefore, our general framework is suitable for the modeling of many
systems exhibiting tip growth. First, we consider the case of turgor-driven growth
for actinomycetes and second the case of fungal hyphae when taking into account
the role of surface traction.
We start with an elastic membrane at rest. Due to a local weakening of the mem-
brane, a soft-spot, the membrane will deform and reach a new steady state. Assuming
that the reference configuration remodels according to the shape of the current con-
figuration, the turgor pressure will further expand the membrane and this process is
repeated as shown in Figure 9.7.
We are particularly interested in answering whether this process of localized
weakening, turgor inflation, and reference configuration remodeling leads to a tip
solution? If so, what is the profile shape? Is it self-similar? Can we further understand
t1 t1+dt
t=0
P P
t=0 (αs(t),αϕ (t))

t=t1 (αs(t+dt),αϕ(t+dt))
(γs(t),γϕ (t))
t=0 t1 t1+dt
Figure 9.7 An initial reference configuration is defined at t = 0. The growth strains (γs , γϕ ) brings
the initial reference configuration to the reference configuration at time t1 . The current (observable)
configuration is computed by applying loads and is characterized by the elastic strains (αs , αϕ ).
The new reference configuration at time t1 + dt is obtained by solving (9.7)–(9.8) with data from
the current configuration (dashed line). The growth process is then iterated.
beading? What is the effect of external friction due to the medium? Is normal growth
a consequence of mechanics and growth?
9.2.6.1 The Soft-Spot Hypothesis: Variable Moduli
As discussed in Section 8.1.4.1, the existence of tip shapes for an elastic membrane
requires a softening of the tip. The softening of cell walls close to the tip can easily
be taken into account using a material-dependent elastic function A = A(S). The
general form of p = P/A that we choose is

P S − S1
p= 1 − tanh + p∞ , (9.10)
2 a1
where P is the turgor pressure and the parameters S1 and a1 describe the length of
the apical extension zone as shown in Figure 9.8. Since p → p∞ as S → ∞, the
parameter p∞ describes the effective pressure in distal regions.
It is important to note that in our model the critical parameters of pressure and
modulus only appear in the ratio P/A. Thus a decreased modulus or increased pres-
sure is, at the mechanical level, indistinguishable. We will return to this point in our
discussion of lysis. The idea of a geometry-dependent elastic modulus is biophysi-
cally plausible as classic results in polymer physics [1247] tell us that the modulus
increases with the degree of polymeric cross-linking, and in the current context this
is consistent with the process of wall rigidification.
9.2.7 Tip Shapes for Filamentous Bacteria
We first consider the case of bacterial hyphae and use information and data from
the model organism Streptomyces coelicolor A3(2). The effective pressure profile in
Figure 9.8 can be fitted from published data on the rate of incorporation of tritiated
markers along the hyphae of Streptomyces coelicolor A3(2) [497] and provides a
Figure 9.8 A typical plot

of the renormalized pressure
p = P/A(S). Close to the
tip (S = 0), the walls are soft
and the elastic coefficient
minimal. In the distal
regions, the walls are set and,
comparatively, the elastic
coefficient A is very large, so
that the effective pressure is
small (equal to p∞ ). 1
(d)
(a)
Figure 9.9 Sequence of growth obtained from an initial spherical shape of radius 2.
(S1 = π/12, P = 1, a1 = 1/8, p∞ = 10−6 ). The corresponding three-dimensional representa-
tions for the profiles (a) and (d) are shown in Figure 9.10.
characteristic length scale for the soft-spot as done in [479]. Here we use the linear
constitutive laws given in Section 8.1.3.1.
The computation of a growth profile is performed as a discrete sequence of
boundary-value problems; time evolution is performed, as described above, by re-
parameterization of the initial profile. A typical computation is shown in Figure 9.9
where it is seen that a tip rapidly emerged through this process. Remarkably, this
process appears very robust as an asymptotic shape is obtained from different initial
profiles and appears to depend only on the elastic parameters and pressure profile.
Before we try to determine the asymptotic shape, we investigate, computationally,
the self-similarity of the process. When viewed in the reference frame of the tip, the
tip shape, once established, is effectively self-similar, that is, at each step the new
shape is a duplicate of the previous one properly translated, as shown in Figure 9.11.
The asymptotic linear increase of arc length and area, shown in the same figure, is
also consistent with a self-similar solution. The self-similar process is obtained after
a combination of elastic expansion followed by a growth process. Therefore, the final
tip shape, r (S), is of the same functional form as the initial reference shape R0 (S).
The self-similar ansatz in this process, as illustrated in Figure 9.12, is simply to set
r (S) = R0 (s(S)). (9.11)
After expansion, the value of the radial variable r for a given material point S is
required to be the value of R0 evaluated at the new position of S, expressed in terms
of arc length, along the shell. The angular deformation variable, λϕ , now takes the
form
(a)
(b)
(c)
(d)
Figure 9.10 Three-dimensional graphics representation for iterates 50, 100, 150, and 200 of
Figure 9.9.
λϕ = R0 (s(S))/r0 (S). (9.12)
This form is well suited for hyphal growth since far from the tip we expect to see,
effectively, no material stretching and hence
λϕ = lim r0 (s(S))/r0 (S) ≈ r0 (S)/r0 (S) = 1. (9.13)

S→∞
A B
2k
+0.0
3.84
0.11 k
1.20+
Figure 9.11 Self-similarity. A. For the same growth sequence as in previous figure, but viewed in
the reference frame of the tip, the increase of total length and area during growth as a function of
the number of growth iterates k are shown. B. After an initial phase, the increases of length and
area are linear with time, as observed in experiments.
Note, however, that the functional relationship (9.11) is nonlocal since the arc length
s(S) is determined by integration of the metric along the curve up to the given point S.
This self-similar ansatz maybe explored analytically by expanding all dependent
variables asymptotically close to S = 0:

N
N
r= ri S i + O(S N +1 ), θ = θi S i + O(S N +1 ), (9.14)
i=1 i=1

N
N
s= si S i + O(S N +1 ), R0 = ρi S i + O(S N +1 ). (9.15)
i=1 i=2
Substitution of these expansions into the system (8.36–8.37) yields a set of algebraic
relationships for the expansion coefficients with the constraint expanded to order N ,
that is, we have
Figure 9.12 Self-similar

growth ansatz. We assume
that after pressurization and S S
growth response, the new s S S
profile is similar to the initial R S S
profile up to a translation.
⎛ ⎞i

N N
N
ρi ⎝ sj S j, ⎠ = ri S i .
i j i
Here, for computational simplicity, we use a simpler renormalized pressure field and
choose a limiting form of (9.10), namely

pmax 0 ≤ S ≤ S1 ,
p= (9.16)
0, S > S1 ,
thereby making the membrane completely rigid outside the apical expansion zone
0 ≤ S ≤ S1 . The outer profile is then a cylinder of radius r (S1 ) and we can solve
for the inner solution. The equations for the coefficients of the asymptotic series are
closed by requiring that the inner and outer solutions match, that is, θ(S1 ) = π/2.
A typical computation of the asymptotic form of the self-similar solution together
with the corresponding numerical solution is shown in Figure 9.13. For the self-
similar solution, there are only two length scales: The size of the hot-spot S1 , and the
ratio A/ pmax . We introduce a dimensionless number that characterizes the relative
effect of the turgor pressure with respect to the elastic stiffness of the soft tip as
pmax S1
ξ= . (9.17)
A
It is now easy to compute numerically the width w and curvature κ of the self-
similar profile as a function of ξ. Examples of such universal profiles are shown in
Figure 9.14.
Figure 9.13 Comparison of

growing membrane profile
with self-similar solution. 1/2
The inner profile (dashed
line) corresponds to the
self-similar solution with the
same parameter values. The
outermost profile is obtained
numerically (S1 = π/6,
P = A = 1, a1 = 1/64,
p∞ = 10−2 ).
Figure 9.14 Self-similar ξ=2

profile for different values of ξ=1
the dimensionless number ξ.
ξ=0.1
0.5
-1.5 -1 -0.5
-0.5
9.2.8 Lysis and Beading
It is now possible to consider two simple situations where modifications of the elastic
properties of cell walls can lead to a change in morphology. First, when the hyphae
are treated with β-lactam antibiotics, the rigidification of the wall is partly prevented
and results in apical swelling. This effect is described as an “...increase internal
hydrostatic pressure acting on a cell wall of reduced rigidity” [1046]; in our setting
we see that both effects are represented by an increase of the effective pressure
parameter p, as shown in Figure 9.15.
Instead, if the hyphae are treated with lysozyme, the hyphae can exhibit sudden
beading, that is, generalized nonlocalized hyphal swelling. Again, this situation can
be easily modeled by a variation of the parameter p along the hyphae. A typical
example of such a profile is shown in Figure 9.16. Here the first dips in p correspond
Figure 9.15 Lysis is

obtained by an increase of
the pressure at the tip (or
equivalently, a softening of
the tip).
A. P/A B.
Figure 9.16 A. A swollen hyphal Streptomyces Coelicolor tip. The tip is approximately 2–5 μm
in diameter and its surface appears roughened due to the lytic action of lysozome (Bar is 1 μm.
Picture courtesy of J. I. Prosser). B. An effective pressure profile capable of reproducing lysozyme
induced beading (top). Beading obtained by allowing variations of p along the hyphae (bottom).
to a large increase in the elastic modulus. Since the beading is believed to be associated
with septation, one would expect a significant increase in structural rigidity in the
neighborhood of the septa that would survive the introduction of lysozyme.
9.2.9 Shear Stress and the Normal Growth Hypothesis
Next, we consider the effect of the environment on the shape of hyphae. The mechan-
ical action of the external medium on the growing tip is to induce shear stress, repre-
A. B.
1 1
0.8 0.8
0.6 0.6
=0.2 =0.75
0.4 0.4
0.2 0.2 =0.5
=0.1
0 0
−0.2 −0.2 =0.25
−0.4 =0 −0.4
−0.6 −0.6
−0.8 −0.8
−1 −1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 0 0.2 0.4 0.6 0.8 1 1.2 1.4
Figure 9.17 Comparison of the final shape of the tip for increasing values of external surface stress
τs . As the external stress increases the curvature at the tip decreases and the tip flattens.
Figure 9.18 Evolution of

the maximum of the 6 maximum
curvature
curvature as τs increases.
5.5
s
4.5
0 0.05 0.1 0.15 0.2
sented by τs in (8.73). Here, we use a large deformation theory with a neo-Hookean

strain-energy density function and an initially spherical shell with a soft-spot.
The effect of the external media on the surface friction is obtained by increasing
the value of τs . In Figure 9.17, we compare tip shapes after 70 iterations with a series
of values of τs with the same material parameters and initial conditions. The results
clearly show a progressive “flattening” of the tip shape as τs is increased as indicated
by looking at the maximum curvature as a function of τ , as shown in Figure 9.18.
We can now test the normal growth hypothesis by tracking the paths of material
points on the surface of the tip and comparing them with what would be the equivalent
normal paths starting with τs = 0. As we see in Figure 9.19, the particle trajectories
do not follow normal paths for τs = 0. Therefore, in the absence of shear stresses,
we do not expect growth to be normal.
A. s=0 B. s=0.2
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0.2 0.4 0.6 0.8 1 1.2 0.2 0.4 0.6 0.8 1 1.2
Figure 9.19 Testing the normal growth hypothesis. We compare the path of a material point on
the tip with the path starting from the same material point, but following a trajectory normal to the
curve for increasing values of τs . As can be seen the material trajectories tend to align with the
normal trajectories.
As τs increases, the particle trajectories become evermore “normal.” Therefore,

we conclude that a possible explanation for the normal growth hypothesis is simply
the result of the shear stress exerted by the media on the surface. On an energetic
level, this is a reasonable assumption as normal growth reduces the motion of material
points in the media by following close to normal trajectory, that is, without having
to work against external forces there is no energetic reason for the tip expansion
to follow a normal path. However, once such forces are included, such paths are
the most work-efficient ones to follow. This result suggests that normal growth is a
consequence of the mechanics rather than being coded positionally at the level of
the membrane.
Chapter 10
Morphoelastic Plates
So far, we have considered relatively simple two-dimensional structures, namely

axisymmetric shells and membranes. It is important to develop a full theory of grow-
ing and remodeling for plates and shells. Indeed, many thin three-dimensional elastic
bodies can be reduced to either elastic plates or shells: Shells are two-dimensional
elastic bodies whose reference shape is not necessarily flat, while plates are a par-
ticular case of an initially flat shell. More generally, morphoelastic shells are elastic
shells that can remodel and grow in time. These idealized objects are suitable models
for many physical, engineering, and biological systems.
Work on classical elastic shells finds its root in the experimental work of Chladni,
Germain, and Kirchhoff on vibrating plates [203, 437, 698], followed by the work
of Love [805] who first obtained a consistent general theory for small strains of
linear elastic shells. The governing equations of elastic shells in terms of stress
and couple-stress tensors were first derived by the Cosserat brothers [235] followed
by a coordinate-free theory by Synge and Chien [1204]. Many different alternative
derivations and formulations of the governing equations for elastic shells were later
proposed by [501, 719, 932, 966].

252 10 Morphoelastic Plates
10.1 Elastic Plates
A number of attempts to generalize the theory of plates to include the effects of

growth have been proposed [93, 323, 705, 851, 852]. Here, we briefly present the
results of Ben Amar and co-workers [275, 277, 923] who presented a generalization
of the classical Föppl-von Kármán equations for growth. To do so, we will require
the basic geometric descriptions of a surface in terms of its curvatures.
10.1.1 Mean and Gaussian Curvatures
We consider here an orientable parameterized surface defined by the position

vectors
x = x(ξ 1 , ξ 2 ) ∈ R3 , (ξ 1 , ξ 2 ) ∈ D ⊆ R2 . (10.1)
We assume that x is at least of class C 2 and such that the tangent vectors
∂x
ri = , i = 1, 2, (10.2)
∂ξ i
are linearly independent for all (ξ 1 , ξ 2 ) ∈ D. Since is orientable, we can define a

normal vector
r1 × r2
n= , (10.3)
|r1 × r2 |
√
where |a| = a · a is the usual Euclidean norm.
By definition {r1 , r2 , n} forms a basis, but this basis is not necessarily orthonormal.
The area of is given by

A= dS = det(M) dξ 1 dξ 2 , (10.4)
D
where Mi j = ri .r j are the components of the metric tensor with matrix representation
[M]i j = Mi j (a full description of tensors will be given in Section 11.11).
Similarly, we can compute the length of a path ρ on by defining a curve r =
r(t), t ∈ I ⊂ R,

L = ds = Mi j ξ˙i ξ˙ j dt. (10.5)
ρ I
Associated with the metric we define the first fundamental form
ds 2 = Mi j dξ i dξ j . (10.6)
10.1 Elastic Plates 253
t
k2 P
k1 kn
Σ
ρ
Figure 10.1 On a surface , we define a curve ρ through a point P. Here, kn is the curvature of the
curve obtained by slicing the surface at P by a plane spanned by the normal and tangent vectors.
The principal curvatures k1,2 are obtained as the extrema of kn at P over all tangent vectors, that is
by rotating the slicing plane.
We are interested in defining curvatures on the surface . We consider a curve ρ

on passing through a point P and parameterized by its arc length s and define t as
the tangent vector of ρ at P as shown in Figure 10.1. As defined by Equation (5.4),
the curvature of the curve C at a point P is obtained as |t |, where the prime denotes
differentiation with respect to s. It is therefore natural to define the curvature vector
dt
k= , (10.7)
ds
and decompose it into two components, the normal curvature vector kn and the
geodesic curvature vector kg
k = kn + kg , (10.8)
where kn = −kn n is along the normal vector, which is

dt dt
kn = −n · , kg = |kg | = t · × n . (10.9)
ds ds
The normal curvature kn is a property of the surface itself and gives the curvature in
a planar slice spanned by the normal and tangent vectors, as shown in Figure 10.1.
The geodesic curvature gives the curvature along the surface and is identically zero
for a geodesic curve.
In terms of the coordinates, the normal curvature is given by
kn = K i j (ξ i ) (ξ j ) , (10.10)
where
∂rj
K i j = K ji = −n · , (10.11)
∂ξ i
which naturally leads to the definition of [K] = [K i j ], the matrix of the extrinsic
curvature tensor. This tensor is naturally associated with the second fundamental
form K i j dξ i dξ j .
From the extrinsic curvature tensor, we define the principal curvatures as the
extremal values of the normal curvature as we vary the tangent vector at a fixed point
P. The principal curvatures k1 and k2 are given by the eigenvalues of the principal
curvature matrix
L = M−1 K. (10.12)
The principal curvatures can be used to define the mean curvature K M and Gaussian
curvature K G as follows
1 k1 + k2
KM = tr(L) = , (10.13)
2 2
K G = det(L) = k1 k2 . (10.14)
The Gaussian curvature is intrinsic to the surface, in the sense that it only depends on
the metric and not on the normal vector. This result is contained in Gauss’ Theorema
Egregium (remarkable theorem) [169, 496]. Note that K G is independent of the
parameterization but that K M can change sign depending on the choice of the normal
vector. It is used to classify surfaces as either elliptic (K G > 0), hyperbolic (K G < 0),
or parabolic (K G = 0). A minimal surface is such that K M = 0 identically at all
points. These surfaces play a particularly important role in a number of significant
problems in mathematics and physics [1001].
10.1.2 Growing Elastic Plates
Next we consider an elastic plate as shown in Figure 10.2. In its reference configura-
tion, the body has thickness H in the transverse direction and lateral sizes measured
by a typical length scale L. The transverse (vertical) direction is aligned with the
Z –axis and, accordingly, the middle surface in lateral directions is along the X and
Y axes. We assume that the thickness is much smaller than the lateral size H/L 1
and that the deformation is sufficiently small so that it can be fully characterized by
the vertical displacement of the middle surface ζ(X, Y ). We further assume that the
curvatures of the plate remain small and that there is no shear through the thickness
eZ
eY
eX
H
(X,Y,0)
LY
eZ
eY
eX
ζ(X,Y)
Figure 10.2 Deformation of a Föppl-von Kármán plate. Top: initial reference configuration. Bot-
tom: deformed configuration. The deformation is fully characterized by the vertical displacement
of the middle surface ζ(X, Y ).
of the shell so that the normal vector remains close to the Z –direction. For the stress,
we assume that the applied loads are sufficiently small and that the Z –components of
the Cauchy stress tensor T vanish identically (see Section 11.2.2 for the general defi-
nition of stresses). It is convenient to introduce the Airy potential χ(X, Y ) to express
the remaining stresses on the middle surface through a single scalar quantity:
∂2χ ∂2χ ∂2χ

TX X = , TY Y = , TX Y = − . (10.15)
∂Y 2 ∂X2 ∂ X ∂Y
Using these assumptions, a systematic expansion of the three-dimensional equa-
tions for an elastic continuum in terms of the small parameter H/L leads to a set of
two equations for the the displacement ζ(X, Y ) and the Airy potential χ(X, Y ):

M − 2H [χ, ζ] = P,
D 2 ζ − K (10.16)

G = 0,
2 χ + E [ζ, ζ] − K (10.17)
where E is the Young’s modulus, D = E H 3 /9 is the bending rigidity of the plate,

and P is the pressure acting on the plate along the normal to the surface. The bracket
appearing in these equations is defined as
1 ∂2 f ∂2g 1 ∂2g ∂2 f ∂2 f ∂2g

[ f, g] = + − . (10.18)
2 ∂ X 2 ∂Y 2 2 ∂ X 2 ∂Y 2 ∂ X ∂Y ∂ X ∂Y
A.
B.
Figure 10.3 A. Shape changes in the algae Acetabularia acetabulum during morphogenesis
(adapted from [1137]). B. Corresponding equilibrium shapes of a growing disc with different growth
rates along its radius and angular directions [275].
In the absence of growth, i.e., K M = K G = 0, these equations were initially

derived formally by Föppl and von Kármán [372, 1286]. Since then, various methods
have been proposed to derive them rigorously [74, 209, 210, 394, 788, 934].
The effect of growth is contained in the two geometric parameters, K M , and K G
that can be interpreted as the mean and Gaussian curvatures of a new reference
middle surface that the plate acquires through growth [323, 324, 705, 1140]. These
two curvatures can be directly obtained from the growth strains. Therefore, growth
can be interpreted as a source of intrinsic curvatures in a plate theory.
The effects of growth through Gaussian curvature can be understood by looking
at the anisotropic growth of a disk as shown in Figures 10.3 and 10.4. If the disk
grows isotropically, it remains flat with vanishing Gaussian curvature. However, if
it grows slower inside than close to the edge, or grows faster in the hoop than in the
radial direction, the disk will buckle to a cup shape with positive Gaussian curvature.
Conversely, if the edge grows faster the disk will develop negative Gaussian curvature
and release partly its residual stress by adopting a shape with wavy edges as shown
in Figure 10.4 [1032, 1033].
The particular relationship between differential growth and Gaussian curvature in
plates can been used to explain how plant leaves develop waviness or control flatness
[830, 937, 1097, 1141, 1142]. Another example of a buckling instability in strips
used to explain the wrinkling of leaves is shown in Figure 10.5.
The Föppl-von Kármán is the simplest model of an elastic plate. Relaxing some of
the model assumptions leads to increasingly more complex models [209, 210]. For
Figure 10.4 A negatively curved surface can be obtained by differential growth. The edge of the
disk grow faster than the central part of the disk [1032].
instance, in the theory of Cosserat and Kirchhoff plates, the possible deformations of
a three-dimensional body are constrained to deformations compatible with the two-
dimensional geometry [33]. This approach can be used to explore the full nonlinear
behavior of the material and does not restrict the analysis to thin bodies. It can
also be generalized to the case of a growing body [851, 852] by directly using the
three-dimensional theory of volumetric growth discussed in the next chapters.
A. B.
Figure 10.5 A. Picture of a grass blade (picture courtesy of Jacques Dumais). B. Instability of
parallel clamped growing strips. Growth induces compression in the strips leading to rippling (from
[277]).
In the case where the reference configuration is a shell rather than a plate, a
general theory based on differential geometry can be used to derive the fundamental
equations of growing shells [1110]. This general theory is particularly interesting as
it shows explicitly the coupling between the intrinsic geometry of surfaces and the
intrinsic geometry of growth as discussed in Section 12.5.
Part IV
Volumetric Growth: A Three-Dimensional
Theory
The sixth age shifts

Into the lean and slipper’d pantaloon,
With spectacles on nose and pouch on side,
His youthful hose, well sav’d, a world too wide
For his shrunk shank, and his big manly voice,
Turning again toward childish treble, pipes
And whistles in his sound. . .
Chapter 11
Nonlinear Elasticity
Before introducing the effect of growth into a general mechanical theory, it is impor-
tant to review the elements of continuum mechanics that describe the response of
a material under loads, and in particular, the theory of elasticity. The traditional
approach to elasticity is to consider the regime of small deformations, where a mate-
rial is slightly perturbed from an unstressed configuration. In that regime, the govern-
ing equations and constitutive relationships are linear. This theory of linear elasticity
has been successfully developed over the last two centuries to address many funda-
mental problems of physics and engineering [761].
Starting in the 1940s, it was found that the theory of linear elasticity was inadequate
to model the response of elastomers such as rubbers in large deformations [1084,
1253]. Similarly, many tissues and organs also operate in large deformations. For
instance, large arteries are typically stretched between 20% and 60% from their
unloaded configurations [597, 635] and their response to loads is drastically different
from the response of elastomers.
A striking example of the difference between rubbers and soft tissues is observed
in experiments first performed in the nineteenth century and shown in Figure 11.1
[822, 1000]. In these experiments, a cylindrical elastic membrane made out of rubber
or soft tissue is pressurized. At a critical pressure, the radius of the rubber cylinder

262 11 Nonlinear Elasticity
A. C.
B.
Figure 11.1 The importance of nonlinearities and large deformations is demonstrated in a pressure
experiment as originally investigated by Mallock in 1890 [822]. A. The experiment consists in
increasing the pressure inside a tube or a sphere while recording the radius of the bulge. B. In
the case of rubber, the experiment of Osborne and Sutherland [1000] shows a sudden limit-point
instability at a critical pressure. Past that pressure, the radius suddenly increases. C. In a similar
experiment performed on a dog bladder, the behavior of the system is qualitatively different and the
instability disappears. In this case, it will be increasingly harder to increase the size of the bulge by
increasing the pressure.
will suddenly jump to a new equilibrium, whereas for soft tissues, the radius will
evolve continuously to an asymptotic radius. Therefore, a continuum theory for the
mechanical response of biological tissues requires the general theory of nonlinear
elasticity, which, by contrast to the theory of linear elasticity, assumes neither small
deformations, a particular choice of constitutive law, nor a particular symmetry of
the material. As we will demonstrate in subsequent chapters, the nonlinear theory of
elasticity is also the natural framework to develop a theory of growth and remodeling.
In this chapter, we briefly review the general theory of nonlinear elasticity, starting
with the description of kinematics and moving to the Cauchy equations governing
the response of a continuum. In the theory of elasticity, these equations are comple-
mented by constitutive laws relating stresses to strains. The theory is presented at an
elementary level and details can be found in the textbooks of Ogden [975], Gurtin,
Fried, and Anand [525, 526], Truesdell and Noll [1251], or Antman [33].
11.1 Kinematics
Our first task consists in defining properly the deformation of a body in a three-
dimensional space. To do so, we first define a body as occupying a region of a three-
dimensional space. After a deformation, the body also occupies a region in space
and the motion of each point of the body, from its initial to its current positions, can
be defined by a mapping. The relative motion of nearby points can be extracted from
11.1 Kinematics 263
Initial Current
configuration configuration
X0
B0 B
E e
3
X0 3 x
e
E 2
2
E1 e
1
Figure 11.2 Basic kinematic of nonlinear elasticity. Two configurations are defined. The defor-
mation is a 1–1 map between points in the reference configuration B0 and points in the current
configuration B.
this mapping through its derivatives in space and time. It allows us to define key
quantities such as strains and stretches. Therefore, we first introduce basic notion of
kinematics in three dimensions.
A body B is a set of material points whose elements can be put into a 1–1 corre-
spondence (bijection) with points in a region B ⊂ E3 . As the body moves or deforms,
this set can change with time t ∈ R. In nonlinear elasticity, one considers multiple
configurations for the description of a body and denotes by Bt (or B, when there is
no possibility of confusion) the configuration of B at time t.
We use B0 as the initial configuration, typically an unloaded configuration, and B
for the current configuration where loads are applied. The initial configuration, B0 ,
is parameterized by material points relative to the position vector X0 with origin O
and the current configuration, B, by the position vector x with origin o.
The basic assumption for the deformation of a continuum is that the body retains
its integrity and that material points do not overlap during a deformation. Therefore,
both B0 and B are bijections of B, and there exists an invertible mapping, pictured
in Figure 11.2, called deformation or motion χ : B0 → Bt such that
x = χ(X, t), ∀ X ∈ B0 and X = χ−1 (x, t), ∀ x ∈ Bt . (11.1)
In the absence of phase boundaries, singularities, or jumps, we assume that this

mapping is twice continuously differentiable in space and smooth in time.
It is convenient to use two orthonormal rectangular Cartesian bases {e1 , e2 , e3 }
and {E1 , E2 , E3 } to represent vectors in the initial and current configuration
x = xi ei , X = X i Ei , (11.2)
where summation over repeated indices is always assumed unless explicitly specified.
The convention in continuum mechanics is to refer to coordinates in the current
configuration, or quantities expressed with these coordinates, as Eulerian or spatial.
Coordinates in the initial configuration, or quantities expressed with these coordi-
nates, are referred to as Lagrangian, referential, or material.
11.1.1 Scalars, Vectors, and Tensors
To describe the deformations of a body B, we attach, at each material point, physical

quantities known as fields, which make continuum mechanics a theory of fields.
These quantities can be scalar fields, such as density, temperature; vector fields, such
as velocity, acceleration, force; or tensor fields, such as deformation gradients or
stress and strain tensors. These different fields can all be understood as tensor fields
of different orders. By definition, a scalar field is a tensor field of order 0 and a vector
field is a first-order tensor field. Higher order tensor fields require the definition of
the tensor product.
The scalar product between two vectors a = ai ei and b = bi ei , in the same
vector space, follows the usual definition:
u · v = ui vi , (11.3)
and is used to define the Euclidean norm

√
|v| = v · v. (11.4)
We can also use the scalar product to define the tensor product. Consider two
vectors u = ui ei and v = vi Ei , not necessarily defined in the same vector space.
Then, the tensor product, u ⊗ v, of these two vectors is a second-order tensor such
that, for an arbitrary vector a = ai Ei ,
(u ⊗ v)a = (v · a)u. (11.5)
This definition implies that the vectors v and a must belong to the same vector space,
but in general, u can belong to a different space. We see from this definition that a
second-order tensor maps vectors from one vector space to vectors in another vector
space.
Explicitly, the tensor product is
u ⊗ v = ui ei ⊗ v j E j = ui v j ei ⊗ E j . (11.6)
When there is no possibility of confusion, we can write the components of the second-
order tensor u ⊗ v in the Cartesian bases {e1 , e2 , e3 } and {E1 , E2 , E3 } as (u ⊗ v)i j =
ui v j , i, j = 1, 2, 3.
11.1 Kinematics 265
Equipped with the tensor product, we can define a general second-order tensor
in the Cartesian bases {e1 , e2 , e3 } and {E1 , E2 , E3 } as
T = Ti j ei ⊗ E j ⇐⇒ Ti j = ei · TE j , (11.7)
which implies that for a vector a = a j E j ,
(Ta) j = Ti j a j . (11.8)
We define the matrix of components of a tensor in Cartesian coordinates by [T]

such that [T]i j = Ti j . We can then extend most definitions and identities of traditional
linear algebra to tensors.
A particularly important class of second-order tensor is the tensors whose com-
ponent matrices are square matrices. For these second-order tensors, the determinant
and trace of a second-order tensor are defined, respectively, as
det T = det([T]), tr T = tr ([T]) = Tii . (11.9)
In particular, in three dimensions, we have
det T = det([T]) = i jk T1i T2 j T3k , (11.10)
where i jk denotes the usual Levi-Civita permutation symbols or permutation sym-

bols, that is, i jk = 1 if (i, j, k) is even permutation of (1, 2, 3), −1 if it is an odd
permutation, and 0 if any index is repeated.
Similarly, the matrix of the transpose of a tensor is the transpose of the matrix,
that is,
[TT ] = [T]T (11.11)
and a tensor is symmetric, TT = T, if and only if Ti j = T ji .

The product of two tensors S and T is only defined when the image of a vector
by T is in the domain of S. Then, for an arbitrary vector a, we have
(ST)a = S(Ta). (11.12)
In such cases, the matrix of the product is the product of the two matrices:
[ST] = [S][T]. (11.13)
A tensor S is an orthogonal tensor if
SST = ST S = 1, (11.14)
where 1 is the identity tensor defined as (1)a = a ∀a. As expected, it follows that
the components of an orthogonal tensor form an orthogonal matrix. The group of
all orthogonal tensors in three dimensions is denoted O(3). A proper orthogonal

tensor is an orthogonal tensor with the additional property det S = 1. The group of
all proper orthogonal tensors in three dimensions is denoted S O(3). Orthogonal and
proper orthogonal tensors are particularly useful to characterize rotations and proper
rotations in a three-dimensional space.
We can also contract two tensors together to obtain a scalar by introducing the
double contraction
S : T = tr(ST) = Si j T ji . (11.15)
If the determinant of a tensor T does not vanish, the matrix of inverse of T is the
inverse of the matrix:
[T−1 ] = [T]−1 . (11.16)
Explicitly, for a tensor T = Ti j ei ⊗ E j , we have
T−1 = ([T]−1 )i j Ei ⊗ e j , (11.17)
so that
TT−1 = 1 = δi j ei ⊗ e j , (11.18)
T−1 T = 1 = δi j Ei ⊗ E j , (11.19)
where δi j is the usual Kronecker delta’s symbol (δii = 1 and δi j = 0 for i

= j).
The tensor product can also be used to define higher order tensors. For instance,
a third-order tensor, Q, and a fourth-order tensor Q, in the basis {e1 , e2 , e3 }
are defined as
Q = Q i jk ei ⊗ e j ⊗ ek , (11.20)
Q = Q i jkl ei ⊗ e j ⊗ ek ⊗ el . (11.21)
However, in this case, the equivalence with linear algebra is lost and identities for
higher order tensors must be obtained by following the rules defining the tensor
products and playing the game of indices. Note that higher order tensors can also be
defined with respect to multiple vector spaces.
11.1.2 Spatial Derivatives of Tensors
Next, we define spatial derivatives of scalar, vector, and tensor fields. We have two
sets of spatial variables, the Lagrangian variables X and the Eulerian variables x. We
can therefore define different types of spatial derivatives depending on the description
of a given quantity.
We first consider the case where φ, u, and T are scalar, vector, and tensor fields,
respectively, over x, that is,
11.1 Kinematics 267
φ = φ(x, t), u = ui (x, t)ei , T = Ti j (x, t)ei ⊗ e j . (11.22)
We define the Eulerian gradient of scalar and vector functions as
∂φ ∂φ
grad φ = = ei , (11.23)
∂x ∂xi
∂u ∂u ∂(ui ei ) ∂ui
grad u = = ⊗ ej = ⊗ ej = ei ⊗ e j . (11.24)
∂x ∂x j ∂x j ∂x j
The gradient is an operation that increases the order of the tensor and is defined, in
general, as the operation
∂( · )
grad( · ) = ⊗ ej. (11.25)
∂x j
It follows from this definition that
grad(φu) = u ⊗ grad φ + φ grad u. (11.26)
Similarly, we define the gradient of a second-order tensor as
∂ ∂Ti j
grad T = Ti j ei ⊗ e j ⊗ ek = ei ⊗ e j ⊗ ek . (11.27)
∂xk ∂xk
The divergence decreases the order of a tensor by contracting indices. For a vector,
we have simply
∂ui
div u = . (11.28)
∂xi
For a second-order tensor, the contraction can take place on the first or second
index depending on the convention. Here, we choose to define the divergence as
a contraction on the first index, that is,
∂Ti j ∂Ti j
divT = e j (ei · ek ) = ej. (11.29)
∂xk ∂xi
With this particular definition of the divergence operator, the divergence theorem,
applied on a domain ⊂ R3 , reads

Tn da = div (TT ) dv. (11.30)
∂
Next, we consider spatial derivatives with respect to Lagrangian coordinates, that

is, , U, and T are now fields over X:
= φ(X, t), U = ui (X, t)Ei , T = Ti j (X, t)ei ⊗ E j . (11.31)

The Lagrangian gradient is then the operation

∂( · )
Grad( · ) = ⊗ Ej. (11.32)
∂X j
Note that we use the lower case “grad” and “div” to describe spatial derivatives
with respect to Eulerian coordinates and the upper case “Grad” and “Div” for spatial
derivatives with respect to Lagrangian coordinates. Explicitly, we have
∂ ∂
Grad = = Ei , (11.33)
∂X ∂ Xi
∂U ∂U ∂(Ui Ei ) ∂Ui
Grad U = = ⊗ Ej = ⊗ Ej = Ei ⊗ E j , (11.34)
∂X ∂X j ∂X j ∂X j
∂ ∂Ti j
Grad T = Ti j ei ⊗ E j ⊗ Ek = ei ⊗ E j ⊗ Ek . (11.35)
∂ Xk ∂ Xk
The divergence is then

∂Ui
Div U = , (11.36)
∂ Xi
∂Ti j ∂Ti j
Div T = e j (Ei · Ek ) = ej. (11.37)
∂ Xk ∂ Xi
Note that these definitions can be appropriately modified for the case where T =
Ti j Ei ⊗ E j by changing e j → E j in the definition of Grad and Div.
11.1.3 Derivatives in Curvilinear Coordinates
It is often convenient to describe a body and a deformation with respect to curvilinear

coordinates. For instance, it is natural to use cylindrical coordinates to describe
simple deformations of a cylinder. We use the curvilinear coordinates {q1 , q2 , q3 }
in the current configuration and {Q 1 , Q 2 , Q 3 } in the reference configuration. These
coordinates are related to the Cartesian coordinates in each configuration through
the position vectors x = x(q1 , q2 , q3 ) and X = X(Q 1 , Q 2 , Q 3 ). Here, we use greek
subscripts to denote quantities defined in non-Cartesian coordinates. For instance,
we associate to each coordinate set, a set of basis vectors
∂x ∂X
eα = h −1
α , Eα = Hα−1 , α = 1, 2, 3, (11.38)
∂qα ∂ Qα
where h α and Hα are scale factors, used to normalize the basis vectors:

∂x ∂X
h α = ,
Hα = ,
α = 1, 2, 3. (11.39)
∂qα ∂Q α
11.1 Kinematics 269
For brevity, we restrict our attention to a set of orthonormal coordinate, so that
eα · eβ = δαβ , Eα · Eβ = δαβ , α, β = 1, 2, 3. (11.40)
We define the gradient, grad T = ∇ ⊗ T, of a tensor T at a point x ∈ B as the tensor

that maps a vector v in the tangent space of B at x onto the infinitesimal variation of
T along a path going through x with tangent v. For any given v, we define a path ,
parameterized by its arc length s, going through x and tangent to v. The operation
of the gradient on a vector v is
dT((s)) T((s + ds)) − T((s))

(∇ ⊗ T)v = = lim
ds ds→0 ds
∂T(x) dx α ∂T(x) dxfi
= = δαβ
∂x α ds ∂x α ds
∂T −1 dqβ
= (h α eα · h β eβ )
∂q ds
α
∂T
= ⊗ h −1
α eα (v), (11.41)
∂qα
where we have used the fact that the tangent to at x is h β eβ (dqβ /ds). Since this
operation applies to arbitrary vectors v, the gradient of a tensor in orthogonal curvi-
linear coordinates is
∂T ∂T
grad T = h −1
α ⊗ eα , Grad T = Hα−1 ⊗ Eα . (11.42)
∂qα ∂ Qα
Similarly, we define the divergence of a tensor field T as div T = ∇ · T, that is,
∂T ∂T
div T = h −1
α eα · , Div T = Hα−1 Eα · . (11.43)
∂qα ∂ Qα
Note that we take the scalar product on the left which corresponds to the contraction
on the first index of T. Choosing Cartesian coordinates {q1 , q2 , q3 } = {x1 , x2 , x3 }
leads to h α = 1 ∀α, and the definitions (11.42) and (11.43) reduce to those of the
previous section.
As an example, consider the choice of polar coordinates {q1 , q2 } = {r, θ} in the
Euclidean plane. The position vector is x = r cos θe1 + r sin θe2 , so that
∂x
er = , h r = 1,
∂r (11.44)
1 ∂x
eθ = , h θ = r.
r ∂θ
Hence, according to (11.42), the gradient of a scalar φ is
1
grad φ = (∂r φ) er + (∂θ φ) eθ , (11.45)
r
and we recover the usual formula of vector calculus. Similarly, if we write a second-
order tensor T in polar representation:
T = Trr er ⊗ er + Tr θ er ⊗ eθ + Tθr eθ ⊗ er + Tθθ eθ ⊗ eθ , (11.46)
then its divergence is the first-order tensor

1 ∂ 1 ∂Tθr Tθθ
div T = (r Trr ) + − er
r ∂r r ∂θ r

1 ∂ 1 ∂Tθθ Tθr
+ (r Tr θ ) + + eθ . (11.47)
r ∂r r ∂θ r
And, the gradient of T is the third-order tensor
∂T 1 ∂T
grad T = ⊗ er + ⊗ eθ
∂r r ∂θ
= (∂r Trr ) er ⊗ er ⊗ er + (∂r Tr θ ) er ⊗ eθ ⊗ er
+ (∂r Tθr ) eθ ⊗ er ⊗ er + (∂r Tθθ ) eθ ⊗ eθ ⊗ er

1
+ (∂θ Trr ) − Tr θ − Tθr er ⊗ er ⊗ eθ
r

1
+ (∂θ Tθr ) + Trr − Tθθ eθ ⊗ er ⊗ eθ
r

1
+ (∂θ Tr θ ) + Trr − Tθθ er ⊗ eθ ⊗ eθ
r

1
+ (∂θ Tθθ ) + Tr θ + Tθr eθ ⊗ eθ ⊗ eθ .
r
11.1.4 Derivatives of Scalar Functions of Tensors
We will also consider scalar functions of tensors and their derivatives with respect
to a tensor. Let A, B, C be second-order tensors with Cartesian components in the
basis {e1 , e2 , e3 } given by Ai j , Bi j , Ci j . Let F = F(A) be a scalar function of A.
The derivative of the scalar function F with respect to the tensor A is a tensor with
Cartesian components
∂ F(A) ∂ F(A)
= , (11.48)
∂A ij ∂ A ji
11.1 Kinematics 271
that is,
∂ F(A) ∂F
= ei ⊗ e j . (11.49)
∂A ∂ A ji
Now, let A = BC and consider the derivative of F with respect to B. In components,

we have

∂ F(A) ∂ F(Bkl Clm )
=
∂B ij ∂ B ji
∂ Akm ∂ F(A) ∂ Bkl Clm ∂ F(A)
= =
∂ B ji ∂ Akm ∂ B ji ∂ Akm
∂ F(A) ∂ F(A)
= δ jk δil Clm = Cim
∂ Akm ∂ A jm

∂ F(A)
= Cim . (11.50)
∂A mj
So that, in general, we can write
∂ F(A) ∂ F(A)
=C . (11.51)
∂B ∂A
Other useful identities are Jacobi’s relations for the first and second derivatives
of a nonvanishing determinant,
∂
(det(A)) = det(A)A−1 , (11.52)
∂A
∂ ∂

tr det A B = det(A) tr A−1 B A−1 − A−1 BA−1 ,
∂A ∂A
(11.53)
where the contraction tr (LA) of a second-order tensor A with a fourth-order tensor L

is defined by (tr (LA))i j = Li jkl Alk . In the last equality, the derivative of the inverse
of a tensor by itself defines a fourth-order tensor such that

∂ −1
tr A B = −A−1 BA−1 . (11.54)
∂A
If A = A(t), the derivative of a scalar function of A with respect to a parameter t

can be obtained by the chain rule, that is,

d ∂ F(A) dA
F(A) = tr . (11.55)
dt ∂A dt
As an example, the first Jacobi relation (11.52) can be used to compute the derivative
of the determinant of a tensor with respect to a parameter

d ∂ dA −1 dA
(det(A)) = tr (det(A)) = det(A)tr A . (11.56)
dt ∂A dt dt
11.1.5 The Deformation Gradient
The central geometric object of nonlinear elasticity that describes, locally, relative
deformations is the deformation gradient. It is obtained as the spatial derivative of
the mapping χ. Given a vector x = xi (X)ei , the deformation gradient tensor is
F = Grad χ. In Cartesian coordinates, it reads
∂ ∂xi
F= (xi ei ) ⊗ E j = ei ⊗ E j ≡ Fi j ei ⊗ E j . (11.57)
∂X j ∂X j
Note that the bases in which the gradient is taken are mixed. Geometrically, F is a
linear map that transforms a vector v in the tangent space T p B0 at a material point
p ∈ B0 to a vector Fv in the tangent space T p B at the same material point but in the
current configuration, as shown in Figure 11.3.
We can also express the deformation gradient in curvilinear coordinates. Let
{q1 , q2 , q3 } and {Q 1 , Q 2 , Q 3 } be the coordinates in the reference and current config-
uration, respectively. The deformation χ in the bases {e1 , e2 , e3 } and {E1 , E2 , E3 } is
Figure 11.3 The deformation gradient maps vectors on the tangent space at a material point in the
initial configuration to vectors in the tangent space in the current configuration at the same material
point.
11.1 Kinematics 273
given by qα = qα (Q 1 , Q 2 , Q 3 ), α = 1, 2, 3. Then, following the definition of the

gradient (11.42), we have
∂x
Grad x = Hβ−1 ⊗ Eβ
∂ Qβ
∂x ∂qα
= Hβ−1 ⊗ Eβ (11.58)
∂qα ∂ Q β
h α ∂qα
= eα ⊗ Eβ , (11.59)
Hβ ∂ Q β
∂x
where we used ∂q β
= h β eβ and the scale factors h α , Hβ given by (11.39). Then, we
conclude that the matrix of coefficients of the deformation gradient F = Fαβ eα ⊗Eβ is
h α ∂qα
[F]αβ = Fαβ = (no summation on indices). (11.60)
Hβ ∂ Q β
As an example, we consider the deformation of a ring in a plane to another ring

shown in Figure 11.4. In the two sets of polar coordinates {q1 , q2 } = {r, θ} and
{Q 1 , Q 2 } = {R, }, this deformation is given by
r = f (R), θ = . (11.61)
For these coordinates, we have h r = H R = 1 and h θ = r , H = R, and the

deformation gradient is

∂r h r ∂r
hr
∂ R H ∂ f (R) 0
[F] = HR
hθ ∂θ h θ ∂θ = f (R) , (11.62)
HR ∂ R H ∂
0 R
that is,
r
F = f (R)er ⊗ E R + eθ ⊗ E . (11.63)
R
Y y
EΘ ER er
eθ
r
R Θ X θ
x
B0 B
Figure 11.4 Deformation of a ring into another ring in the plane.

11.1.6 Volume, Surface, and Line Elements
The deformation of a body may change the relative size of material elements. First,
consider a set of material points in the reference configuration 0 ⊆ B0 . This set
evolves in time and is deformed to a new volume ⊆ B in the current configuration.
The new volume is related to the reference volume by

dv = J dV0 , (11.64)
0
where
J (X, t) = det F (X, t) , (11.65)
is the Jacobian of the transformation that represents the local change of volume. That
is, the image of an infinitesimal volume element dv at a material point p is
dv = J dV, (11.66)
as shown in Figure 11.5.

Since volume elements are positive and cannot vanish in a deformation, we require
that J > 0 in all deformations, which ensures the invertibility of the deformation
gradient, that is, there exists a second-order tensor F−1 mapping vectors from B to
B0 , such that F−1 F = 1. Explicitly, this tensor is
F−1 = grad X(x, t). (11.67)
A transformation that conserves locally every volume element, that is J = 1, is said

to be isochoric.
Similarly, we define an area element by considering a material area element normal
to a given vector N. Then, it is standard to show [975] that the surface integral
transforms as
χ dv=JdV
dV x(X)
X
B0 B
Figure 11.5 Transformation of volumes: In a deformation, an infinitesimal volume element dV at

a material point evolves by a factor J to a new volume dv = J dV .
11.1 Kinematics 275
χ
N dA
n da=JF -TNdA
dA X x(X)
da
B0 B
Figure 11.6 Transformation of areas: In a deformation, an infinitesimal area is transformed accord-

ing to Nanson’s rule.

n da = J F−T N d A, (11.68)
∂ ∂0
where n (x, t) and N (X, t) are outward unit normals, d A and da are the area ele-
ments at a given point as shown in Figure 11.6, that is, an infinitesimal element of
area defined in the reference configuration by a normal N and surface area dA is
transformed into another element of area in the current configuration defined by a
vector n with area da and related to the reference one by Nanson’s formula:
n da = J F−T N dA. (11.69)
Finally, consider a local infinitesimal vector dX tangent to a material line in B0 at

a material point p. Then its image is dx = FdX as shown in Figure 11.7. If M is the
unit vector along dX, then
dX = M dS and dx = m ds, (11.70)
where dS = |dX| and ds = |dx|.

This last identity implies that m ds = FM dS. Now take the norm of each side:
|ds|2 = (FM · FM)|dS|2 = (FT FM) · M|dS|2 . (11.71)
Equivalently, we can write

ds
= (FT FM) · M, (11.72)
dS
where ds/dS is the change of length of a material line in the direction M. This last
relationship is used to define the stretch, λ = λ(M) of a material line in the direction
M as
λ(M) = (FT FM) · M. (11.73)
dX χ dx=FdX
dX’
dx’=FdX’
x(X)
X
B0 B
Figure 11.7 Transformation of material lines: In a deformation, an infinitesimal line element is

mapped by the deformation gradient to a new line element.
Since we are interested in characterizing elastic materials, we need to characterize

deformations that change the relative length of line elements. Therefore, stretches
provide a natural measure of strain in a material. A material is said to be unstrained
in the direction M if and only if λ(M) = 1.
In (11.71), we see the appearance of an important tensor in the description of
strain for a three-dimensional body, namely the right Cauchy–Green tensor
C = FT F. (11.74)
A material is unstrained at a given point if it is unstrained in all directions, that is,

λ(M) = 1 ∀M. In terms of the right Cauchy–Green tensor, this implies C = 1.
Geometrically, the tensor C can be interpreted as a metric on B as it provides a
way to measure distances and angles on the new body (see Section 12.5.1).
11.1.7 Polar Decomposition Theorem
The action of the deformation gradient F on a vector M can be decomposed into a

rotation about a direction m, followed by a stretch of size λ(M). This decomposition
into a stretch and a rotation can be applied directly to the deformation gradient
through the polar decomposition theorem stating: For a second-order tensor F such
that det F > 0, there exist unique positive-definite symmetric tensors U, V and a
unique proper orthogonal tensor R such that
F = RU = VR.
The positive symmetric tensors U and V are called the left and right stretch tensors,
respectively. Their squares can easily be obtained from F as follows:
FT F = U2 ≡ C, the right Cauchy–Green tensor,

FF = V ≡ B,
T 2
the left Cauchy–Green tensor.
11.1 Kinematics 277
Since V = RURT , U and V have the same eigenvalues {λ1 , λ2 , λ3 }. The principal
stretches can be obtained conveniently as the square roots of the eigenvalues of either
C or B. Note that since U and V are positive symmetric, the principal stretches are
positive and real and the corresponding eigenvectors {u1 , u2 , u3 } and {v1 , v2 , v3 } of
U and V form two bases. Therefore, the stretch tensors can be written as

3
U= λi ui ⊗ ui , (11.75)
i=1

3
V= λi vi ⊗ vi . (11.76)
i=1
Since J = det F = det(RU) = (det R)(det U) = det U = det V, we have J =

λ1 λ2 λ3 and we can rewrite F as

3
F= λi vi ⊗ ui . (11.77)
i=1
11.1.8 Velocity, Acceleration, and Velocity Gradient
The motion associated with a deformation x = χ(X, t), X ∈ B0 , is associated

with change in time t. Since X is the position of a material point, the velocity and
acceleration of this material point are, respectively,
∂
v(x, t) = χ(X, t) ≡ χ̇(X, t), (11.78)
∂t
∂2
a(x, t) = 2 χ(X, t) ≡ χ̈(X, t). (11.79)
∂t
In general, we define the material time derivative d/dt as a total time derivative
with respect to a fixed material coordinate X. For a scalar field φ = φ(x, t), the
material derivative is

d dφ ∂φ
φ≡ ≡ φ̇ ≡ + (gradφ) · v, (11.80)
dt dt X ∂t
and we define the derivative of a vector field u = u(x, t) similarly as
d ∂u
u= + (grad u) v. (11.81)
dt ∂t
Another important kinematic quantity is the velocity gradient tensor, defined as
∂vi
L = grad v, Li j = , L = L i j ei ⊗ e j . (11.82)
∂x j
Since, in general the chain rule gives Grad u = (grad u)F, we have
Grad v = (grad v)F = LF, (11.83)
but also,
∂ ∂F
Grad v = Grad ẋ = Grad x = = Ḟ, (11.84)
∂t ∂t
so that the evolution of the deformation gradient can be expressed in terms of the
velocity gradient tensor as
Ḟ = LF. (11.85)
Taking the determinant of each side of this last equality and applying Jacobi’s for-
mula (11.56) for the derivative of the determinant of a nonsingular matrix
∂
det F = (det F) tr(F−1 Ḟ) = (det F) tr(L), (11.86)
∂t
leads to an equation for the evolution of the determinant:
J˙ = J tr(L) = J div v. (11.87)
In particular, since J
= 0, we note the well-known relationship between conservation
of volume during motion and the vanishing of the divergence of the velocity:
div v = 0 ⇐⇒ J˙ = 0. (11.88)
11.2 Balance Laws
We have obtained a complete description of the deformation of a body. Starting

with a mapping χ, we defined the deformation gradient F. This tensor contains all
information on relative deformations of a body, such as local changes of volume,
area, and stretch. It was used to define secondary quantities, such as the left and right
Cauchy–Green tensor, B and C, that contain information on the strain developed
during a deformation. Now that we have a complete kinematic description of the
body, we can define physical fields at each point and used fundamental laws of
physics to find local equations between these fields.
11.2 Balance Laws 279
The governing equations of continuum mechanics are obtained by considering

the balance of physical quantities: mass, linear momentum, angular momentum, and
energy. The traditional approach to derive local laws consists in stating a balance
law on an arbitrary subset of the body and, under suitable conditions, obtaining local
relationships between physical quantities expressed as differential equations. We
illustrate this process first on the balance of mass.
11.2.1 Balance of Mass
To describe the properties and response of a material, we attach physical quantities

at each point of the body B. First, we define a scalar field ρ = ρ(x, t) and the
volume density (mass per unit current volume) at each point of the body in the
current configuration. We assume that the mass of any subset of the body ⊆ B is
conserved in time, that is,
d
ρ (x, t) dv = 0. (11.89)
dt
The problem is that the position of a material subset evolves with time, and the
time derivative cannot be directly applied to the integrand. Therefore, we first map
the integral to the reference configuration

d
ρ (x(X, t), t) J dV = 0, (11.90)
dt 0
where we have used the transformation of a volume element (11.66).

Second, since the domain 0 is fixed, we can write

d d
J ρ (x, t) dV = (J ρ) dV = 0. (11.91)
dt 0 0 dt
Third, since we wish to obtain a balance law in the current configuration, we map
the integral back, that is,

d d
(J ρ) dV = (J ρ) J −1 dv = (ρ̇ + ρ div v) dv = 0, (11.92)
0 dt dt
where we have used (11.87).

Fourth, assuming that the integrand is continuous, the vanishing of an integral
on an arbitrary domain implies that it vanishes pointwise, which leads to the usual
continuity equation for the evolution of density in the current configuration [525]
ρ̇ + ρ div v = 0. (11.93)
The procedure consisting in mapping an integral relationship between configurations

and localizing this relationship to obtain a local differential equation is a two-step
process called the Maxwell transport and localization procedure, respectively.
In the first step, integrals in the current configuration are transformed into integrals
in the initial configuration that are expressed on a fixed domain. In this configuration,
the balance law can be written as a single integral over a domain. Once this expression
has been obtained, the integral can be mapped back to the current configuration.
In the second step, a local differential equation is obtained from the integral by
assuming that it holds on an arbitrary subset and that the integrand is continuous.
Note that the localization procedure can be also applied directly to the first integral
appearing in (11.92), that is,
d
(J ρ) = 0. (11.94)
dt
If we define the reference density ρ0 (X, t) = J (X, t) ρ (x(X, t), t), then the mass
conservation in the reference configuration is simply
∂
ρ0 = 0. (11.95)
∂t
11.2.1.1 Transport Formulas
We will be making systematic use of the Maxwell transport to obtain balance laws.
It is therefore important to obtain general transport relationships for any scalar φ
and vector field u associated with the moving body in the current configuration.
Following the same steps as in (11.90–11.92), we obtain [977] the useful transport
formulas:
d
φ dv = (φ̇ + (div v)φ) dv, (11.96)
dt

d
u dv = (u̇ + (div v)u) dv, (11.97)
dt
where ⊆ B is an arbitrary subset.
11.2.2 Balance of Linear Momentum
The balance of linear momentum expresses the fundamental relationship between

the rate of change of linear momentum of a body as a result of the force applied to
the body. When applied to a rigid body, it simply leads to the well-known Newton’s
second law. However, in the case of a deformable body, the body also experiences
internal forces that need to be taken into account.
In the current configuration, the total linear momentum on any part of the body
⊆ B is simply
ρ(x, t)v(x, t)dv, (11.98)

where ρ(x, t) is the density and v(x, t) is the velocity of a point at (x, t).
The total force acting on a point x ∈ includes a body force density b, represent-
ing the contributions of external forces and a contact force density tn , representing
the force per unit area resulting in contact. Therefore, the total force acting on is

ρ(x, t)b(x, t)dv + tn da. (11.99)
∂
Euler’s first law of motion then states that the rate of change of the linear momentum
on any part of the body ⊆ B is equal to the sum of the forces acting on [1251].
Therefore, in our context, this law reads

d
ρ(x, t)v(x, t)dv = ρ(x, t)b(x, t)dv + tn da . (11.100)
dt
∂

rate of change of linear momentum sum of body and contact forces
We apply the transport formula (11.97) by mapping the rate of linear momentum to
the reference configuration

d d d
ρv dv = ρv J dV = (ρv J ) dV
dt dt 0 0 dt

= (ρv̇ + ρ̇v + ρvdiv v) J dV
0
=0, per continuity

= ρv̇ dv, (11.101)

where we have used the identity (11.87).

In order to apply the localization procedure, we need to express all the quantities as
a single integral. However, the last integral in (11.100) is a surface integral. Therefore,
it needs to be expressed to a volume integral. The standard way to transform a surface
integral to a volume integral is to use the divergence theorem. However, the integrand
does not have the required form for a direct application of the divergence theorem.
This difficulty prompts us to reexpress the contact force density tn in a tensorial form.
We first use the Cauchy’s stress principle stating that the contact force density
depends continuously on the unit normal n. Then, Cauchy’s tetrahedral argument
[975] can be used to show that the contact force density depends linearly on the unit
normal, so that
tn = Tn, (11.102)
where T is a second-order tensor independent of n. This last identity has the correct
form for the application of the divergence theorem (11.30):

tn da = Tn da = div (TT ) dv. (11.103)
∂ ∂
The tensor T is the Cauchy stress tensor, a central quantity describing forces per unit
area in a material. Using this last equality, Euler’s law (11.100) simplifies to

ρv̇ dv = ρb dv + div (TT ) dv, (11.104)

and the localization procedure leads to the first Cauchy equation:
div (TT ) + ρb = ρv̇. (11.105)
11.2.3 Balance of Angular Momentum
In a continuum both forces and torques have to be balanced. The corresponding

balance for torque is Euler’s second law of motion stating that the rate of change of
angular momentum of an arbitrary material subset ⊆ B with respect to a given
point is equal to the sum of all torques acting on with respect to the same point.
The total angular momentum of is

ρx × v dv, (11.106)

where, without loss of generality, we choose to express the angular momentum with
respect to the origin.
If we assume here that the material under consideration is nonpolar, that is, the
body is not subject to extra body or contact torques and cannot support couple stresses,
then the total torque due to body and traction forces acting on , with respect to the
origin, is
ρx × b dv + x × tn da. (11.107)
∂
Then, Euler’s second law can be written as

d
ρx × v dv = ρx × b dv + x × tn da. (11.108)
dt
∂
rate of change of angular momentum torques due to body and traction forces
The transport procedure and the continuity equation can be used to simplify this
expression to
ρx × (v̇ − b) dv = x × Tn da, (11.109)
∂
where we have expressed the contact force in terms of the Cauchy stress tensor,
tn = Tn. Cauchy’s first equation (11.105) can be used to transform the right-hand
side of this expression to obtain

x × div (T ) dv =
T
x × Tn da. (11.110)
∂
By application of the divergence theorem and the localization procedure, the last
integral implies that the Cauchy stress tensor is symmetric, that is,
TT = T. (11.111)
We can now simplify (11.105) to obtain the standard form of Cauchy’s equation, also
known as the equation of motion for a continuum:
div T + ρb = ρv̇. (11.112)
11.2.4 Many Stress Tensors
The Cauchy stress tensor is the natural measure for contact forces measured in the
current configuration per unit area in the current configuration. The contact forces
acting on the boundary ∂ of a region can be extracted from T by considering
different directions with respect to the boundary. Let n be the normal vector to ∂
at a point p in the body. Then, the normal stress, that is, the force per area normal to
∂, applied at p is
n · tn = n · (Tn). (11.113)
Considering a vector m tangent to ∂ at p (that is, m · n = 0), the product
m · tn = m · (Tn). (11.114)
is a shear stress acting on at p along m.

While the Cauchy stress is a natural measure for forces acting on a continuum
it is not always a convenient quantity for computation since the area in the current
configuration changes during the deformation. Therefore, it is often useful to measure
contact forces with respect to areas measured initially in the reference configuration.
To do so, we apply Nanson’s formula n da = J F−T N dA to the traction vector to
obtain the contact force on a material area element:
tn da = Tn da = (J TF−T )N dA = ST N dA, (11.115)
where
S = J F−1 T, (11.116)
is the nominal stress tensor. Its transpose, ST , is the first Piola-Kirchhoff stress
tensor. It is also called the engineering stress tensor, as it is a convenient quantity
for experimental measurements.
Since T is symmetric, we have
ST FT = FS. (11.117)
11.2.5 Balance of Energy for Elastic Materials
The equations for the stress and mass density derived so far are valid for a large class of
continuum bodies independently of their specific material characteristics, including
solids and fluids. To close the system of equations, constitutive relationships between
stress, deformation gradient, rate of deformation, and density must be imposed to
characterize the particular body under consideration. The balance of energy provides
restriction on the form of these constitutive relationships. Here, we turn our attention
to elastic materials.
The general principle for the balance energy states that for any part of a body
⊆ B, the rate of change of the total mechanical energy E is balanced by the power
of the forces P. If we ignore heat dissipation, the total energy for an elastic material
is the sum of the kinetic energy and an internal elastic energy, that is

1
E= ρv · v dv + J −1 W dv, (11.118)
2

kinetic energy internal energy
where W is the internal elastic energy density per unit reference volume.
The power of the forces acting on is given by

P= ρ b · v dv + tn · v da. (11.119)
∂
The balance of energy is then
dE
= P. (11.120)
dt
The transport and localization procedure applied to the energy principle together
with the equations of continuity and motion lead to a local form of energy balance
[975] for the stress power
dW
= tr(SḞ). (11.121)
dt
From a thermodynamic point of view, the combination of S and F in the stress power
tr(SḞ) identifies these two tensors as being work conjugate, that is, they form a
conjugate pair of stress and deformation tensors.
11.3 Constitutive Equations for Hyperelastic Materials
We further assume that the material is hyperelastic. That is, the internal energy density
W is a function of F alone. Explicitly, we posit that
W (X, t) = W (F(X, t), X), (11.122)
in which case, W is referred to as the strain-energy function of the system.

Using (11.55), the time derivative of W is

d ∂W
W (F) = tr Ḟ , (11.123)
dt ∂F
so that the energy balance (11.121) reads now

∂W
tr − S Ḟ = 0. (11.124)
∂F
Since this identity must be true for all motions, we conclude that
∂W
S= , (11.125)
∂F
where we used the derivative of a scalar W with respect to the second-order tensor F
defined with respect to Cartesian bases in the reference and current configurations, by

∂W ∂W ∂W ∂W
= Ei ⊗ e j , = . (11.126)
∂F ∂ F ji ∂F ij ∂ F ji
Written in terms of the Cauchy stress this identity provides a constitutive relationship
relating the Cauchy stress to the deformation gradient:
∂W
T = J −1 F . (11.127)
∂F
11.3.1 Internal Material Constraints
If we consider a material where the possible deformations are constrained during

all motions, extra internal material constraints must be satisfied. We consider the
case where these constraints take the form C(F) = 0 where C(F) is a smooth scalar
function of the deformation gradient. For instance, in the case of an incompressible
material, we assume that all deformations must preserve volume, which implies
det(F) = 1. In this case, C(F) = det(F) − 1.
A simple way to ensure that a constraint holds is to introduce a Lagrangian mul-
tiplier p = p(X, t) and modify accordingly the energy density W → W − pC so
that Equation (11.124) reads now

∂
tr (W − pC) − S Ḟ = 0, (11.128)
∂F
which leads to
∂W ∂C
S= −p . (11.129)
∂F ∂F
In terms of the Cauchy stress, we have
∂W
T = J −1 FS = J −1 F − pN, (11.130)
∂F
where
∂C
N = J −1 F , (11.131)
∂F
is the reaction stress enforcing the constraint. In particular, for incompressible mate-
rials, we have
∂C
= (det F) F−1 = J F−1 , (11.132)
∂F
that is, N = 1. The constitutive relationship for an incompressible hyperelastic
material is then
∂W
T=F − p1. (11.133)
∂F
Recalling that a hydrostatic pressure is a stress that is a multiple of the identity, we
can identify the reaction stress in (11.133) with a hydrostatic pressure. Physically,
we see that a pressure p is required to enforce locally the conservation of volume.
11.3 Constitutive Equations for Hyperelastic Materials 287
For a given W = W (F), the Cauchy stress for compressible or incompressible

materials can be written in the general form
∂W
T = J −1 F − p1, (11.134)
∂F
where J = 1 for an incompressible material and p = p(x, t) must be determined.
If the material is unconstrained, then p = 0.
11.4 Summary of Equations
We can now collect the different equations from the previous chapters to obtain a
closed set of equations
ρ̇ + ρdiv v = 0, continuity equation (11.135)

div T + ρb = ρv̇, equation of motion (11.136)
T = T,
T
symmetry of Cauchy stress tensor (11.137)
∂W
T = J −1 F − p1, constitutive law (11.138)
∂F
Since the elements of F are related to the motion χ by F = Gradχ, and v =
∂t χ(X(x, t), t), there are ten unknowns in this system: the scalar field ρ, the vector
field χ, and the six components of the symmetric tensor T for ten equations (excluding
the third equation that reduces the number of unknowns in T).
It is also sometimes convenient to write these equations with respect to the refer-
ence variables [975]:
ρ̇0 = 0, continuity equation (11.139)

Div S + ρ0 B = ρ0 v̇ equation of motion (11.140)
S F = FS,
T T
symmetry of Cauchy stress tensor (11.141)
∂W
S= − pJF−1 , constitutive law (11.142)
∂F
where the divergence and gradient are now taken in the initial reference config-
uration, ρ0 = J (X, t)ρ(x(X, t), t) is the reference density at a material point,
B = b(x(X, t), t) is the body force acting at the same point, and v̇ = v̇(x(X, t), t)
is the acceleration of a material point.
11.5 Boundary Conditions
Equilibrium and static solutions are obtained by setting v(X, t) = 0 for all X ∈ B0
and for all time t. The equilibrium solutions must satisfy the conditions imposed on
the boundary. Depending on the setting, different types of boundary conditions can
be applied and it is well appreciated that the solutions will depend crucially on these
conditions. The two main types of boundary conditions are dead loading and rigid
loading.
In dead loading, a traction is prescribed and maintained constant throughout the
deformation, i.e., the prescribed traction is independent of the deformation. A typical
example of dead loading is hydrostatic loading where a constant pressure P > 0 is
applied at the boundary in the current configuration. In this case, tn = −Pn, so that
T = −P1 on the boundary.
In rigid loading, fixed displacements are prescribed at the boundary.
In mixed loading, a surface traction tb and deformation xb are prescribed at the
boundary:
Tn = tb for X ∈ ∂B t0 (11.143)
x(X) = xb for X ∈ ∂B d0 (11.144)
where ∂B t0 and ∂B d0 are parts of the body boundary such that ∂B t0 ∪ ∂B d0 = ∂B 0 and
∂B t0 ∩ ∂B d0 = ∅. Note that tb can be a function of the deformation gradient as well
as of the position.
11.6 Objectivity and Material Symmetry
The functional form of the elastic energy can be restricted by combining a funda-
mental principle, the principle of objectivity, together with symmetry properties of
the material.
The principle of objectivity or material-frame indifference [1251] states that mate-
rial properties are independent of superimposed rigid-body motions. For hyperelastic
materials, the principle of objectivity implies
W (QF) = W (F), ∀ Q ∈ S O(3), (11.145)
where S O(3) is the set of all proper orthogonal tensors. The principle of objectivity
implies that W only depends on F through C, so that we can write W (F) = W̄ (C).
Here, to simplify the notation we will drop the overbar and simply write W (F) =
W (C).
Next, we consider the implication of material symmetries. A material is said to
be symmetric with respect to a linear transformation if the reference configuration is
mapped by this transformation to another configuration which is mechanically indis-
11.6 Objectivity and Material Symmetry 289
tinguishable from it. The set of all such linear transformations constitutes a symmetry
group Q ⊆ S O(3). The symmetry condition for a hyperelastic material is [975]
W (FQ) = W (F), ∀ Q ∈ Q. (11.146)
11.7 Isotropic Materials
The maximal possible symmetry group is S O(3), which defines an isotropic material.
The case of isotropic materials is particularly important both for its simplicity and
its wide applicability. Isotropy implies that the strain-energy function depends on F
only through V, where V is the symmetric second-order tensor appearing in the polar
decomposition F = VR. Indeed, choosing Q = RT in (11.146) leads to
W (FQ) = W (FRT ) = W (VRRT ) = W (V). (11.147)
Combining isotropy with objectivity, we have
W (QFQ̃T ) = W (FQ̃T ) = W (F) = W (V), ∀ Q, Q̃ ∈ S O(3), (11.148)
and choosing Q̃ = QR in W (QFQ̃T ) leads to the conclusion that the strain-energy

function satisfies
W (V) = W (QVQT ), ∀ Q ∈ S O(3). (11.149)
The property (11.149) defines W as an isotropic function of V. By inspection,

it can readily be observed that the determinant and trace are simple examples of
isotropic functions of a second-order tensor. More generally, an isotropic function
of V can only depend on V through its three principal invariants
1
{tr(V), tr(V)2 − tr(V2 ) , det(V)}. (11.150)
2
However, since V is a symmetric positive-definite tensor, it is often more convenient
to express W through the principal invariants of the left Cauchy–Green tensor B =
V2 = FFT :
I1 = tr(B) = λ21 + λ22 + λ23 , (11.151)

1 2
I2 = I1 − tr(B2 ) = λ22 λ23 + λ23 λ21 + λ21 λ22 , (11.152)
2
I3 = det(B) = λ21 λ22 λ23 . (11.153)
Equivalently, it implies that W only depends on F through its principal stretches λ1 ,

λ2 , and λ3 (the square roots of the principal values of B). With a slight abuse of
notation, we write either W = W (I1 , I2 , I3 ) or W = W (λ1 , λ2 , λ3 ).
For an isotropic compressible material, we have J = 1, which implies that λ3 =

1/(λ1 λ2 ). Therefore, W can be either expressed in terms of {λ1 , λ2 } or {I1 , I2 }.
The explicit form of the Cauchy stress tensor for a compressible material in terms
of the invariants and their derivatives is
T = w0 1 + w1 B + w2 B2 , (11.154)
where the functions wi depend on the invariants and are given explicitly by
∂W
w0 = 2J − p, (11.155)
∂ I3
∂W ∂W
w1 = 2J −1 + 2J −1 I1 , (11.156)
∂ I1 ∂ I2
∂W
w2 = −2J −1 . (11.157)
∂ I2
As before we choose p = 0 for compressible materials and J = I3 = 1 for incom-

pressible materials. If the reference configuration is assumed to be stress-free, then
we must have T(F = 1) = 0, that is the functions wi = wi (I1 , I2 , I3 ) satisfy
w0 (3, 3, 1) + w1 (3, 3, 1) + w2 (3, 3, 1) = 0. (11.158)
A convenient alternative representation, the Rivlin–Ericksen representation, is obt-

ained from (11.154) by using Cayley–Hamilton’s theorem [525] for B. In terms of
the invariants, the Cayley–Hamilton theorem in three dimensions reads
B3 − I1 B2 + I2 B − I3 1 = 0. (11.159)
Substituting B2 = I1 B − I2 1 + I3 B−1 in (11.154) gives
T = β0 1 + β1 B + β−1 B−1 , (11.160)
where
∂W ∂W
β0 = 2J + 2J −1 I2 − p, (11.161)
∂ I3 ∂ I2
∂W
β1 = 2J −1 , (11.162)
∂ I1
∂W
β−1 = −2J . (11.163)
∂ I2
11.7 Isotropic Materials 291
11.7.1 Adscititious Inequalities
The principle of objectivity together with isotropy leads to a representation of the

strain-energy function in terms of three invariants. This formulation still leads to
many choices for a suitable functional form of W . If we want to establish general
results independent of the particular choice of strain-energy function, we can impose
certain desired behaviors [1291]. For instance, we may require that, in simple exten-
sion, an isotropic elastic body extends rather than shrinks. These conditions take
the form of inequalities either on the coefficients of the constitutive relations, or on
the principal stresses and strains. They are called adscititious inequalities as they
come from outside the theory and are derived empirically from everyday experience
or experiments [1253]. We mention here three standard inequalities often used in
elasticity and suitable for most elastomers.
• Baker–Ericksen inequalities. The Baker–Ericksen inequalities follow from the

requirement that the greater principal stress occurs in the direction of the greater
principal stretch [51] which implies
λi
= λ j ⇒ (ti − t j )(λi − λ j ) > 0, i, j = 1, 2, 3, (11.164)
where {t1 , t2 , t3 } and {λ1 , λ2 , λ3 } are the principal stresses and principal stretches
obtained by the spectral decomposition

3
3
V= λi vi ⊗ vi , T= ti vi ⊗ vi . (11.165)
i=1 i=1
Condition (11.164) imposes the following restrictions on the coefficients

in (11.160):
λi2 λ2j β1 > β−1 , if λi

= λ j , (11.166)
λi4 β1 ≥ β−1 , if λi = λ j . (11.167)
For a hyperelastic body under uniaxial tension, the deformation is a simple exten-
sion in the direction of the (positive) tensile force. The ratio of the tensile strain
to the strain in the perpendicular direction is greater than one if and only if the
Baker–Ericksen inequalities hold [836].
• The ordered-forces inequalities. Similar to the Baker–Ericksen inequalities, the

ordered-forces inequalities state that the greater stretch occurs in the direction of
the greater force. While similar, the two sets of inequalities do not imply each other
[1251, p. 157]. However, it can be shown that if two of the three principal stresses
are nonnegative, then the Baker–Ericksen inequalities follow from the ordered-
forces inequalities, and if two of the three principal stresses are non-positive, then
the ordered-forces inequalities are implied by the Baker–Ericksen inequalities.
• Empirical inequalities. Based on experimental observations in elastomers, the

following empirical inequalities on the coefficients of (11.160) have been postu-
lated [901]:
β0 ≤ 0, β1 > 0, β−1 ≤ 0, (11.168)
for the compressible case. For the incompressible case, only the last two inequali-
ties are considered. These inequalities directly imply the Baker–Ericksen inequal-
ities (11.166) but not conversely.
11.7.1.1 Example: Pure Shear of an Elastic Cube
The adscititious inequalities can be used to establish general qualitative trends. For
instance, consider a homogeneous isotropic hyperelastic cube subject to a pure shear
stress on its top face as shown in Figure 11.8. In Cartesian coordinates, this stress
can be written as T = T (e1 ⊗ e2 + e2 ⊗ e1 ), where T > 0 is constant. Equivalently,
in matrix form, it reads ⎡ ⎤
0 T 0
[T ] = ⎣ T 0 0 ⎦ . (11.169)
0 0 0
Since T is constant, in the absence of body force, the equation of motion is

identically satisfied. The corresponding deformation is homogeneous, that is, the
deformation gradient is independent of the position. For a pure shear stress, it reads

x = aX + b2 − a 2 Y, y = bY, z = cZ . (11.170)
It consists of a triaxial stretch, a pure strain deformation, combined with a simple

shear in the direction of the shear force if and only if the Baker–Ericksen inequalities
hold. Therefore, the Baker–Ericksen inequalities guarantee that the shear strain is in
the same direction as the shear force.
(d,b,c) (a+d,b,c)
(0,1,1) (1,1,1) T
(d,b,0) (a+d,b,0)
(0,1,0) (1,1,0)
(0,0,1) (1,0,1) (0,0,c) (a,0,c)
(0,0,0) (1,0,0) (0,0,0) (1,0,0)
√
Figure 11.8 A cube deformed under pure shear stress T applied to the top face (d = b2 − a 2 ).
If the Baker–Ericksen inequalities are not satisfied, the material would shear in
the direction opposite to the direction of the shear stress. This behavior would be
unrealistic, even though there is no fundamental principle that would rule it out. It
is only through our own experience of everyday materials that we infer its physical
impossibility.
For the deformation (11.170), obtained under pure shear, we define the Poynting
effect as 0 < b
= 1 [901]. Specifically, if a cube is deformed under pure shear, the
positive Poynting effect occurs when b > 1 [99], i.e., the sheared faces spread apart,
whereas the negative Poynting effect is obtained when b < 1, i.e., the sheared faces
are drawn together. It can be shown that the validity of the empirical inequalities (with
β−1 < 0) is a necessary and sufficient condition for the positive Poynting effect as
found in rubber materials [888]. However, experimental observations suggest that
semi-flexible polymer gels exhibit a negative Poynting effect [663], which implies
β−1 > 0. Therefore, it appears that the last of the empirical inequalities may not hold
for some biological materials and should be replaced by the generalized empirical
inequalities [889] which simply state β0 ≤ 0 and β1 > 0.
11.7.2 Choice of Strain-Energy Functions
The choice of strain-energy functions W = W (F) for particular applications is a

controversial and difficult problem. Methods based on a statistical analysis of the
microstructure have been proposed [119, 368]. However, typically, phenomenologi-
cal models are used to capture the essential features of a material such as its behavior
under shear or its strain-hardening or strain-softening properties [617, 620, 1106],
while respecting basic material properties such as convexity and objectivity [1324].
Note that the words model and material are used exchangeably to describe these
particular strain-energy functions. For instance, a neo-Hookean material is a mate-
rial described by the neo-Hookean model, that is, a hyperelastic material with the
particular form of strain-energy density function given below. These models can be
calibrated and fitted to uniaxial or biaxial experiments [603, 604, 751, 768, 1108,
1109, 1273, 1283]. Here, we limit our presentation to a few key popular models
for incompressible materials that capture specific features and are widely used in
applications. A summary is given in Table 11.1.
• Neo-Hookean materials. The simplest model, and the starting point of many
theories, is the neo-Hookean model [370]:
C1
Wnh = (I1 − 3). (11.171)
2
This strain-energy function can be derived from statistical mechanics as a macro-
scopic limit of the energy density of an amorphous cross-linked network of poly-
meric molecules [367, 368, 1245, 1246]. Each molecule in this network is a freely
jointed chain with the same number of monomer units and their end-to-end dis-
tances follow a Gaussian distribution.
For small deformations, the macroscopic parameter C1 can be identified with the
shear modulus μ and is proportional to the product kB T of the Boltzmann constant
with the absolute temperature. The Young’s modulus is then related to C1 by
E = 3μ = 3C1 . (11.172)
A simple generalization of neo-Hookean materials is obtained by assuming that

the strain-energy function is only a function of the first invariant, W (F) = W (I1 ).
This class of models is known as generalized neo-Hookean materials [13, 1327],
for which general results can easily be established [800].
• Mooney–Rivlin materials. The neo-Hookean model can be interpreted as the

lowest approximation of a strain-energy function with respect to the strain tensor
E = (FT F−1)/2, the so-called second-order elasticity approximation. This model
is a good descriptor for many elastomers in tension or compression but it often
fails to capture quantitatively behaviors associated with shear or torsion. The next
order approximation for incompressible isotropic elasticity gives rise to third-order
elasticity [267, 281, 486]. Expressed in terms of invariants, it takes the form of
the Mooney–Rivlin strain-energy function
C1 C2
Wmr = (I1 − 3) + (I2 − 3). (11.173)
2 2
The combination C1 + C2 = μ can be identified again as the shear modulus.
Therefore, we can write
1 1
C1 = μ( + α), C2 = μ( − α). (11.174)
2 2
The Baker–Ericksen inequalities imply that α ∈ [−1/2, 1/2]. This model can
also be derived from statistical mechanics arguments by relaxing some of the
assumptions that lead to the neo-Hookean model [392].
• Ogden materials. A general approach for material modeling, originally proposed

by Ogden [974, 976, 978], consists in considering a general expansion with N
terms of the form
N
μi αi
WogN = λ1 + λα2 i + λα3 i − 3 . (11.175)
i=1
αi
Each parameter μi and αi is a material constant to be determined. These constants

are related to the shear modulus μ of small deformations by

N
μi αi = 2μ. (11.176)
i=1
In practice, the number of terms is limited to N ≤ 6. The possibility of having a

large number of parameters provides a systematic way to explore many different
behaviors and can be used to fit experimental data of both elastomers and biolog-
ical tissues [628, 887, 1305, 1307].
• Fung–Demiray materials. Many soft tissues and elastomers exhibit strong strain-
hardening properties. That is, in simple extension, it becomes increasingly difficult
to further extend the material. Examples of generalized neo-Hookean materials that
have been used to capture this effect in soft tissues are the Fung and Gent models.
The Fung model [404, 405, 766, 1202], in its simplest form, reads
μ
Wfu = [exp(β(I1 − 3)) − 1], (11.177)
2β
where β > 0 controls the strain-hardening property. In the limit β → 0 the Fung
model reduces to the standard neo-Hookean model. This particular form of the
Fung model was first proposed by Demiray in 1972 [270, 271].
• Gent materials. Another popular model is the Gent model that has finite-chain
extensibility enforced by a singular limit of the strain-energy function [432, 618,
620]
Table 11.1 A list of phenomenological strain-energy functions for isotropic incompressible mate-
rials. Note that the materials have been written so that they share the same infinitesimal shear
modulus μ. The limits β → 2 in Wog1 and, β → 0 in Wfu and Wge all lead to the neo-Hookean
strain-energy function. Estimates for 1-term Ogden are from [109, 1145], Gent [432, 435, 617,
619], Fung [269, 601].
Name Definition Soft tissues Elastomers
neo-Hookean Wnh = C21 (I1 − 3)
Mooney–Rivlin Wmr = C21 (I1 − 3)+ C22 (I2 − 3)
2μ β β β
Ogden 1 Wog1 = 2 (λ1 + λ2 + λ3 − 3) β≥9 β≈3
β
μ
Fung Wfu = [exp β(I1 − 3) − 1] 3 < β < 20
2β
μ
Gent Wge = − log[1 − β(I1 − 3)] 0.4 < β <3 0.005 < β < 0.05
2β
μ
Wge = − log[1 − β(I1 − 3)]. (11.178)
2β
The neo-Hookean model is obtained in the limit β → 0.
For a compressible material, there is an extra dependence on the invariant I3 = J 2 .

Again, a choice must be made to take into account the energy associated with local
changes of the volume. Typically, an additive choice is made for which the strain-
energy function is separated into an incompressible part and a compressible part
[614, 801] so that
W = Winc (I1 , I2 ) + Wcomp (I3 ) (11.179)
and a model from Table 11.1 is adopted for Winc (I1 , I2 ). Possible choices for
Wcomp (I3 ) include μc (I3 − 1), μc (J − 1)2 , μc ln I3 , μc ln J , where μc is a mater-
ial parameter related to the bulk modulus.
11.8 Examples
11.8.1 A Simple Homogeneous Deformation
Homogeneous deformations are specified by a constant deformation gradient so that
x = FX + c, (11.180)
where F and c are constant.

For a homogeneous isotropic compressible material, this choice implies that T is
constant, and in the absence of a body load, the equilibrium equations are therefore
trivially satisfied. The solution is then fully specified by the boundary conditions and
the constitutive law [1105].
Consider for example, the diagonal deformation of a cuboid into another cuboid
shown in Figure 11.9 and described by
xi = λi X i , i = 1, 2, 3 (no summation over i). (11.181)
The corresponding deformation gradient is [F] = diag(λ1 , λ2 , λ3 ) and, conse-

quently, the Cauchy stress tensor is also diagonal: [T] = diag(t1 , t2 , t3 ). The defor-
mation and stresses are then obtained from the constitutive law
λi ∂W
ti = , i = 1, 2, 3 (no summation over i). (11.182)
J ∂λi
For instance, in a hydrostatic uniaxial extension, the stress is prescribed in one

direction, that is, t3 = N and t1 = t2 = 0. By symmetry, we have λ1 = λ2 and
11.8 Examples 297
H1 h1
H2 h2
E2 e2
E1 e1
E3 e3 h3
H3
Figure 11.9 A simple diagonal homogeneous deformation transforming a cuboid into another
cuboid. The deformation is characterized by the three constants λi = h i /Hi , i = 1, 2, 3.
1 1
t2 = W2 = 0, t3 = 2 W3 = N . (11.183)
λ2 λ3 λ2
These equations constitute a system of two equations for two unknowns. If we further
specialize these equations to a compressible neo-Hookean material with a strain-
energy function of the form
μ1 μ1 μ2
W = (I1 − 3) − (I3 − 1) + (I3 − 1)2 . (11.184)
2 2 4
This particular form is chosen so that all stresses vanish at λ1 = λ2 = λ3 = 1.
Equations (11.127) then read
1
0 = μ1 ( − λ22 λ3 ) + μ2 λ22 λ3 (λ42 λ23 − 1), (11.185)
λ3
λ3
N = μ1 ( 2 − λ22 λ3 ) + μ2 λ22 λ3 (λ42 λ23 − 1). (11.186)
λ2
The Young’s modulus E is obtained for small deformations as the ratio of uniaxial
stress to the uniaxial stretch, that is

∂ N (λ2 , λ3 ) ∂ N (λ2 , λ3 ) ∂λ2
E= +
∂λ3 λ2 =λ3 =1 ∂λ2 ∂λ3 λ2 =λ3 =1
2μ1 − 3μ2
= 2μ1 . (11.187)
μ1 − 2μ2
If we now consider the same deformation but for an incompressible material

with a neo-Hookean strain-energy function W = μ(I1 − 3)/2, we have λ1 = λ2
again by symmetry, but λ3 λ22 = 1 by incompressibility. We replace the constitutive
law (11.133) by
∂W
ti = λi −p i = 1, 2, 3 (no summation over i). (11.188)
∂λi
The boundary conditions lead to

3
μ λ3 − 1 μ
p= , N (λ3 ) = , (11.189)
λ3 λ3
which defines a Young’s modulus

∂ N (λ3 )
E= = 3μ. (11.190)
∂λ3 λ3 =1
11.8.2 The Half-Plane in Compression
As a second example, we consider an incompressible hyperelastic half-space with a

free surface, characterized by W = W (λ1 , λ2 , λ3 ), under pure homogeneous static
deformation with principal stretch ratios λ1 , λ2 , λ3 [103, 504]. We take e2 = E2
normal to the surface with the half-space located in X 2 > 0, so that λ2 is the
stretch ratio in the direction normal to the free surface. We consider homogeneous
loadings so that the Cauchy stress tensor, in Cartesian coordinates, can be written
[T] = diag(t1 , t2 , t3 ) with deformation gradient [F] = diag(λ1 , λ2 , λ3 ).
There are three typical types of loading which are of interest as shown in
Figure 11.10:
• Equibiaxial strain. An equibiaxial strain is associated with deformations with

equal strains in the plane, that is λ1 = λ3 . In this case, the half-space is compressed
with equal forces in the 1- and 3-directions, and expands freely in the 2-direction,
so that
t1 = λ1 W1 − λ2 W2 , t2 = 0, t3 = t1 , (11.191)
where Wi = ∂λi W .
A. B. C.
X2<0 t2=0 t2=0 t2=0
t3=t1
X2>0 λ3=1
t1 t1 t1 t1 t1 t1
t3=0
t3=t1
Equibiaxial Plane strain Uniaxial

-2 -1 -1/2
λ3=λ1, λ2=λ1 λ3=1, λ2=λ1 λ2=λ3, λ2=λ1
Figure 11.10 Three typical deformations of a half-space. A. Equibiaxial strain. B. Plane strain. C.
Uniaxial strain.
11.8 Examples 299
• Plane strain. Plane strain corresponds to the condition λ3 = 1. It corre-

sponds to a half-space compressed in the 1-direction and prevented from expand-
ing/contracting in the 3-direction. Thus, it expands in the 2-direction, normal to
the free surface. This deformation is maintained by applying the tractions
t1 = λ1 W1 − λ2 W2 , t2 = 0, t3 = λ3 W3 − λ2 W2
= t1 . (11.192)
• Uniaxial strain. A uniaxial strain is defined by t3 = 0. That is, the half-space is

free to expand in the 2- and 3-directions, and
t1 = λ1 W1 − λ2 W2 , t2 = 0, t3 = 0. (11.193)
It follows from the incompressibility condition that λ1 λ2 λ3 = 1 and therefore the

three cases can be written in general as
⎧
⎨ n = −2 equibiaxial,
λ2 = λn1 with n = −1 plane strain, (11.194)
⎩
n = − 21 uniaxial.
Therefore, we can use λ1 to fully characterize the deformation. For a given strain-
energy function, the stresses developed as a function of λ1 can be computed by direct
evaluation of the relations (11.191–11.193).
Once the stresses are known, a natural problem is to look for possible bifurcations,
that is, we wish to identify a critical value of λ1 such that the half-space develops
surface wrinkles as one would expect when compressing a large rubber block. We
will consider this problem in Section 11.10.
11.8.3 The Inflation–Extension of a Tube
As an example of nonhomogeneous static deformations, we consider the problem of

an incompressible hyperelastic cylindrical shell subject to combined extension and
inflation [311, 621, 699] in the absence of body loads. We consider a simple thought
experiment in which the tube is capped at both ends and subject to an axial extension
ζ due to an internal pressure P and to a total axial load N on the top cap. The tube
of initial inner radius A and outer radius B > A and height H is then deformed into
a tube with radii a, b and height h as shown in Figure 11.11. We consider a finite
deformation in which the cylinder is allowed to inflate and extend while remaining
cylindrical at all times regardless of possible stability issues [96, 197, 484, 552,
553].
For this problem, the deformation x = χ(X, t), in cylindrical coordinates {r, θ, z}
and {R, , Z } reads
2B
H
ez
χ
h
-P
2b N
Figure 11.11 Inflation–extension of a tube. The tube is inflated with an internal pressure P (which
is equivalent to an external pressure −P as shown) and an axial load N resulting from an applied
load and the pressure acting on the end caps.
r = r (R), θ = , z = ζ Z , (11.195)
where ζ is the constant axial stretch of the cylinder such that h = ζ H . The position
vectors in the reference and current configurations are
X = RE R + Z E Z , x = r (R)er + ζ Z ez , (11.196)
where {E R , E , E Z } and {er , eθ , ez } are the two standard cylindrical bases. Follow-
ing the identity (11.60), the deformation gradient F = Grad(χ) with respect these
coordinates is given by
r
F = r er ⊗ E R + eθ ⊗ Eθ + ζez ⊗ E Z , (11.197)
R
where the prime denotes differentiation with respect to R. Equivalently, we can write
r
[F] = diag(r , , ζ) ≡ diag(λr , λθ , λz ), (11.198)
R
which defines the three principle stretches {λr , λθ , λz }.
The incompressibility condition det(F) = 1 = λr λθ λz gives
R
r r = , (11.199)
ζ
which, together with r (A) = a, leads to

R 2 − A2
r= a2 + . (11.200)
ζ
Then, λ = λθ is given by
11.8 Examples 301

r 1 R 2 − A2
λ= = a2 + . (11.201)
R R ζ
Therefore, the deformation is fully specified by two parameters: the axial stretch ζ
and the radial stretch of the inner wall λa = a/A so that

b 1 A2
λb = = 1 + 2 ζλa2 − 1 . (11.202)
B ζ B
Since the deformation is diagonal in cylindrical coordinates and only depends on

R, it follows from Equation (11.154) that the Cauchy stress tensor is also diagonal
in these coordinates so that
[T] = diag(tr , tθ , tz ) ⇔ T = tr er ⊗ er + tθ eθ ⊗ eθ + tz ez ⊗ ez . (11.203)
This particular form of the Cauchy stress tensor implies that the Cauchy equation
div T = 0 in cylindrical coordinates reduces to a single scalar equation
dtr 1
+ (tr − tθ ) = 0. (11.204)
dr r
This equation can be integrated once over r to obtain:
r
tθ − tr
tr (r ) = tr (a) + dr, r ∈ [a, b]. (11.205)
a r
We now examine the boundary conditions. First, due to inflation, the jump in pressure
between the inner and outer sides of the cylinder is P = tr (b) − tr (a). Without loss
of generality, we choose
tr (a) = −P, tr (b) = 0, (11.206)
which in (11.205) implies

b
tθ − tr
P= dr. (11.207)
a r
Second, the boundary conditions on the two caps of the tube, defined as the two disks
at z = 0 and z = h, can be written as
tz (z = 0) = Nz , tz (z = h) = Nz . (11.208)
However, since a constant axial stretch ζ cannot be used to fit a constant Nz , we

replace this pointwise condition on the caps of the cylinder by an integral condition
for the total axial load applied on the cap [1085, 1086]
b
2π Nz r dr = N ≡ F + χPπa 2 , (11.209)
a
thereby eliminating the explicit dependence on the variable r . The total axial load N
is further decomposed into an external applied load F, which corresponds to pulling
or compressing the tube, and the load created by the internal pressure acting over the
cap. That load is simply the pressure multiplied by the projected area of the cap. The
coefficient χ is 1 for a capped cylinder, and 0 for an infinite cylinder [207, 484].
For incompressible materials, this last expression is not the most practical one
as the term tz will contain an arbitrary pressure. An equivalent expression can be
obtained by adding and subtracting Tr to obtain
b b
2π Tz r dr = 2π (Tz − Tr + Tr )r dr. (11.210)
a a
The last term can be integrated by parts, and use of the balance law (11.204) gives
b b
2π Tz r dr = π (2Tz − Tr − Tθ )r dr + Pπa 2 , (11.211)
a a
which implies
b
π (2Tz − Tr − Tθ )r dr = F + (χ − 1)Pπa 2 , (11.212)
a
and the last term vanishes for a capped cylinder.

To close the system, we use the constitutive law:
∂W ∂W ∂W
tr = λr − p, tθ = λθ − p, tz = λz − p, (11.213)
∂λr ∂λθ ∂λz
and substitute the functions of (λr = 1/(λζ), λθ = λ, λz = ζ) in (11.207–11.209)

so that tθ − tr = Q(r, λa , ζ).
The semi-inverse problem consists in finding the values of (λa , ζ) corres
ponding to the two external loads (F, P) through the analysis of the two equa-
tions (11.207–11.209). Note that once (λa , ζ) is known, the stress tr (r ) can be com-
puted as r
tr (r ) = −P + Q(r, λa , ζ) dr, (11.214)
a
and the remaining stresses are obtained from (11.213).

11.8 Examples 303
11.8.3.1 A Toy Model for an Artery
As an example, we consider a toy model for an artery subject to pressure and tensile
stretch. The artery is modeled as a capped tube (χ = 1) made of a Fung material
with values taken from Table 11.1. The system is subjected to a fixed axial force F
and varying pressure P [1275]. We solve Equations (11.207)–(11.209) for (λa , ζ)
and plot P as a function of ζ, so that if the pressure is controlled, one can determine
the amount of axial stretch in the tube.
First, we consider the case of a neo-Hookean material. The curve P(ζ) shown
in Figure 11.12 is non-monotonic and presents a maximal value of pressure after
which unbounded extensional growth follows. This behavior is a drawback of the
neo-Hookean model which is not well defined for arbitrarily large deformations and
stresses.
Second, we consider the effect of the strain-stiffening parameter β. In Figure 11.13,
we see that for small values of β, a non-monotonic behavior is observed in moder-
ate deformations, followed by a rapidly increasing pressure for larger deformations.
For larger values of the strain-stiffening parameter, the behavior is monotonic and
increasingly large pressures are needed for further small incremental extensions.
We conclude that the behavior of the neo-Hookean model is qualitatively different
from the behavior of the Fung model even for very small values of the strain-stiffening
parameter β. In particular, typical axial stretches in arteries are around 1.3 to 1.6, in
the region where the response of the structure clearly depends on the choice of the
material model. This simple computation clearly demonstrates the importance of the
choice of the material model for a given problem and the need to use nonlinear rather
than linear elasticity to study these problems.
P
P1
0.12
0.10
0.08
0.06
0.04
0.02
1.0 1.5 2.0 2.5 3.0
Figure 11.12 Behavior of a tube under pressure and axial force for a neo-Hookean material ( A =
1, B = 1.2, μ = 1, F = 0.1, χ = 1). The neo-Hookean model is ill-defined past a critical
pressure P1 .
P
0.35 =0.1
0.30 =0.09
=0.08
0.25
=0.07
0.20 =0.06
=0.05
0.15 =0.04
=0.03
0.10 =0.02
=0.01
0.05
1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Figure 11.13 Behavior of a tube under pressure and axial force for a Fung material (β > 0).
Parameters for all figures:A = 1, B = 1.2, μ = 1, F = 0.1, χ = 1.
The non-monotonic behavior shown in Figure 11.14, for intermediate values of

β, is the well-known limit-point instability that already appeared in the experiment
by Osborne and Sutherland (compare Figure 11.1 with Figure 11.14) for spherical
and cylindrical shells under internal pressure [8, 17, 198, 433, 922, 974]. For certain
materials, the pressure–stretch curve for spherical shells may present a maximum,
followed by a minimum. In that case, once the maximum is reached, and the pressure
is increased, the stretch will “jump” to a significantly higher value. In spherical shells,
several authors have shown that this limit-point instability disappears as the strain-
hardening parameter is increased [72, 109, 635, 975].
It is tempting to associate certain pathologies such as aneurysms appearing in
arteries to a possible mechanical instability. However, the stabilizing effect due to
strain-stiffening leads us to conclude that a simple explanation for the formation of
aneurysms in terms of limit-point instability is not plausible [262, 746]. The same
general trend is exhibited by more realistic models of arteries, involving multiple
layers, fiber anisotropy, and residual stress discussed in Section 16.5. Nevertheless,
various authors have looked at the interesting possibility that aneurysms could be
triggered by a local mechanical instability resulting in a localized, but stable, bulged
configuration that would then evolve slowly and remodel [398–400, 1021, 1022,
1092]. Aneurysms are such complex progressive diseases that it is unlikely that their
formation could be explained simply as a mechanical phenomenon. Still, it is now
appreciated that mechanics and mechanical feedback play an important role in the
proper function of arteries and in the formation of aneurysms [425, 1083].
11.8 Examples 305
P
0.160 =0.05
0.155
0.150
0.145
P1
P2
0.140
0.135
0.130
3 1 2
1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Figure 11.14 Limit-point instability of a tube under pressure and axial force for a Fung material
(β = 0.05). The curve P(ζ) has a minimum and a maximum. If the internal pressure of the tube
is raised to P1 , a sudden extension occurs through a jump from ζ1 to ζ2 . When the pressure is then
decreased to P2 , a second jump occurs to ζ3 (lower arrow).
11.9 Universal Deformations for Isotropic Materials
One of the advantages of the general formalism of nonlinear elasticity is the pos-
sibility of solving boundary-value problems for arbitrary strain-energy functions.
However, it is not clear that a given class of deformations can exist for any strain-
energy function in the absence of body loads. A semi-inverse problem consists in
specifying a class of deformations (for instance the inflation of a sphere) with sev-
eral unknown functions or constants that are determined through the equilibrium
equations. The question is then to determine all transformations that can be effected
through boundary traction in every homogeneous isotropic hyperelastic material, and
in the absence of body forces, the so-called universal deformations for which the
semi-inverse problem is well defined [71, 342, 343].
For compressible materials, Ericksen [343] proved that the homogeneous defor-
mations given in Section 11.8.1 are the only possible universal deformations. That
is, without further restricting the class of strain-energy functions or applying body
loads, homogeneous deformations are the only ones that can be sustained for arbitrary
strain-energy functions [1047].
For incompressible materials, a number of interesting universal deformations are
known, but the general problem of determining all such possible deformations is
still open [71, 373]. In addition to homogeneous deformations (classified as Family
0), there are five known families of universal deformations. Each family of solutions
exists for all strain-energy functions and suitable boundary tractions. Since the defor-
mation is known, the boundary tractions needed to maintain a given solution can be
Figure 11.15 Family 1. Bending, stretching, and shearing of a rectangular block. When the para-
meter c vanishes, the deformation corresponds to the bending/stretching of a block. The parameter
c is associated to shearing out of the bending plane.
found by evaluating the Cauchy stress tensor at the boundary. Doing so, one is able
to relate the parameters appearing in the solution to the loads required to maintain
them.
• Family 0. Homogeneous deformations of a rectangular block:
xi = Fi j X j , i = 1, 2, 3, (11.215)
where (X 1 , X 2 , X 3 ) and (x1 , x2 , x3 ) are the Cartesian coordinates of a material

point in the reference and current configurations, respectively. The deformation is
specified by the nine constants appearing in the deformation gradient F.
• Family 1. Bending, stretching, and shearing of a rectangular block as shown in

Figure 11.15. This deformation is defined with three arbitrary constants (a, b, c)
(ab
= 0) by
√ Z
r= 2a X , θ = bY, z = − bcY, (11.216)
ab
where (X, Y, Z ) and (r, θ, z) are the Cartesian and cylindrical coordinates of a
material point in the reference and current configurations, respectively.
In these coordinates, the matrix of components of the deformation gradient reads

(for a > 0):
11.9 Universal Deformations for Isotropic Materials 307
Figure 11.16 Family 2. Straightening, stretching, and shearing of a sector of a cylindrical shell.
When the parameter c vanishes, the deformation corresponds to the straightening of the cylinder.
The parameter c is associated to extra shearing of the resulting block.
⎡√ √ ⎤
a/ 2X √0 0
[F] = ⎣ 0 b 2a X 0 ⎦. (11.217)
0 −bc 1/(ab)
• Family 2. Straightening, stretching, and shearing of a sector of a cylindrical shell,

depicted in Figure 11.16, and defined for three arbitrary constants (a, b, c) (ab
=
0) by
1 2 2 Z
x= ab R , y= , z = −c , (11.218)
2 ab b ab
where (R, , Z ) and (x, y, z) are the cylindrical and Cartesian coordinates of a
material point in the reference and current configurations, respectively.
d=e d e d e
Figure 11.17 Family 3. Inflation, bending, torsion, extension, and shearing of a cylindrical shell.
The parameters d and e control the torsion and the shearing along the axis.
In these coordinates, the matrix of components of the deformation gradient is

given by ⎡ 2 ⎤
ab R 0 0
[F] = ⎣ 0 1/(ab R) 0 ⎦ . (11.219)
0 −c/(ab R) 1/b
• Family 3. Inflation, bending, torsion, extension, and shearing of a sector of

an annular wedge, shown in Figure 11.17, defined by six arbitrary constants
(a, b, c, d, e, f ) with the constraint a(c f − de) = 1:

r= a R 2 + b, θ = c + d Z , z = e + f Z , (11.220)
11.9 Universal Deformations for Isotropic Materials 309
Figure 11.18 Family 4. Inflation of a spherical shell to another spherical shell. The deformation is
controlled by a single parameter. A possible choice is the ratio of current to reference inner radii.
where (R, , Z ) and (r, θ, z) are the cylindrical coordinates of a material point in
the reference and current configurations, respectively. The shearing is both axial
(torsional) and azimuthal, that is, the region at constant Z becomes a helicoidal sur-
face.
given by
Figure 11.19 Family 5. Inflation, bending, extension, and azimuthal shearing of an annular wedge.
Note that the deformation is essentially planar since the deformation in the z–direction is homoge-
neous.
⎡ ⎤
√ aR 0 0
a R 2 +b √
⎢ √ ⎥
[F] = ⎣ 0 c a R 2 +b
d a R2 + b ⎦ . (11.221)
R
0 e/R f
• Family 4. Inflation (+) or inversion (−) of a sector of a spherical shell, shown in

Figure 11.18 with a single constant a:

r= ±R 3 + a, θ = ±, φ = ,
3
(11.222)
where (R, , ) and (r, θ, φ) are the spherical coordinates of a material point in
the reference and current configurations, respectively.

given by ⎡ ⎤
2
± 3 R 2/3 0 0
⎢ ( R +a ) √ ⎥
[F] = ⎢⎣ 0
3 3
± RR +a √0
⎥.
⎦ (11.223)
3 3
R +a
0 0 ± R
• Family 5. Inflation, bending, extension, and azimuthal shearing of an annular

wedge, shown in Figure 11.19, with five constants (a, b, c, d, e) constrained by
the condition a 2 ce = 1:
r = a R, θ = c + d log(b R), z = eZ , (11.224)
where (R, , Z ) and (r, θ, z) are the cylindrical coordinates of a material point in
the reference and current configurations, respectively.

given by ⎡ ⎤
a 0 0
[F] = ⎣ ad ac 0 ⎦ . (11.225)
0 0 e
11.10 Bifurcation, Buckling, and Instability
For given material and reference configurations with known boundary conditions, the
deformations and stresses at equilibrium can be computed, in principle, by solving
the system (11.135–11.138). This finite deformation solution represents a mechani-
cal equilibrium whose stability can be probed by the theory of bifurcation. That is,
we consider a wider class of deformations and establish the possibility of a bifurca-
tion of a base state by looking for adjacent equilibrium solutions. We note that the
11.10 Bifurcation, Buckling, and Instability 311
F(0) εF(1)
Figure 11.20 Bifurcation: A small incremental deformation is added over a known large-
deformation equilibrium solution in order to obtain a new solution valid near the base solution.
The dotted line represents an addition on top of the deformation and should be understood as the
mapping 1 + F(1) .
correspondence between mechanical stability and bifurcation is far from trivial as

noted by many authors [95, 195, 197, 548]. Here, we take the traditional viewpoint
of buckling theory: The existence of a bifurcation offers the possibility of an insta-
bility that can be further analyzed either by looking at the dynamics associated with
new solutions, as done in the case of rods (Section 6.5), or by analyzing the second
variation of the energy, if applicable [257, 713, 785].
In this type of analysis, one starts with a known family of solutions with several
parameters. Once the equilibrium homogeneous solution has been obtained, one can
study the existence of nearby equilibrium solutions with respect to change in the
parameters and identify the value of the parameters where such new solutions first
appear.
The general procedure consists in implementing a perturbation expansion in which
the new deformations, superimposed on the finite deformation, are assumed to be
arbitrarily small as shown in Figure 11.20. This type of deformation in nonlinear
elasticity is often referred to as incremental deformations [103, 402, 975] or small-
on-large deformations [555]. Since the incremental deformations are assumed to be
small, the full nonlinear system can be linearized and bifurcation points are obtained
as the values at which the linearization fails to be injective.
In our setting, we assume that a family of solutions for the Cauchy stress T(0)
and deformations χ(0) is known for given boundary conditions and is controlled by
several parameters. We refer to this solution as the base solution. We seek solutions
close to the base solution by formally linearizing the governing equations around the
base state. Here, for simplicity, we assume that the material is incompressible and
hyperelastic and, for this discussion, neglect body force. A detailed discussion of
the general bifurcation scheme under arbitrary kinematic constraints and growth is
given in Section 14.8. Here, following [465], we present the method by decomposing
it into four steps to produce a closed set of linear equations. Then, we illustrate the
method on a simple example.
Step 1: Kinematics of incremental deformations. We start with the first-order

kinematics by considering a finite deformation χ(0) on which we superimpose an
incremental deformation χ(1) as
χ = χ(0) + χ(1) , (11.226)
where is an arbitrary parameter controlling the amplitude of the perturbation. It

follows that the deformation gradient is

F = Grad(χ) = 1 + F(1) F(0) , (11.227)
where
F(1) = grad (χ(1) ), (11.228)
is expressed in the current configuration. Explicitly, if we write χ(0) = x(X) and

χ(1) = u(x(X)), we have
Grad u(x(X)) = (grad u(x)) Grad x(X) = (grad u) F(0) . (11.229)
Step 2: Linearized equations. Second, we obtain the linearized equations of motion

by expanding the Cauchy stress tensor to first order:
T = T(0) + T(1) + O(2 ). (11.230)
Then, the Cauchy equation (11.136), div T = 0, written here in the absence of body
force, implies both
div(T(0) ) = 0, (11.231)
and
div(T(1) ) = 0. (11.232)
Step 3: Linearized constitutive equations. Third, we relate the first-order Cauchy

stress tensor to the deformation gradient by expanding the constitutive relationship
∂W
T=F − p1, (11.233)
∂F
to first order:
T(0) = F(0) WF(0) − p (0) 1, (11.234)

T (1)
=L:F (1) (1) (0)
+F F WF(0) (1)
− p 1, (11.235)
where p = p (0) + p (1) , L is the fourth-order tensor defined by


(0) (1) (0)
L : F(1) = F(0) WFF : F F , (11.236)
and WF(0) , WFF

(0)
are, respectively, the first and second derivatives of W with respect
to F evaluated on F(0) . In Cartesian coordinates, these expressions read
∂W (F(0) )
WF(0) = , (11.237)
ij ∂ F ji(0)
∂ 2 W (F(0) )
(0)
WFF = (0)
, (11.238)
m jnl ∂ F jm ∂ Fln(0)
and
L : F(1) ij
= Li jkl Flk(1) = Fim
(0) (0)
WFF Flk(1) Fkn
(0)
. (11.239)
m jnl
The components of the fourth-order tensor L are the instantaneous elastic moduli
which are defined as
(0) (0) (0)
Li jkl = Fim WFF Fkn . (11.240)
m jnl
These moduli characterize the linear behavior of a material at a finite value of the
deformation. For isotropic materials, their form can be given explicitly [975, p. 412].
Let {λ1 , λ2 , λ3 } be the principal values of F(0) . Then for an isotropic hyperelastic
material the components of L on the Eulerian principal axes associated with F(0) are
∂2 W
Lii j j = λi λ j , (11.241)
∂λi ∂λ j

∂W ∂W λi2
Li ji j = λi − λj , i
= j, λi
= λ j , (11.242)
∂λi ∂λ j (λi − λ2j )
2
∂W
Li j ji = L jii j = Li ji j − λi , i
= j. (11.243)
∂λi
If λi = λ j for some i
= j, then

1 ∂W
Li ji j = Liiii − Lii j j + λi , (11.244)
2 ∂λi
Liiii = L j j j j , Liikk = L j jkk , (11.245)
Likik = L jk jk , Likki = L jkk j . (11.246)
Step 4: The boundary-value problem. The stability analysis proceeds by solving

the equation for mechanical equilibrium (11.136) to first order in , that is,
div T(1) = 0, (11.247)

together with the incompressibility condition to first order. The incompressibility

condition is det F = 1, which, using Jacobi’s identity (11.52), reads to first order:
tr F(1) = div u = 0. (11.248)
In some cases, the geometry of the problem and the deformations considered are
simple enough that the condition for instability related to the existence of solu-
tions for Equation (11.247) can be written in terms of W and the principal stretches
{λ1 , λ2 , λ3 }.
11.10.1 Example: Bifurcation of the Half-Plane
As an example, we consider the problem of a half-plane under compression and look

for the possibility of bifurcations. We assume that the half-space is a incompressible
hyperelastic half-space with a free surface under pure homogeneous static deforma-
tion as studied in Section 11.8.2. The half-space is located in X 2 > 0 and λ2 is the
stretch ratio in the direction normal to the free surface. We recall that the principal
stretch λ1 can be used to characterize the deformation so that the components of
Cauchy stress tensor can be written as
[T(0) ] = diag(t1 , t2 , t3 ), (11.249)
and the deformation gradient is
[F(0) ] = diag(λ1 , λ2 , λ3 ). (11.250)
The three cases of interest are

⎧
⎨ n = −2 equibiaxial,
λ2 = λn1 with n = −1 plane strain, (11.251)
⎩
n = − 21 uniaxial,
and λ3 = λ−n−1
1 . Since t2 = 0, we have p = λ2 W2 and
t1 = λ 1 W 1 − λ 2 W 2 , t3 = λ 3 W 3 − λ 2 W 2 , (11.252)
where Wi = ∂λi W . Therefore, all the components of the stress tensor and the defor-
mation gradient tensor are functions of the control parameter λ1 .
Step 1: Kinematics. We look for values of λ1 such that the system supports nonho-
mogeneous solutions in the form of creases that are uniform in the e3 –direction as
shown in Figure 11.21, that is, we look for solutions of the form
χ(0) εχ(1)
t2=0
t1 t1 t1 t1
t3=0
Figure 11.21 Bifurcation of a half-space under uniaxial compression. The finite deformation com-
presses the half-space. For the same loading, there exists another solution with surface creasing.
u = [u(x1 , x2 ), v(x1 , x2 ), 0], (11.253)

p (1) = q(x1 , x2 ), (11.254)
T(1) = T(1) (x1 , x2 ). (11.255)
Step 2: Linearized equations. The two nonvanishing equations of equilibrium and

the incompressibility condition form a system of three equations for {u, v, q}:
(1)
∂T11 ∂T (1)
+ 21 = 0, (11.256)
∂x1 ∂x2
(1)
∂T12 ∂T (1)
+ 22 = 0, (11.257)
∂x1 ∂x2
∂u ∂v
+ = 0. (11.258)
∂x1 ∂x2
Step 3: Linearized constitutive equations. The components of the linearized stress

tensor are
(1) ∂u ∂v ∂u
T11 = L1111 + L1122 −q + (t1 + p), (11.259)
∂x1 ∂x2 ∂x1
(1) ∂u ∂v ∂u
T12 = L1221 + L1212 + (t2 + p), (11.260)
∂x2 ∂x1 ∂x2
(1) ∂u ∂v ∂v
T22 = L2211 + L2222 −q + (t2 + p), (11.261)
∂x1 ∂x2 ∂x2
(1) ∂u ∂v ∂v
T21 = L2121 + L2112 + (t1 + p). (11.262)
∂x2 ∂x2 ∂x1
Step 4: Solving the boundary-value problem. The stress tensor is constant with
respect to (x1 , x2 ). Therefore, System (11.256–11.258) is a linear, homogeneous,
system of PDEs with constant coefficients. It can be verified by direct substitution
that it admits exponential solutions of the form
{u, v, q} = eikx1 +iksx2 {U, V, ik Q}. (11.263)
For our problem, the relevant solutions are periodic along x1 and exponentially
decaying for increasing x2 , that is, we choose
k > 0, Im(s) > 0. (11.264)
Substitution of (11.263) in (11.256–11.258) gives a system of three homogeneous

linear equations for {U, V, Q} and the condition for the existence of a solution is
a0 + 2a2 s 2 + a4 s 4 = 0, (11.265)
where
a0 = L1212 , (11.266)
2a2 = L1111 + L2222 − 2L1122 − 2L1221 . (11.267)
a4 = L2121 . (11.268)
In terms of λ1 , these coefficients are
λ1 W 1 − λ2 W 2 2
a0 = λ1 , (11.269)
λ21 − λ22
λ1 W 1 − λ2 W 2
2a2 = λ21 W11 + λ22 W22 − 2λ1 λ2 W12 − 2λ1 λ2 , (11.270)
λ21 − λ22
λ1 W 1 − λ2 W 2 2
a4 = λ2 , (11.271)
λ21 − λ22
where we now regard W as a function of λ1 and λ2 , that is, W̄ (λ1 , λ2 ) = W (λ1 , λ2 , λ3

(λ1 , λ2 )), Wi = ∂λi W̄ and Wi j = ∂λ2i λ j W̄ .
We need conditions on these coefficients to identify the roots of (11.265) that
satisfy (11.264). Here, we use the so-called strong-ellipticity conditions. These con-
ditions naturally arise when one considers the propagation of a plane homogeneous
wave superimposed on the compression of the half-space. If we require that the speed
of propagation is real, conditions on the parameters can be obtained [280, 296]. In
our case, these conditions are given by
√
a0 > 0, a4 > 0, a2 + a0 a2 > 0. (11.272)
These conditions imply that there exist two roots of
a0 + 2a2 s 2 + a4 s 4 = 0 (11.273)
with Im(s) > 0. Let s1 and s2 be these two roots and {U1 , V1 , Q 1 }, {U2 , V2 , Q 2 },
the corresponding solutions. Then, the incremental solution is an arbitrary linear
combination of these solutions, that is,

u = eikx1 C1 U1 eiks1 x2 + C2 U2 eiks2 x2 , (11.274)

v = eikx1 C1 V1 eiks1 x2 + C2 V2 eiks2 x2 , (11.275)

q = eikx1 C1 Q 1 eiks1 x2 + C2 Q 2 eiks2 x2 . (11.276)
The bifurcation condition is obtained by imposing the linearized boundary condition

T(1) · e2 = 0. The compressed half-space becomes unstable and develops surface
instability for critical principal stretch ratio λ1 /λ2 such that
2
λ1 a2 λ1
+2 1+ − 1 = 0. (11.277)
λ2 a4 λ2
Explicitly, in terms of W and its derivatives, the bifurcation condition is

λ2
λ2 W1 + (2 − )W2 + λ21 W11 − 2λ1 λ2 W12 + λ22 W22 = 0. (11.278)
λ1
Next, we look at this condition for various strain-energy functions. We start with
the classic elastomer modeled by the Mooney–Rivlin energy Wmr from Table 11.1.
In this case, the criterion (11.278) leads to a universal condition (independent of μ
and ν) [280]:
λ1
Tangential biaxial compression (λ =λ3)
1
0.6
Uniaxial compression (λ =1)

0.5 3
0.4
0.3
0.2
0.1
β
0 0.1 0.2 0.3 0.4 0.5
Figure 11.22 Critical values of the stretch ratio λ1 for the instability of a Fung elastic half-space
characterized by a strain-hardening parameter β.
λn+2
1 + 3λ2n+1
1 − λ3n
1 + λ1 = 0.
3
(11.279)
Depending on n, we obtain the classic values for the critical compression ratio of
instability for a half-plane made of a material with the Mooney–Rivlin (or neo-
Hookean) strain-energy function. Green and Zerna [504] found (λ1 )cr = 0.66614
under biaxial compression (n = −2); Biot [102] found (λ1 )cr = 0.54369 under plane
strain (n = −1) and (λ1 )cr = 0.44375 under uniaxial compression (n = −1/2).
Next, we turn to the popular Fung strain-energy function for soft tissues. We
take W = Wfu in (11.278) and obtain after simplification the following bifurcation
condition:
λn+2
1 + 3λ2n+1
1 − λ3n
1 + λ1 + 2β(λ1 − 2λ1
3 5 3+2n
+ λ14n+1 ) = 0. (11.280)
For n = −1 or n = −2, there is a critical value α−1 = 1/2, α−2 ≈ 0.1644

after which the bifurcation criterion has no positive real root, see Figure 11.22. We
conclude that a Fung elastic half-plane under either axial or tangential compression
is always stable for realistic physiological values of the parameters. For n = −1/2
(normal biaxial compression), the criterion has a positive real root for all α, which
however decreases rapidly toward zero. Hence, for β > 3, the half-space can be
compressed by more than 97% before the bifurcation criterion is met. Next, we
consider the Gent strain-energy function, originally proposed for rubber [432], but
also used for the modeling of strain-hardening soft tissues. We take W = Wge in
(11.278) and obtain after simplification of the following bifurcation condition:
λ1
Tangential biaxial compression ( λ1 =λ 3 )
0.7
Uniaxial
0.6 compression
(λ 3 =1)
0.5 Normal biaxial compression
(λ 2 =λ 3)
0.4
0.3
0.2
0.1
β
0
0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Figure 11.23 Critical values of the stretch ratio λ1 for the surface instability of a Gent elastic
half-space characterized by a strain-hardening parameter β.
λ1
0.6
Tangential biaxial compression (λ 1 =λ 3)

0.5
0.4
Uniaxial
0.3
compression (λ3 =1)
0.2
Normal biaxial
compression (λ2 =λ3)
Neo-Hookean case β
0.1
5 10 15 20 25 30
Figure 11.24 Critical values of the stretch ratio λ1 for the instability of a Ogden elastic half-space
characterized by a strain-hardening parameter β.
β(3λn−1
1 − 2λn+1
1 − 11λ2n
1 − λ1
4n−2
+ 9λ2n−2
1 + λ5n−3
1 + λ3n−1
1 − 3λ3n−3
1 + 3)
+λn−1
1 + 3λ2n−2
1 − λ13n−3 + 1 = 0. (11.281)
For n = −1, n = −2, and n = −1/2, there is a critical value β−1 ≈ 0.122,
β−2 ≈ 0.066, and β−1/2 ≈ 0.170 after which the bifurcation criterion has no positive
real root, see Figure 11.23. We conclude that a Gent elastic half-plane under axial,
tangential, or normal compression is always stable for realistic physiological values
of the parameters.
Finally, we use the one-term Ogden model, for which the bifurcation condition
reads

λ1 n+β + λ1 nβ+1 − λ1 n(1+β) − λ1 1+β + β λ1 1+β + λ1 nβ+1 = 0. (11.282)
The left-hand side of this equation is equal to −1 for λ1 = 0 and to 2β for λ1 = 1.

Therefore, it admits a real root for all positive values of β and for all values of n.
The bifurcation curves are shown in Figure 11.24.
We comment that no finite body is a half-space. The instability found in the half-
space is relevant when finite size effects can be ignored. Otherwise, we expect a
body to become unstable through other deformation modes such as the ones found
in buckling or barreling. We will encounter such instabilities when we study the
deformations of cylinder and spheres induced by growth.
11.11 Anisotropic Materials
Many elastic biological tissues have highly anisotropic mechanical properties [408,
p. 500]. These anisotropic properties are determined primarily by the presence of
fibers [386, 1368]. For instance, for soft tissues such as tendons, arterial wall, aortic
valve, myocardium, and pericardium, anisotropy is determined in the first place by
the arrangement of the collagen fibers [408, 599, 635, 686, 769] in a mostly isotropic
elastin matrix.
Collagen is a fibrous protein, which comprises 25% of the total protein mass in
mammals [15, p. 1184] and is the most abundant protein in vertebrates. Collagen is
present in tissues in various forms, in particular, in the form of fibrils, which consist
of many cross-linked collagen molecules, that are from 50 to 500 nm in diameter
and can further organize into fascicles. Special collagen type, geometry, density, and
arrangement endow tissues with anisotropic mechanical properties. Collagen fibers
in tendons are parallel and aligned in the direction of loading, while in the arterial
wall, a significant fiber dispersion around two preferred fiber directions is observed.
It also provides the tissue with strongly nonlinear mechanical responses as illustrated
by the phenomenon of stretch locking in rabbit skin [769] shown in Figure 11.25.
In plant tissues the same mechanical role is played by cellulose microfibrils based
on sugar chains, and with Young’s modulus around 130 GPa, reinforcing an isotropic
matrix composed of hemicellulose and lignin molecules with respective Young’s
modulus around 40 MPa and 2 GPa [148]. It is the fine control of fiber geometry
A. B.
Figure 11.25 A. Langer’s lines showing line of anisotropy in the skin. B. Uniaxial tension test
on rabbit skin shows different behaviors in direction along or across Langer’s lines (reproduced
from [144], based on data [769]). Solid lines represent the computational simulation based on an
eight-chain model [40]. The dotted lines are the so-called locking stretches after which no further
stretch is possible (picture courtesy of Ellen Kuhl).
11.11 Anisotropic Materials 321
A. B. C.
Figure 11.26 Examples of fiber reinforcement in biological tissues. A. Microtubule alignment in

the exocarp cells of Cardamine hirsuta. The microtubules wind helically around the cells (pic-
ture courtesy of Angela Hay [582]). B. Microfibrils in green algae. The X-ray analysis shows the
characteristic X structure indicating a mostly helical fiber [389]. C. Fiber wall structure of the spo-
rangiophore of the fungus Phycomyces showing spiraling fibers in the cytoplasm together with the
same characteristic X-ray signature [886].
and density that provides plants with their mechanical property and their ability to
interact with their environment [317, 387, 797].
Similarly, in the fungal kingdom, the cell wall is, typically, constructed of chitin
microfibrils embedded in an elastic matrix of amorphous material composed of chi-
tosan and chitin [178, 625]. Such a composite material naturally lends itself to mod-
eling in terms of an elastic matrix with fiber reinforcements [483] (Figure 11.26).
As the internal structure of a tissue determines its mechanical properties, it is rea-
sonable to include it in a constitutive model. The theory of nonlinear fiber-reinforced
elastic composites, developed by Rivlin, Spencer and others [9, 1183, 1184, 1248],
asserts that the strain-energy function is in general expressed through a set of defor-
mation invariants, whose number depends on the symmetry exhibited by the material.
In this theory, fibers are modeled by continuous fields, that is, they are represented
by local directions of anisotropy rather than actual physical fibers. A popular, a
priori, assumption in constitutive models is that the total stress generated by the
whole tissue is the sum of stresses generated by its constituents [604, 767, 768].
This assumption can be used to incorporate quantitative data characterizing a tis-
sue’s structure directly into the constitutive relation, e.g., fiber volume fraction or
orientation-dependent density can be included as multiplicative factors in the appro-
priate term. In addition, structural approaches allow for a formulation of phenom-
enological laws for fiber remodeling and the study of the dynamics of mechanically
induced fiber reorientation [303, 540, 865, 870].
11.11.1 One Fiber
A material reinforced by fibers that are perfectly aligned in one direction is an example
of a transversely isotropic material, i.e., a material with one distinguished direction.
The fiber direction is specified by a unit vector M in the reference configuration as
shown in Figure 11.27. For this fiber, we define the structure tensor
H = M ⊗ M, (11.283)
which contains all information about material anisotropy. We assume that the strain-
energy function of the fiber-reinforced material depends on H, and as a consequence,
the requirement of isotropy W (QT CQ) = W (C) is not identically satisfied for an
arbitrary proper orthogonal second-order tensor Q. Instead, it must be satisfied for
all proper orthogonal tensors Q such that QM = ±M. This condition is enforced
by considering that the strain-energy function is a function of both C and H, that is
W = W (C, H), and demanding that
W (C, H) = W (QT CQ, QT HQ), ∀Q ∈ S O(3). (11.284)
A strain-energy function for such materials depends in general on a deformation

tensor through five scalars, which consists of the three usual isotropic deformation
invariants I1 , I2 , and I3 defined by
1 2
I1 = tr B, I2 = [I − tr (B2 )], I3 = det B = J 2 . (11.285)
2 1
and two extra pseudo-invariants, which are related to the strains in the fibers when
deformed.
I4 = M · (CM) = C : H, I5 = M · (C2 M) = C2 : H, (11.286)
m=FM
M x(X)
X
Figure 11.27 A one-fiber material. In an anisotropic one-fiber material, the response of the material
depends on the deformation gradient F and the direction of a distinguished material line represented
by a unit vector M. In a deformation, this direction is mapped to a vector m = FM.
where we used the double contraction between second-order tensors
A : B = tr(AB) = Ai j B ji . (11.287)
The invariant I4 has a natural interpretation as the square of the fiber stretch in the
current configuration, that is, the norm of m = FM.
The general form for the Cauchy stress tensor (11.134) in terms of W remains
valid. Using the identities
∂ I4
F = 2FM ⊗ FM = 2m ⊗ m, (11.288)
∂F
∂ I5
F = 2F(M ⊗ FCM + CM ⊗ FM)
∂F
= 2(m ⊗ Bm + Bm ⊗ m), (11.289)
we obtain an explicit expression for the Cauchy stress tensor:
T = J −1 [ − pI + 2W1 B + 2W2 (I1 B − B2 ) + 2I3 W3

+ 2W4 m ⊗ m + 2W5 (m ⊗ Bm + Bm ⊗ m)] , (11.290)
where Wi = ∂W/∂ Ii , i = 1, . . . , 5.
11.11.2 Two Fibers
Many biological tissues are reinforced in two directions and, accordingly, can be
modeled by a material with two fibers. The mechanical advantage of a tissue with
two fibers appears when the tissue is in tension. With a single fiber, an extension
in any direction away from the fiber always produces shear. With two fibers, the
shear produced by each fiber can be balanced by the other fiber when they have the
same material response and their average direction matches the maximal principal
direction. Further, the angle between the two fibers can be tuned to change the overall
stiffness of the material.
Reinforcing by an additional family of fibers further reduces the symmetry of
the material, but it extends the set of invariants from five to nine. These four extra
scalars account for the strains in the second family and the coupling between the
two fiber families [1184]. We use the unit vectors M and M to define two preferred
directions in the reference configuration B0 . The energy for an unconstrained material
with two families of fibers is a function of eight invariants. These are the principal
invariants I1 , I2 , I3 together with two pseudo-invariants I4 , I5 that depend on M, two
pseudo-invariants I6 , I7 that depend on M and two coupling terms I8 , I9 defined by
I4 = M · (CM), I5 = M · (C2 M),

I6 = M · (CM ), I7 = M · (C2 M ), (11.291)
I8 = (M · M )M · (CM ), I9 = (M · M )2 .
Note that the last invariant is not a function of the deformation and will only appear
as a constant in the strain-energy function. It will therefore be ignored from the
analysis at the expense of a possible redefinition of the strain-energy function. The
strain-energy function W is now a function of all first eight invariants so that we
write
W = W (I1 , I2 , I3 , I4 , I5 , I6 , I7 , I8 ) . (11.292)
For an incompressible material we have I3 = 1, and the explicit expression for the
Cauchy stress tensor is
T = − pI + 2W1 B + 2W2 (I1 B − B2 )

+ 2W4 m ⊗ m + 2W6 m ⊗ m
+ 2W5 (m ⊗ Bm + Bm ⊗ m) + 2W7 (m ⊗ Bm + Bm ⊗ m )
+ W8 (m ⊗ m + m ⊗ m)(M · M ), (11.293)
where m = FM, m = FM , and Wi = ∂W/∂ Ii for i = 1, . . . , 8.
11.11.3 Example: The Fiber-Reinforced Cuboid
We consider the homogeneous deformation of a hyperelastic, incompressible cuboid

B0 = [0, L 1 ] × [0, L 2 ] × [0, L 3 ], subjected to constant normal external loads t1 ,
t2 , t3 measured as force per unit area in the deformed configuration. The cuboid is
made out of a homogeneous material reinforced by two families of fibers, which are
aligned symmetrically in the E1 − E2 plane, as shown in Figure 11.28. The directions
of the fiber families are M = (cos , sin , 0), M = (cos , − sin , 0).
We further restrict our attention to a simple form for the isotropic and anisotropic
responses, the so-called standard fiber-reinforcing model [272, 615, 877–879, 1052,
1248]
μ μγ

W = Wiso + Waniso = (I1 − 3) + (I4 − 1)2 + (I6 − 1)2 . (11.294)
2 2
The constitutive equation (11.293) for this strain-energy function is

T = − p1 + μF 1 + 2γ(I4 − 1)M ⊗ M + 2γ(I6 − 1)M ⊗ M FT . (11.295)
Due to the particular choice of fiber alignment and strength, the two equal and
opposite fibers with angle ± in the E1 –E2 plane are mapped into two equal and
t2
t3
t1 t1
t3
t2
Figure 11.28 A cuboid is subjected to external hydrostatic loading, maintaining constant normal
stress on each face of the cuboid. The material is reinforced by two families of fibers, which are
in-plane, aligned symmetrically, and make an angle with the E1 direction.
opposite fibers with angle ±θ as shown in Figure 11.29. We also have I4 = I6 =

λ21 cos2 + λ22 sin2 , where λi is the stretch in the X i direction, and
[M ⊗ M + M ⊗ M ] = 2diag(cos2 , sin2 , 0). (11.296)
Hence, we have

T = − p1 + μB 1 + 4γ(λ21 cos2 + λ22 sin2 − 1)diag(cos2 , sin2 , 0) .
For a homogeneous deformation, the Cauchy stress is homogeneous as well and

the Cauchy equation is identically satisfied. The particular form of the fiber stiffness
and alignment and the boundary conditions
T(X i = 0) = T(X i = L i ) = ti ei , (11.297)
imply that [T] = diag(t1 , t2 , t3 ). From (11.11.3), we see that F, in these coordinates,
must also be diagonal, that is, [F] = diag(λ1 , λ2 , λ3 ). Thus, (11.295) together with
the incompressibility condition become
M
Θ
χ m=FM
θ
Figure 11.29 Deformation of two equal and opposite fibers.


t1 = − p + μ λ21 + 4γ(λ21 cos2 + λ22 sin2 − 1)λ21 cos2 , (11.298)

t2 = − p + μ λ22 + 4γ(λ21 cos2 + λ22 sin2 − 1)λ22 sin2 , (11.299)
t3 = − p + μλ23 , (11.300)
λ1 λ2 λ3 = 1, (11.301)
where the four unknowns λi and p can be determined from the loads ti and the fiber
orientation angle . The last two relations (11.300) and (11.301) give λ3 = λ−1 2 λ1
−1
and p = μλ3 − t3 , so that, after substitution in the first two equations, the problem
2
can be reduced to finding (λ1 , λ2 ) as a function of t1 , t2 , t3 , that is,
λ21 − λ22
A= + (λ21 cos2 + λ22 sin2 − 1)(λ21 cos2 − λ22 sin2 ), (11.302)
4γ
−1 λ21 + λ22
B= + + (λ21 cos2 + λ22 sin2 − 1)(λ21 cos2 + λ22 sin2 ),
2γλ21 λ22 4γ
(11.303)
where
t1 − t 2 t1 + t2 − 2t3
A= , B= . (11.304)
4γμ 4γμ
Geometrically, Equations (11.302–11.303) define, for given A and B, two level sets,
and the solution lies at their intersection.
In many situations, fibers embedded in a matrix may not support compressive
loads as they would buckle under compression [604, 1189]. Therefore, we further
restrict the study of our model to the fiber-tensile region, defined by
λ f = λ21 cos2 + λ22 sin2 > 1, ∈ [0, π/2]. (11.305)
Fibers are unstrained on the boundary of the region (11.305), and are in compression
when λ21 cos2 + λ22 sin2 < 1.
We first compute the effect of fiber alignment and stiffness by computing an
effective Young’s modulus. Young’s modulus is only defined for isotropic materials,
but in a given direction, one can define a modulus as the ratio of the uniaxial tension
N by the stretch in the same direction in the limit of small deformations, that is,
the effective Young’s modulus in a direction is defined as the gradient of N in that
direction evaluated at the stress-free state. For example, for an uniaxial tension in
the e1 direction t1 = N , t2 = t3 = 0, we define

∂ N
E eff ≡ , (11.306)
∂λ1 λ1 =1
Eeff
μ(3+8γ)
Compressive fibers
μ(3+8γ)/(1+2γ)
3μ Θ
0 Θm π/2
No compressive fibers
Figure 11.30 Effective Young’s modulus for a fiber-reinforced sheet as a function of the fiber
angle.
and, after linearizing (11.302–11.303) around the unstressed configuration, we find
4 [3 + 5γ + 3γ cos(4)]
E eff = μ . (11.307)
4 + 3γ − 4γ cos(2) + γ cos(4)
As shown in Figure 11.30, the effective Young’s modulus has a maximum for
= 0 when the fibers are aligned with the axis:
E eff ( = 0) = (3 + 8γ)μ. (11.308)
The minimum effective Young’s modulus E eff ( = m ) = 3μ is attained for

√
m = tan−1 2 = m ≈ 54.74◦ , (11.309)
where m is the magic angle that appears in several interesting applications. It will
be discussed in detail in Section 11.11.5.
For m < < π/2, I4 < 1 and the fibers are in compression. If we allow for
compressive fibers, E eff increases to the locally maximal value of
3 + 8γ
E eff ( = π/2) = μ . (11.310)
1 + 2γ
In the absence of compressive fibers, the material behaves isotropically and the
effective modulus remains at 3μ.
Remarkably, we see that the apparent stiffness of a material in a given direction
can be tuned to any value between the matrix stiffness and the fiber stiffness by
choosing the appropriate fiber angle and fiber density. The modulus ratio between
the two extremal values is 1 + 3γ/8 and a typical ratio of γ is between 10 and 1000.
This simple effect is at work in fungi, plants, and animal tissues where, typically, the
isotropic elastic matrix remains mostly unchanged while fibers are constantly turned
over to maintain the appropriate level of homeostatic stress, the tissue stiffness or
for healing [303, 540, 635, 871].
As an example, in the aorta an elastin matrix reinforced with two equal and oppo-
site fibers helically wrapping around a cylindrical geometry, the effective Young’s
modulus is around 90 kPa at the ascending aorta and 10 kPa at the femoral bifurcation,
which correlates with the elastin content [137].
In the nonlinear regime, we observe an interesting behavior. As expected, the
stretch along the tensile stress always increases as shown in Figure 11.31. Similarly,
the stretch perpendicular to the direction of the applied load in the fiber plane always
decreases as we would expect for an isotropic material. However, in the direction
normal to the fibers, the material thickness determined by λ3 can increase, decrease,
or first decrease then increase. This non-monotonic behavior depends on the balance
between the nonlinear material response and the anisotropy.
We identify in the parameter space, (1/γ, ), these different behaviors by solving
Equations (11.302–11.303) under the condition λ3 = 1, that is, λ1 = λ2 . This extra
condition leads to two curves in the plane (1/γ, ) given by [864]
1 1
= sin2 (2), = sin2 (2) − 4 sin4 . (11.311)
γ γ
These curves are shown in Figure 11.32 together with the corresponding points of
Figure 11.31.
11.11.4 Example: The Fiber-Reinforced Cylinder
We consider a tube made out of an incompressible material with initial inner radius
A = 1, outer radius B > A, and height H deformed into a tube with radii a and b
and height h. We assume that the tube is allowed to inflate, extend, and twist while
remaining cylindrical at all time. This problem is the classic inflation–extension–
torsion problem for the cylinder. It is a particular case of Family 3 of semi-inverse
problems (11.220). In the usual cylindrical coordinates {r, θ, z} and {R, , Z } the
deformation is given by

R 2 − A2
r= a2 + , (11.312)
ζ
θ = + τζ Z, (11.313)
z = ζ Z, (11.314)
A. λ3
(b)
1.015
1.010
(c)
1.005
1.000
0.995
0.990
0.985 (a)
0.2 0.4 0.6 0.8 1.0
0.980
Ν/ μ
B. λ1, λ2
1.20 λ1 (c)
1.15
(a)
1.10
1.05 (b)
1.00
λ2
(b)
0.95
(a)
0.90
0.85 (c)
0.2 0.4 0.6 0.8 1.0
Ν/ μ
Figure 11.31 The three principal stretches as a function of the tension. For some values of γ and
, the thickness of the cuboid always decreases, always increases, or first decreases, then increases.
Parameter values: (a): = π/8, γ = 4/3; (b): = π/8, γ = 4; (c): = π/4, γ = 4.
where the axial stretch ζ and the twist τ are constant.

The position vectors are, respectively,
X = RE R + Z E Z , (11.315)
x = λRer + ζ Z ez . (11.316)
Using identity (11.60), the deformation gradient, F = Grad x, in cylindrical coordi-

nates is given by
1/ γ
1
(a)
I III
1
2
(b) (c)
II
Θ
0
π π 3π π
0
8 4 8 2
Figure 11.32 Parameter space for inversion: Depending on the relative fiber stiffness γ and the
fiber direction , in tension, the thickness of a either decreases (Zone I), increases (Zone II), or first
decreases, then increases (Zone III). Insert shown for different values of the tension. The stretch
profiles as a function of the tension for specific points (a, b, c) are given in Figure 11.31.
⎡ 1 ⎤
0 0
λζ
[F] = ⎣ 0 λ ζτr ⎦ , (11.317)
0 0 ζ
where we have used the incompressibility condition det F = 1 and

r 1 R 2 − A2
λ= = a2 + . (11.318)
R R ζ
Therefore, a single parameter fully describes the radial profile of the deformation.
Setting λa = a/A, it follows that

b 1 A2
λb = = 1 + 2 ζλa2 − 1 . (11.319)
B ζ B
The anisotropic response of the cylinder is modeled by two families of embedded

fibers M and M . For simplicity, we will refer to a family of distributed fibers simply
as a fiber. Both fibers wind helically around the axis and may induce a rotation of the
cylinder under extension depending on their strengths and angles. The components
of the direction vectors with respect to the basis (E R , E , E Z ) are
Figure 11.33 Geometry of B

the fibers. The angle A
denote the direction of the
first fiber with respect to the
cross section (counted
counter-clockwise) and the M’
angle is the angle of the
second fiber (counted
clockwise). M
H Ψ
Φ
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
MR 0 M R 0
⎣ M ⎦ = ⎣ cos ⎦ , ⎣ M ⎦ = ⎣ − cos ⎦ . (11.320)
MZ sin M Z sin
Here we have assumed that the fibers remain locally tangent to the cylinder. Following
Figure 11.33, the angles between the fibers and the circumferential direction are
denoted by and . Note that we have chosen the angle so that when the angles
are equal = , the two fibers make the same angle with the axis, and are said to be
opposite.Opposite fibers are a common occurrence and they are found, for instance,
in the popular toys shown in Figure 11.34.
Under a deformation F, the orientation of the fiber characterized by a vector M
with angle in the reference configuration is mapped, in the current configuration,
to the vector
⎡ ⎤ ⎡ ⎤
mr 0
m = ⎣ m θ ⎦ = FM = ⎣ λ cos + r ζτ sin ⎦ . (11.321)
mz ζ sin
Therefore, the new fiber angle is

ζ sin
φ = arctan . (11.322)
λ cos + r ζτ sin
A. m’
B.
m’
Figure 11.34 Two simple toys demonstrate the principle of opposite fibers and the change in fiber
orientation in compression and tension. Top: The boing-boing rocket stores elastic energy under
compression by changing the fiber angle. The rocket can be released by quickly removing one
finger. Bottom: The finger trap. In extension, the cylinder radius decreases and traps the fingers.
For these deformations, the only nonvanishing component of the Cauchy equation
div T = 0 is
dTrr 1
+ (Trr − Tθθ ) = 0. (11.323)
dr r
This equation can be integrated once over r to obtain
b
Trr − Tθθ
Trr (r ) = dr, a ≤ r ≤ b. (11.324)
r r
We consider a simple thought experiment in which the tube is capped at both ends
and subject to an axial extension ζ due to an internal pressure P and to a total axial
load N on the top cap. The tube is also subject to an external moment M leading to
a twist represented by τ .
First, we express the boundary condition in the radial direction. Taking the radial
component of the Cauchy stress tensor T to vanish at the outer boundary, we have
Trr (r = b) = 0. On the inner wall, the boundary condition associated with the
pressure is Trr (r = a) = −P. Therefore, the two conditions on the pressure can be
used to simplify (11.324) to
b
Tθθ − Trr
P= dr. (11.325)
a r
Second, the condition on the two end caps is a combination of an external axial
stress superimposed on the pointwise stress due to the internal pressure acting on the
end cap:
Tzz (z = 0) = Nz , Tzz (z = h) = Nz . (11.326)

Following the discussion of Section 11.8.3, we replace these pointwise conditions

by an integral condition relating the total application of forces and moments on the
caps of the cylinder:
b
2π Tzz r dr = N = F + χPπa 2 , (11.327)
a
where the total axial load N is decomposed into an external applied load F and the
load created by the internal pressure acting over the cap. Following the argument in
Section 11.8.3, this last condition can be replaced by
b
π (2Tzz − Trr − Tθθ )r dr = F + (χ − 1)Pπa 2 , (11.328)
a
and the last term vanishes for a capped cylinder, the case considered here.
Third, when τ
= 0, we have to take into account the possibility of applying a
moment on the ends. This loading can be expressed also as an integral condition
relating the total moment acting on the tube axis to the axial stress. That is,
b
Tθz r 2 dr = M. (11.329)
a
Therefore, the three boundary conditions are

b
Tθθ − Trr
C1 : dr = P, (11.330)
a r
b
C2 : π (2Tzz − Trr − Tθθ )r dr = F, (11.331)
a
b
C3 : 2π Tθz r 2 dr = M. (11.332)
a
The semi-inverse problem consists in finding the values of (λa , ζ, τ ) corresponding

to the three external loads (F, M, P) through the analysis of equilibria.
The standard fiber-reinforcing model with fiber extension. We further restrict our
attention to the standard fiber-reinforcing model, W = Wiso + Wfib , with
μ1
Wiso = (I1 − 3), (11.333)
2
μ4 μ6
Wfib (I4 , I6 ) = (I4 − 1)2 + (I6 − 1)2 , (11.334)
4 4
where the material parameters μi > 0 have the dimension of a pressure.
From the strain-energy function, we compute the Cauchy stress tensor
∂W
T=F − p1, (11.335)
∂F
where p is the Lagrangian multiplier associated with incompressibility. In our case,
these equations simplify to
T = 2W1 B + 2W4 m ⊗ m + 2W6 m ⊗ m − p1, (11.336)
with Wi = ∂ Ii W . The nonvanishing components of the Cauchy stress tensor are

given by
Trr = − p + 2W1 ζ −2 λ−2 ,

Tθθ = − p + 2 λ2 + r 2 ζ 2 τ 2 W1
+2(λ cos + r ζτ sin )2 W4 − 2(λ cos − r ζτ sin )2 W6 ,
Tzz = − p + 2ζ 2 W1 + 2ζ 2 sin 2 W4 + 2ζ 2 sin 2 W6 ,
Tzθ = Tθz = 2ζ [r ζτ W1 + sin (λ cos + r ζτ sin )W4
− sin (λ cos − r ζτ sin )W6 ] .
Since the constitutive relationships are written in terms of {λ, ζ, τ }, we rewrite the
three boundary conditions in terms of integrals over λ using the identity
dr (1 − ζλa2 )1/2
=A , (11.337)
dλ (1 − ζλ2 )3/2
which yields the equivalent boundary conditions

λb
Trr − Tθθ
C1 : dλ = P, (11.338)
λa λ(λ 2 ζ − 1)
λb
1 − ζλa2
C2 : πA 2
λ(2Tzz − Trr − Tθθ ) dλ = F, (11.339)
λa (1 − ζλ )
2 2
λb
(1 − ζλa2 )3/2 2
C3 : 2π A3 λ Tzθ dλ = M. (11.340)
λa (1 − ζλ )
2 5/2
While explicit expression for the three integrals for (M, N , P) for the particular
choice (11.333–11.334) can be obtained, they are far too cumbersome to be useful.
Membrane limit. We can take advantage of the assumption that the tube is thin and
expand the three integrals (M, N , P) in the thickness of the tube. Without loss of
generality, we measure all lengths with respect to the inner reference radius, that is,
we set A = 1. Then, we introduce by B = 1 + and expand
M = M (1) + M (2) 2 + . . . , (11.341)

(1) (2) 2
F = F + F + ..., (11.342)
P = P (1) + P (2) 2 + . . . . (11.343)
Explicitly, to first order these expressions read
M (1) = 2πλ [ζλμ1 τ + μ4 J4 sin()(ζλτ sin() + λ cos()) (11.344)

+ μ6 J6 sin()(ζλτ sin() − λ cos())] ,

π μ1
F (1) =− 1 + ζ 4 λ2 λ2 τ 2 − 2 + ζ 2 λ4 (11.345)
ζ ζ λ 2 2

+ μ4 J4 ζ sin() ζ λ2 τ 2 − 2 sin() + 2λ2 τ cos() + λ2 cos2 ()

+ μ6 J6 ζ 2 λ2 τ 2 − 2 sin2 () − 2ζλ2 τ sin() cos() + λ2 cos2 () ,

1 μ1 4 4 2
P (1) = ζ λ τ + ζ 2 λ4 − 1 (11.346)
ζ ζ 2 λ4

+ μ4 J4 (ζτ sin() + cos())2 + μ6 J6 (cos() − ζτ sin())2 ,
where
J4 = (I4 − 1)

= ζ sin() ζ λ2 τ 2 + 1 sin() + 2λ2 τ cos() + λ2 cos2 () − 1,
J6 = (I6 − 1)

= ζ 2 λ2 τ 2 + 1 sin2 () − 2ζλ2 τ sin() cos() + λ2 cos2 () − 1.
11.11.4.1 A Cylinder with Equal Fibers but Different Orientations
As a first example of the possible behaviors in this system, we consider the problem
of rotation under pressure in a tube with two fibers with equal strength, μ6 = μ4 ,
but of varying angle. If we fix one fiber, we can vary the angle of the other fiber
and ask whether a change in pressure will lead to a reversal in rotation from a left-
handed rotation to a right-handed rotation. The condition for inversion is obtained by
considering the change of twist τ with respect to pressure τ = τ (P) and identifying
the points at which this relationship is stationary. The condition τ (P) = 0 leads to
μ4 sin( − ) [6μ1 (2 cos( + ) + cos(3 + ) + cos( + 3))

(11.347)
+ 8μ4 sin2 ( + )(cos() cos() − 2 sin() sin()) = 0.
We show in Figure 11.35 the inversion curves in the parameter space (, ), for
μ1 μ4 and μ4 μ1 . The two limits are easily obtained analytically from (11.347)
by taking μ1 = 0 or μ4 = 0. We see the role of the magic angle introduced in
Section 11.11.3 and discussed below, as being the distinguished value at which an
Φ2
π/2
τ>0 (RH)
μ1>>μ4 (stiff matrix)
τ<0 (LH)
Φ*m
τ<0 (LH) μ1<<μ4 (stiff fibers)
π/6
τ>0 (RH)
Φ1
0
0 π/6 Φ*m π/2
Figure 11.35 Twist induced by inflation for a cylindrical membrane in small deformations as a
function of the orientation of the two fibers.
inversion of rotation appears for systems with stiff fibers. In this case, if we start with
one fiber oriented at the magic angle and vary the other one there will just be a single
inversion of rotation (when the two are equal). For other values, there will be two
inversions, one when the two angles are equal and the second at different values of
the angle, showing the interesting property of no net rotation (in small deformations)
despite the tube being clearly anisotropic.
11.11.5 Application: The Hydrostatic Skeleton
A simple cylindrical structure, known as a cylindrical hydrostatic skeleton with

reinforced fibers is found in many ectothermic and soft-bodied organisms such as
nemertean and turbellarian worms. Of particular note is the work of Harris and
Crofton [543] and Clark and Cowey [214]. Building on the concept of a hydrostatic
skeleton [690, 691, 1104], they model the worm body as a cylindrical membrane
2πP
2πP
D
Φ Φ
2πR
Figure 11.36 A geometric model of hydrostatic skeleton obtained by assuming that the skeleton
can have a cylindrical shape, with inextensible fibers winding helically around the cylinder.
reinforced with a lattice of crossed, inextensible, fibers winding around the membrane
with helical geometry, and they characterize the possible extension and retraction
of the lattice network as being analogous to that of lazy tongs or a trellis. Their
mathematical model, which is purely geometric, proved to be remarkably successful
in explaining their detailed experimental studies of the locomotion and flattening of
worms.
Consider a cylinder of height 2π P and radius R with a helical inextensible fiber
of fixed length D as shown in Figure 11.36. The volume enclosed is simply
D3
V = 2π 2 R 2 P = sin2 cos . (11.348)
4π
The maximal volume is attained when ∂ V = 0, that is, for
π √
∗m = − tan−1 2 ≈ 35.26◦ , (11.349)
2
as shown in Figure 11.37. For any given volume 0 < V1 < Vmax , there exist two
angles 1 and 2 in (0, π/2) that give rise to a cylinder of volume V1 . Therefore,
in principle, a worm could change its extension from H1 = D cos 1 to H2 =
D cos 2 > H1 at constant volume. This change in extension can be obtained by
contracting fiber muscles, a transformation that can take place by flattening the cross
section to conserve the volume. In the initial and final configurations, the fibers
have length D and are at rest. Therefore, in this range of parameters, the hydrostatic
skeleton can adopt two shapes of different lengths and maintain these shapes without
muscular contraction.
V/D
0.4 V
0.3
0.2
V
0.1
Φ1 Φ2 Φ
Φ*m 3
8 4 8 2
Figure 11.37 The magic angle gives the maximal enclosed volume of a cylinder reinforced by
inextensible fibers of fixed length D. For V < Vmax there exist two angles at fixed volume.
We refer to the reciprocal of this angle, m = π/2 − ∗m , as the magic angle
since it seems to appear, as if by magic, in many different settings in mechanics and
in solid-state nuclear magnetic resonance [484, 569, 600, 1117]. For instance, this
angle is believed to be key in understanding the elongation of notochords [6, 716,
717]. Further, an inanimate analog of the model of Clark and Cowey can be found
in the McKibben actuator which consists of a flexible tube surrounded by a sheath
of braided families of inextensible fibers helically wound in opposing directions.
This design is the basis for so-called pneumatic artificial muscles used in robotics,
prosthetics, and orthotics [256, 796]. These actuators are typically pressure controlled
and their precise functionality is determined by the weave of the fibers. As with the
Clark and Cowey model, the fiber winding angle of 35.26◦ plays a special role in the
actuator design.
We will now show that the magic angle also appears naturally as a special limiting
case of a nonlinear elastic model. We start our analysis with the simple case of
a cylinder with two families of fibers of equal strength (μ6 = μ4 ) and opposite
orientation ( = ). This corresponds to the classic case of the McKibben actuators,
arteries, and other hydrostats. Under extension or inflation, the couples created by
the two fibers cancel out and there is no net couple associated with the deformation
and thus no rotation or twist (τ = 0).
We are particularly interested in identifying inversion due to internal change of
pressure. We look at the condition under which the radial strain does not vary under
a change a pressure. A local analysis for small strains [484] leads to the condition
3μ1 + μ4 sin2 ()(1 − 3 cos(2)) = 0. (11.350)

λ>1 (radial extension)

ζ<1 (axial reduction)
Φm
*
λ<1 (radial reduction) λ>1 (radial extension)

ζ>1 (axial extension) ζ>1 (axial extension)
1/18
Figure 11.38 Parameter space for the radial and axial expansion of a thin tube under pressure.
Depending on the relative stiffness of the fiber versus the matrix and the fiber angle, a capped tube
under pressure can extend radially and axially (bottom left), extend radially but shrink axially (top),
or extend axially but shrink radially (bottom right).
The condition for an inversion in the axial strain is = ∗m

By denoting μ = μ1 /μ4 as the ratio of matrix modulus to the fiber modulus, we
obtain a complete description of the possible inversions under a change in pressure
in the parameter space (μ, ) as shown in Figure 11.38.
Depending on the design criterion, one can consider different tube constructions
by varying the fiber angle. For fiber angles larger than ∗m the tube contracts under
increased pressure and this behavior provides a model for pneumatic muscles. For
tubes with fiber angle close to ∗m , the deformation of the tube in the axial direction
is minimal. For fiber angles less than ∗m , the tube extends maximally. Note that this
analysis is only valid for small enough P. For larger pressure, we expect the tube to
increase eventually in length and radius.
11.11.6 Fiber Dispersion
In addition to having strong anisotropy dictated by several fiber directions, many

tissues also exhibit a fair amount of dispersion around these principal directions. For
instance, collagen fibers in the arterial wall show a significant fiber dispersion around
two preferred fiber directions as demonstrated experimentally [164] by polarized light
microscopy. It is therefore necessary to consider continuous distributions of fiber
direction as first proposed by Lanir [767] who computed the total stress generated by
fibers as an integral over a range of fiber angles. This model was further simplified
[388, 426] by the introduction of the notion of generalized structure tensors. Models
based on this simplified approach were applied to many soft tissues, including the
human aortic valve [388], the arterial wall [604], the cornea [518, 1010], the articulate
cartilage [358], and the bat wing membrane [1296].
First, we consider the case of a single principal direction around which mechani-
cally equivalent fibers are dispersed. We further assume that the strain-energy func-
tion can be split into an isotropic component and an anisotropic contribution due to
the fibers [864] so that
W = Wiso + Wfib , (11.351)
with !
Wfib = ρ(M)wfib (If (M)) dω, (11.352)
U2
where U2 = {M ∈ R3 , s.t. |M| = 1} is the unit sphere, dω is the surface element

area in the reference direction, IM = M⊗M : C is the squared stretch in the direction
M, and ρ(M) is an orientation density function defined as a probability distribution
with the usual normalization condition
!
ρ dω = 1. (11.353)
U2
Note, however, that other authors have used different conventions [426, 1011]. With-
out loss of generality, it is assumed that ρ(M) = ρ(−M) holds, since M and −M
define the same fiber.
In the case of m families of distributed fibers, the fiber contribution to the strain-
energy function is
m !
(i)
Wfib = ρ(i) (M)wfib (If (M)) dω, (11.354)
i=1
U2
(i)
where the ith fiber family is characterized by a strain-energy function wfib (M) and
an orientation density function ρ(i) = ρ(i) (M). The Cauchy stress is then given by
∂Wiso T ∂Wfib T
T = − p1 + Tiso + Tfib = − p1 + 2F F + 2F F . (11.355)
∂C ∂C
For the strain-energy function (11.354), we obtain
⎛ ⎞
m !
(i)
∂wfib (If )
Tfib = 2F ⎝ (i)
ρ (M) M ⊗ M dω ⎠ FT . (11.356)
i=1
∂ If
U2
This angular integration method [228, 1011, 1170] is a rather cumbersome approach
from a computational point of view as a full angular integration has to be performed
at each material point to obtain the Cauchy stress tensor.
The main idea to simplify this computation is to perform an angular averaging.

We define the averaging operator
!
•ρ = ρ(M) • dω, (11.357)
U2
where • denotes any tensorial quantity. Its normalized counterpart is

&
ρ(M) • dω
•ρ U2
[•]ρ = = & . (11.358)
1ρ ρ(M) dω
U2
Next, we compute the Taylor expansion of wfib (If ) about If = [If ]ρ

wfib (If ) = wfib ([If ]ρ ) + wfib ([If ]ρ )(If − [If ]ρ ) + O (If − [If ]ρ )2 . (11.359)
By omitting higher order terms and by applying [•]ρ to both sides we obtain

[wfib (If )]ρ ≈ wfib ([If ]ρ )[1]ρ + wfib ([If ]ρ )[(If − [If ]ρ )]ρ . (11.360)
Using [1]ρ = 1 and [If − [If ]ρ ]ρ = 0, we obtain

If ρ
wfib (If )ρ ≈ 1ρ wfib . (11.361)
1ρ
This approximation forms the basis of the generalized structure tensor approach. The
generalized structure tensor is defined as
H = M ⊗ Mρ . (11.362)
The right-hand side of (11.361) can then be written as

If ρ H:C
1ρ wfib ( ) = tr(H) wfib ,
1ρ tr(H)
&
whereas the left-hand side of (11.361) equals U2 ρwfib (If ) dω. Therefore, (11.361)
approximates the total angularly integrated fiber strain-energy function by a strain-
energy function depending directly on H.
(i)
In a material reinforced by fiber families with strain-energy functions wfib and
(i) (i)
orientation density functions ρ , we define for each fiber family, H = M⊗Mρ(i) .
These generalized structure tensors are objective and symmetric tensors. They do not
depend on the deformation and serve as an alternative representation of structural
data. The total contribution of the fibers to the strain-energy function is then

(i) (i) H(i) : C
Wfib = tr(H )wfib , (11.363)
i
tr(H(i) )
and the Cauchy stress is

' (
(i) H(i) : C
(i)
Tf = 2F H wfib FT . (11.364)
i
tr(H(i) )
The approach that uses (11.362) together with (11.363–11.364) is the generalized
structure tensor method. Note that if the orientation density functions ρ(i) sat-
isfy (11.353), the term tr H(i) = 1 can be omitted.
The angular integration and the generalized structure tensor methods are equiv-
alent in the case of spherical deformations and in the absence of dispersion (i.e.,
perfect fiber alignment) [228]. From a computational point of view, the generalized
structure tensor method is vastly superior since the angular integration has to be
performed only once, to compute H, which is independent of the deformations, as
long as ρ does not change. Further, the derivative of the fiber strain-energy function
wfib needs to be evaluated only once for a given value of the deformation tensor C.
Transversely isotropic fiber density functions. Of particular interest are trans-
versely isotropic, or axisymmetric, orientation density function where for each fiber,
we have
ρ(M) = ρ̃(θ) = ρ̃(−θ), θ = arccos(M · M̄), (11.365)
which defines a fiber dispersion around the mean fiber directions M̄. An example of
this fiber distribution is shown in Figure 11.39A, where there are two main mean
fiber distributions around which there is a transversely isotropic orientation density.
The generalized structure tensor for such an orientation density function is
H = κ1 + (1 − 3κ)M̄ ⊗ M̄, (11.366)
where π
1
κ=π ρ̃(θ) sin3 θ dθ ∈ [0, ], (11.367)
0 2
is the dispersion parameter [426]. The complete alignment in the direction M̄, the
isotropic distribution, and the complete alignment in the plane perpendicular to M̄
correspond to the values κ = 0, 1/3, 1/2, respectively [604].
Apart from the case κ = 0 and 1/2 different distributions may lead to the same
parameter κ. For example, the value κ = 1/3 corresponds to the case when the fiber
distribution forms a cone, whose generatrix and axis make an angle
√
∗ = arctan 2, (11.368)
A. B.
(1) (1)
M (2) M
M (2)
M
Figure 11.39 A. A material reinforced by two families of mechanically equivalent fibers is in

uniaxial tension, subject to the axial load N . Both fiber families are transversely isotropic and
described by the von Mises distribution. The mean fiber directions M(i) are aligned symmetrically
with respect to the principal directions. If φ
= 0, π/2, the material is orthotropic. B. Orthotropically
symmetric fiber-reinforcement in a material with four families of fibers. Each family has a reference
direction M̄(i) and orientation density function ρ(i) .
i.e., ρ̃(θ) = δ(∗ − θ), where we recognize again the ubiquitous magic angle m
discussed in the previous sections.
The simple form of the generalized structure tensor in (11.366) allows us to express
the mean square of fiber stretch through the invariants of a transversely reinforced
material, I1 = trC, I4 = M̄ ⊗ M̄ : C,
I f ρ = H : C = (1 − 3κ)I4 + κI1 . (11.369)
Therefore, the standard reinforced model for transverse isotropic fiber distribution
can be obtained by the following replacement I4 → κI1 + (1 − 3κ)I4 , and I6 →
κI1 + (1 − 3κ)I6 , that is,
μ1
W = (I1 − 3)
2
μ4
2
+ (κI1 + (1 − 3κ)I4 )2 − 1
2
μ6
2
+ (κI1 + (1 − 3κ)I6 )2 − 1 , (11.370)
2
and we see that the fibers now also contribute to the isotropic part of the strain-energy
function.
Fiber exclusion. The generalized structure tensor provides a simple way to model
the exclusion of compressed fiber by annihilating the anisotropic part of H if the
stretch in the main direction is compressive (M̄ ⊗ M̄ : C − 1 < 0), as suggested by
Gasser et al. [426, 604]. That is, for each fiber we define
)
κ1 + (1 − 3κ)M̄ ⊗ M̄, M̄ ⊗ M̄ : C > 1,
H= (11.371)
κ1, M̄ ⊗ M̄ : C ≤ 1,
and Equations (11.363) and (11.364) should be used as before.

Holzapfel and Ogden [604] observed that values of κ from (1/3, 1/2) yield neg-
ative pressure in an inflated thin-walled tube and concluded that these values are
therefore nonphysical.
This problem can be circumvented by defining the generalized structure tensor
for transversely isotropic fiber distribution as follows [864]:
⎧
⎪
⎨κ1 + (1 − 3κ)M̄ ⊗ M̄, (1 − 3κ) M̄ ⊗ M̄ : C − 1 > 0,
H = κ1, κ ≤ 13 , M̄ ⊗ M̄ : C ≤ 1, (11.372)
⎪
⎩
(1 − 2κ)1, κ > 13 , M̄ ⊗ M̄ : C > 1.
This modified generalized structure tensor has now all the desired properties. In
particular, it provides a mechanism of exclusion of compressed fibers, so that an
isochoric deformation always results in an average extension of fibers (H(i) : C ≥ 1)
for any values of κ. With this definition, it is not necessary to perform an integration
for each deformation. The integration can be performed only once to compute the
dispersion parameter κ.
Chapter 12
The Kinematics of Growth
In the previous chapter, we have defined general kinematics for the deformations of
three-dimensional bodies in Euclidean spaces. We can generalize and adapt these
methods to describe growth processes changing the shape and volume of a body and
quantify the stresses generated through these processes.
We saw in Section 1.1.3 that a kinematic description of volumetric growth can
be achieved by specifying at each point of a body how an infinitesimal sphere is
transformed into an infinitesimal ellipsoid as shown in Figure 12.1.
The problem of growth is that neighboring points in the body may grow in a way
that is incompatible for the body to retain its integrity. Due to growth, parts of the
body may overlap or be torn apart. In an elastic body, as points are pushed together,
or pulled apart, they develop elastic strains to maintain the body’s continuity and
integrity.
One of the central ideas of morphoelasticity is that the body keeps its integrity
in growth by developing elastic strains. Elastic growth can be thought of as a two-
step process. The first step is a purely geometric deformation, prescribing how each
point of the body is affected by growth. The second step is an elastic accommodation
necessary to maintain both the integrity of the body as a continuum and its mechanical
equilibrium in the current configuration.
The focus of this chapter is the geometric description of growth deformations.
In particular, we define the notion of incompatibility and different tensorial objects
346 12 The Kinematics of Growth
G(p1)
p1
G(p )
2
p
2
B0
Figure 12.1 A local element inside a body can expand or shrink anisotropically. This process is
characterized locally by a growth tensor G, transforming a sphere into an ellipsoid. Two neighboring
points may grow such that they would overlap after growth creating incompatibility.
needed to quantify incompatibility in a growing body. We first show how growth

tensors naturally arise from the description of residual stresses.
12.1 A Thought Experiment
In order to understand the nature of stresses generated through growth, we first

perform a thought experiment on an elastic body in a three-dimensional space. This
body could be part of a soft organ such as an artery or a heart, it could be the branch of
a tree, a solid tumor, or any other body that could be reasonably modeled as an elastic
continuum. For simplicity, we assume that it is not currently growing or moving: The
body is in state of static equilibrium and has been brought to our thought laboratory
ready for a thought experiment that will allow us to describe its stress state and its
material response. We have two problems. First, we need to quantify these internal
stresses. Second, recalling that the basic assumptions of nonlinear elasticity is to
define a constitutive relationship by considering the strains away from a stress-free
configuration, we need to obtain a state where all stresses have been removed. The
two problems are closely related.
First, we remove all possible external applied loads and body forces by releasing
attachments, removing pressure fields, and turning off gravitation. Gravitation can
be removed by either sending our thought experiment in space, by placing it in a
neutrally buoyant enclosure, or by realizing that the typical stresses generated by
the gravitational field can be neglected in comparison to other stresses acting on the
body.
The second step consists in evaluating the internal stresses that may have arisen
as a consequence of growth. Indeed, in the absence of applied loads and body forces,
our body may still experience residual stresses. Without knowledge of the elastic
properties of our elastic material, the existence of residual stress can be demonstrated
12.1 A Thought Experiment 347
by a destructive experiment that consists in cutting the body and realizing that, in
the process, pieces of the material change shape as elastic stresses are relieved. For
instance, residual stresses are observed by slicing arteries or rhubarbs as discussed
in Section 2.2. The hope here is that by successive slicing, we obtain a configuration
where all residual stresses have been removed. This stress-free configuration could
then be used to obtain a constitutive law and by carefully tracking all changes in
shape, we could then compute the residual stresses in the material.
To demonstrate the existence of residual stress, we consider the following simple
physical experiment. Take an unstressed elastic tube and turn it inside out to trans-
form it into a different tube. Even in the absence of external loads, this new tube
is stressed. The residual stresses can be easily relieved by a single longitudinal cut
that would allow the tube, after relaxation, to regain its original unstressed tubular
shape. However, without a priori knowledge of how the residual stress was built in
the structure, it is not clear how we should cut the tube as to relieve all stresses.
For instance, an alternative way of cutting the tube would be to slice it into many
rings. At each cut the internal diameter of the ring would change indicating again the
presence of resiudal stress. But these rings are still stressed. We could then slice each
ring along a circle as to create two new rings. Each new ring could be divided again
into two new rings and so on. After each cut, we would observe some variation in the
diameters of the rings indicating the release of residual stress. The cutting process
would need to be repeated indefinitely since each new cut reveals the existence of
residual stress. We conclude from this simple example that even, in the case where
a single cut may be enough, without prior knowledge of the residual stress field,
a destructive experiment, if not performed judiciously, may require infinitely many
cuts to relieve all stresses.
There may be also cases where a finite number of well-placed cuts may not be
sufficient to relieve all stresses, even with a perfect knowledge of the residual stress
field. In these cases, we would need to repeat the cutting process indefinitely, until we
reach a state where all material pieces are completely stress-free. This state which is
determined by a collection of pieces or points is particularly important as the elastic
properties of our material are defined with respect to a stress-free configuration.
Performing suitable mechanical experiments on each piece would yield a constitutive
law relating stresses to strains. These pieces or material points can then be put back
into the original body by first elastically straining each piece so that they fit together
and then by reapplying loads and body forces.
Rather than starting with the grown body, a different approach for the problem of
growth is to go back in time and assume that, before growth takes place, the body is
in a stress-free configuration. In this initial configuration, all constitutive laws can be
defined. Then the grown configuration is obtained from the collection of stress-free
pieces by changing their size through growth, shrinking, and rigid-body motion. As
we will see, the action of growth defined a new state that can be used to compute the
residual stresses.
Our first goal in this chapter is to show that this process of repeated slicing will
converge and that we can reach a state where each individual, possibly infinitesimal,
piece is stress-free. Our second goal is to make sense of the configuration obtained
by this repeated process. The third goal is to understand the geometry associated
with these different configurations and understand the differences between the case
of global incompatibility where a finite number of cuts may completely relieve all
residual stresses, and the case of local incompatibility, where an infinite number of
cuts are needed. Finally, we will complete this picture by defining the configuration
obtained by growth and shrinking of each individual piece to reconstruct a stress-free
body. That is, we will demonstrate that the process responsible for incompatibility
can be described by a growth tensor and that there exists an initial stress-free reference
configuration from which the current state of the body can be described through a
combination of growth and elastic deformations. The view in this chapter is purely
static. That is, the body is assumed to be in an elastic static equilibrium and any
growth dynamics has been frozen.
12.2 Relieving Stresses
We now formalize the process of unloading a stressed body as described in the previ-
ous section. We consider a body B in a configuration B in the three-dimensional
Euclidean space E3 . The body is composed of hyperelastic nonpolar material
presently in a state of elastic equilibrium, that is, in balance with body forces
b and applied loads. The stresses are characterized by the Cauchy stress tensor
T = T(x), x ∈ B, that obeys the equations of motion in the current configuration
div T + ρb = 0, TT = T, (12.1)
where, as before, x denotes a position in the body of a material point p ∈ B, and

ρ = ρ(x) is the current mass density. The boundary conditions are specified by a
combination of displacement xb and surface tractions tb as explained in Section 11.5.
We wish to analyze the nature of the stresses in this body and how they were
generated. The first step is to remove all applied loads, displacements and body
forces so that b = 0 in B and tn = 0 everywhere on ∂B. By removing loads and
tractions, the body is elastically deformed into a residually stressed configuration
Br , where the subscript r refers to the residually stressed configuration with Cauchy
stress Tr that satisfies
divr Tr = 0, TTr = Tr , (12.2)
where divr is the divergence with respect to xr ∈ Br . The stress further satisfies the
condition that no surface traction is applied on the body
Tr (xr )n(xr ) = 0, ∀ xr ∈ ∂Br , (12.3)
where n(xr ) is the normal vector at a material point p located at xr on the boundary
of Br (Figure 12.2).
The loading process is the reverse deformation, that is a deformation from Br to
B denoted by χl so that xl = χl (xr ). The deformation gradient tensor from Br to B is
12.2 Relieving Stresses 349
Figure 12.2 The removal of the traction forces and body forces is a deformation from Bl to Br .
The inverse deformation is the unloading process and is characterized by the invertible map χl with
gradient tensor Fl (xr ).
Fl (xr ) = gradr χl (xr ). (12.4)
If this unloaded configuration is stress-free, then we have a reference configuration

from which all geometric and physical quantities can be evaluated as described in the
previous chapter. Instead, we will assume that the new unloaded configuration still
exhibits stress. There are two fundamental questions related to the residually stressed
configuration. First, assuming that the residual stress Tr and the elastic response of
the unstressed elastic material are known, how can we compute the current state of the
body for given loads and body forces? The main problem comes in the formulation
of a constitutive law that would enable us to compute the Cauchy stress from (12.1).
This constitutive relationship would naturally depend on the residual stress Tr and
on the deformation. That is, it must take the form
Tr = H(Fl , Tr ). (12.5)
For particular configurations [522] or in the limit of small deformations, this con-
stitutive relationship for a stressed material is given by a linear superposition of the
stress related to the loads and body forces, and the residual stress. However, for large
deformations, the constitutive relationship that describes the relationship between
the Cauchy stress T and the deformation gradient Fl cannot be easily obtained [669].
Here, the theory of invariants can be used to obtain a general form of this constitutive
relationship in the same spirit that leads to the formulation, in Section 11.7, of strain-
energy functions in terms of the invariants of the right Cauchy–Green tensor [880,
881, 962]. However, this formulation includes many new invariants and unknown
functions that are difficult to specify without prior knowledge of the stresses. In par-
ticular, it requires knowledge of the residual stress, which is precisely the object that
we are trying to relate to a microscopic growth process.
An alternative method to obtain a constitutive law is to unload the body along the
principal direction of stresses to obtain a stress-free configuration from which the
total elastic strains can be computed [669]. Therefore, it relies on the existence of a
stress-free configuration. But, the existence of such a configuration has not yet been
established.
The second question is then, how do we obtain a stress-free state from the resid-
ually stressed configuration? If such a configuration could be obtained, then the
traditional methods of elasticity would apply.
The existence of such a state through the repeated slicing of the body relies on
Signorini’s mean-stress theorem [456, p. 100]: In a simply connected body B of
volume V , in mechanical equilibrium, subject to an applied load tb and body force
b, we define the mean-stress T̄ to be

T̄ = V −1 T dv. (12.6)
B
If the stress T is differentiable, then

1
T̄ = V −1 ρ (b ⊗ x + x ⊗ b) dv + (tb ⊗ x + x ⊗ tb ) da . (12.7)
2 B ∂B
Therefore, in the absence of body force and traction, the mean stress on a body
vanishes identically. In particular, if we follow a point x ∈ Br , through repeated cuts,
the volume of the sub-bodies decreases. Since the Cauchy stress is continuous and
differentiable at x, it converges, in the limit of infinitely small bodies, toward the null
average. In this limit, the stress vanishes identically.
Therefore by a repeated series of cuts sketched in Figure 12.3, the initial body is
divided into a collection, possibly infinite, of sub-bodies on which the Cauchy stress
vanishes identically.
x x x
x x
Ar
Figure 12.3 The residually stressed body Br is repeatedly cut, relieving stress. As the number of
cuts n → ∞, the Cauchy stress at a point x converges towards the average stress.
12.2 Relieving Stresses 351
Figure 12.4 In the Residually stressed configuration

unstressing procedure, the
residually stressed body is
cut in sub-bodies which are
stress-free. At each point, the
deformation tensor Ar
characterizes the strains
relieved by cutting. The
resulting state is a stress-free
dislocated body.
Ar
Virtual stress-free state
After every cut, stress is relieved and the evolution of a basis of tangent vectors in
the tangent space can be followed. The change in these bases are defined by the strain
relieved at each step and it can be used to define a deformation tensor. More precisely,
the change in strain is defined locally by a two-point tensor that maps tangent vectors
at p to tangent vectors in the new unstressed configuration. We denote by A−1 r the
local elastic deformation tensor mapping the tangent space T p Br at a material point
p ∈ B to the tangent space of the same material point in a stress-free configuration
as depicted in Figure 12.4. The union of all images of the tangent spaces of Br under
A−1
r forms a tangent bundle TV, that is,
Ar : TV → TBr . (12.8)
At each point, this mapping is between tangent spaces. If the configuration Br

is not stress-free, then Ar cannot be obtained as the gradient of a deformation over
the entire body, otherwise, by contradiction, all stresses would have been removed
during unloading. Therefore, the actual position in space of a point whose tangent
space is T p V cannot be defined by the mapping Ar .
Two possible equivalent definitions can be adopted here. The first possibility is to
consider that the tangent bundle TV is the tangent bundle to a global differentiable
manifold V. The nonexistence of a global invertible differentiable map from Br to
V implies that V is not an Euclidean manifold and its geometry can be described
through the standard quantifiers of differential geometry. This point of view will be
further explored in Section 12.5.
The second possibility is to identify the placement of points p in the two states.
Then, the virtual state V in the Euclidean space can be seen as a dislocated body, that
is, an incompatible state defined by a collection of stress-free, possibly infinitesimal,
sub-bodies with possible overlaps or gaps. For the remainder of this section, we
adopt this point of view and consider the virtual state V as a dislocated body. The
virtual state is not realized as a configuration in Euclidean space, but a state on which
kinematic measures such as strains can be defined. More importantly, V is a stress-
free state of a hyperelastic material suitable for the computation of stresses in both
the residual and final configurations.
We recall that a deformation is a global differentiable one-to-one map between
the tangent bundle of two body manifolds. Whereas the virtual state is not defined as
a configuration, its tangent bundle TV is well defined. We use the letter A to denote
elastic deformations that may be incompatible and we use F to denote a compatible
deformation, that is, deformations that can be written as the gradient of a mapping.
It is important to note that the deformation tensor A is not necessarily unique or
defined independently of the cutting procedure. Assume that two different cutting
procedures lead to two different configurations V1 and V2 defined by two different
deformation tensors A1 , A2 . We restrict our attention to strictly convex strain-energy
functions W = W (C) with respect to the right Cauchy–Green tensor C. A scalar
function W is strictly convex with respect to C if
2(W (C2 ) − W (C1 )) > (WC + WCT )(C2 − C1 ), ∀ C2 = C1 . (12.9)
This condition implies that the material can only have a single stress-free state.
Therefore, it can be shown [524] that for a hyperelastic material with a strictly
convex strain-energy function W = W (C), the two deformation tensors A1 and A2
are related by
A1 = A2 Q, Q ∈ S O(3). (12.10)
Therefore, in the unstressing process, the two right Cauchy–Green tensors
C1 = AT1 A1 , C2= AT2 A2 , (12.11)
are related by
C2 = QT C1 Q. (12.12)
Furthermore, if there exists a mapping ξ between the two states V1 and V2 such that
Q = ∇ξ, then it must be a rigid-body deformation, that is, a rotation followed by
a translation [525, p. 49]. This observation allows us to define the virtual state as
a class of equivalence, where any two members in the same class are related by a
proper rotation.
A natural problem regarding the non-Euclidean manifold V and the incompatible
deformation tensor Ar is to characterize how these objects fail to be, respectively, flat
and compatible. Before doing so we introduce another configuration that will be used
to connect explicitly the existence of a residual stress field to a growth mechanism.
12.3 The Conceptual Hypothesis of Morphoelasticity 353
12.3 The Conceptual Hypothesis of Morphoelasticity
The existence of residual stress in biological materials is closely associated to growth

or to an internal reorganization of the reference configuration. As different elements
of a body change in size or relative position, they create stresses even in the absence
of applied loads. So far, we have assumed that the body B under consideration has
a residual stress field. This assumption led us to the definition of a virtual state V.
The central postulate of morphoelasticity is that the residual stresses are solely
created by a local growth deformation tensor. That is, the deformation gradient can
be decomposed multiplicatively as
F = AG, (12.13)
where the tensor G describes the change of shape and volume at all points in the
body due to growth.
The corresponding conceptual hypothesis, that anelastic contribution can be taken
into account through a multiplicative decomposition of the deformation gradient,
follows from early work in different communities [1111]: In polymer swelling it was
first discussed by Flory (1956) [368]; in the theory of defects by Bilby, Gardner, and
Stroh (1957) [98]; in elastoplasticity by Kröner (1958) [729], [732, p. 100] [730,
p. 286]; and later popularized by Lee [777] (1969). In the theory of thermoelasticity,
it was properly formalized by Stojanović et al. (1964) [1196, 1197] and in the context
of biological tissues, the multiplicative decomposition was independently proposed
in Russia by Kondaurov and Nikitin (1987) [721] and in Japan by Takamizawa et
al. (1987) [1216–1218] who used it to characterize residual stresses in arteries. The
same conceptual ideas can also be found in the work of Tranquillo and Murray on
wound healing [1243, 1244].
However, this approach only became a central concept of biomechanics following
the seminal work of Rodriguez, McCulloch, and Hoger (1994) [1091] who showed
how to translate growth processes in terms of the tensor G.
The general principle states that growth in an elastic tissue can be modeled by an
anelastic process which is the result of a combination of a local growth deformation
from a compatible configuration followed by a local elastic deformation.
A configuration of a body B is said to be compatible if there exists an isometric
embedding of the body B in E3 . That is, we postulate the existence of an initial
compatible reference configuration B0 ⊂ E3 and a growth tensor G(X0 , t) defined
at every material point p ∈ B with coordinates X0 ∈ B0 at time t, as shown in
Figure 12.5. The growth tensor maps the tangent space T p B0 at every point p to the
linear space T p V of a virtual state V for the same material point: that is, we have
G : TB0 → TV. (12.14)
We note that for a given virtual state V, the growth tensor G and the initial
configuration B0 are not unique. For instance, it is easy to picture that a rectangle
Intial reference configuration Virtual stress-free state
Figure 12.5 The action of the growth tensor. The growth tensor maps vectors in the tangent space
of the initial configuration to vectors in the tangent space of the virtual state. The initial configuration
and the virtual state are stress-free.
can be obtained as the isotropic growth of a smaller rectangle, or the anisotropic

growth of a square along one direction. Nevertheless, the Cauchy stress and shape of
the body in the current configuration is independent of the choice of both the initial
configuration and growth tensor as long as their tangent bundle can be identified.
We now consider the product of the two deformations taking the initial reference
configuration to the residually stressed configuration as shown in Figure 12.6:
Fr (X0 , t) = Ar (X0 , t)G(X0 , t). (12.15)
Since Fr is a local deformation tensor between two compatible configurations of the

same body B, it is the gradient of an invertible differentiable map between the initial
reference configuration and the final configuration:
xr = χr (X0 , t), (12.16)
where xr denotes the position of a point p in the current unloaded configuration with
position X0 in the initial reference configuration, as shown in Figure 12.6. Physically,
the inverse of the growth process can be thought of as the local shrinking, growing,
and rigid-body motion of the dislocated sub-bodies associated with V so that they
again form a compatible configuration.
We can now slightly shift our point of view. Initially, we started with a body
with residual stress and, by a series of thought experiments, first unloaded it, then
unstressed it to define a virtual stress-free state amenable to a treatment in nonlinear
elasticity. The next step was to postulate that this virtual state is itself the result of a
local growth process defining the tangent bundle TV. Doing so, we went backward
in the process that created the shape and stresses in the current configuration.
Now, to respect causality, we look at the forward growth process: We start with
an initial body and study its evolution through growth by specifying a growth law.
12.3 The Conceptual Hypothesis of Morphoelasticity 355
Residually stressed configuration Fl
χl
Fr
χr
Ar
Loaded and residually
stressed configuration
G
Intial reference configuration Virtual stress-free state
Figure 12.6 The creation of residual stress is obtained by considering a growth process followed
by an elastic response. The last deformation consists in loading the body with traction and body
forces.
Accordingly, we start from a known initial stress-free configuration B0 of a body B

in E3 and study those growth processes in an elastic body that can be expressed as
the result of a growth deformation followed by an elastic deformation. The growth
tensor G takes the initial configuration to a virtual stress-free state that may be
incompatible. Then, a local elastic tensor A restores compatibility of the body and
enforces the boundary conditions and body forces so that the body is in a compatible
configuration in mechanical equilibrium, that is, we have
F(X0 , t) = A(X0 , t)G(X0 , t), (12.17)
as shown in Figure 12.7. Note that both A and G have positive determinants. With
respect to the previous discussion it should be clear that A = Fl Ar .
12.4 Example: The Growing Ring
As a paradigm, we consider the simple, but illuminating problem of a growing circular

ring with diagonal growth. The growing ring is modeled as the section of three-
dimensional tube. In cylindrical coordinates, the growth tensor is
F
0
G A
Initial stress-free
reference configuration Loaded and residually
stressed configuration
Virtual stress-free state
Figure 12.7 The multiplicative-decomposition: Starting from an initial stress-free configuration

in Euclidean space, a local deformation G is applied to each material point, creating a virtual
stress-free state which is further deformed by A, to recover integrity of the body, into a residually
stressed configuration in Euclidean space. The intermediate virtual state is needed for the definition
of constitutive laws.
G = γr E R ⊗ E0,R + γ E ⊗ E0, + γz E Z ⊗ E0,Z , (12.18)
where γr , γθ , and γz are strictly positive and function of the radial variable R0 only.
If we further restrict growth so that γz = 1, each cross section of the tube are subject
to the same deformation. Therefore, we can identify the two bases at each point so
that E R = E0,R , E0, = E , E0,Z = E Z .
To understand the effect of growth, we first consider the case where γr , γθ constant.
If γθ = 1 and γr = 1, then growth is purely in the radial direction, with each radial
fiber expanding if γr > 1 and contracting if γr < 1. We will refer to this case as radial
growth. Similarly, the case of hoop growth defined when γr = 1, γθ = 1 corresponds
to circumferential fibers gaining or loosing mass, as shown in Figure 12.8.
If growth is chosen to be transversely isotropic, γr = γθ , the new ring is a pure
dilation of the original one, that is, R = γr R0 and = 0 . The tensor G is therefore
the gradient of a deformation and, following the reasoning of the previous section,
no residual stress is created in the growth process.
If we consider anisotropic homogeneous growth (γr = γθ ), then residual stresses
are created in a ring, but the actual values of these stresses will depend on the
equations of equilibrium. We will establish these equations in the next chapter and
compute the stresses generated by constant growth coefficients in Section 13.8.3.
12.4 Example: The Growing Ring 357
y
EΘ
ER
γr >1
R0 Θ0
x
γθ >1
Figure 12.8 Growth of ring elements. A ring sector remains a ring sector when either grown
through hoop only growth and/or radial growth.
The residual stress generated during the growth of a ring can be directly observed
by cutting the ring and observing that it relaxes to a sector as shown schematically in
Figure 12.9. In the present case, we assume that the body is homogeneous and that
there is no deformation in the axial direction. We further assume that the open body
has the shape of a ring sector and that it is stress-free. Then, the residual stress can
be fully described by the geometry of this open configuration and a suitable growth
tensor can be identified up to a multiplicative constant. Indeed, in this case the virtual
state is locally compatible and is, for positive opening angles, globally compatible
too. Therefore, we can find a set of coordinates describing the body and an invertible
mapping between the initial and virtual state given by
R = R0 , = (1 − ϕ/π)0 , Z = Z0, (12.19)
and choose, without loss of generality, A = A0 , B = B0 . As discussed above, a

given residually stressed material can be obtained through different growth processes
starting with a different initial configuration. The choice A = A0 and B = B0 is one
of such possibilities.
B=R(B0) A=R(A0)
b
a
Figure 12.9 Schematic of a radial cut in a ring and the corresponding opening angle ϕ.
The growth tensor is then easily identified, in cylindrical coordinates, as
[G] = [R] diag(1, ϕ/π, 1). (12.20)
where R is the rotation matrix around the Z -axis of angle ϕ/π. Once the growth
tensor is known, the stress field of the closed ring under load can be calculated from
the equilibrium equations.
The existence of a locally compatible stress-free grown configuration allows
for an explicit construction of the residually stressed state based on explicit map-
ping obtained through a series of cutting and gluing moves that are reminiscent
of Volterra’s dislocation theory [175, 1175]. From this perspective, the problem
of residual stress consists in finding an appropriate locally compatible stress-free
configuration. Equivalently this stress-free configuration can be obtained by a defor-
mation whose growth tensor is compatible. For instance, in the simple case of the
ring studied here, the open ring can be obtained from the current configuration by a
single cut or from an initial configuration by the growth tensor (12.20).
12.5 The Problem of Incompatibility
The unloading and unstressing process creates a virtual state characterized, locally,
by a linear space T p V at a material point p ∈ B. This decomposition can easily be
performed in terms of deformation tensors. However, a difficulty emerges when one
tries to associate a deformation to these deformation tensors so that they are deforma-
tion gradient tensors bona fide. The impossibility to define such mappings, globally
or locally, is the problem of incompatibility. We wish to quantify this incompatibility
both geometrically and analytically. Geometrically, it means that we need to define
a measure of how V fails to be Euclidean. Analytically, we need to quantify the lack
of existence of a mapping whose gradient would be A.
A tensor G(X) is said to be locally compatible if it is the gradient of a deformation,
that is, if there exists a smooth mapping y = y(X) such that G = Grad y. It is globally
compatible if it is a diffeomorphism over the body B0 defined as an open set in E3 . A
natural question is to establish whether a given growth tensor is locally compatible
as it would ensure the existence of a local set of coordinates on the virtual state that
can be used to solve the elastic problem and compute the residual stress. Here, we
assume that the body B0 is simply connected. An excellent generalization to the case
of non-simply connected bodies has been developed by Yavari [1351]. We draw from
results in the theory of continuous dislocations [180], elasto-plasticity [524], and the
geometry of differentiable manifolds [259] to establish the main results.
12.5 The Problem of Incompatibility 359
12.5.1 A Differential Geometry Perspective
We first discus the geometry of Euclidean and Riemannian manifolds to describe the
virtual state. The basic idea is to quantify the geometry in the virtual state V based
on geometric quantifiers defined on the initial configuration B0 . Since many of the
expressions in this section are best defined in component form, we will use some
of the standard notation of differential geometry, assume summation over repeated
indices. In this chapter, we distinguish between upper and lower indices, so that any
geometric quantity can be written in an arbitrary basis.
12.5.1.1 The Growth Metric and the Riemannian Curvature Tensor
The growth tensor G : TB0 → TV maps, at each point p, the tangent space of the
initial configuration to the tangent space of the virtual state at the same material
point p. We assume that the initial configuration is a simply connected Euclidean
manifold with the usual definition of a Euclidean metric. That is, for a given set of
Cartesian coordinates {x i } in a basis {Ei } the initial position vector is X0 = X0,i E0,i .
The metric tensor M0 has components
[M0 ]i j = E0,i · E0, j = δi j , (12.21)
where, we have assumed the standard inner product.

At a given point p, we wish to introduce a notion of distance and angle between
vectors on T p V. One way to do so is to consider the right Cauchy–Green tensor
associated with G, that is,
M = GT G. (12.22)
The tensor M provides a measure of distance and angle between vectors that are the
image of vectors in the initial configuration, as explained in Section 11.1.6. Explicitly,
if V0 = V0i Ei is a vector in TB0 , its image is V = G V0 and its lengths after
deformation is
j
|V| = |G V| = Mi j V0i V0 . (12.23)
j
We see that the quadratic form Mi j V0i V0 characterizes the natural lengths and angles
of vectors defined in the tangent plane of a point in the virtual state in terms of
quantities defined in the initial configuration. That is, the metric M is the pull-back
of the initial configuration metric associated with the tensor G [216]. This quadratic
form, defines on V, a metric referred to as the growth metric.
An important characterization of a metric is its flatness. A metric is said to be flat
if there exists a local system of coordinates where the metric tensor reduces to the
identity, otherwise, the metric is non-flat.
Initial stress-free
reference configuration
F
G
Stress-free configuration Virtual stress-free state
Figure 12.10 If the metric is flat there exists a stress-free configuration B̃ such that the map F̃−1 G
is locally a rotation.
We first consider the case of a flat metric shown in Figure 12.10. That is, there
exists a mapping ϕ : B0 → B̃ ⊂ E3 from the initial configuration to a new stress-free
configuration B̃ such that the associated metric in this new configuration reduces to
the identity,
F̃−T MF̃−1 = 1. (12.24)
where F̃ = Grad0 (ϕ). This last relation can be written as
M = F̃T F̃, (12.25)
which implies that the growth metric is the right Cauchy–Green tensor of a smooth
deformation. Since, by definition we have M = GT G, the identity GT G = F̃T F̃
implies that the tensors G and F̃ differ at most by a rotation at each point as can
easily be established by using the polar decomposition of F̃. Therefore at each point
p ∈ B, we have
G( p) = R( p)F̃( p), (12.26)
where R ∈ S O(3). A few comments are in order.

• When the metric tensor M is flat, the tensor M is a compatible strain tensor: M is
equal to the right Cauchy–Green strain tensor of a smooth deformation.
• The existence of a compatible strain tensor M does not imply that the growth ten-
sor G is compatible. Indeed, if the rotation R is not a rigid-body rotation, then the
compatibility of the map G does not follow from the compatibility of F̃. If M is
compatible, and G incompatible, then the body is in a state of contorted aleotropy
in the nomenclature of Noll [963]. See also [331, 336, 337] for examples and
interesting discussions.
• The mapping ϕ is an immersion, that is, the mapping is differentiable and its gra-
dient is invertible at all points. An immersion ensures the existence of F̃ and its
inverse at all points, but it does not guarantee the existence of a globally invertible
map ϕ. Consider for instance the growing ring studied in Section 12.4. In the case
of positive constant hoop growth γr = 1, γθ > 0, the growth tensor can be written
locally as a deformation gradient of a mapping, but this mapping is not globally
invertible.
• By construction the state V is stress-free. Since the configuration B̃ is related to V

by a local rotation and stress is insensitive to rotation, the configuration B̃ is also
stress-free.
We conclude that if the growth metric is flat, there exists a configuration B̃ from
which stresses can be evaluated. The next question is to find an effective method
to determine whether a given growth metric is flat. The answer is provided by a
basic theorem from differential geometry that relates the flatness of the metric to
the vanishing of the associated Riemannian curvature tensor. Curvature tensors are
fourth-order tensors that play a central role in the description and properties of Rie-
mannian manifolds. They are defined in general through the concept of connections
studied in the next section. However, there is a particular curvature tensor, the Rie-
mannian curvature tensor, that can be defined directly from the metric tensor. First,
we define the Christoffel symbols associated with the metric M:
◦ 1
i jk = ∂ j Mik + ∂i M jk − ∂k Mi j , (12.27)
2
◦ ◦
ki j = M kl i jl , (12.28)
where ∂ j = ∂∂X 0, j and [M kl ] = [Mi j ]−1 is the inverse of the metric M in component
form. Then the Riemannian curvature tensor associated with the metric M is
◦ ◦ ◦ ◦ ◦ ◦ ◦
Rli jk = ∂ j ikl − ∂k i jl + i j klm − ik jlm .
m m
(12.29)
We can now state the fundamental theorem related to the Riemannian curvature
tensor [211]: Let be a connected and simply-connected set in R3 equipped with
a metric M ∈ C 2 (). The metric M is flat if and only if its Riemannian curvature
tensor vanishes identically. that is,
◦
Rli jk = 0, ∀i, j, k, l. (12.30)
Note that due to symmetries with respect to indices, only six components of the Rie-
mann curvature tensor are independent in three dimensions. These six independent
components can be extracted from the curvature tensor by constructing either the
Ricci curvature tensor or, equivalently, the second-order symmetric Einstein tensor
or incompatibility tensor
◦ 1 ikl ◦
S = Rklmn ,
ij jmn
(12.31)
4
where i jk denotes the usual Levi-Civita or permutation symbols, that is, i jk = 1 is

(i, j, k) an even permutation of (1, 2, 3), −1 if it is an odd permutation, and 0 if any
◦
index is repeated. Condition (12.30) for flatness can then be written as S = 0.
This last result gives us a simple algorithmic way to test the flatness of the growth
metric without the need to construct an immersion ϕ: Given a growth tensor G,
◦ ◦
compute M and S . If S = 0, then the metric is flat and M is a compatible strain
tensor. Otherwise, the metric is non-flat. In such a case, no immersion exists but the
virtual state can be described as a Riemannian manifold with metric M.
Our geometric analysis identifies two different sources of residual stresses. In the
first case, the growth tensor defines a growth metric that is flat. The nonexistence of
a mapping between the initial configuration and the grown configuration is due to a
global constraint. Putting aside for the time being the situation where the body could
be in a state of contorted aleotropy, the growth tensor is locally integrable, but there
is no global mapping such that it is a gradient and therefore it fails to be globally
integrable.
In the second case, the deformation gradient is not locally integrable and the
growth metric is non-flat. There is no local mapping that will immerse the grown
stress-free body in the Euclidean space. That is, there are regions of the stressed body
where infinitely many cuts will be necessary to relieve all stresses.
Historical note. Within the theory of elasticity, the problem of incompatibility can
be seen as the nonlinear version of the Saint-Venant compatibility problem in linear
elasticity. In 1864, Saint-Venant derived six independent conditions for the linear
strains to be the symmetric part of the gradient of a displacement field. More, pre-
cisely, given a second-order linear strain tensor E, the Saint-Venant equations provide
conditions for the existence of a displacement field u such that
1
E= ∇u + ∇uT . (12.32)
2
These conditions were subsequently derived by Beltrami in 1886 in the context of
differential geometry [30, 321]. The connection between differential geometry and
anelasticity has been first put forward by Eckart in 1948 [321] and by Kondo in 1949
[722] who writes: “The Riemannian curvature has however been introduced into the
theory of elasticity in connection with incompatibilities and we can realize without
difficulty the existence of a Riemannian state of the elastic body, observing that the
essential properties of elastic bodies with initial stresses can best be described by
means of Riemannian geometry.” These ideas were further formalized by Kondo
[723], Bilby et al. [97] and Kröner [730, 731] and first applied in the context of
biomechanics in [1218] (see also [418, 702, 703, 1350]).
12.5.1.2 Growth Metric and Curvature for a Growing Ring
We consider a growing ring as a section of a growing tube. Since we do not consider

deformations in the axial direction, the problem is two-dimensional, but all formulas
can be readily adapted to this case. We describe the stress-free configuration B0
with respect to the cylindrical coordinates {R0 , 0 , Z 0 } so that, a material point has
coordinates
X0 = R0 E R0 = (R0 cos 0 )E1 + (R0 sin 0 )E2 + Z 0 E3 . (12.33)
The basis vectors are E R0 = ∂ R0 X0 , E0 = ∂0 X0 , and E Z 0 . Accordingly, the matrix
of the metric tensor M0 = M0,αβ Eα ⊗ Eβ in these coordinates is simply given by
⎡ ⎤ ⎡ ⎤
E R0 · E R0 E R0 · E0 E R0 · E Z 0 1 0 0
[M0 ] = ⎣ E0 · E R0 E0 · E0 E0 · E Z 0 ⎦ = ⎣ 0 R02 0 ⎦ . (12.34)
E Z 0 · E R0 E Z 0 · E0 E Z 0 · E Z 0 0 0 1
We consider the particular case where the growth tensor is diagonal and only a
function of the radial variable, that is,
⎡ ⎤
γr (R0 ) 0 0
[G] = ⎣ 0 γθ (R0 ) 0 ⎦ . (12.35)
0 0 γz (R0 )
The growth metric is the right Cauchy–Green tensor associated with G and is given
by M = GT G which, in components, reads
⎡ ⎤
γr2 0 0
[M] = [G]T [M0 ] [G] = ⎣ 0 R02 γθ2 0 ⎦ . (12.36)
0 0 γz2
Once the components of the metric tensor are known, it is a tedious, but easy exercise
to compute a table of the 27 possible Christoffel symbols. In our case, the only
nonvanishing components are
◦ γr ◦ 1 −R0 γθ (γθ + R0 γθ )
111 = , 22 = , (12.37)
γr γr2
◦ ◦ 1 γ
212 = 221 = + θ, (12.38)
R0 γθ

◦ ◦ γ
331 = 133 = z , (12.39)
γz
◦ γz γz
313 =− . (12.40)
γr
It is now straightforward to generate the 81 components of the curvature tensor. Doing

so, we realize that the only nonvanishing components of the Riemann curvature
tensors are
◦ γθ
R2112 = R0 γr (2γθ + R0 γθ ) − γθ γr − R0 γθ γr , (12.41)
γr
◦ ◦ ◦ ◦
R1212 = R2121 = −R1221 = − R 2112 , (12.42)
◦ γz
R1313 = (γr γz − γz γr ), (12.43)
γr
◦ ◦ ◦ ◦
R1331 = R3113 = −R3131 = −R1313 , (12.44)
◦ γθ γz
R2332 = R0 2 γz (γθ + R0 γθ ), (12.45)
γr
◦ ◦ ◦ ◦
R2323 = R3232 = −R3223 = −R2332 . (12.46)
◦
The condition R= 0 leads to
R0 γθ = Cγr − γθ , (12.47)
γz (γθ + R0 γθ ) = 0, (12.48)
γr γz − γz γr = 0, (12.49)
where C is an arbitrary constant. The second equation leads either to γz = 0 or

γθ + R0 γθ = 0.
In the first case, the third equation is identically satisfied and the condition for
local compatibility is
R0 γθ = Cγr − γθ . (12.50)
If, for instance, we choose γr = γθ , this condition leads to γr = γθ = μR ν with μ, ν

constant and the corresponding growth tensor does not create locally residual stress.
In the second case, γθ + R0 γθ = 0 and C = 0. We have
γθ = C1 /R0 and γr γz = γz γr , (12.51)

that is, γz = Cγr for a new arbitrary constant C. If we choose, for instance γr = γz ,
we have γr = K exp(C R0 ). Again, this profile of growth will not create residual
stress unless global constraints, such as the closure of the ring, are applied.
In these examples, if there is global incompatibility due to the material overlapping
on itself during growth, a finite number of cuts would relieve all stresses. Similar
examples of continuous distributions of defects leading to so-called impotent stresses
can be found in [1352, 1354].
12.5.1.3 A Heuristic Definition of the Growth Connection
The vanishing of the Riemannian curvature tensor guarantees the compatibility of

the strain tensor, but it does not guarantee that G is the gradient of a deformation,
merely that there exists a rotation matrix R at each point p ∈ B such that RG is
compatible. Therefore, the analysis of the Riemannian curvature is not sufficient
to establish the compatibility of the growth tensor. However, we can associate the
growth tensor with other geometric quantities that will enable us to provide a finer
description of compatibility.
The growth metric gives us a way to characterize lengths and angles between
vectors at a given point in the virtual state and identify the flatness, or lack thereof, of
V. We now wish to compare vectors at different points in V which, in turn, leads to the
idea of parallel transport, derivatives, and connections. Connections provide a way
to take derivatives of vector fields over manifolds and can be defined intrinsically on
manifolds. We first restrict our attention to a particular connection associated with a
deformation tensor. In our case, the growth tensor gives us a natural way to compare
vectors and define a derivative. We say that two vectors in TV at two different points
are equivalent if they are the images of the same vectors. More precisely, as shown
in Figure 12.11, equivalence for a vector field v at two different points p and q with
coordinates X and Y in B0 implies that
G(X0 )v(X0 ) = G(Y0 )v(Y0 ). (12.52)
0
G
G(X0 )v(X 0 )
X0 v( X0 )
G( Y0 )v( Y0 )
Y0 v(Y0 )
Intial reference Virtual stress-free

state
Figure 12.11 Defining a notion of equivalence for vector fields in V at different points in V .
Figure 12.12 The parallel

transport of a vector on V
0
can be performed by using v(X0)
the initial configuration B0 . X (0)
0
A vector field v(X 0 ) is
parallel transported on all X
0
points from G(X0 (0)) to
G(X0 (1)) on V if it satisfies
Equation (12.53) along the
path in B0 .
X (1)
0
If this equality holds for every pair of material points in the body B, the vector field
v should be regarded as constant on V. In particular, for q close to p, we can localize
the equivalence relation and write the differential condition for a vector field to be
constant, namely:

Grad0 v(X0 ) + G−1 Grad0 G v(X0 ) = 0. (12.53)
This last relation can be used to define a notion of parallel transport of a vector:
Starting at a point with coordinates X0 (0) ∈ B0 , a vector v(X0 ) is parallel transported
to a point X0 (1) ∈ B0 along a path if the differential equation (12.53) is satisfied on
this path (Figure 12.12).
The notion of parallel transport is central to the theory of differential geometry
as it provides a natural notion of equivalence between vector fields. In our heuristic
construction, the tensor G endows V with a structure that allows us to compare
vectors at different points. The central quantity associated with this construction is
the growth connection defined by
= G−1 Grad G. (12.54)
The connection is not coordinate independent, but tensorial quantities can be

extracted from its components. To do so, we now turn to a formal description of
the Riemannian and affine geometry of the virtual state.
12.5.1.4 Riemann–Cartan Manifolds
Connections are fundamental quantities defined on manifolds that allow for the proper
definition of derivatives. It is appropriate at this point to give a full definition of
basic geometric quantities associated with manifolds. A linear affine connection on a
manifold V is an operation ∇ : (u, v) ∈ X (V)×X (V) → ∇u v ∈ X (V), where X (V)
is the set of vector fields on V, and such that ∀ f, f 1 , f 2 ∈ C ∞ (V), ∀ a1 , a2 ∈ R:
i) ∇ f1 u1 + f2 u2 v = f 1 ∇u1 v + f 2 ∇u2 v, (12.55)

ii) ∇u (a1 v1 + a2 v2 ) = a1 ∇u (v1 ) + a2 ∇u (v2 ), (12.56)
iii) ∇u ( f v) = f ∇u v + (u f )v. (12.57)
We call ∇u v, the covariant derivative of v along u. A parallel transport of the vector

field v is obtained along u if ∇u v = 0. In a local set of coordinates {X i }, we can give
an explicit form to the connection
∇∂i ∂ j = k i j ∂k , (12.58)
in terms of k i j , the Christoffel symbols of the connection and where ∂i = ∂∂X i form
a natural basis for the tangent space corresponding to the choice of coordinates {X i }.
A linear connection is said to be compatible with a metric M of the manifold if
∇u (v · M w) = ∇u v · (M w) + v · (M ∇u w) . (12.59)
In general, the geometry of a manifold V is fully specified by a metric M and a con-

nection ∇. Here, we restrict our attention to Riemann–Cartan manifolds (V, ∇, M)
which have the properties of being metric-compatible, that is ∇M = 0. In compo-
nents this condition reads
∂ Mi j
− l ki Ml j − l k j Mil = 0. (12.60)
∂ Xk
The general case where the metric is not compatible with the connection is related to
the notion of point defects in the theory of dislocation [458, 1353]. We can attach to a
manifold equipped with a connection two fundamental quantities the torsion tensor
T : X (V)×X (V) → X (V) and the curvature tensor R : X (V)×X (V)×X (V) →
X (V). They are defined as
T (u, v) = ∇u v − ∇v u − [u, v], (12.61)

R(u, v)w = ∇u ∇v w − ∇v ∇u w − ∇[v,v] w. (12.62)
A geometric interpreation of the torsion tensor is given in Figure 12.13. In compo-

nents with respect to the local coordinates {X i }, we have
T i jk = i jk − i k j , (12.63)
∂ kl
i
∂ jl i
Ri jkl = − + i jm m kl − i km m jl . (12.64)
∂X j ∂ Xk
Note that these two tensors should not be confused with the Frenet torsion and
curvature of a curve as defined in Part II, even though the curvature tensor does
contain information about the Frenet curvature when reduced to a one-dimensional
manifold.
∇v u
[ u,v] ∇u v
ur
T (u,v )
r //
ur
//
vq
vp vq
p up q
Figure 12.13 The torsion tensor can be interpreted as a failure of closure under parallel transport.
Consider two vector fields u and v, and compute the difference between the parallel transports of
each vector field along the other one. If this difference does not vanish, the manifold has non-zero
torsion. Note that q and r are assumed to be infinitesimally near p so that their tangent spaces can
be identified (adapted from [558]).
There are two important special cases of Riemann–Cartan manifolds when either
the torsion or the curvature vanishes as shown in Figure 12.14.
Riemannian manifolds. Given a metric M on a manifold V, there is a unique

◦
linear connection ∇ that is compatible with G and is torsion-free, that is,
◦ ◦
∇ u v − ∇ v u = [u, v]. (12.65)
This connection is the Levi-Civita connection defined by the following Christoffel

symbols
◦ M kl
ki j = ∂ j Mik + ∂i M jk − ∂k Mi j . (12.66)
2
This connection is identical to the one defined by (12.27). In this special case where
the connection is induced by the metric, the curvature tensor is the Riemannian
curvature tensor associated with the metric M given by (12.29).
It is important to make the distinction between the Riemannian curvature tensor
defined on a Riemannian manifold by the metric and the curvature tensor defined on
a general Riemann–Cartan manifold through the connection. These two objects only
coincide when the torsion vanishes.
The existence of a connection induced by the metric and leading to a torsion-free
manifold is a fundamental result of differential geometry and the starting point for
the theory of general relativity [383].
Weitzenböck manifolds. If the curvature tensor vanishes identically, the

Riemann–Cartan manifold is called a Weitzenböck manifold. Whereas the Levi-Civita
Riemann-Cartan Manifolds (V, ∇, G)

∇M= 0, T = 0, R = 0
Riemannian Manifolds (V, ∇, M) Weitzenböck Manifolds (V, ∇, M)

∇M= 0, T = 0, R = 0 ∇M= 0, T = 0, R = 0
Flat(Euclidean)Manifolds (V, ∇, M)
∇M= 0, T = 0, R = 0
Figure 12.14 Special cases of Riemann–Cartan manifolds. The case of vanishing torsion leads to
a Riemannian manifold whereas the case of a vanishing curvature leads to a Weitzenböck manifold.
These two cases further reduce to a Euclidean manifold where both curvature or torsion vanish.
connection gives us a direct construction of a torsion-free manifold, it is not clear

how to create a curvature-free manifold. However, the particular choice of a connec-
tion defined through a deformation tensor G provides such a construction. Indeed,
consider an invertible second-order tensor G of class C 2 and its associated metric
M = GT G. In a given coordinate chart {X i }, we define the Weitzenböck connection
by
i jk = G−1 l i ∂ j (G)l k . (12.67)
With respect to this connection, it can be checked that the curvature tensor (12.64)
vanishes identically. The torsion tensor associated with the growth tensor G is defined
in an orthonormal basis as the third-order tensor [3, 1352]
T = G−1 Skw(Grad G), (12.68)
where Skw(·) is the skew-symmetric part of a tensor, defined by Skw(T )ki j = Tki j −
Tk ji .
12.5.1.5 Example: The Torsion of a Growing Ring
We consider again a growing cylinder with growth tensor [G(R0 )] = diag(γr , γθ , γz )

in cylindrical coordinates as defined by (12.18). A direct computation of the torsion
tensor gives (denoting by [Tk.. ] the matrix components of the second-order tensor for
a given k)
⎡ γθ
⎤ ⎡ γ
⎤
γθ −γr
0 γθ R0
− γθ
0 0 0 γzz
⎢ γθ ⎥ ⎢ ⎥
[T1.. ] = 0, [T2.. ] = ⎣ γr −γθ + 0 0 ⎦ , [T3.. ] = ⎣ 0 0 0 ⎦ . (12.69)
γθ R0 γθ γ
0 0 0 − γzz 0 0
and the vanishing of the torsion tensor leads to
γθ − γr
γz = 0 and = γθ . (12.70)
R0
Comparing these conditions with Equations (12.47–12.49), we see that the condition
for the compatibility of the growth tensor are stricter (C = 0 and γz constant) than
for the compatibility of the metric, as expected.
12.5.1.6 Summary on Geometry and Incompatibility
From a practical point of view, we can use the information provided by the geometry
of the growth deformation to establish important properties of the deformation with
respect to local compatibility.
We recall that a deformation tensor is locally compatible if there exists a smooth
mapping X = ϕ(X0 ) such that G = Grad0 ϕ. The strain tensor M = GT G is said
to be locally compatible if there exists a compatible deformation tensor F̃ such that
M = F̃T F̃.
Given the growth tensor G, by construction the curvature tensor R vanishes iden-
tically. We compute the two fundamental objects for compatibility: the Riemannian
◦
curvature tensor R (G) and the torsion tensor T (G). We then have the following
cases:
◦
• Case 1: Local compatibility. If T (G) = 0, then R (G) = 0 and both the strain
tensor M and the growth tensors G are locally compatible. From a mechanical
perspective, if a body is residually stressed due to a growth characterized by such
a tensor G, then these stresses are only created by global incompatibility and
they can be completely relieved by a finite number of cuts. This situation arises,
for instance, for a growing ring with constant γr and γθ . If γθ > γr , then the
deformation would create an overlap in the circumferential direction that would
result in compressive stress. A finite number of cuts in the radial direction will
completely relieve the stress. Note, however, that the ring is not simply connected
but that a nongrowing inner disk can be added to the ring to restore connectedness
without changing its local compatibility.
◦
• Case 2: Contorted aleotropy. If R (G) = 0 but T (G) = 0, then the strain tensor
M is locally compatible but G is not. However, there exists a rotation tensor R
such that RG is compatible. The difficulty in this case is that the rotation R can
vary from point to point. Note that despite the fact that the growth tensor is not
compatible, there is no stress created through the growth process (apart from the
possibility of global incompatibility).

An example of contorted aleotropy is provided by the growth of a small enough
sector of a growing ring of initial radius and such that R0 γθ = Cγr −γθ with C = 1
and C = 0. From a modeling perspective, this situation is somewhat artificial and
points to the fact that the growth tensor has not been properly defined. Indeed
for isotropic material, there is a freedom in the decomposition of the deformation
gradient as
F = AG = ART RG, ∀ R ∈ S O(3). (12.71)
Therefore, replacing the growth deformation G by RG where R is chosen to restore

compatibility does not change the description of mass addition or the residual stress
field for isotropic materials, but may change the computation of residual stress for
anisotropic materials.
◦
• Case 3: Local incompatibility. If R= 0 then T (G) = 0. In this case, both strain
and deformation tensors are locally incompatible. Mechanically, it implies that, in
the regions of the body where the Riemannian curvature tensor does not vanish,
an infinite number of cuts would be necessary to relieve all stresses. A simple
example of this situation is provided by a sector of a growing ring with γθ = 1
and γr = R0 .
12.5.2 An Analytic Perspective
Incompatibility can also be understood from a purely analytic perspective based

on Stokes’ theorem [3]. Recall that in a simply-connected domain a differentiable
vector field is the gradient of a differentiable function if and only if its curl vanishes
identically. It is therefore appropriate to introduce a similar notion for tensor fields
[180, 524]. The curl of a differentiable second-order tensor T is the second-order
tensor defined by [526]
(curl T)c = (curl TT c) for all constant vectors c. (12.72)
In a Cartesian basis, the curl can be expressed as
∂T jl
(curl T)i j = ikl , (12.73)
∂xk
where ikl is the Levi-Civita permutation symbol. Note the particular choice of the
transpose of the tensor in (12.72) that guarantees the following identity for all twice
continuously differentiable vector fields v:
curl (grad v) = 0. (12.74)

The Stokes’ theorem naturally generalizes to

T dx = (curl T)T n dx, (12.75)
∂S S
where S is an oriented surface in a simply-connected region of R3 with unit normal

n and boundary ∂ S. Similarly if curl(T) = 0 in a simply-connected region ⊂ Bt
then by using the Stokes’ theorem the integral of T vanishes on all arbitrary paths
in and we conclude that [32]: In a simply-connected domain for a continuously
differentiable tensor T, curl(T) = 0 if and only if there exists a twice continuously
differentiable vector field ϕ such that T = grad ϕ.
We apply these ideas to the growth tensor. We define Curl0 (G) to be the curl of
the tensor G, where the derivatives are taken with respect to the initial coordinates. It
follows from our discussion that if G is the gradient of a vector field, then Curl0 (G)
vanishes identically. We can therefore use Curl0 (G) or curl(A−1 ) to quantify incom-
patibility. Different functions of Curl0 (G) and G have been proposed for this purpose
in the theory of dislocations. The original proposal is Nye’s dislocation tensor [969]:
α = (curl A−1 )T . (12.76)
This tensor can be directly related to the cumulative Burger’s vector b, that measures
the presence of dislocations on a loop as the closed path integral around a surface S
in B:
b= A−1 dx. (12.77)
∂S
By using Stokes’ theorem, we have

b= α n dx, (12.78)
S
where n is a unit vector normal to S. The traditional geometric interpretation of the

cumulative Burger’s vector is that its magnitude and direction quantifies the failure
of the closure of the image of the path integral around a point. However, particular
care must be taken when considering path integral in the virtual state as discussed
by Ozakin and Yavari [1003].
Another interesting quantifier of incompatibility is the true incompatibility tensor
proposed in [180, 1224] and defined as
K(G) = JG−1 G Curl0 G, (12.79)
where JG = det(G). This tensorial function K has the remarkable property that it
is invariant under arbitrary differentiable variations of the configuration: If Q is the
gradient of a deformation, then
K(GQ) = K(G). (12.80)
Since G = A−1 F and F is, by definition, a deformation gradient, it naturally follows

that K(G) = K(A−1 ), that is
K(G) = JG−1 G Curl0 G = JA A−1 curl A−1 = K(A−1 ). (12.81)
Further, it was shown in [180] that K is closely related to the density of Burger’s
vectors in the theory of dislocations and that all functions that are invariant under a
compatible change of configuration are necessarily functions of K, which places the
true incompatibility tensor as a central measure of incompatibility.
Finally, we note that the tensor K is closely related to the torsion tensor associated
with the growth tensor G. In components, K i j = [K(G)]i j , the two tensors are related
by
1
K i j = JG G ik klm (GT ) jlm . (12.82)
2
12.5.2.1 The True Incompatibility Tensor for a Growing Cylinder
We consider again our example of a growing cylinder with growth tensor [G(R0 )] =
diag(γr , γθ , γz ) in cylindrical coordinates. The true incompatibility tensor can be
readily evaluated as
⎡ ⎤
0 0 0
1 ⎢0 ⎥
[KG ] = ⎣ 0 −γθ γz ⎦ , (12.83)
γr γθ γz 0 γz γθ −γr
− γθ 0
R0
and we conclude that G is locally compatible if and only if
γθ − γr
γz = 0 and = γθ . (12.84)
R0
As expected, we recover the conditions (12.70) obtained from the geometric argu-
ment.
Chapter 13
Balance Laws
We adopt the view that growth is modeled by a multiplicative decomposition of the

deformation gradient from an unstressed configuration to the reference configuration
through a virtual state suitable to describe the stress at each material point. To sum-
marize the previous chapter, we start from a known initial stress-free configuration
B0 of a body B ⊂ E3 and consider a growth deformation G at each material point,
mapping locally the tangent space T p B0 at a point p in the initial configuration to a
linear space T p V in the virtual state V as shown in Figure 12.7. The study of stresses
is done by considering an elastic deformation A mapping T p V to T p B. The com-
position of A and G defines a map from the initial to the current configuration with
deformation gradient F at all material points:
F = AG, (13.1)
where both A and G have strictly positive determinants. The elastic tensor A restores
the integrity of the body under the constraint that the physical balance laws in the
current configuration B are respected. We now discuss balance laws for a growing
continuum. We start with the discussion of a crucial assumption based on time scales
in growing media.

376 13 Balance Laws
13.1 The Slow-Growth Assumption
The decomposition (13.1) is instantaneous. It applies continuously with respect to

time, but in general, elastic and growth processes do not operate on the same time
scales. In typical biological growth processes or in the swelling of soft materials, we
can identify five important time scales:
• Elastic time scales. These time scales, denoted τelas , are associated with inertial
effects such as vibration or the propagation of waves in a tissue. Since most tissues
and organs have a typical size of 10 cm and typical waves propagate with velocities
in the order of 10–100m/s, we choose τelas ≈ 10−2 s.
• Viscous time scales. Inertial effects are damped in elastic tissues due to internal
friction. Depending on the softness of the tissue and its viscosity, a typical viscous
time scales is τvis ≈ 1 s.
• Stress relaxation time scales. In some systems, the residual stresses built up dur-
ing growth are relaxed due to internal rearrangement. In a morphogenetic process,
this rearrangement can act on the same scale as growth itself. An example of this
relaxation is the evolution of the intrinsic shape found in loaded plants and dis-
cussed in Section 6.3. In physiological regimes, however, residual stresses develop
in order to carry specific functions and are therefore associated with a plastic-like
relaxation of stresses due to internal rearrangement with a typical time scale τrel .
• Loading time scales. These time scales, τload are associated with typical strain
rates due to loading such as the cell cycle, blood flow, the forces applied on mus-
cles during exercise, or the circadian rhythm which may vary from seconds to days.
• Growth time scales. Typically, growth acts on time scales τgrowth longer than
most of the above processes. Even when we consider the fastest growth process
found in nature, they are typically orders of magnitude slower than other effects
acting on the system. For instance, the fastest doubling time found in bacteria
is of the order of 10 minutes for some thermophilic anaerobic bacteria found in
the hot hydrothermal vent of Lake Tanganyika [330]. It has been argued that a
doubling time found in Escherichia coli of about 20 min to 1 hour is close to the
thermodynamic limit [332]. Estimates of doubling times for cancer cell in vivo are
typically of the order of 24–60 hours [1186], but the doubling times for tumors
varies between 100 days for fast growing tumors to years for slow ones [36]. For
terrestrial vertebrates, Case gives an extensive list of postnatal growth rates with
typical doubling times in the range of years during early growth years [174]. In
plants, maximum growth rates vary with doubling times ranging from 3 hours in
unicellular algae to 4–7 years in some trees [950]. Moso bamboos are often cited
as the fastest growing plants with linear growth of more than 100 cm in 24 hour
[396].
13.1 The Slow-Growth Assumption 377
These various considerations lead to the following ordering of time scales for
typical growth processes in biological systems:
τelas τvisc < τload τgrowth . (13.2)
The relaxation time scales is absent from this list as it defines different types
of growth. Therefore, an explicit time scale will enter the evolution laws for the
tensor G. If growth is mostly elastic as found in mature organs, there is little or
no relaxation of the internal stresses and τgrowth > τrel so that relaxation effects can
be safely ignored. However, in development there is a quick rearrangement of cells
within the tissue that dissipates rapidly the mechanical effect of mass addition. The
system then behaves more in a fluid-like manner and τrel is comparable to τgrowth .
Yet, in other systems, such as in the growth of stems discussed in Section 6.3, there
is a balance between relaxation and stresses so that as growth proceeds the system
maintains a given level of stresses while still adding mass.
In the rest of this chapter, we focus on the physical balance laws in a growing
continuum. We consider a body in the current configuration subject to loads, and
generalize the balance laws obtained in Section 11.2 to include the possibility of
mass addition, change in density, as well as fluxes of linear and angular momenta
due to new material entering the body. These new fluxes are associated with the time
scale of growth. Based on the ordering of time scales, we will use the slow-growth
assumption of morphoelasticity, that is, we will assume that the time scales of growth
processes are much larger than any other time scales and respect (13.2). Therefore,
the time scale of growth and elastic response are widely separated and it is reasonable
to assume that for growth processes, the elastic dynamics of the material is dependent
on the growth only through the change of geometry and not through the dynamics
of the fluxes associated with growth processes. That is, at any given time and on
short-time scales, we assume that the material is hyperelastic. The generalization to
a theory with possible viscoelastic components follows naturally, but it will not be
discussed here. In order to see how the slow-growth assumption affects the balance
equations, we will first include these terms formally before neglecting them. This
process allows for generalizing the theory if necessary. Further discussion and details
on the derivation of balance laws in anelasticity can be found in several articles and
books [339–341, 809, 811].
13.2 Balance of Mass
We follow the evolution of a subset ⊆ Bt of a body B. Assuming that growth can

occur either through volumetric growth or through material flux across the boundary,
the rate of local increase by volumetric growth is captured by the growth rate function
ργ(x), and the flux of material through the boundary ∂ is described by the vector
(x). We note that the body B is fixed, i.e., we do not have surface growth on the
boundary of the body. The flux (x) describes a possible effect of diffusion of matter
378 13 Balance Laws
into itself [340]. The balance of mass, applied to an arbitrary volume element with
outward normal boundary vector n, reads now

d
ρ dv = ργ dv + n da. (13.3)
dt ∂
We follow the Maxwell transport and localization procedure given in Section 11.2.1.
That is, we apply the divergence theorem and the transport formula (11.96). Then
assuming that the integrands are all continuous and using the fact that is arbitrary,
we use the localization procedure to obtain the local version of the mass balance
equation for a growing continuum:
ρ̇ + ρ div v = ργ + div . (13.4)
Using the slow-growth assumption, we neglect fluxes through the boundary and their
associated momenta, that is, we set = 0 so that
ρ̇ + ρ div v = ργ. (13.5)
We can relate the growth tensor G to the growth rate function γ(x) appearing in the
mass balance equation (13.5). To do this, we first note that expressed in the initial
configuration, the mass equation is
d
(J ρ) = J ργ. (13.6)
dt
Next, we write J = J A JG , where J A = det(A) and JG = det(G) measure the change

in volume due to the elastic deformation and growth deformation, respectively. Let
ρG be the density with respect to the virtual state, and consider an element of mass
in the virtual state, dM = ρG dV . Since mass is not added in the elastic deformation,
the mass in the current state is
dm = ρ dv = dM = ρG dV, (13.7)
and since dv = J A dV , we have ρJ A = ρG . Using this relationship along with Jacobi’s

identity (11.86) for G:

∂ −1 ∂G
(detG) = (detG)tr G , (13.8)
∂t ∂t
Equation (13.6) can be written
ρ̇G + ρG tr(G−1 Ġ) = ρG γ. (13.9)

13.2 Balance of Mass 379
There are two important limits when either the density or the volume does not change
with growth.
• Constant density. In this case, ρ̇ = 0, and the relation between the growth tensor
G and the rate of growth is simply
tr(G−1 Ġ) = γ. (13.10)
• Constant volume. If there is no change in volume, then JG = 1, and the evolution

of the density can be written
ρ˙G = ρG γ. (13.11)
That is, γ is a rate of densification, adding mass to the system without changing
its volume.
We can also relate the growth tensor and growth rate to the experimental kinemat-
ics of growth presented in Section 1.4. Assuming that there is no elastic deformation,
that is, A = 1, we have G = Grad χ and we define the growth velocity as
∂χ
vG (x, t) = ẋ = . (13.12)
∂t
The growth velocity gradient tensor is then
LG = grad vG . (13.13)
Using the chain rule, we have Grad vG = (grad vG ) G = LG G, but since Grad vG =
Grad ẋ = ∂t Grad x, it follows that
Ġ = LG G. (13.14)
Using again identity (13.8) and tr LG = div vG , we have
J˙G = JG div vG , (13.15)
which in the case of constant density specializes to Equation (1.15). We note that if
G is not a multiple of the identity (isotropic growth), there is much more information
in G than in γ, so that in general for the description of anisotropic growth, all
components of the tensor G may be needed.
380 13 Balance Laws
13.3 Balance of Linear and Angular Momenta
The balance of linear momentum is modified to take into account the possible con-
tributions coming from the addition of mass through the boundary. We have two
types of contributions. First, newly added material may have the same properties as
the material at the same point. We call these types of sources compliant, using the
terminology of [339]. Typically, these contributions balance out and do not appear in
the local statement of the balance laws. Second, the new material may have sources
that are different from the pre-grown material, having for instance locally a higher
energy density than the surrounding tissue, or due to microscopic organization, a
lower entropy. We call these sources non-compliant and we denote them with an
overbar. They appear as extra contributions in the local statements. For instance, in
the case of the balance of linear momentum for a grown material, we have

d
ρv dv = Tn da + ρb dv + ργv dv + p̄ dv . (13.16)
dt
∂

body and traction forces compliant non-compliant
The compliant source is a direct result of the mass uptake. As the mass is added to
the material, it carries a momentum that needs to be included in the overall balance
of linear momentum. The non-compliant source p̄ is more delicate. It corresponds
to other possible sources of linear momenta acting on the system due to growth. It
can be used to model internal microscopic processes that would add momenta in the
open system.
The use of the divergence theorem on the terms integrated over ∂ as well as the
transport formula (11.97) lead to

(ρv̇ + ρ̇v + ρ(div v)v) dv = (div T + ρb + γv + p̄) dv. (13.17)

Inserting the expression for ρ̇ from (13.4), applying the transport formula (11.97),
and using the arbitrary nature of , (13.17) simplifies to
ρv̇ = div(TT ) + ρb + p̄. (13.18)
Again, using the slow-growth assumption, we set p̄ = 0 and recover the usual balance
of linear momentum
ρv̇ = div(TT ) + ρb. (13.19)
The balance between change in angular momentum and applied torques with respect
to the origin can be stated as
13.3 Balance of Linear and Angular Momenta 381

d
ρx × v dv = x × (T · n) da + ρx × b dv
dt
∂

torques due to body and traction forces

+ ργx × v dv + x × p̄ dv. (13.20)

compliant non-compliant
Since the non-compliant linear momentum source p enters the angular momentum
balance, we require no additional non-compliant angular momentum source. We
assume again that the material is non-polar and that there are no sources of internal
torques acting on the system, compliant or otherwise. The transport and localization
procedure leads to the standard condition that, even in the presence of growth fluxes,
the Cauchy stress tensor T is symmetric:
TT = T. (13.21)
13.4 Energy Balance
The energy balance is the first law of thermodynamics. It relates the change in the
total energy of the body and contributions from mechanical power, heat, and extra
sources of energy due to new material entering the system that can be either compliant
or non-compliant:
d
(E + K) = P + Q + S + S̄, (13.22)
dt
where the total energy has been decomposed into an internal energy E (i.e., elastic
and chemical energy) and a kinetic energy K. The mechanical power P is the work
per unit time that the body is exposed to through volumetric forces b and surface
tractions t. The heating Q is composed of a volumetric heat source r (for radiation)
and a surface flux q called the Fourier heat flux. In addition, we have the compliant
energy contribution S and the non-compliant contribution S̄ due to non-compliant
energy source density ε and the power associated with the non-compliant source of
linear momentum p. The energy balance on an arbitrary domain reads

d 1
ρ ε + |v|2 dv = ρb · v dv + n · Tv da
dt 2
∂
E()+K() P()

+ ρr dv + q · n da
∂
Q()
382 13 Balance Laws

1
+ γ ε + |v|2 dv + (ε + p · v) dv .
2

S: compliant S̄: non-compliant
(13.23)
A tedious, but otherwise uneventful, process of transport and localization using the
other balance laws leads to the local statement of energy balance
ρε̇ = tr (TL) + ρr − grad q + ε̄, (13.24)
where L is the velocity gradient tensor introduced in Section 11.1.8. We observe

again that apart from possible non-compliant sources of energy, this balance is the
one that we would expect for a non-growing material.
13.5 Imbalance of Entropy
For a non-growing medium, entropy typically measures the disorder induced by

microscopic fluctuations, and entropy increases with disorder [390]. The second law
of thermodynamics states that the rate of change of internal entropy of a system
is larger than or equal to the flow of entropy transferred to that system. Locally, it
implies that the change in entropy density (per unit mass), η, measuring the system’s
disorder is greater than the local entropy transfer given by the heat supply Q, defined
in the previous section, divided by the temperature θ > 0. In the case of a growing
material, we also need to include the non-compliant energy source ε and new possible
non-compliant entropy contribution −h which takes into account both non-compliant
entropy fluxes and sources [737]. This law, known as the Clausius–Duhem inequality
takes the form

d ρr q·n ε̄ − h̄
ρη dv ≥ dv − da + γη dv + dv . (13.25)
dt θ θ θ
∂
entropy due to heat supply compliant non-compliant
The local version of the Clausius–Duhem inequality in the current configuration,

assuming all fields are smooth [526, p. 186], is given by
ρr q
ε̄ − h̄
ρη̇ ≥ − div + . (13.26)
θ θ θ
The entropy balance together with the energy balance can be used to derive another
inequality by applying a Legendre transform to the specific free-energy density func-
tion
ψ = ε − θη. (13.27)
13.5 Imbalance of Entropy 383
In the current configuration, this inequality reads
1
ρψ̇ ≤ tr(TL) − ρη θ̇ − q · grad(θ) + h , (13.28)
θ
where L = ḞF−1 is the velocity gradient tensor. We can rewrite this inequality by
introducing an important quantity, the dissipation as
1
− θ = ρ(ψ̇ + η θ̇) ≤ tr(TL) − q · grad(θ) + h ≤ 0. (13.29)
θ
The term θ represents the energy dissipation per unit volume. In the present context,
a re-statement of the second law of thermodynamics, is the dissipation inequality,
that states that the dissipation in the system is positive, that is, ≥ 0.
13.6 Elastic Constitutive Laws
The imbalance of entropy can be used in a general context to derive constitutive laws
between stresses and strains. This can be done systematically by using the Coleman–
Noll procedure [222] of rational thermodynamics. We will use this procedure in the
next chapter to extract information on both the growth laws and the elastic constitutive
laws.
Here, we simply assume that the material is hyperelastic and isothermal when
considered on elastic time scales defined in Section 11.3. Therefore, using both the
hyperelastic and the slow-growth assumptions, there exists an internal energy density
that depends only on the elastic deformation tensor A, so that
1

= W, (13.30)
JAρ
where, as before, W = W (A) is the internal strain-energy density per unit volume
(of the virtual state). That is, the material is hyperelastic with respect to the virtual
state viewed as an unstressed configuration. The stresses are then given by
∂W
T = J A−1 A − p1. (13.31)
∂A
If the material is elastically incompressible, J A = 1 and p is the hydrostatic pressure;
for an elastically compressible material p = 0. All kinematic quantities associated
with F such as the invariants, the principal stretches, and the elastic strain tensors,
needed to obtain different explicit forms of this constitutive law, can now be written
in terms of the elastic deformation tensors A instead of F.
384 13 Balance Laws
13.7 Summary of Volumetric Morphoelasticity
In our approach to growth, we start with an initial unstressed configuration B0 . We

assume that a volume element at any material point is free to grow. Locally, this
growth process transforms an infinitesimal sphere into an infinitesimal ellipsoid.
The size and orientation of this ellipsoid is fully specified by the growth deformation
tensor G. The application of G transforms the initial configuration into a virtual state
V. This state is also unstressed and is used to compute the elastic strains that ensure
both the integrity of the continuum body and the elastic stresses that balance the
applied body loads and the boundary conditions. These strains are defined through
an elastic deformation tensor A that maps the virtual state to the current configuration
B. The deformation gradient tensor F that defines the local deformation between B0
and B is therefore decomposed as F = AG.
In the current configuration B, the balances of mass, linear and angular momenta
lead to the usual Cauchy equations for the Cauchy stress tensor T and mass density
ρ. The stresses are related to the elastic strains for a hyperelastic material from a
strain-energy function W . Altogether, we have
F = AG (13.32)
ρ̇ + ρ div v = ργ, (13.33)
ρv̇ = div(T) + ρb, (13.34)
TT = T, (13.35)
∂W
T = J A−1 A − p1, (13.36)
∂A
where b is the body force density, ρ the density, v the velocity, and J A = det(A). As
before, J A = det(A) = 1 for an incompressible material and p = 0 for a compress-
ible material.
The key difference with the classic theory of elasticity is that geometry and elas-
ticity are divorced. That is, the stresses are computed with respect to the elastic
deformation tensors A and not the entire deformation tensor F. The constitutive
equations give T as a function of A which can be rewritten as a function of FG−1
and Equation (13.34) can be solved as usual with respect to the initial or current
coordinates.
Note that we still need to specify G constitutively. Before doing so, we consider
a couple of examples where G is specified.
13.8 Simple Examples 385
13.8 Simple Examples
13.8.1 A Growing Cuboid
To understand the steps involved in solving a problem of elastic growth, we first con-
sider the simple problem of an isotropic incompressible cuboid constrained between
two blocks and allowed to grow along its axis as shown in Figure 13.1. That is, the
growth deformation transforms a cuboid initially of length L 0 into a longer one of
length L and the boundaries along this axis prevents this expansion so that its length
remains l = L 0 . This growth process is compatible and does not generate residual
stress. Therefore, it could easily be solved by the methods of nonlinear elasticity
following the steps of Section 11.8.1 by considering the equivalent problem of com-
pressing a block of length L to a block of length l. The purpose here is to understand
the steps involved in the multiplicative decomposition.
Using the same Cartesian basis {e1 , e2 , e3 } for the three configurations and assum-
ing uniform growth in the e3 -direction, the three deformation tensors are
[F] = diag(λ1 , λ2 , λ3 ), [A] = diag(α1 , α2 , α3 ), [G] = diag(1, 1, γ 2 ),

(13.37)
where γ 2 = L/L 0 . The multiplicative decomposition implies that λ1,2 = α1,2 and
λ3 = α3 γ 2 . Since the block is isotropic and there is no load applied in the directions
normal to growth, we have α1 = α2 = α which together with the incompressibility
condition, det(A) = 1, implies α3 = 1/α2 . The rigid boundary condition along the
e3 -axis gives λ3 = 1, that is, α = γ.
The Cauchy equations for the Cauchy tensor [T] = diag(t1 , t2 , t3 ) are identically
satisfied since the deformation is homogeneous. The constitutive equations simplify
to
t1 = αW1 (α, α, 1/α2 ) − p, (13.38)

t2 = αW2 (α, α, 1/α ) − p,
2
(13.39)
Figure 13.1 Growth of a e2

block constrained between e1
two plates. Stress will e3
develop in the growth
process.
2
l=L0
386 13 Balance Laws
1
t3 = W3 (α, α, 1/α2 ) − p, (13.40)
α2
where Wi = ∂αi W . Since t1 and t2 are constant and vanish at the boundary, we have
p = αW2 (α, α, 1/α2 ) and the stress generated during growth is
1
t3 = W3 (γ, γ, 1/γ 2 ) − γW2 (γ, γ, 1/γ 2 ). (13.41)
γ2
(γ) = W (γ, γ, 1/γ 2 ), we obtain

If we introduce the auxiliary function W
γ
t 3 = − ∂γ W . (13.42)
2
13.8.2 Two Growing Cuboids
Next, consider a block made out of two layers characterized by a strain-energy func-
tion W± and growing independently with γ3,± = γ± as sketched in Figure 13.2. A
similar example has been considered in [423]. Initially, the layers have the same
width, height, and length and they can slide along each other apart in the e3 -direction
where they are constrained by two lubricated platens that can move with the defor-
mation. The problem is then to find the stretch ζ = λ3 of this sandwich structure.
Following the same steps as above but without the restriction ζ = 1, we find that the
axial stress in each layer is
α

t3,± = − ± √
∂α W . (13.43)
2 α= γ± /ζ
Figure 13.2 Growth of a e2

two-layer block. Each block
e1
has different growth and is
attached to the other one so e3
that their lengths remain
equal in the current
configuration.
l= L0
tz
30
20
10
0.5 1.0 1.5 2.0
-10
-20
Figure 13.3 Stress-strain curves for the growing structure shown in Figure 13.2. Here, both strain-
energy functions are neo-Hookean with μ± = 1 and γ = γ+ , γ− = 1.
The total stress exerted on the platens is given by
α+
2
t3,+ + α−
2
t3,−
tz = . (13.44)
α+ + α−
2 2
Examples of the total stress tz as a function of ζ are shown in Figure 13.3 where it is
observed that a residual stress is obtained for tz = 0. The extension of the structure
in the absence of stress tz = 0 leads to the condition
γ+ t3,+ + γ− t3,− = 0, (13.45)
that can be solved for ζ for a given strain-energy function. In this deformation the
cuboids are residually stressed due to their connection with the platens, one is in
compression and the other one is in tension.
We can generalize this problem to a block with a continuous growth distribution
γ = γ(Z 0 ) in the e3 -direction and such that stresses are transmitted only through
the interaction of the platens, that is, by neglecting all possible shear stresses in the
block. In this case, the same steps lead to the following equation for ζ

0=
dZ 0 γ 3 ∂α W √ . (13.46)
α=γ/ ζ
388 13 Balance Laws
13.8.3 A Growing Ring
A detailed analysis of the growth of a cylindrical shell is given in Chapter 16. Here, we
consider the simpler case of the incompressible isotropic growing ring. The growing
ring is a section of a cylindrical shell (a tube) that is only allowed to deform in a
plane perpendicular to its axis. We assume that it can only grow in the radial and hoop
directions, as depicted in Figure 12.8. In this simple case, assuming that there is no
deformation in the axial direction, the deformation from the initial configuration B0
to the current configuration B is given by the specialization of (11.196) with ζ = 1.
This deformation defines, in cylindrical coordinates, the deformation gradient
[F] = diag(r (R0 ), r/R0 , 1). (13.47)
The elastic strain and the growth tensors are assumed to be diagonal:
[A] = diag(αr , αθ , 1), [G] = diag(γr (R0 ), γθ (R0 ), 1). (13.48)
As described in Section 12.4, γr corresponds to radial growth, and γθ to hoop growth.

Since F = AG, we have
r r
αr = , αθ = . (13.49)
γr R0 γθ
Material incompressibility implies det A = 1, that is, αr αθ = 1 which implies r dr =

R0 g(R0 )dR0 , that is, R0
r 2 = a2 + 2 g(ρ)ρ dρ, (13.50)
A0
where g(R0 ) = det G = γr γθ .

In the reference configuration, the equilibrium conditions have not been modified
from those derived in Section 11.8.3. They are still given by the only non-vanishing
Cauchy equation
dtr 1
+ (tr − tθ ) = 0, (13.51)
dr r
and the applications of the boundary condition
b
tθ − tr
dr = P, (13.52)
a r
where P is the internal pressure. The difficulty here is that the bounds of the integrals
a and b are functions of the growth terms γi . It is therefore simpler to reformulate
these integrals in the initial configuration. We first define
R0
tθ − tr
τ (A0 , R0 ; a) ≡ g(R0 )R0 dR0 , (13.53)
A0 r 2 (R0 )
where r (R0 ) is given by (13.50). The inner radius a can then be found to find as a
function of P by solving
τ (A0 , B0 ; a) = P. (13.54)
We use the constitutive law

∂W ∂W
tr = αr − p, tθ = αθ − p, (13.55)
∂αr ∂αθ
and the relationship

R0 γθ r (R0 )
αr = , αθ = , (13.56)
r (R0 ) R0 γθ
so that after using (13.55–13.56), Equation (13.54) is a definite integral on R0 only,

with a single unknown constant a that can be obtained for a given pressure P. Once
a is known, the radial and hoop stresses are obtained as
tr (R0 ) = −P + τ (A0 , R0 ; a), (13.57)

∂W ∂W
tθ (R0 ) = tr (R0 ) + αθ − αr . (13.58)
∂αθ ∂αr
13.8.3.1 Homogeneous but Anisotropic Growth
We consider first the case where γr and γθ are both constant as described in
Section 12.4 and illustrated again in Figure 13.4. For illustration purposes, we assume
that the material is neo-Hookean: It is characterized by the strain-energy function
μ 2
W = (α + αθ2 − 2), (13.59)
2 r
with μ > 0. The first problem is to compute the stress profile in the absence of load.
This is done by explicitly integrating (13.57) and solving tr (A0 ) = 0 for the unknown
a when P = 0 (Figure 13.4).
If growth is chosen to be transversely isotropic (γr = γθ ), no residual stress is
created and the new ring is a dilation of the original one, that is, r = γr R0 . If, how-
ever, we consider anisotropic homogeneous growth (γr = γθ ), then residual stress
is created as shown in Figure 13.5A where excess of radial growth with respect to
hoop growth induces a compressive residual radial stress in the material. In terms
of the hoop stress, the inner radius is in compression while the outer radius is in
tension. Circumferential growth, or equivalently radial resorption, creates a tensile
radial stress (not shown).
390 13 Balance Laws
Figure 13.4 Growth of ring

elements. A ring sector
remains a ring sector when
either grown through hoop
and/or radial growth.
Next, we consider hoop stress in the material when P < 0 (internal pressure) as
shown in Figure 13.5B. In the absence of growth, one observes a steep profile as a
function of R0 , whereas with suitable growth this profile flattens.
It was this simple, but crucial, observation that provided the first indication that
residual stress could play an important role in physiology. Indeed, in arteries, the
residual stress acts in the same way suggesting that differential growth in arteries
might take place in order to reduce gradients of hoop stress preventing material
failure associated with tissue separation [22, 208, 406].
The residual stress created through this homogeneous growth process can be
visualized by cutting the ring, creating either an overlap (in the case of compressive
hoop stress) or an opening (in the case of tensile hoop stress) as shown in Figure 12.9.
Both cases lead to locally compatible strain tensors as discussed in Section 12.5.1.2.
Therefore, these stress fields could have been computed by considering a sector of a
ring, and by computing the stresses developed in the deformation necessary to close it
into a ring. Nevertheless, the decomposition of the deformation into a growth process
and an elastic process allows to obtain systematically the stresses and deformations
generated through growth, within a single modeling framework.
A. t
0.4
0.2
R0
1.2 1.4 1.6 1.8 2.0

- 0.2
t
r
- 0.4
- 0.6
- 0.8
B. t
0.28
0.26 no growth
0.24
with growth, r=1.037
0.22
0.20
0.18 R
1.2 1.4 1.6 1.8 2.0
Figure 13.5 Residual stress due to growth. A. In the absence of loads (P = 0, γr = 3, γθ = 2),
radial and hoop growth result in a compressive radial stress and non-vanishing hoop stress. B. Under
pressure (P = 1/10) and in the absence of residual stress (γr = γθ = 1), large hoop stress gradients
are observed. With appropriate growth (γr = 1.037, γθ = 1), the hoop stress profile flattens out. In
both cases μ = 1 is chosen without loss of generality.
13.8.3.2 Non-homogeneous but Isotropic Growth
We now consider the possibility that the growth deformation is locally incompatible
by choosing an isotropic but homogeneous growth profile: γθ = γr = γ with
γb − γa
γ= (R0 − A) + γb , γa,b > 0. (13.60)
B0 − A0
This choice corresponds to a simple model for the growth of a tissue where growth
of a material element decays (or increases) linearly with respect to the initial radius.
In this case, condition (12.47) is not satisfied and the residual stresses generated by
the growth process could not be removed by a finite number of cuts. The computation
392 13 Balance Laws
b> a
1.0
t
0.5
tr
R
1.2 1.4 1.6 1.8 2.0
0.5
Figure 13.6 Homogeneous, non-uniform growth in a ring of radii A0 = 1, B0 = 2. Here, growth

is larger on the outer surface γa = 1, γb = 1.2.
proceeds as before by solving τ (A0 , B0 ; a) = 0 for a and computing the stresses. As

expected, this growth process for γb = γa creates non-vanishing residual stresses as
shown in Figures 13.6 and 13.7.
13.9 Mixture Models
Most of the emphasis in this monograph is on the growth and remodeling of a con-
tinuum composed of a single constituent. An example of an additional constituent
is the fiber contribution, included as a single anisotropic contribution in the the-
ory of morphoelastic materials. However, biological tissues and organs are much
more complicated structures. The load-bearing components are composed of mul-
tiple proteins bathed in biological fluids subject to multiple diffusing fields such as
small ions, oxygen, and glucose. Growth and remodeling is caused by the production
and removal of constituents and are the direct consequences of biochemical reactions
[409]. Therefore, if we wish to understand the relative effect of different constituents
in the growth and remodeling of tissues, we need to take into account the contribution
of all these constituents. We briefly review some mixture models.
13.9 Mixture Models 393
a> b
0.04
t
0.02 tr
R
1.2 1.4 1.6 1.8 2.0
0.02
0.04
0.06
0.08
Figure 13.7 Homogeneous, non-uniform growth in a ring of radii A0 = 1, B0 = 2. Here, growth

is larger on the inner surface γa = 1.2, γb = 1. Note the complex stress profile created by the
incompatible growth.
13.9.1 Classical Mixture Theory
At the continuum level, a material composed of different constituents can be modeled

by the theory of mixtures. This theory finds its origin in classical work of poroelasticity
and multiple diffusion by Fick and Darcy and has been formalized by Truesdell [1250,
1252]. A mixture is seen as the weighted superposition of different constituents, all
defined at each point of the domain. To build a theory, Truesdell uses three basic
principles [538]:
1. All properties of the mixture must be mathematical consequences of the properties
of the constituents;
2. So as to describe the motion of a constituent, we may virtually isolate it from the
rest of the mixture, provided we allow properly for the actions of other constituents
upon it;
3. The motion of the mixture is governed by the same equations as a single body.
This last assumption is important as it states that, even though the structure is hetero-
geneous at the microscale, it should be considered as a homogeneous material at the
continuum level. In particular the mixture respects the traditional balance equations
394 13 Balance Laws
for mass, linear and angular momenta, and energy. Therefore under a body force b,
it must obey
∂ρ
+ div (ρv) = ργ, (13.61)
∂t
div T + ρb = v̇, (13.62)
where, as before, ρ is the mass density of the homogenized mixture, v its velocity,
and T the Cauchy stress tensor. If we consider an open system, we allow for the possi-
bility of an overall mass increase or decrease through γ as described in Section 13.2.
Similarly, the N constituents of the mixture obey the same balance laws individually.
Using a Greek superscript to denote each component, we have [638]:
∂ρα
+ div (ρα vα ) = ρα γ α , α = 1, . . . , N , (13.63)
∂t
div Tα + ρα (bα − v̇α ) = pα , α = 1, . . . , N , (13.64)
where for each constituent, ρα is the density, vα denotes the velocity, and bα the body
force acting on the constituents.
The exchange of mass and linear momentum between the different constituents
is contained in the term γ α and pα , respectively. These terms must be specified
constitutively. Note here that the upper dot denotes the material derivative taken with
respect to time as measured by an observer moving with constituent α [636],

ρ= ρα , ρv = ρα vα , T= Tα . (13.65)
α α α
Despite great advances in the thermodynamics and mechanics of mixtures [118,

346, 503], a number of difficult fundamental and computational problems remain.
For instance, it is still not clear that the total stress can be simply decomposed into
the sum of stresses or, rather, weighted by the volume fraction [340]. Similarly, there
is no definite way to split the boundary tractions between the different constituents:
Without detailed knowledge of the constituents, how should the external load be dis-
tributed among constituents? Another key issue is to specify the relative contribution
of each constituent to mass and linear momentum exchanges. Without a model for
the microstructure, there is no systematic way to derive constitutive laws for the inter-
action between constituents. While a general theory is still lacking, there have been
a number of successful applications of the theory to soft tissues and other biological
problems.
The simplest two-constituent models for soft tissue incorporate just solid and fluid
constituents: a porous, elastic solid matrix (representing the extracellular matrix and
cells), and an interpenetrating fluid phase (representing interstitial fluid). When the
solid constituent is elastic, there are two formulations of these models: poroelas-
tic and mixture-based. Poroelastic models were first developed in the field of soil
mechanics and in particular by Biot and Terzaghi [101]. These poroelastic models
were originally derived from constitutive assumptions about the coupling between
applied stress, pore pressure, and volume changes for a fluid filled, porous, elastic
medium, but can be obtained either through homogenization theory [151, 1027], or
from mixture theory by assuming that the mixture is saturated and both solid and
fluid constituents are intrinsically incompressible [1165].
Two-constituent models were developed to model the behavior of articular carti-
lage [245, 596, 821, 919] and gel-like tissues [595]. An important generalization of
these models takes into account the contribution of charged electrolytes in dilute con-
centrations, leading to three or more constituents [519, 650, 749, 841, 1201]. Such
models are referred to as triphasic or quadriphasic models. They typically include
an extra net charge attached to the solid phase, a fixed charge density, which may
be relevant to biological tissues when the extracellular matrix can block the motion
of large negatively charged biomolecules as sketched in Figure 13.8. These models
allow for the coupling between tissue deformation, fluid flux, and solute movement
due to both osmotic and hydrostatic pressure gradients. The case of a mixture with
a single growing solid phase with application to cartilages has been considered by
Klisch and co-workers [707, 708, 710, 711].
The original triphasic model [749] considered a mixture of a linear elastic solid,
a saturating fluid and species of positive and negative ions. To close the model,
constitutive laws must be chosen for the drag between constituents; typically solid-
fluid drag is assumed to follow Darcy’s law and fluid-solute drag satisfies Fickian
diffusion while other interactions are neglected. Apart from cartilages, biphasic,
triphasic, and quadriphasic models have also been used to investigate the swelling
of brain tissues [297, 327, 762, 763, 931].
Figure 13.8 A quadriphasic

model is composed of a solid
phase, fluid phase, and
solutes (positive (+) and
negative (−) ion species, and
non-permeating solutes (n)).
Some components of the
solid phase have a net
negative charge (fixed charge
density), that induces ion
concentrations to satisfy
charge neutralization
396 13 Balance Laws
In tissue engineering, a typical method to grow new tissue is to use a porous scaf-
fold seeded with cells and continuously fed with nutrients. Assuming that the stresses
generated by the solid constituents are sufficiently small, the theory of mixtures (usu-
ally referred to as multiphase modeling) can be used to model the overall growth of the
composite by coupling a fluid and cells constituents [971]. These reduced models
naturally lend themselves to analytic studies and computational simulations [972,
1279]. The same approach has been used to model tumor [157, 1099] and bone
growth [781].
13.9.2 Constrained Mixture Theory
The previous applications mostly focus on the fluid and solute constituents flowing
and diffusing through a solid matrix. In many soft tissues the solid, load-bearing,
structure is itself composed of different constituents. For instance, it is known that
the response of arteries under loads is due to muscle cells, elastin proteins, and
different collagen molecules. If we want to know how residual stress arises through
the interaction, production, and removal of these different components, they must
be followed individually. For solid constituents, an additional problem arises: Each
new constituent may have a different reference configuration. As new fibers are
added to the system, they may be added either without stresses or with a certain
level of stress to match the existing stress in the system. Therefore, the reference
configuration of a certain fraction of a constituent added at a certain time must be
tracked through the evolution of the system, as sketched in Figure 13.9. Essentially,
a multiplicative decomposition of the deformation gradient for each constituent and
for each time must be performed. However, since the focus is now mostly on the
solid phase, the theory of mixture can be simplified by assuming that all solid phases
have the same velocity. This constrained mixture theory, put forward by Humphrey,
Rajagopal and co-workers [638–641, 1271], is particularly relevant to explore growth
and remodeling in collagen-rich tissues in which there is a continual turnover of
collagen, such as arteries, ligaments, and tendons [447, 448]. One of the difficulties
of this approach is that it becomes rapidly analytically intractable as even the simplest
one-dimensional examples require a full numerical approach and dedicated finite-
element methods must be developed [255].
From a mathematical point of view, the constrained mixture theory can be seen
as the generalization of the multiplicative decomposition to a case where the solid
is composed of a finite number of constituents each described by a different decom-
position at each point in the domain and at each time in the evolution of the system.
The overall stress at any given point is therefore not only a sum over all constituents
but also an integral over the entire history of each constituent weighted by a kernel
that describes the removal and addition of material with a given state of stress [1055,
1056]. Therefore, mixture theory is not an alternative to multiplicative decomposi-
tion but a generalization to multiple evolving constituents, each of which must be
described by a different reference configuration for each time. If the rules of pro-
Constituent 1
Constituent 2
Constituent N
Figure 13.9 In a constrained mixture theory, each solid constituent has a different reference config-
uration at each time τ at which they are added to the system. For growth and remodeling, constrained
mixture theory accounts for different properties of the different constituents.
duction and removal of each constituents, their interactions, and the level of stress
at which they are produced can be fully elucidated through dedicated experiments,
constrained mixture theory holds the promise to characterize fully the intricate rela-
tionships between constituents that leads to homeostasis and pathology. The devel-
opment of such a theory represents a formidable challenge as these rules are typically
not known even for a single constituent.
Chapter 14
Evolution Laws and Stability
The growth tensor adds a kinematic descriptor to the theory of elasticity. Therefore,
it requires a corresponding set of constitutive laws. Since, we think of growth as
a dynamical process, we refer to these laws as evolution laws. The prescription
of evolution laws for growth, in particular in their interaction with the physical
environment, is a fundamental problem of biology and, mostly an open field of
research. Within our framework, the question is to specify the dynamics of the growth
tensor G in terms of other biological or physical quantities.
These laws could be dictated by the system’s ability to provide new material, be
a function of the available nutrient, or be mediated by stresses acting on the system.
Formally, we can write an equation for the relative growth rate:
ĠG−1 = G(T, F, G, μ; t, x, X0 ), (14.1)
where the function G is a function of the stress tensor, the geometric deformation
F, the growth tensor G itself, the initial or current position, or any other chemical,
biochemical, or physical fields μ.
Before we study the properties of the growth tensor and discuss simple growth
laws, it is of interest to review the various proposals for the functional form of the
relative growth rate.

400 14 Evolution Laws and Stability
• Constant growth. The simplest choice for the tensorial function G is to take
it as a constant in time. If we further assume that it is constant in space in a
given coordinate system, analytical results can be obtained [80, 196, 709] and the
residual stress can be computed. In the theory of material defects, the choice of a
constant tensor G is equivalent to the problem of specifying constant eigenstrains
[925, 1356, 1357]. These pre-strains are also known in the literature as initial
strains [722], inherent strains [1266], and transformation strains [348].
• Position dependent growth. Many growth processes depend on the location in the
material. This effect is sometimes referred to as differential growth to indicate that
some parts of a tissue grow faster than others. In morphoelasticity, it implies that G
is a function of either X0 or x [463, 464]. Both situations are of interest. In the first
case, growth is a function of the material position X0 in the initial configuration
and this dependence assumes that the material is made out of points that grow at
different rates and keep growing differentially as time goes by. In the second case,
the ability of a group of cells to grow depends on its location x in space at any
given time. This is the case, for instance, when cell reproduction depends on the
availability of nutrients that diffuse through the boundary. At any given time, the
amount of nutrients may be described by the distance to the boundary as in the
case in the spheroid growth studied in Section 4.3.2.
• Stress-dependence. As discussed in the main Introduction, it has been recognized

experimentally and theoretically in many systems, such as aorta, muscles and
bones, that one of the main biomechanical regulators of growth is stress [409,
627, 1054, 1091, 1206, 1208, 1212]. It has even been suggested that stresses on
cell walls play the role of a pacemaker for the collective regulation of tissue growth
[1154]. Accordingly, in these systems, the growth rate tensor should be a function
of the Cauchy stress tensor, which could also vary according to the position of
tissue elements in the reference configuration. The study of the coupling between
stress and growth remains largely unexplored and only simple functional forms
have been proposed [409].
Before proposing possible forms for the evolution laws, we discus first the basic
properties of the growth tensor.
14.1 Symmetry of the Growth Tensor
In many problems, it is useful to restrict the form of the growth tensor. Similar
to the decomposition used to describe material symmetries, we can use different
tensorial forms to describe growth processes occurring with a prescribed symmetry.
We first restrict our attention to the case of isotropic materials with a strain-energy
function W = W (A). There is an ambiguity in the multiplicative decomposition: If
F is known, the elastic and growth tensors A and G are not uniquely prescribed.
Indeed, if A and G are such that F = AG then the tensors
14.1 Symmetry of the Growth Tensor 401
Ã = AR, and G̃ = RT G, R ∈ O(3) (14.2)
also provide a possible decomposition since F = ÃG̃.

Using the polar decomposition theorem discussed in Section 11.1.7, we can
decompose G = RG UG where UG is a symmetric tensor and RG is a proper orthog-
onal tensor. For isotropic materials, the energy density satisfies W (A) = W (AR) for
any proper orthogonal tensor R, and we have
W (A) = W (FG−1 )
= W (FU−1
G RG )
T
= W (FUG ). (14.3)
Therefore, we conclude that W (FG−1 ) = W (FUG ). That is, the elastic energy of a
growing isotropic material only depends on the symmetric part of the growth tensor.
Note that this property does not hold for anisotropic materials. In that case, even if
the deformation gradient can be decomposed into an elastic deformation tensor and
a symmetric growth tensor, the elastic energy will depend on the particular choice
of the rotation matrix.
The non-uniqueness of the multiplicative decomposition has been discussed at
length in the context of finite strain plasticity [176, 933] and is often cited as an argu-
ment against using this approach for the modeling of physical phenomena. Instead,
we view this ambiguity as a possible gauge that allows us to restrict the form of
the growth tensor by choosing it to be symmetric [585]. Within this class, we fur-
ther characterize different types of growth laws [733, 811, 873], following similar
descriptions in elasto-plasticity [120].
• Orthotropic growth. A rather useful and general way to describe growth is to
identify two directions in which growth can take place differentially. That is, in
the initial configuration, we identify two unit vectors γ 1 and γ 2 and introduce the
particular, but rather large, class of growth tensors
G = g0 1 + (g1 − 1)γ 1 ⊗ γ 1 + (g2 − 1)γ 2 ⊗ γ 2 , (14.4)
where g0 represents the isotropic contribution to the growth process and g1 , g2

its anisotropic contributions. Note that, in this description, we are using the same
vectors γ i in both the initial and virtual states. The advantage of such a represen-
tation is that it reduces the number of growth descriptors to three functions for the
increase in volume and two unit vectors for the directions in which growth takes
place anisotropically. For semi-inverse problems with a diagonal deformation gra-
dient in a given set of coordinates, this particular form is suitable for analytic
treatment. Indeed, a diagonal growth tensor in the same set of coordinates leads
to an anelastic system that can be solved explicitly for all isotropic strain-energy
functions.
• Transversely isotropic growth. In transverse isotropic growth, we assume that,

locally, growth takes place isotropically in the directions normal to a unit vector
γ so that
G = g0 1 + (g1 − 1)γ ⊗ γ. (14.5)
The growth stretch in the direction γ is obtained by applying the growth tensor on
γ, that is,
Gγ = (g0 + g1 − 1)γ. (14.6)
This growth stretch shows explicitly the contribution of both isotropic and
anisotropic expansions in that direction, while the growth stretch in a direction
normal to γ is just g0 . This description of growth is particularly useful for fiber-
reinforced systems with growth along the fibers. In this case, γ can be chosen to
coincide with the fiber direction. The determinant of G is
JG = g02 (g1 + g0 − 1), (14.7)
and, with the prescription JG = 0, its inverse can be directly computed from the
Sherman–Morrison formula [1146]:
(g1 − 1)
G−1 = g0−1 1 − γ ⊗ γ. (14.8)
g0 (g1 + g0 − 1)
• Pure fiber growth. If we further restrict growth along a single fiber we obtain
G = 1 + (g − 1)γ ⊗ γ, (14.9)
and following the previous argument we can identify γ with the growth stretch in
the γ-direction. This description of growth is particularly useful for describe the
growth of cardiac and skeletal muscle fibers [449, 450].
• Area growth. The particular case of in-plane growth obtained as a reduction

of (14.5) is interesting as it provides a simple characterization for problems of
growing plates and membranes [591, 1014] and is given by
√ √
G= g1 + (1 − g)γ ⊗ γ. (14.10)
By taking g different from the identity only in a small layer near the boundary and
γ as the normal vector to the boundary, this particular form for the growth tensor
is a suitable model for growing boundaries.
• Isotropic growth. The simplest nontrivial form for the growth tensor is to take it
as a multiple of the identity, that is,
G = g1, (14.11)
and we have JG = g 3 representing the isotropic change of a volume element.

In this case, the growth function g can be related directly to the growth rate γ
through (13.9), which for constant density reads simply
14.1 Symmetry of the Growth Tensor 403
γ
ġ = g. (14.12)
3
We further study this special case in connection to gel swelling in the next section.
14.2 Isotropic Growth and Gel Swelling
The simplest growing materials are polymeric gels. Polymeric hydrogels are cross-
linked, hydrophilic polymer networks. When mixed with water, they can exhibit large
volumetric swelling in response to different chemical and physical stimuli. These
gels can be used as microfluidic actuators as well as scaffolds for engineered tissues
[325, 581, 999]. They have also been used to study pattern formation in growing
and shrinking materials as they display a variety of interesting shapes following
buckling-type instabilities [115, 116, 839, 1220, 1221]. Most gels are isotropic
due to the random nature of the network and, as a material, they behave elastically
before rupture. They are also mostly incompressible due to their high water content.
Despite the fact that they exhibit unusual properties compared to elastomers, such as
a negative Poynting effect [663], they can be modeled as incompressible hyperelastic
materials when viscous effects are neglected [888].
Upon addition of water, gels can swell and a theory of growth is appropriate
for their description. However, since both growth and materials can be modeled
as isotropic, the kinematics of swelling can be further simplified as first proposed
by Flory [369]. Indeed, the machinery of the multiplicative decomposition becomes
superfluous in the sense that the effect of growth can be completely taken into account
through a proper definition of the determinant of F and a rescaling of the strain-energy
function [288, 610, 1023, 1024, 1343]. The key is to realize that since G = g1,
the deformation gradient F can be written as F = gA. Assuming that the gel is
hyperelastic and incompressible, a suitable choice for W is then
W S (A, g) = m(g)W1 (A), (14.13)
with m(g) > 0 and m(1) = 1 so that W S (A, 1) = W1 (A) represents the strain-energy
function of a gel before swelling. The function m(g) takes into account the remodeling
of the gel during swelling. If the same material is added during swelling then m(g) =
1. However, if swelling is due to the addition of water, the elastic energy density, per
unit unstressed volume, decreases as water cannot store elastic energy and we have
m(g) < 1. A simple modeling choice for this effect is
m(g) = g q , (14.14)
where q < 0 is chosen to model softening during swelling. Therefore, the constitutive
law (13.31) for this material can now be written
∂W1 (A)
T = m(g)A − p1
∂A
∂ ∂F
= m(g)(g −1 F) W1 (A(F)) − p1
∂F ∂A
∂
= m(g)F W1 (g −1 F) − p1
∂F
∂
= F W S (g −1 F, g) − p1. (14.15)
∂F
This last expression is now only a function of F and g. No further reference to the
virtual state is needed.
For example, if the strain-energy function of the material before swelling is given
in terms of its invariant W1 = W (I1 , I2 ), then we have
W S (g −1 F, g) = m(g)W (g −2 I1 , g −4 I2 ), (14.16)
and the explicit expression for the Cauchy stress (11.156) in terms of the derivatives
of W with respect to the invariants is still valid (with J = 1). Note that in our case,
the unstressed state is the free isotropic swollen state given by F = g1. With this new
constitutive law, all standard equations given in Section 11.4 still apply and can be
solved for any given geometry and strain-energy functions.
14.3 Discrete Growth Steps
The decomposition F = AG is clear as a one-step process, but proper care must

be taken when looking at the cumulative effect of growth and long-time evolution
[706]. Assume for instance that growth takes place in time in such a way that it
depends on the state of stress in the current configuration. Then the tensor G(X0 , t)
depends continuously on the configuration Bt at time t. Moreover the growth tensor
is defined from an unstressed configuration B0 , but it takes values that depend on the
stressed configuration Bt . The virtual state also depends on time and evolves with the
current configuration according to Equation (14.1). Following the original proposal
of Rodriguez et al. [1091], the evolution of growth can be understood by discretizing
the evolution law forward in time to obtain a sequence of growth increments
G(t + δt) = G(t) + δt G(T, F, G, μ; t, x, X0 ), (14.17)
where all arguments in the left hand side of this relation are evaluated at time t.
That is, initially, the growth tensor depends on the configuration B0 , and we define
G1 = G(δt), corresponding to a small time increment δt. Given G1 , we compute
the residual stress field and elastic deformations due to external loading to obtain a
geometric deformation tensor F1 mapping the initial configuration B0 to B1 . From
the configuration B1 , we define a new growth tensor
14.3 Discrete Growth Steps 405
F
F2
F1
B0 B1 B2 B
G1 A1
A2
V1 A
G2
V2
V
G
Figure 14.1 Evolution of the virtual grown state. The reference configuration evolves incrementally
with time. To compute G2 , we need the information contained in the state V2 and B2 as indicated
by the dashed lines.
G2 = G(2δt) = G1 + δt G. (14.18)
Once the virtual state V2 is obtained, a new elastic deformation tensor A2 can be
computed. The process is iterated as shown in Figure 14.1. The cumulative effect of
all growth increments can be computed to obtain the overall growth process.
We see from this incremental decomposition that the growth law gives the evolu-
tion of the reference configuration by updating the growth tensor based on the values
of the stress and geometric deformation tensors. Through this process, the continuous
limit is well defined. Note that at each incremental time, a nontrivial boundary-value
problem needs to be solved. Then, the geometric deformation tensor is known and
the reference configuration updated before a new boundary-value problem is solved.
However, in the case where the boundary-value problem is sufficiently simple further
progress can be achieved. In this case, one can use directly the differential equation
to obtain a continuous description of growth.
14.4 The Thermodynamics Perspective
The problem of determining a suitable form for the growth law remains. The general
approach in rational thermodynamics to find restrictions on evolution laws is to use
the Clausius–Duhem inequality to constrain the functional form of those laws. The
standard procedure, known as the Coleman–Noll procedure [222], has been used effi-
ciently to derive constitutive laws for systems with coupled fields such as the theory
of thermoelasticity [526]. Nevertheless, as is the case with most fundamental ther-
modynamics concepts, the use and interpretation of the Clausius–Duhem inequality

remain topics of debate in the rational mechanics literature [774, 842].
We first recall the form of the Clausius–Duhem inequality (13.28), derived in
Section 13.5, for the free-energy density
1
ρψ̇ ≤ T : L − ρη θ̇ − q · κ + h , (14.19)
θ
where L = ḞF−1 is the velocity gradient tensor, κ = grad θ, and −h is the non-
compliant entropy contribution to the process. We now view the free-energy density
ψ = ψ(A, θ, κ), the entropy density η = η(A, θ, κ), the Cauchy stress tensor T =
T(A, θ, κ) and the heat flux q = q(A, θ, κ) as functions of the elastic tensors A as
well as the temperature θ and temperature gradient κ.
The basic assumption in the Coleman–Noll procedure is that the inequality (14.19)
is expected to hold for all admissible thermodynamic processes, which means that it
must hold for arbitrary deformation gradients and temperature fields. This condition
puts restrictions on the possible functional forms for the fields T and q.
We first expand ψ̇ and (T : L) by using the chain rule and the substitution F = AG
to obtain
∂ψ ∂ψ ∂ψ
ψ̇ (A, θ, κ) = : Ȧ + θ̇ + · κ̇, (14.20)
∂A ∂θ ∂κ
T : L = T : ḞF−1 = (TA−T ) : Ȧ + (AT TA−T ) : (ĠG−1 ). (14.21)
We substitute these two expressions in (14.19) to obtain

∂ −T ∂ψ ∂ψ
ρ − TA : Ȧ + ρ + η θ̇ + ρ · κ̇
∂A ∂θ ∂κ
1
≤ AT TA−T : ĠG−1 − q · κ + h. (14.22)
θ
Since the rate of change Ȧ, θ̇ and κ̇ can vary independently, this inequality must hold
for each term independently, that is,
∂ ∂ψ ∂ψ
T=A , η=− , = 0. (14.23)
∂A ∂θ ∂κ
We recover from this procedure, the thermodynamic connection between entropy
and the free-energy function as well as the independence of the free-energy density
from temperature gradients.
If we introduce W = J A ψ, the internal energy density per unit reference volume,
the first relationship gives the traditional form of the constitutive law
∂W
T = J A−1 A . (14.24)
∂A
14.4 The Thermodynamics Perspective 407
Taking into account (14.23), (14.22) simplifies to
1
0 ≤ AT TA−T : ĠG−1 − q · κ + h. (14.25)
θ
We define from this relationship the stress tensor
M = J A−1 AT TA−T . (14.26)
This stress tensor, known as the Mandel stress tensor, maps vectors in the virtual
state to vectors in the same state. It is therefore a natural descriptor for growth laws
as noted by different authors [22–24, 654]. Note that if A and T are coaxial (that is,
AT = TA) then M = J A−1 T.
If we further neglect temperature gradient by considering those processes with
spatially homogeneous temperature fields, then κ = 0 and
M : G + J A h ≥ 0, (14.27)
where G is the growth law (14.1).

Many authors have emphasized that growth laws should be consistent with the
laws of thermodynamics [22, 516, 654], an indisputable fact since, tautologically, all
physical processes must obey the laws of physics. However, this last relationship does
not provide much information since, within the confines of a macroscopic theory of
open systems, the entropy sink −h is unknown and, a priori, the form of evolution
laws cannot be constrained.
This issue is best illustrated on a simple example. We consider the case of a
growing incompressible cuboid deformed along its axes and only allowed to grow
along a single axis, say e3 , as detailed in Section 13.8.1. Since the Cauchy stress
tensor and the deformation gradients are diagonal, the Mandel stress is equal to
the Cauchy stress tensor. We assume a simple model where growth takes place to
maintain a state of homeostatic tension t3∗ > 0. For tension t3 close to t3∗ , a plausible
model is given by
[Ġ][G−1 ] = diag(0, 0, g3 (t3 − t3∗ )), (14.28)
where g3 > 0 to ensure that the homeostatic stress is stable in time as shown explicitly
in Section 14.6.1.1. The dissipation inequality (14.27), in this case, reads
g3 (t3 − t3∗ )t3 + h ≥ 0. (14.29)
If initially, t3 > t3∗ , then this inequality is satisfied for h = 0. This process is analo-
gous to the physical plastic extension of a bar under tension given in Section 4.4.5.
However, if we start with t3∗ > t3 > 0, then the inequality requires a sink of entropy
h. Does it mean that we should rule out this process? Does it imply that no biolog-
ical system could follow this evolution law? The requirement of the existence of a
non-vanishing entropy sink only implies that a closed physical system could not be
described by such a law. Starting, for instance with fixed end-to-end displacement,
the shortening of a cuboid from its homeostatic state would result in a shrinking of
the cuboid along its axis, a process that would need reorganization of matter at the
microscopic level.
We conclude that the rational thermodynamics approach, apart from recovering
the usual constitutive laws for stress and entropy, cannot restrict the form of evolution
laws for general growth processes. Nevertheless, it allows us to identify the Mandel
stress as a key quantity (or equivalently, the closely related Eshelby stress tensor [22,
1231]) in the formulation of growth laws. It also allows us to describe two types of
growth processes:
• Passive growth processes. During these processes, the dissipation inequality is
satisfied even in the absence of entropy sinks. This situation is typical for physical
systems. It arises, for instance, in plasticity, thermoelasticity, or gel swelling, where
a non-compliant entropy contribution is not required for the process to take place
[1076]. The observed macroscopic response is then slaved to the overall dissipation
mechanism. There could be some active processes involved at the microscopic level
but, in principle, this type of growth is not incompatible with physical processes.
• Active growth processes. In this case, an additional sink of entropy must be

included to satisfy the dissipation inequality. This situation arises in many biolog-
ical systems where the cell, through its genetic information and internal energy con-
tribution, can alter the entropy of a system by forcing a preprogrammed response
to external stimuli, against the physical increase of entropy. Therefore, active
processes at the microscopic level must be at work, and indeed, are fundamental
for the formation of highly organized structures in a dissipative environment.
Detailed analyses of the Clausius–Duhem inequality involving multiple species,
chemical potentials, and higher-order gradients have been proposed by different
authors [22, 341, 422, 424, 804]. It provides a comprehensive thermodynamic pic-
ture of growth and remodeling processes by identifying the different contributions
to energy and entropy and the key players involved in the formulation of constitu-
tive laws. Unfortunately, despite many claims, the thermodynamic approach fails to
restrict universally the functional form of evolution laws for growth.
14.5 Phenomenological Laws and Homeostatic Stress
In the absence of fundamental physical principles guiding the formulation of suitable

functional forms for a growth law, we consider plausible phenomenological mod-
els based on general physiological or biological principles. It is important to make
the distinction between developmental and physiological growth processes. During
development, the role of growth is to rapidly expand the mass, mostly by cellular
division, and to shape tissues and organs. During regular physiological functions,
growth is used to maintain the proper mechanical environment within a tissue or an
organ so that it functions adequately in response to external loads.
14.5 Phenomenological Laws and Homeostatic Stress 409
The homeostatic stress assumption states that there exists a particular preferred
value of the stress field that is regulated through growth and remodeling during
regular physiological conditions. Homeostasis, a concept first developed by Walter
Bradford Cannon in 1926 [165, 227], is the ability of a living organism to regulate
its internal state to maintain constancy, such as the ability of mammals to maintain
a stable internal temperature. Within mechanobiology, the concept of a homeostatic
stress has its origin in the early work of Woods, Thoma, Davis, and Wolf on the
heart, arteries, soft tissues, and bones as discussed in Section 2.1. This idea was only
formulated as a generic concept following the study of Fung and Liu on residual
stress in arteries [411, 795].
In our framework, the homeostatic stress assumption implies the existence of a
function of the Cauchy stress tensor T ∗ (T) at which no growth takes place. This
function could be a scalar (the overall pressure for instance), a vector (e.g., a traction
at the boundary), or a tensorial field.
In the simplest case, the homeostatic stress is a tensor T∗ such that, excluding
other stimuli for growth, we have
G(T∗ , F∗ , G∗ , μ; t, x, X0 ) = 0, (14.30)
where G∗ and F∗ are the growth and deformation tensors that brought an initial
configuration to its current state of stress T∗ . We further require that the homeostatic
state must be stable so that under small perturbations growth takes place to recover
the preferred state of stress, that is, for T close enough to T∗ , we have
(T − T∗ ) → 0. (14.31)
t→∞
This condition cannot be easily formulated in terms of a growth law G due to the
complexity of the dynamics. However, close to a state of homeostasis, it is reasonable
to assume that the dynamics can be linearized so that we obtain the linear homeostasis
growth law:

ĠG−1 = K : M − M∗ , (14.32)
where K is a fourth-order tensor and M∗ represents the homeostatic value of the Man-
del stress tensor. Following our discussion on thermodynamics, we have expressed
this law in terms of the Mandel stress tensor M = J A−1 AT TA−T so that all relevant
quantities are expressed in the virtual state. Various forms of this linear growth law
have been used to model growth in tumors [26], arteries [18, 1054, 1212], muscles
[1208], and the heart [1206, 1211].
The concept of homeostasis is central to many aspects of mechanobiology. Yet, it
is not clear that it holds as a general principle. While it seems reasonable and well-
documented in a normal physiological state, the concept may not apply directly in
pathological processes such as tumor growth or wound healing, or in morphogenesis
where the mechanical environment and material properties of the tissues change
rapidly. To this end, it has been proposed that homeostatic stress itself should evolve
in these processes [1057, 1209, 1210, 1344] following some early experimental
work and the theory of hyper-restoration of Beloussov in embryogenesis [76, 77,
79].
14.6 Dynamics of Homogeneous Deformations
It is particularly useful to consider homogeneous elastic and growth deformations to

understand the overall dynamics of growth. We first collect here all relevant equa-
tions describing the dynamics of the morphoelastic process. Assuming that the mass
density remains constant, that the body is in elastic equilibrium at all times and in
the absence of body force, we are looking for solutions x = χ(X, t) for the system
F = Grad χ, (14.33)
F = AG, (14.34)
ĠG−1 = G(T, A, G), (14.35)
T = H(A), (14.36)
div T = 0. (14.37)
These equations are complemented by the initial conditions
χ(t = 0) = 0, (14.38)
F(t = 0) = A(t = 0) = G(t = 0) = 1, (14.39)
T(t = 0) = 0, (14.40)
and the boundary conditions
F (F|∂B , T|∂B ) = 0. (14.41)
Equations (14.33) to (14.37) form a system of 36 partial algebraic-differential equa-

tions to solve for the three components of the vector χ and the nine components
of each of the tensors F, A, G, and the six components of T (due to the symmetry
TT = T). No obvious analytical progress can be made without further assumptions.
We consider the simple, but informative, case where the deformation gradients
are homogeneous. Then, the Cauchy equations (14.37) are trivially satisfied and
the boundary conditions provide a global relationship between the fields. By using
Equations (14.34) and (14.36), all dependence on F and T is removed and the system
can be written as a set of 18 equations for A and G, that is,
G),
Ġ = G(A, (14.42)
0=F (A, G), (14.43)
14.6 Dynamics of Homogeneous Deformations 411
where
G) = G(H(A), A, G)G,
G(A, (14.44)
(A, G) = F (AG, H(A)).
F (14.45)
Therefore, the problem is to solve, for given growth and constitutive laws a set of
algebraic-differential equations.
For an incompressible system, we introduce the extra variable p, so that the system
reads now
G, p),
Ġ = G(A, (14.46)
0=F (A, G, p), (14.47)
0 = det A − 1, (14.48)
where
G, p) = G(H(A, p), A, G, p)G,

G(A, (14.49)
(A, G, p) = F (AG, H(A), p).
F (14.50)
14.6.1 Diagonal Deformations
We further focus on the case where the material is incompressible and the deformation
is diagonal so that, in the Cartesian basis {e1 , e2 , e3 }, we have
[F] = diag(λ1 , λ2 , λ3 ), [A] = diag(α1 , α2 , α3 ), [G] = diag(γ1 , γ2 , γ3 ).

(14.51)
Note that the Cauchy stress tensor is also diagonal in the same basis
[T] = diag(t1 , t2 , t3 ), (14.52)
and therefore equal to the Mandel stress.

Viewed as a dynamical system, the system (14.46–14.48) can be reduced to a sys-
tem for two sets of variables, the vector γ = (γ1 , γ2 , γ3 ) and α = (α1 , α2 , α3 , α4 ),
where α4 = p. Then the algebraic-differential system of seven variable simply reads
γ̇ = g(γ, α), (14.53)

0 = f(γ, α). (14.54)
We assume the existence of a homeostatic state of elastic strain (α∗ , γ ∗ ) associated

with a state of elastic stress t∗ and such that
g(γ ∗ , α∗ ) = 0, f(γ ∗ , α∗ ) = 0. (14.55)
In the simplest case, the stability of this state is obtained by linearizing the equations
around the equilibrium. That is, we write
γ = γ ∗ + εγ, α = α∗ + εα, (14.56)
and expand all equations to first order in ε to obtain
γ̇ = Dγ g γ + Dα g α, (14.57)
0 = Dγ f γ + Dα f α, (14.58)
where the notation

∂g
Dγ g ≡ , (14.59)
∂γ (α∗ ,γ ∗ )
stands for the Jacobian matrix of g with respect to γ evaluated at the equilibrium
point (α∗ , γ ∗ ). Assuming that Dα f is invertible, we have

γ̇ = Dγ g + (Dα g) (Dα f)−1 (Dγ f) γ
≡ K γ. (14.60)
Therefore, it follows from the linear stability theorem [520] that the stability of the
equilibrium (α∗ , γ ∗ ) state depends on the eigenvalues of the matrix K as long as none
of them have zero real part. However, due to underlying symmetries in the system,
typically, some eigenvalues vanish identically and no information can be obtained
from the linear stability theorem. In this case, a reduction of the system to a smaller
dimensional system can often be obtained as we show in the following examples.
14.6.1.1 A One-Dimensional Example
We first revisit the case of a growing rectangular cuboid subject to constant com-
pression along one of its axes and allowed to grow in the two other directions as
originally studied in [1091] and depicted in Figure 14.2. The growth evolution is a
Figure 14.2 Growth of a

block constrained between a
wall and a plate on which a F
constant compressive force is
applied. The block grows in
response to stress in the two
directions normal to the
force.
function of the stress and we assume the existence of a homeostatic stress t3∗ along
the compression e3 -axis. Therefore, the material will grow or resorb along the e1
and e2 directions until equilibrium is restored. We assume that the growth and elastic
deformation tensors are both diagonal and that the material is hyperelastic, incom-
pressible, and under uniaxial loading, so that in the Cartesian basis {e1 , e2 , e3 }, we
have
1 1
[F] = diag(λ, λ, ), [A] = diag(α, α, 2 ), [G] = diag(γ, γ, 1), (14.61)
α 2 α
In components, the Cauchy stress tensor is given by
t1 = αW1 (α, α, 1/α2 ) − p, (14.62)

t2 = αW2 (α, α, 1/α ) − p,
2
(14.63)
1
t3 = 2 W3 (α, α, 1/α2 ) − p, (14.64)
α
where Wi = ∂αi W . We adopt here the evolution law
γ̇ = kγ(t3 − t3∗ ), (14.65)
where t3∗ is a given homeostatic stress. We impose as boundary conditions a com-

pressive force on the faces normal to the e3 direction and no traction on the other
four faces. Since the deformation is homogeneous, we have
t1 (t) = t2 (t) = 0, (14.66)

F = A(t)t3 = A(0)λ1 λ2 t3 = α2 γ 2 t3 , (14.67)
where A(t) is the surface area at time t of the face normal to e3 and A(0) = 1 is chosen
without loss of generality. Note that since t1 (t) = t2 (t) = 0, we have p(t) = αW1
which implies
α
t3 = h(α) = − W (α), (14.68)
2
where W (α) = W (α, α, 1/α2 ). By using the constitutive relationships the equa-
tions for (α, γ) are
γ̇ = kγ(h(α) − t3∗ ), (14.69)

0 = F − γ 2 g(α), (14.70)
where g(α) = α2 h(α). These last equations are a particularly simple example of a
reduction of the general relationships (14.53–14.54). Following the same steps, we
obtain a single linearized equation for γ
g(α∗ )
γ̇ = μγ, μ = −2k(α∗ )2 h (α∗ ) . (14.71)
g (α∗ )
The local dynamics near the fixed state can be directly inferred from this equation.
Assume that F ∗ is chosen so that the equation h(α∗ ) = t3∗ has at least one solution.
This fixed point α∗ is an equilibrium solution that corresponds to a growth

1 F∗
γ∗ = ∗ . (14.72)
α t3∗
The stability of this fixed point is determined by the sign of the linear stability expo-
nent μ. We know from the Baker-Ericksen and the ordered-force inequalities that an
increase in compressive force or stress reduces the length of the bar. These inequali-
ties therefore imply that h(α) > 0 and both h (α) < 0 and g (α) < 0. Therefore, we
conclude that μ < 0 for k > 0 and the equilibrium point is asymptotically stable.
14.6.1.2 Two Growing Beams in Parallel
Next, we revisit the example shown in Figure 13.2 where two beams grow in parallel
and are connected by a platen through which forces balance. We assume that both
beams have a homeostatic stress t±∗ with a growth law
γ̇± = k± γ± (t3,± − t±∗ ), (14.73)
where k± > 0 and α

t3,± = h ± (α± ) = − ±
∂α W , (14.74)
2 α=α±
similar to Equation (13.44). In the absence of external force applied on the platen,
we have
0 = α+
2
t3,+ + α−
2
t3,− , (14.75)
γ+ α−
2
= γ− α+
2
. (14.76)
Depending on the values of t±∗ , we have different dynamics. Let α±

∗
be the stretches
∗ ∗
associated with the homeostatic stress so that h ± (α± ) = t± . In the phase plane
(α+ , α− ), the dynamics is constrained to lie at all time on a curve given by the
force balance
0 = α+2
t3,+ + α−
2
t3,− . (14.77)
We start the dynamics at α± = 1 and consider different cases depending on the

homeostatic stress:
• Case I. First, we assume that the homeostatic stresses are compatible with the
force balance, that is,
∗ 2 ∗ ∗ 2 ∗
(α+ ) t+ + (α− ) t− = 0, (14.78)
as shown, for instance, by the circle I in Figure 14.3. Then, one of the homeostatic
stresses must be compressive and the other one must be tensile. Starting from the
initial state α± (t = 0) = 1, the system evolves monotonically (see arrow): One
beam grows while the other shrinks until the system reaches the equilibrium state.
• Case II. Assume now, that

∗ 2 ∗ ∗ 2 ∗
(α+ ) t+ + (α− ) t− = 0, (14.79)
and t+ t− < 0 as illustrated by the circle II in Figure 14.3. Taking without loss of
generality t−∗ > 0, we have initially t3,− − t−∗ < 0, that is, γ̇− < 0, and t3,+ − t+∗ >
0, that is, γ̇+ > 0 as indicated by the segment IIa in the figure. The minus beam
∗
shrinks until the minus stretch α− crosses α− after which point, both beams grow
∗
(IIb) until the stretch of the plus beam crosses α+ . After this point the minus beam
grows and the plus beam shrinks without bounds (IIc) as the homeostatic balance
can never be established.
∗ 2 ∗ ∗ 2 ∗
• Case III. Assume now, that (α+ ) t+ + (α− ) t− = 0 but t+ t− > 0. Assume that
both homeostatic stresses are tensile. Then initially one beam shrinks while the
other grows (IIIa) until the growing beam reaches its homeostatic stretch after
which both beams will shrink indefinitely (IIIb), unable to achieve a state of home-
ostatic stress. If both homeostatic stresses are compressive, both beams will grow
indefinitely.
This simple example already demonstrates the possible complexity of growth dynam-
ics which includes non-monotonic behavior and unstable dynamics. Clearly, as
growth proceeds without bounds, both the elastic model and the evolution laws
loose their validity and other mechanisms must be at work to prevent infinite growth,
compression, or extension.
A simple evolution law, linear in the stress difference with respect to the homeo-
static stress may only be valid close to that state. Such a law assumes, implicitly, that
the homeostatic stress can be reached. However, mechanical or geometric constraint
may prevent this state to be reached leading to unstable dynamics.
14.6.2 A Two-Dimensional Example
We consider again a growing cuboid constrained to grow and deform along its axes,
such that the deformation and Cauchy stress tensors are diagonal and subject to the
applied compressive forces
Fi = Ai (t)ti = λ j λk ti , i, j, k = 1, 2, 3, i = j < k = i, (14.80)
where Ai (t) is the area of the face perpendicular to the ei –axis and chosen to be equal
to one at time t = 0, that is, Ai (0) = 1, i = 1, 2, 3. We consider here the particular
case where α3 = α2 ≡ α. From the incompressibility condition, we have α1 = α−2 .
We assume that γ3 = γ2 and a growth law for γ1 and γ2 of the form
3.0
IIc
2.5
2.0 II III
1.5
IIb
IIa
1.0 IIIa
IIIb
I I
0.5 IIIb
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Figure 14.3 Dynamics of growing parallel beams shown in Figure 13.2. The curve represents the
force balance that constrains the growth process. At any given time, the system evolves along this
curve. Here, both strain-energy functions are neo-Hookean with μ− /μ+ = 10.

2
γ̇i = γi K i j (t j − t ∗j ), i = 1, 2, (14.81)
j=1
where K is a constant matrix and t1∗ , t2∗ , are prescribed homeostatic stresses. The
constitutive relationships can be written
t1 = − p + α−2 (∂α1 W ), (14.82)

t2 = − p + α(∂α2 W ), (14.83)
so that the two remaining boundary conditions can be written as a single equation
after eliminating p, that is,
α
γ1 F1 − α3 γ2 F2 = γ1 γ22 α2 h, h(α) ≡ − ∂α W , (14.84)
2
where W = W (α−2 , α, α). Assuming that det(K) = 0, the equilibrium solution is

given by

∗ F2 t1∗ 1 F1
h(α ) = t1∗ − t2∗ , γ1∗ ∗ 2
= (α ) ∗ , γ2∗ = ∗ . (14.85)
t2 F1 α t1∗
Close to such an equilibrium point, the dynamics for γ i = γi − γi∗ reduces to

2
γ̇ i = γ i Mi j γ i . (14.86)
j=1
The determinant and trace of the matrix M are given by
α∗ h (α∗ )t1∗ t2∗

det(M) = −2 det(K) , (14.87)
α∗ h (α∗ + 2t1∗ + t2∗
tr(M) =
α∗ h (α∗ )(K 12 +2K 21 + K 22 )+2t1∗ t2∗ (K 11 + K 12 +2K 21 + 2K 22 )
− . (14.88)
α∗ h (α∗ ) + 2t1∗ + t2∗
Interestingly, the position of the equilibrium point depends on the applied force, but
its stability does not. Since we are interested in compressive homeostatic stresses,
we have ti∗ < 0 and h (α∗ ) < 0. Therefore, a necessary condition for stability is
det(K) < 0.
To illustrate the possible dynamics in this system, we further restrict our atten-
tion to the case t1∗ = t2∗ = −1 which implies α∗ = 1. We take h (1) = −6 and the
particular dynamics
1 β
[K] = , (14.89)
0 −1
where β is a control parameter. Independently of the form of the strain-energy func-

tion and the applied forces, the stability of the equilibrium state (γ1∗ , γ2∗ ) is fully
determined by the parameter β:
1 √
For β ≤ (2 − 3 3) (γ1∗ , γ2∗ ) is a stable node;
2
1 √
For (2 − 3 3) < β < 1 (γ1∗ , γ2∗ ) is a stable spiral;
2
For β > 1 (γ1∗ , γ2∗ ) is a unstable equilibrium.
√
The dynamics has a bifurcation point at βcrit = 21 (2 − 3 3) as shown in Figure 14.4.
This simple example demonstrates the complexity of possible dynamics. The
coupling between growth and the nonlinear response of the material implies that
growth dynamics is highly nonlinear.
However, with a particularly simple choice for the evolution laws in terms of the
stresses, this dynamics is simple as it only allows for a single equilibrium point (when
the matrix K is not singular). The stability of such a fixed point and the behavior of the
1.6
=-2 1.6
=-1
1.4 1.4
1.2 1.2
1.0 1.0
0.8 0.8
0.8 1.0 1.2 1.4 1.6 1.8 2.0

0.8 1.0 1.2 1.4 1.6 1.8 2.0
Re( )
Im( ) 2
6 4 2 2 4 6
=3/4
2.0
1.5
1.0
0.5
0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 14.4 Real√ and imaginary parts of the eigenvalues M as a function of β. A bifurcation appears
at βcrit = 21 (2 − 3 3) and stable oscillations are present for values of βcrit < β < 1, indicated by
vertical dashed lines.
solutions close to this point depends nontrivially on the parameters of the system and,
in particular, allows for oscillating dynamics. Oscillations during growth are found
in many biological systems [78, 123, 354, 728, 799]. However, a proper modeling
of the phenomenon of growth oscillation requires the coupling with other important
fields, such as calcium concentration for pollen tubes [188, 360, 590]. The goal of
this example is merely to show the possibility of interesting dynamics related to the
structure of the equations for elastic growth.
14.7 Remodeling
Apart for the addition of mass, an important feature of evolving tissues is their
capacity for remodeling. Remodeling is defined as a change in a tissue or organ
that takes place without addition of mass. An evolution of material stiffness as it
ages is an example of a remodeling process. In this case, the different parameters
appearing in the strain-energy function are allowed to vary over the slow growth
time. The anisotropic response of a material can also evolve in time. For instance,
in many fiber-reinforced tissues, there is a constant renewal of fibers based on the
mechanical environment and the fibers strength and alignment can evolve according
to external stimuli in order to change the effective properties of the material. These
changes include growth, deposition, cleavage of fibers, changes in fiber diameter
and orientation, and cross-linking of fibers [872, 873]. In mammalian tissues, these
processes take place during both development and adulthood, when they either help
maintain proper tissue function or lead to pathologies.
Since remodeling in biological tissues occur in response to mechanical stimuli,
it is believed to be closely related to the optimization of mechanical properties.
For instance, as discussed in Section 2.1.5, Wolff’s law states that bones adapt to
external loads by reinforcement in order to resist these loads [883]. This process can
be modeled by a local change in density depending on the stress. This example of
density growth is accompanied by a local change of volume (hence growth), but it can
also be seen a remodeling process for the entire structure as it affects its anisotropic
response as seen in Figure 14.5. Similarly, it has been suggested that growth in the
arterial wall optimizes its load-bearing functions [1213].
In the particular case of collagen-reinforced tissues, it has been observed in vitro
that fibroblast respond to mechanical loading by changing the orientation of collagen
fibers [320]. This reorientation may arise from either realignment of existing fibers
or collagen turnover as cells are able to synthesize highly aligned collagen [1293].
Mechanically-induced fiber reorientation is believed to play an important role in the
proper function of aortic valves [302–304] and arterial walls [305, 540, 736].
Within our framework, remodeling can take place at two different levels. First,
it can appear kinematically by considering growth processes whose deformation
tensor G does not change volume, that is det (G) = 1. It corresponds locally to a
reorganization of mass in some preferred directions. This remodeling process is a
particular case of the evolution laws discussed in this chapter and is described by
2 1
3
1 3
2
Figure 14.5 Illustration of remodeling in bone through density growth. Left: Illustration from
Wolff (1892). Right: Computation of bone evolution [734] (picture courtesy of Ellen Kuhl).
either the evolution of the tensors G as given by Equation (14.1) or, equivalently, by
an equation for the evolution of the growth directions γ i appearing in the description
of the growth tensors given in Section 14.1. Since a growth direction γ is prescribed
by a unit vector, its evolution in time is fully determined by a spin vector ϕ such that
∂γ
= ϕ × γ. (14.90)
∂t
Second, in an anisotropic material, remodeling can take place to change the direc-
tions of anisotropy. For instance, in a fiber-reinforced material, the reference fiber
direction M can evolve in time as a function of the applied loads or strains. Again,
since M is a unit vector, its evolution is fully specified by a spin vector ω such that
∂M
= ω × M. (14.91)
∂t
In both instances of remodeling, the central problem is to find a suitable law for
the rotation vectors ϕ and ω as a function of stresses or strains. A standard approach
is to assume that fiber reorientation follows either the directions of positive principal
strains [302], maximum principal stretch, maximum principal stress [577, 735, 870],
or a direction that depends on the stretch or stress [304, 305, 540, 736]. In the case
of a system with fiber dispersion, the evolution of the probability density function
for fiber alignment also needs to be specified [303, 304].
14.7 Remodeling 421
A simple evolution law [870] that is consistent with the realignment of a fiber in
the direction M in a given direction N is
∂M 1
= (N − (N · M)M) , (14.92)
∂t τ
where both M and N are taken in the reference configuration and τ is the typical
relaxation time. Since M is initially a unit vector, we verify that M · ∂t M = 0, that
is, M remains a unit vector during the dynamics. This formulation is equivalent
to (14.91) with
1
ω = M × N. (14.93)
τ
In a theory of maximum stretch remodeling, we choose N to be the direction of largest
stretch, then (14.92) implies that the fibers reinforce the most stretched direction by
reorienting toward it and that M → N as t → +∞. In that limit, the stress and strain
tensors are coaxial, which is a necessary condition for the strain-energy to reach an
extremum [577, 1280]. Note that, if the maximum principal stretch is not unique, it
is assumed that ∂t M = 0. An example of a similar strain-driven fiber reorientation
embedded in a three-dimensional model is shown in Figure 14.6.
14.7.1 Fiber Remodeling of a Cuboid in Tension
To gain insight into the dynamics of remodeling, we consider a fiber-reinforced

cuboid under fixed tensional loading as depicted in Figure 11.28. We assume that
the assumptions on material properties, geometry, and boundary conditions, given in
Section 11.11.3, hold with the added assumption that the fiber angle θ is not fixed in
time, but that it evolves on a slow time scale according to the strain-driven remodeling
assumption.
Following [865], we assume that the fibers are initially oriented in the X 1 − X 2
plane and remain equal and symmetric in the same plane during reorientation. Then,
the reorientation law (14.92) can be expressed as an evolution law for the fiber
orientation angle
⎧
⎪− 1 sin , λ1 > λ2 ,
d ⎨ τ
= 0, λ1 = λ2 , (14.94)
dt ⎪
⎩ 1
τ
cos , λ1 < λ2 .
Therefore, the problem can now be cast as the evolution of three dynamical vari-
ables {λ1 (t), λ2 (t), (t)} for the dynamical system consisting of Equation (14.94)
together with the equilibrium conditions (11.303)–(11.304) that we repeat here for
convenience
Figure 14.6 Remodeling of an anisotropic cylindrical tendon based on strain. Initially, fibers have
random orientation and the tendon is in tension. As time evolves, the fibers align with the strain
[735]. Picture courtesy of Ellen Kuhl.
14.7 Remodeling 423
λ21 − λ22
A= + (λ21 cos2 + λ22 sin2 − 1)(λ21 cos2 − λ22 sin2 ), (14.95)
4γ
−1 λ21 + λ22
B= + + (λ21 cos2 + λ22 sin2 − 1)(λ21 cos2 + λ22 sin2 ),
2γλ21 λ22 4γ
(14.96)
where A = (t1 − t2 )/4γ, B = (t1 + t2 − 2t3 )/4γ, and we have set, without loss of
generality, μ = 1.
The initial conditions for the system (14.94–14.96) are given by a choice of
(0) = 0 and λi (0) = λi0 , where λi0 , 0 must satisfy (14.95–14.96).
In phase space the trajectories lie at the intersection of the level set A(λ1 , λ2 , ) =
A1 and B(λ1 , λ2 , ) = B1 . On this curve the dynamics is controlled by (14.94) as
shown in Figure 14.7. Since we assume that fibers do not support compression, the
dynamics of the system is further restricted by requiring that the systems remains in
the fiber-tensile region, defined by
λ21 cos2 + λ22 sin2 ≥ 1, ∈ [0, π/2]. (14.97)
Fibers are unstrained on the boundary of region (14.97), and the material behaves as
a regular isotropic neo-Hookean material when λ21 cos2 + λ22 sin2 < 1.
The key question is to understand the asymptotic dynamics of the system. Starting
with given boundary tensions, specified through A and B, what is the asymptotic
state of the system?
We start from the observation that System (14.94–14.96) has only three possible
steady states:
• = 0, λ1 ≥ λ2 ( achieves its minimum and λ1 is the dominant stretch);
• = π2 , λ1 ≤ λ2 ( achieves its maximum and λ2 is the dominant stretch);
• = ∗ ∈ (0; π2 ), λ1 = λ2 = λ (there is no dominant stretch in X 1 − X 2 plane).
Equilibria of the first and second type are called extreme equilibria, whereas the third
type are intermediate equilibria.
The system is symmetric with respect to the change of variables: λ1 → λ2 , λ2 →
λ1 , → π2 − , A → −A. Thus, we only consider extreme equilibria with λ1 > λ2
and intermediate equilibria with A ≥ 0, the other cases following via this symmetry.
The existence and stability of the different equilibria depend on the geometry of
the level sets and their intersections in phase space. A detailed analysis [865] reveals
that there is always a unique stable equilibrium when the loads place the fibers in
tension for some orientations. Moreover, this equilibrium point does not depend
on the initial conditions. Rather, it corresponds either to a complete alignment of
fibers along the direction of maximum principal stretch, or an intermediate alignment
between two principal stretches. In the case of intermediate alignment, the asymptotic
fiber orientation depends critically on the applied loads.
As can be seen from Figure 14.8, if the difference between the two loads in the
fiber plane is large enough, then the fibers gradually rotate toward the direction of
B=B 1
A=A 1
Figure 14.7 The remodeling trajectory in the phase space (λ21 , λ22 , ) is an oriented curve in phase
space located at the intersection of the level sets of A(λ21 , λ22 , ) = A1 and B(λ21 , λ22 , ) = B1
(shown: γ = 2, μ = 1, A1 = 1/2, B = 1). The solution shown is within the fiber-tensile region.
the greater load, asymptotically approaching it. As a result of fiber reorientation, the
specimen contracts in this direction.
For smaller differences in loads, fibers initially reorient toward the direction of
current maximum stretch and reinforce it. Thus, the maximum principal stretch is
decreasing, while the remaining stretch in the fiber plane is increasing. Eventually,
reorientation stops when the two stretches are equal. For some applied loads, there
is no tensile state and therefore no fiber remodeling. This simple remodeling law
creates an interesting dynamics of fiber through which the strains, hence the loads,
balance each other dynamically.
14.8 Growth Induced Instability 425
A. B.
B B
1
8
0.5
4
0 A A
1 2 3 0 0.5
Figure 14.8 Possible asymptotic dynamics for the remodeling of a fiber-reinforced cuboid for
A > 0 and B > 0. A. γ < 3/4. B. γ > 3/4. For values of A and B above the solid line, the dynamics
settles on a stable intermediate equilibrium, whereas for values of A and B below that line, only
the extreme equilibrium is reached. (shown: γ = 2, μ = 1, A1 = 1/2, B = 1). The dashed line
denotes the possibility of compressed fibers: Above the line, the equilibrium fibers are always in
tension whereas below the line, they may be in compression (depending on the angle).
14.8 Growth Induced Instability
In the previous sections, we studied the dynamics and stability of simple configura-
tions. Next, we consider the possibility of buckling induced by growth and expand
the bifurcation method developed in Section 11.10 to the case of a growing elastic
body.
We restrict our attention to the case where the growth tensor G is given and does
not evolve through the deformation. That is, growth is driven from outside and is not
influenced by the state of the system such as stresses or strains. Therefore, the growth
tensor can be seen as a control parameter that changes the geometry of the body and
its residual stress, but it is not affected by the deformation itself. We are interested
in the loads and growth parameters for which a bifurcation occurs. We assume the
existence of a base solution and look for the existence of small incremental solutions
superimposed on the finite deformation.
As before, we assume that a family of solutions, the base solution, for the Cauchy
stress T(0) and deformations χ(0) are known for given boundary conditions and
controlled by several parameters, such as external loading or prescribed growth.
We seek solutions close to the base solution by formally linearizing the governing
equations around the base state. We assume that there is no body force and that the
elastic response of the material is governed by a strain-energy function W = W (A)
subjected, possibly, to a single kinematic constraint C(A) = 0. We further focus on
the incompressible case for which C(A) = det(A) − 1. We start by considering the
first-order kinematics.
Step 1: Kinematics of incremental deformations. Assuming a known finite defor-

mation χ(0) , we introduce incremental deformations as follows
χ = χ(0) + χ(1) , (14.98)
where has been introduced as a small parameter that characterizes the size of the
perturbation superimposed on the finite deformation. Accordingly, we have

F = Grad(χ) = 1 + F(1) F(0) , (14.99)
and, since the growth tensor G is assumed to be constant, the elastic deformation
tensor is simply
A = 1 + A(1) A(0) , (14.100)
F (1)
A(0)
Final
Incremental
A(1) A(0) deformation
F (0)
F (1) F (0) G
F A
A(0)
(0)
F
Finite
deformation
Reference
Figure 14.9 Decomposition of the finite and incremental deformation tensors. A solid line repre-
sents a composition and a dashed line represents the addition of a small an incremental deformation
on top of a finite deformation. For instance, the total strain tensor needed to reach the final state
from the reference configuration can be decomposed as F = GA(0) + F(1) F(0) .
where A(0) = F(0) G−1 and A(1) = F(1) . The peculiar form of the expansions for F and
A with F(0) as a factor is chosen so that the incremental deformation is expressed in
the current configuration rather than the reference configuration. Figure 14.9 presents
a diagram of the decomposition of the different deformation tensors.
The kinematic constraint C(A) = 0 is expanded to first order as
C(A) = C(A(0) + A(1) A(0) )

= C(A(0) ) + tr CA(0) A(1) A(0) + O(2 ). (14.101)
That is, to first order, we have the condition

tr CA(0) A(1) A(0) = 0, (14.102)
where CA(0) denotes the second-order tensor obtained as the derivative of C with
respect to A evaluated at A(0) . In the particular case of incompressibility, we have
CA(0) = det(A(0) ) (A(0) )−1 , (14.103)
so that
tr(A(1) ) = 0. (14.104)
Step 2: Linearized equations. Since both the incremental deformation F(1) and the
boundary conditions are expressed in the current configuration we only consider the
Cauchy stress tensor and assume that it can be expanded in :
T = T(0) + T(1) + O(2 ). (14.105)
To first order, in the absence of body loads, the Cauchy equation reads
div(T(1) ) = 0. (14.106)
Step 3: Linearized constitutive equations. We can now relate the incremental

stresses to the incremental strains by the constitutive relationship. For a single scalar
constraint, C(A) = 0, the constitutive relationship is given by Equation (11.132).
That is,
−1 ∂W ∂C
T= J A −q . (14.107)
∂A ∂A
To leading and first orders, this constitutive relationship becomes

T(0) = A(0) WA(0) − q (0) CA(0) , (14.108)
and
T(1) = F(1) T(0) − q (1) A(0) CA(0) + L : F(1) − q (0) C : F(1) , (14.109)
where q = q (0) + q (1) , and L, C are the fourth-order tensors defined by the operation
(:) on a second-order tensor

(0) (1) (0)
L : F(1) = A(0) tr WAA F A , (14.110)

(0) (1) (0)
C : F(1) = A(0) tr CAA F A , (14.111)
and WA(0) , WAA

(0)
are the first and second derivatives of W with respect to A evaluated
on A(0) . The components of the fourth-order tensor L are the instantaneous elastic
moduli and their explicit form for isotropic materials is given below.
In the incompressible case, C = det(A) − 1, so that (14.108) now reads
T(0) = A(0) WA(0) − q (0) 1, (14.112)
and we can use Equation (11.53) to show that
C : F(1) = −F(1) . (14.113)
The incremental Cauchy stress tensor then becomes
T(1) = L : F(1) + F(1) A(0) WA(0) − p (1) 1. (14.114)
Linearized equations for incompressible materials. We substitute (14.114)

in (14.106) to obtain
div(L : F(1) ) + (F(1) )T grad (q (0) ) − grad(q (1) ) = 0. (14.115)

T
where we have used T(i) = T(i) , div(T(0) ) = 0 and
div(F(1) ) = 0. (14.116)
This last relation comes from the application of Nanson’s formula to the transforma-
tion of a volume element so that
div(J −1 F) = 0, (14.117)
which implies both div(J −1 F(0) ) = 0 and (14.116).
Step 4: The boundary-value problem. The boundary conditions are prescribed by

fixing χ(1) on some part of the boundary and T(1) n on the remainder of the boundary,
where, as before, n is the unit outward vector normal to the current boundary.
Relation (14.115) is the central equation for the stability analysis of the config-
uration associated with the deformation χ(0) . For given boundary conditions, the
existence of solutions to this system of equations indicates a bifurcation and the
possibility of an instability.
Instantaneous elastic moduli. The elastic moduli are similar to those given in
Section 11.10 with the main difference that we have now a mixture of elastic strains
(0)
and geometric first-order deformations. For W = W (A), we define A = WAA , the
elastic moduli evaluated in the reference configuration. The instantaneous elastic
moduli are the elements of the fourth-order tensor L defined by the relation
L : F(1) = A(0) tr(AF(1) A(0) ). (14.118)
In Cartesian components, the elastic moduli are given by
∂W
Ai jkl = , (14.119)
∂ A ji ∂ Alk
and the instantaneous elastic moduli are

(0)
Li jkl = Aim Am jkn A(0)
nl . (14.120)
Let αi be the principal values of A, then for an isotropic hyperelastic material the
components of L on the Eulerian principal axes associated with A are
∂2 W
Lii j j = αi α j , (14.121)
∂αi ∂α j

∂W ∂W αi2
Li ji j = αi − αj , i = j, (14.122)
∂αi ∂α j (αi2 − α2j )
∂W
Li j ji = L jii j = Li ji j − αi , i = j. (14.123)
∂αi
If αi = α j for some i = j, then

1 ∂W
Li ji j = Liiii − Lii j j + αi , (14.124)
2 ∂αi
Liiii = L j j j j , Liikk = L j jkk , (14.125)
Likik = L jk jk , Likki = L jkk j . (14.126)
In the two following chapters, we will use this general theory to study the possibility
of buckling induced by growth.
Chapter 15
Growing Spheres
Sphere-like biological and physical objects appear in many different sizes and con-
texts. In Figure 15.1, we show examples of spheres either shrinking by evaporation
or growing by cell expansion and division. In these processes, stresses develop in the
material due to either inhomogeneous or anisotropic addition of mass. In turn, these
stresses can lead to buckling or transform the effective properties of the structure.
This particular geometry is well adapted to a mathematical treatment as the deforma-
tions are sufficiently simple to allow for a full quantitative analysis. In this chapter,
we first consider an incompressible hyperelastic spherical shell that preserves its
spherical symmetry during the deformation. Second, we compute the stresses gener-
ated through different growth processes. We will show how limit-point instabilities,
cavitation, and singularity formation can be triggered by growth. Third, we look at
the possibility of buckling by considering nonspherical perturbations superimposed
on a residually stressed spherical shell.
15.1 The Growing Shell
Throughout this chapter, we use spherical coordinates to describe possible defor-

mations. Briefly, the spherical coordinates in the current configuration (r, θ, φ) are

432 15 Growing Spheres
Figure 15.1 Top row: A drying droplet of colloidal suspension on a hot plate. As the drop shrinks,
it looses its spherical shape due to the creation of internal stresses [835, 1254] (scale bar 0.5 mm.
Picture courtesy of Nicolas Tsapis). Bottom row: a. A scanning electron micrograph of cellular and
extracellular patterns during blastulation and gastrulation in the sea urchin, Lytechinus variegatus
[907]; b. Scanning electron micrograph of a human hepatoma cell (HepG2) (scale bar, is 10 μm). c
Electron microscopy cross-section of a V-79 Chinese hamster lung cell spheroid [1007] (scale bar,
is 250 μm).
defined with respect to the Cartesian coordinates (x1 , x2 , x3 ) as
x1 = r cos θ sin φ, (15.1)

x2 = r sin θ sin φ, (15.2)
x3 = r cos φ, (15.3)
where r ∈ [0, ∞), θ ∈ [0, 2π), ϕ ∈ [0, π] and let {er , eθ , eφ } be the associated
orthonormal basis. Similarly, we define (R0 , 0 , 0 ), (R, , ) and their associated
bases in the initial and virtual configurations.
15.1.1 Kinematics of Growing Spheres
We consider the symmetric growth and deformation of a spherical shell with radii
R0 = A0 and R0 = B0 in the initial configuration into a spherical shell of radii r = a
and r = b in the current configuration. We allow for the possibility where both A0 and
a vanish (a solid growing sphere); the case where A0 = 0 but a > 0 (the cavitation
15.1 The Growing Shell 433
R0 r(R 0)
A0 B0 a
b
Figure 15.2 Radial deformation of a shell with inner and outer radii A0 and B0 to a shell with
radii a and b.
of a sphere); as well as the case whereby A0 > 0 and a = 0 (the anti-cavitation of a

spherical shell).
In a spherically symmetric deformation, the shell expands and any point located
at (R0 , 0 , 0 ) in the reference configuration is moved to the point (r (R0 ), 0 , 0 )
in the current configuration, as shown in Figure 15.2. Explicitly, the deformation,
x = χ(X0 ), reads
r = r (R0 ), θ = 0 , φ = 0 , (15.4)
so that the position vectors are, respectively,
r (R0 )
X0 = R0 E R0 , x = r (R0 )er = X0 . (15.5)
R
Due to the symmetry of the deformation, we can identify the basis vectors so that
E R0 = er , E0 = eθ , E0 = eφ .
To compute the deformation gradient, we use the method described in
Section 11.1.5. We have two sets of spherical coordinates {qα } = {r, θ, φ} and
{Q α } = {R0 , 0 , 0 }. For these coordinates, the scale factors, defined by (11.39),
are
h r = 1, H R0 = 1, (15.6)
h θ = r, H0 = R0 , (15.7)
h φ = r, H0 = R0 . (15.8)
Following Equation (11.60), the deformation gradient is

r r
F = r er ⊗ er + eθ ⊗ eθ + eφ ⊗ eφ ,
R0 R0
which we write
[F] = diag(r , r/R0 , r/R0 ), (15.9)
where the primes denote derivatives with respect to R0 .

Next, we assume that this geometric deformation is the result of both a growth
deformation and an elastic deformation. We assume that the growth deformation
preserves the spherical symmetry so that in the virtual state V, the growth deformation
can be written as
G = γr er ⊗ er + γθ eθ ⊗ eθ + γθ eφ ⊗ eφ , (15.10)
or
[G] = diag(γr , γθ , γθ ), (15.11)
where γr and γθ are functions of R0 and, again, due to symmetry, we used the current
configuration basis, that is, E R = er , E = eθ , E = eφ .
We distinguish three types of growth:
• Isotropic growth. Isotropic growth is achieved when γr = γθ . In this case, the
grown sphere is a dilation of the original sphere.
• Area growth. If γr = 1 and γθ = 1, growth is purely along the two angles, increas-
ing the surface area of any given shell and, is referred to as area growth. If γθ > 1
material is added and removed otherwise.
• Fiber growth. If γθ = 1 and γr = 1, fiber growth or radial growth takes place in the
radial direction and each radial material fiber in the shell increases (or decreases)
in length [709]. Both area and fiber growth are illustrated in Figure 15.3.
The key parameter in this growth process is the ratio γ = γr /γθ . Equilibrium
solutions with radial growth are equivalent to solutions with hoop resorption (γ > 1)
up to an isotropic growth and a rescaling of an effective applied pressure
Peff = P/γ. (15.12)
It follows from the multiplicative decomposition, F = AG, that the elastic strain
tensor can be written in the same spherical basis as
[A] = diag(αr , αθ , αθ ). (15.13)

15.1 The Growing Shell 435
Figure 15.3 Transformation

of a spherical volume Fiber growth
element under radial and γ1>1, γ2=1
hoop growth.
Area growth
γ1=1, γ2>1
The incompressibility condition, det A = αr αθ2 = 1, together with the decomposi-

tion, F = AG, has two consequences. First, we can define a unique strain α ≡ αθ ,
so that αr = α−2 and
γr r
r = 2 , = αγθ . (15.14)
α R0
Second, the deformation is completely determined by
r2
r = γr γθ2 , (15.15)
R02
that is, after integration

R0
r 3 − a3 = 3 γr (ρ)γθ2 (ρ)ρ2 dρ. (15.16)
A0
where a = r (A0 ) and b = r (B0 ) denote the inner and outer radii of the deformed
shell, respectively. The hoop stretch α is then given by
R 1/3
a 3 + 3 A00 γr (ρ)γθ2 (ρ)ρ2 dρ
α(R0 ) = . (15.17)
R0 γθ (R0 )
15.1.2 Stresses in a Growing Sphere
Next, we compute the static stresses in the absence of body loads in terms of the
Cauchy stress tensor. By symmetry, the only nonvanishing spherical components of
the stress tensor are its diagonal components, so that
[T] = diag(tr , tθ , tθ ), (15.18)
where tr is the radial stress and tθ is the hoop stress. Using this particular form for the
elastic and stress tensors, the stress–strain relationships (13.37) for a strain-energy
function W = W (αr , αθ , αφ ) read
∂W
tr = αr − p, (15.19)
∂αr
∂W
tθ = αθ − p. (15.20)
∂αθ
The only nonvanishing Cauchy equation for mechanical equilibrium (13.35) in the
current configuration is
∂tr 2
+ (tr − tθ ) = 0, (15.21)
∂r r
and a closed equation for tr is obtained:
∂tr α
= ∂α W , (15.22)
∂r r
where we have introduced the auxiliary function W (α) = W (α−2 , α, α). This last
equation can be written as a function of the initial variable R0 as
∂tr γ .
= ∂α W (15.23)
∂ R0 R 0 α2
Once the radial stress and the deformation α = α(R0 ) are known, the hoop stress is
given by
α
tθ = tr + ∂α W . (15.24)
2
For the case of a shell in the current configuration, the boundary conditions
for (15.23) in the case of hydrostatic loading with pressure P are such that P =
tr (A0 ) − tr (B0 ). For A0 > 0, we take, without loss of generality,
tr (A0 ) = 0, P = −tr (B0 ). (15.25)
If P > 0, the shell is subject to an external hydrostatic pressure P and if P < 0 the
shell is inflated with an internal pressure −P. A negative radial stress corresponds
to a compressive stress and a positive radial stress is tensile. For a given P, the
problem is to determine the inner radius a introduced in (15.16). Once a is known,
the deformation is fully specified and so are the stresses. In the case of a solid
sphere under hydrostatic loading in the current configuration, we have a = 0 and
tr (r (B0 )) = −P.
15.2 Examples 437
15.2 Examples
15.2.1 Anisotropic Growth
We are interested in computing the stresses and the radial deformation that can
be created through constant but anisotropic growth (constant γr and γθ ). In this
case, (15.17) takes the following form:

A30 a3
α = γ 1 − 3 + 3 3.
3
(15.26)
R0 γθ R0
Defining αa = α(A0 ) = a/(A0 γθ ) and αb = α(B) = b/(B0 γθ ), we have

3
A3 A0
αb3 = γ 1 − 03 + αa3 , (15.27)
R0 B0
and using (15.26), Equation (15.23) can be written as a closed equation for tr (α):
∂tr
γr ∂α W
= . (15.28)
∂α γr − α (γθ + R0 γθ )
3
Since both γr and γθ are constant, this last equation can be readily integrated as
α
∂α W
tr (α) = γ dα. (15.29)
αa γ − α3
The boundary conditions (15.25) lead to

αb
∂α W
−P = γ dα, (15.30)
αa γ − α3
and since αb is related to αa by (15.27), this last relation is an equation for P as a

function of αa . For a given P, one can invert this relation to find the strain at the
boundary αa caused by the traction. Once αa is known, the position of the inner
radius is known and so is the deformation r = r (R) through (15.26).
In Figure 15.4, we show the radial stress as a function of αa for a neo-Hookean
material in the absence of growth (γr = γθ = 1). Neo-Hookean materials are char-
acterized by the strain-energy function
μ 2
W = (α + αθ2 + αφ2 − 3), (15.31)
2 r
= μ (α−4 + 2α2 − 3) and
which implies W 2
μ −1
tr (αb ) = 4αb + αb−4 − 4αa−1 − αa−4 . (15.32)

2
tr(αb)
1
0
αa
-P
(αa,−P)
Figure 15.4 The radial stress at the outer boundary for a non-growing neo-Hookean material as
a function of the inner boundary strain αa . For a given P, the strain α(R) and the deformation
r = r (R) are specified by finding the value of αa such that tr (αb ) = −P.
In Figure 15.5, we plot the stresses created by radial and hoop growth for an
isotropic neo-Hookean shell with no external loading (P = 0). Both radial growth
and hoop resorption create a compressive residual stress (i.e., a negative radial stress)
in the material, whereas hoop growth and radial resorption create a tensile radial
stress (i.e., a positive radial stress). For constant growth, the radial stress vanishes
only at the boundaries and its sign depends on whether γ is less than or greater than
one. If we consider nonconstant growth, the effect of the γθ term in (15.28) can be
sufficiently important so as to create alternating regions of compressive and tensile
stresses [196].
15.2.2 Differential Growth
Another possible mechanism to create residual stress is through differential growth.

We consider two cases: a shell shrinking differentially and a shell growing inside an
infinite elastic medium. We assume that the differential growth process is a function
of the current radius and the result of many incremental growth steps as described in
Section 14.3 and sketched in Figure 15.6.
We first define an incremental deformation tensor
Finc = Ainc Ginc , (15.33)

15.2 Examples 439
A.
1.6 )
t θ(R 0
0.8
0 1.2 1.4 1.6 1.8 2

R0
-0.8 )
t r( R 0
-1.6
-2.4
-3.2
Compressive radial stress
-4
-4.8
-5.6
B.
1.6
1.2 tθ (
R
0)
0.8 Tensile radial stress
0.4
tr ( R )
0
R0
0
1.2 1.4 1.6 1.8 2
-0.4
-0.8
√ Stresses generated by A. fiber growth (γr = 2, γθ = 1) and B. area growth (γr =

Figure 15.5
1, γθ = 2) of a neo-Hookean shell μ1 = 1, A = 1, B = 2. Radial growth creates a compressive
radial stress. The hoop stress is tensile on the outer radius and compressive inside. In hoop growth
the situation is reversed and similar to fiber resorption.
F
Incremental
F1 steps F2 Finc Fk
Stress-free
G
A
Cumulative
growth
Figure 15.6 The process of incremental and cumulative growth.
where the growth tensor Ginc is assumed to be isotropic and a function of the radial
position r in the current configuration:
Ginc = ginc (r )1. (15.34)
The decomposition (15.33) represents an incremental deformation step. Once

(slow) growth and (fast) elastic response take place, the shell continues its growth,
and after successive incremental steps the total cumulative deformation is of the form
F = AG, G = g(r )1. (15.35)
As an example, we choose a simple incremental growth of the form
ginc (r ) = 1 + f (r − a), (15.36)
with f (0) = 0, so that there is no growth at the inner boundary. The problem is then
to find a suitable form for the cumulative growth function g(r ). Since the solution for
a and r as a function of the initial radius R0 at each step depends on the solution of
a boundary-value problem, there is no analytic form for g(r ) as a function of ginc (r ).
However, a numerical study of the incremental growth process, shown in Figure 15.7,
reveals that, for a simple linear incremental law, cumulative growth is also accurately
modeled by a linear profile even for large variations of volumes. Therefore, we take
g(r ) = 1 + ν(r − a), (15.37)
where ν is positive for growth and negative for resorption.

15.2 Examples 441
A.
g(r)
1.8 linear fit 275%
226%
1.6 205%
185%
1.4 168%
168% cumulative
growth
153%
1.2 137%
124%
113% r
1
1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8
g(r) B.
1
89%
0.9
80%
72%
0.8
65%
linear fit 59%
0.7 53%
48%
44%
0.6
cumulative 40% r
growth 36%
0.8 1 1.2 1.4 1.6 1.8 2
Figure 15.7 The cumulative growth function (solid line) in the current frame g = g(r ) together with
a linear fit (dashed) for the case of A. growth and B. resorption. The percentage values correspond
with the overall volume change of the sphere.
As before, the incompressibility condition determines both the deformation and

the strain. However, it is advantageous to formulate the problem in terms of the
current radius, for which we have
r
ρ2 r
R03 = A30 +3 dρ, α(r ) = , (15.38)
a g 3 (ρ) g(r )R0 (r )
from (15.14) and (15.15). As in the previous case, the value of a is obtained from
the boundary conditions on the Cauchy stress.
In the absence of an external or internal medium, the radial component of the
Cauchy stress for a material characterized by a strain-energy function Wsh is given
by r
α
tr (r ) = ∂α Wsh (α) dρ, (15.39)
a ρ
with boundary conditions tr (b) = −P, and tr (a) = 0.

When growth occurs inside a medium, the Cauchy stress inside an infinite medium,
characterized by a strain-energy function, Wme is
r
α
tr (r ) = ∂α Wme (α) dρ, (15.40)
b ρ
where it is assumed that g = 1 for r > b. The boundary condition is now tr (∞) = 0
and equality of radial stresses at r = b.
Once the radial stress is known, the deformation is completely determined and
the hoop stress in the shell is given by
α
tθ = tr + ∂α Wsh (α). (15.41)
2
Typical profiles of radial and hoop stresses for vanishing external pressure are
given in Figures 15.8–15.9. The constant ν is fixed by the overall increase in volume
b3 − a 3
V = . (15.42)
B03 − A30
Resorption creates a compressive residual stress whereas growth in an elastic medium

creates both compressive contact stress (close to the outer boundary) and tensile
residual stress (around the inner boundary).
15.3 Limit-point Instability and Inflation Jump
The classic theory of rubber materials predicts that for particular choices of strain-
energy functions and parameters, a limit-point instability may occur in spherical
shells as the internal pressure is increased [8, 17, 198, 433, 922, 974]. This effect is
triggered by the loss of monotonicity of the function tr (αb ) as a function of αa , that is,
the pressure–stretch curve has a local maximum (see Figure 15.4) and the resulting
instability is known as a limit-point instability, as discussed in Section 11.8.3.1.
For example, consider the Mooney–Rivlin strain-energy function
μ ν
W = (α2 + αθ2 + αφ2 − 3) + (α−2 + αθ−2 + αφ−2 − 3). (15.43)
2(1 + ν) r 2(1 + ν) r
15.3 Limit-point Instability and Inflation Jump 443
tr,tθ
ΔV=1/2
Tensile ess
1
p str
h oo
1 1.2 1.4 1.6 1.8 2

0
R0
stress
radiak
-1
Compressive
χ
R
B r b
-2 A
a
Figure 15.8 Radial and hoop stresses due to shrinking in a vacuum (V = 1/2, i.e., ν≈ − 0.39)
for a neo-Hookean elastic shell with A0 = 1, B0 = 2 and μsh = 1.
The pressure stretch curve tr (αb ) as a function of αa may present a maximum fol-
lowed by a minimum at finite stretch as shown in Figure 15.10. Therefore, under
controlled pressure, the stretch may jump for increasing pressure and present a hys-
teresis loop when the pressure is reduced leading to an inflation jump. This jump is
more pronounced in thinner shells.
To understand the effect of thickness on this instability, we follow [465, 975] and
consider
α (α) αb (α)
W W
tr (α) = dα, and −P = dα. (15.44)
αa 1 − α3 αa 1 − α3
We use the mean-value theorem and expand (15.44) to second order in

δ = (B0 − A0 )/A0 to obtain
(α)
W δ 2 α3 − 2
tr (α) = δ + 4 W (α) − (α − 1)W (α) + O(δ 3 ),
3
(15.45)
α2 2α α
where α is now the position of the inner radius. Since the wall thickness is assumed
to be small, this relation also describes the stress field at every point in the shell.
tr,tθ
ΔV=2
2 hoo
ps
Tensile tres
s
1 radial stres
s
1 1.2 1.4 1.6 1.8 2

0
R0
-1
Compressive
χ
-2
-3 A0 a
B0
b
Figure 15.9 Radial and hoop stresses due to growth inside an elastic medium (V = 2, i.e.,
ν≈0.378) for a neo-Hookean elastic shell with A = 1, B = 2 and μsh = μme = 1.
A limit-point instability occurs for αcr such that tr (αcr ) = 0. Writing tr = 0 to
order O(δ), we recover the condition for the critical hoop stretch for thin shells [551],
that is,
W (α) = 0.
(α)α − 2 W (15.46)
For instance, in the case of a neo-Hookean strain-energy function, we have

μ 2 = μ (α−4 + 2α2 − 3),
W = (α + αθ2 + α32 − 3), W (15.47)
2 r 2
and a limit-point instability appears at α = 71/6 .

To explore the dependence of the critical stretch with thickness, we expand αcr to
first order in δ as
αcr = α0 + α1 δ + O(δ 2 ), (15.48)
where α0 is chosen as the smallest solution of
W (α0 ) = 0,
(α0 )α0 − 2 W (15.49)
which is larger than one.

Using tr = 0 to order O(δ 2 ), and making use of (15.49), we find that α1 is given
by
t1(αb)
1
P2
P1
1/2
αa
α1 α2 α3 α4
0
2 6 10 14
-1/2
-1
Figure 15.10 Inflation jump: For a Mooney–Rivlin material with μ = 1.886,

ν = −0.05, the curve tr (αb ) as a function of αa has a minimum and a maximum
(A = 1, B = 2, γr = γθ = 1). If the internal pressure of the shell is raised to P2 , a sudden
inflation occurs through a jump from α2 to α4 . When pressure is decreased to P1 , a second jump
from α3 to α1 occurs.
α03 − 2α1 α02 − 1 = 0. (15.50)
It follows that
α03 − 1
αcr = α0 + δ + O(δ 2 ). (15.51)
2α02
The first correction (15.51) shows that universally (independent of the constitutive
relation), the critical stretch for the limit-point instability increases with thickness
for thin shells. In other words, thicker membranes are always more stable against
limit-point instabilities during inflation.
For a given model, we can identify the critical value of the parameters where this
instability disappears. We start with the strain-energy function (15.43) for a Mooney–
Rivlin material. The limit-point disappears as ν increases to νcr when the curve tr (α)
becomes strictly monotonic. This critical point is found by solving tr = tr = 0, which
gives
√
2 11 − 3 √
νcr = √ ≈ 0.21446, αcr = (19 + 6 11)1/6 ≈ 1.84073. (15.52)
5(19 + 6 11)1/3
4
t1(α)
β=0.1
β=β cr
3
2 β =β cr 2
/
1 β=0
Neo-Hookean case
α
0
1 2 3 4 5
Figure 15.11 Pressure–stretch curve for a Fung material. The limit β = 0 corresponds to a neo-
Hookean material. When 0 < β < βcr , the system exhibits a limit-point instability and an inflation
jump (see dotted line when β = βcr /2). For β ≥ βcr , the limit-point instability disappears.
The situation is similar for Fung materials

μ
Wfu = [exp β(I1 − 3) − 1], (15.53)
2β
as shown in Figure 15.11, where we can readily identify the critical values of the
parameters:
√ 2/3 √
1 92 + 12 65 3 + 65
βcr = √ ≈ 0.06685, (15.54)
48 57 + 7 65
1 √
αcr = √ (92 + 12 65)1/6 ≈ 1.69355. (15.55)
2
Hence, when β > 0.067, as is the case for soft biological tissues, there are no
limit-point instabilities [72], in accordance with Osborne’s early observations [1000]
depicted in Figure 11.1.
For Gent materials,

μ
Wge = − log[1 − β(I1 − 3)], (15.56)
2β
we find that the limit-point instability disappears when β > βcr , given by
√ 2/3 √
1 10 + 93 3 + 93
βcr = √ √ 2/3 √ ≈ 0.05669, (15.57)
3
315 + 33 93 − 10 + 93 3 + 93
and the corresponding critical stretch is

√ 1/6
αcr = (10 + 93) ≈ 1.64262. (15.58)
For elastomers, Gent [433] found experimentally limit-point instabilities (and infla-
tion jumps) for inflated shells with β = 0.01 and β = 0.03. However, Horgan and
Saccomandi [619] estimated that β ≈ 0.44 for the aorta of a 21-year-old male and
that β ≈ 2.4 for the (stiffer) aorta of a 70-year-old male. In these instances, no limit-
point instability is possible as the strain-stiffening property of the material prevents
them from occurring.
The behavior for a one-term Ogden material
2μ β β β
Wog1 = (λ + λ2 + λ3 − 3), (15.59)
β2 1
is slightly different as shown in Figure 15.12. Here again, the limit-point instability
disappears at βc = 3, below any realistic physiological values. The asymptotic limits
for tr (α) as α → ∞ are, however, different (0, 2, and ∞ for 0 < β < 3, β = 3, and
β > 3, respectively). Note finally that there is no inflation jump for any value of
β > 3.
We conclude that for soft biological tissues, the critical parameter values are far
below any typical range of physiological values so that the limit-point instability is
unlikely to be observed. As noted repeatedly by Humphrey and coworkers [262, 544,
635], this observation should be kept in mind when a strain-energy function is chosen
in numerical simulations of soft tissues, and when designing artificial soft tissues for
experiments. The choice of a rubber-like strain-energy function in the former case,
or of an elastomer in the latter case, might lead to instabilities which do not actually
exist in the naturally occurring soft tissue.
3.0 t1(α)
β=3.2
2.5
β=3
2.0
1.5
β=2.5
1.0
0.5
Neo-Hookean case (β=2)
α
0.0
1 2 3 4 5
−0.5
−1.0
Figure 15.12 Pressure–stretch curve for a one-term Ogden material. The neo-Hookean material
corresponds to β = 2. For 2 ≤ β < 3, the system exhibits a limit-point instability, and tr (α) → 0 as
α → ∞. For β = 3, tr (α) → 2 as α → ∞. For β > 3, tr (α) → ∞ as α → ∞ and the limit-point
instability disappears.
15.3.1 The Effect of Growth on the Limit-point Instability
Growth can change the property of a material to allow or prevent inflation jumps.
As an example, consider a nongrowing material and values of the parameters where
there is no instability, that is, the strain is a continuous monotonic function of the
applied pressure as shown in Figure 15.13A.
Now, for the same elastic material with radial growth, shown in Figure 15.13B, an
inflation jump occurs if γ is large enough. To understand how growth, geometry, and
elasticity control this effect, we compute the derivative of the pressure as a function
of αa from (15.29) to obtain
αa dP (αb ) ∂αa W
∂α W (αa )
( − αa−2 ) = b 2 − . (15.60)
γ dαa αb αa2
We conclude that limit-point instabilities are controlled by the function (∂α W (α))/α2
as found in Equation (15.45) and that the only dependence on growth enters in the
strain through relation (15.27). Therefore, the effect of constant radial or hoop growth
15.4 Singularities in Growing Solid Spheres 449
t r (αb) A. (no growth) 10 t r (αb) B. (with growth)

1.2
9.9
1
0.8 9.8
0.6
9.7
0.4
9.6
0.2
αa αa
0 9.5
1 1.5 2 2.5 1 1.5 2 2.5
Figure 15.13 Limit-point instability for a growing Mooney–Rivlin material with μ = 5/3, ν =
−1/6, A0 = 1, B0 = 2. A. The curve tr (αb ) is monotonic for a non-growing material (γ1 = γ2 =
1). B. For a growing material (γ1 = 8, γ2 = 1) the curve tr (αb ) has a minimum and a maximum
and the strain becomes a discontinuous function of the pressure.
is to effectively change the geometry of the shell, that is, a thick shell with radial
growth effectively behaves as a thin shell.
We can quantify the combined effects of growth and thickness for neo-Hookean
materials undergoing anisotropic constant growth. The critical value of the stretch
where such an instability occurs can be computed by taking a double expansion
of (15.60) in the thickness δ and growth γ, to obtain
αa = 71/6 {1
√ √
δ 415513 − 157039 7 + 113960 − 43079 7 (γ − 1)
+ √
14(55 7 − 148)2

+O(δ 2 , (γ − 1)2 ) , (15.61)
≈ 1.383 {1 + δ [0.311 − 0.189(γ − 1)]} + O(δ 2 , (γ − 1)2 ), (15.62)
where δ = (B0 − A0 )/A0 measures the thickness of the shell.
15.4 Singularities in Growing Solid Spheres
We consider growth within a solid sphere (a = A0 = 0) under an external pressure

P. In particular, we are interested in the behavior of the stress at the origin. We
start by considering constant anisotropic growth of the form (15.11) controlled by
γ = γr /γθ . The integration of (15.16) with a = A0 = 0 gives
r 3 = γr γθ R 3 , (15.63)
which together with the definition of α = r/(R0 γθ ) gives the stretch α = γ 1/3 . Then,
Equation (15.23) reads
∂tr β
= 1/3 .
, where β = γ 1/3 ∂α W (15.64)
∂ R0 R0 α=γ
With the boundary condition tr (B) = −P, the stress profile is

R0
tr = −P + β ln . (15.65)
B
Assuming that the strain-energy function W has a single minimum for γ = 1, we

have β = 0, and the radial stress exhibits a logarithmic singularity. Note, however,
that the principal stretches and hence the strain-energy function are finite everywhere.
The presence of a singularity in elasticity may seem surprising and unphysical.
However, it is known that anisotropy can lead to stress singularities even for bodies
with smooth boundaries [780, 1074]. For instance, the study of radially symmetric
equilibrium configurations of transversely isotropic solid cylinders and balls under
constant pressure on their boundaries demonstrates the existence of a pressure sin-
gularity above a critical applied pressure [34, 939]. Singularities due to material
anisotropy in cylindrical and spherical geometries were also observed by several
authors [47, 413]. Similarly, anisotropic eigenstrains are responsible for singulari-
ties [1354, 1356].
The existence of a singularity is intimately related to the behavior of the growth
tensor at the origin. Indeed, for an arbitrary growth tensor, since γr (R0 ) and γθ (R0 )
are both bounded and strictly positive, we can define
γr (R0 )
γ = lim . (15.66)
R0 →0 γθ (R0 )
Then, we have
∂tr β
= + O(1), 1/3 ,
β = γ 1/3 ∂α W (15.67)
∂ R0 R0 α=γ
and, we conclude again that if γ = 1, the radial stress exhibits a logarithmic singu-
larity at the origin.
15.5 Cavitation
Elastic cavitation is the phenomenon in which a cavity appears in the interior of an

elastic body under tension. Consider for instance the poker chip experiment shown
in Figure 15.14 where a cylinder of rubber with Young’s modulus E is glued to two
15.5 Cavitation 451
A. B.
tz
2r
Figure 15.14 A. The poker-chip experiment. A cylinder of rubber is glued to two plates that are
pulled apart. The deformation is supported by a nearly spherical tensile field at the center of the
cylinder creating the conditions under which cavitation is predicted to happen. B. Experimental
data for the poker-chip experiment (circles) compared to the theoretical prediction (dashed line) for
a neo-Hookean material [371].
rigid plates that are pulled apart with stress tz . The deformation creates a large tensile
isotropic stress field at the center of the sample which is equivalent to a radial outward
pressure P applied to a sphere. Experimental data show that the critical pressure at
which a cavity first appears scales linearly with the Young’s modulus.
This problem has received considerable interest in the literature, starting with
the pioneering experimental work of Gent and Lindley [434] and the seminal math-
ematical analysis of Ball [53]. Gent and Lindley considered solid elastic spheres
under uniform hydrostatic loads and found that the critical pressure at which radially
symmetric configurations with an internal cavity emerge from the undeformed con-
figuration is Pcrit = −5/6E for a neo-Hookean material with Wnh = E/6(I1 − 3).
Since the work of Ball, cavitation has received substantial attention and has been
studied for various situations—compressibility versus incompressibility, anisotropy,
and composite layers have all been analyzed in different combinations and for various
material models [546, 613, 616, 800, 1028, 1040, 1168, 1169]. Further experimental
and numerical works have validated the theoretical approaches of some of these
theories [204, 290, 294, 371, 1263].
Within our framework, cavitation can easily be understood by looking for a solu-
tion of the boundary-value problem of a sphere (A0 = 0) under tensile pressure
such that the inner radius in the current configuration is strictly positive a > 0. The
smallest value of the pressure for which such a solution exists is the critical pressure.
15.5.1 Cavitation Induced by Tensile Loading
We first integrate the equation for the radial stress (15.22) in the absence of growth:
∂tr
∂α W
= , tr (αa ) = 0, tr (αb ) = −P, (15.68)
∂α 1 − α3
to obtain αb
∂α W
−P= dα. (15.69)
αa 1 − α3
In the limit A0 → 0, we have αa → ∞ and αb → 1 so that

∞
∂α W
Pcrit = dα. (15.70)
1 1 − α3
This integral is improper and the existence of a finite critical pressure depends on
the behavior of W as α → ∞. If, in that limit, W
∼ αn then

∂α W
∼ αn−4 , (15.71)
1 − α3
and the integral converges if n < 3 [616]. For instance, in the neo-Hookean case,
taking the limit αa → ∞ and αb → 1 in (15.32) with tr (αb ) = −P, μ = E/3 gives
Pcrit = −5E/6 as shown in Figure 15.14.
The fact that cavitation depends both on the material response and on the partic-
ular geometry; that it changes the topology of the material; and that it relies on the
assumption that elasticity remains valid at the microstructure suggest that cavitation
may not be a robust material feature. However, cavitation is an idealized concept.
If there exists a microvoid in the material, cavitation is replaced by the problem of
cavity opening, which does not suffer the topological instabilities encountered in
pure cavitation. If the cavity is small enough, the bifurcation is essentially indistin-
guishable from the ideal cavitation problem and no experiment could differentiate
between microvoid opening and pure cavitation.
15.5.2 Cavitation Induced by Growth
We showed that cavitation can occur as a consequence of external mechanical loading.

However, growth or swelling can also induce cavitation [850, 1023, 1024] as we will
now demonstrate. We study the possibility of growth-induced cavitation following
[469] by looking at the existence of nontrivial solutions for given growth functions
γr and γθ . We consider two complementary cases, first the case of anisotropic and
homogeneous growth and second, the case of isotropic and inhomogeneous growth.
15.5 Cavitation 453
We restrict our attention to a neo-Hookean energy function and scale all forces by
taking the elastic modulus to be μ = 1, which corresponds to a Young’s modulus of
E = 3.
In the first case (anisotropic growth), we take, without loss of generality, an outer
radius B0 = 1 and γr = 1. Then, (15.23) can be integrated explicitly to obtain an
implicit relationship between a and γθ . In Figure 15.15, we graph the solution a =
a(γθ , A) for various initial inner radii.
The cavitated solution from the solid sphere corresponds to the choice A0 = 0. A
local analysis of the solution close to γθ = 1 leads to
√
√ 3π 66 γθ 2 + 183 γθ − 114
a ∼ 3 exp − , (15.72)
γθ →1 18 36 (5 γθ + 1) (γθ − 1)
which clearly establishes the existence of a cavitated solution whose inner radius is
exponentially small close to γθ = 1, but nonvanishing for all values of γθ > 1. This
situation is in contrast to the classic cavitation case discussed previously for which
the base solution exists for all pressure and a large value of the external traction is
required to obtain a cavitated solution. Physically, the two problems are different due
to the particular boundary conditions associated with the residual stress field. The
effect of anisotropic growth and residual stress is to create large radial tension close
to the inner radius. In the classic case, a large elastic energy is necessary to balance a
large gradient of radial stress close to the inner boundary which jumps from a finite
value for the base solution to zero on the bifurcated solution.
In Figure 15.15, the radial stress is plotted for some cavitated solutions. Notice
that the magnitude of the stress and its gradient close to the origin are larger in the
solution with the smaller cavity radius.
In the case of inhomogeneous growth), we set γr = γθ = 1 + βr with constant β.
Then, a cavity opens for β > βcrit ≈ 0.8971637 as shown in Figure 15.16. Here the
radial tension at the origin increases with β up to the critical point where sufficient
elastic energy has been built in the system to trigger cavitation. A similar mechanical
environment could be created by considering an elastic sphere surrounded by a
growing spherical shell. Again, for sufficiently large growth, the tension created by
the shell pulling on the sphere would be enough to open a cavity.
Therefore, we conclude that growth, either inhomogeneous or anisotropic, could
be a simple and universal mechanism to open cavities in elastic materials. Its potential
relevance to biological matter and the opening of plant stem will be discussed in
Section 16.2.
15.6 Instability Due to Anisotropic Growth
Buckling computations for pressurized spherical shells in the absence of growth have
a long history going back to the work of Wesolowski [1313] and Wang and Ertepinar
A. 0.25
a(γθ)
0.2
0.15
0.1
II
0.05 I
γθ
0.0
1.0 1.1 1.2 1.3 1.4
B.
I
II
Figure 15.15 A. Growth-induced cavitation of a sphere of initial radius 1 with constant anisotropic
growth (γr = 1, γθ ≥ 1). Dashed curve: asymptotic solution. Solid curves correspond to cavity
opening of a shell of initial inner radius A = 0.01, 0.05, 0.1, 0.2 due to growth. B. Radial residual
stress corresponding to the two highlighted solutions. The dashed curve corresponds to the solution
with smaller cavity radius. In both cases the sphere is in radial tension, but as the cavity opens the
magnitude of the radial tension decreases.
15.6 Instability Due to Anisotropic Growth 455
A. a(β)
III
I II β
B. tr
II
III
R0
Figure 15.16 Growth-induced cavitation of a sphere of initial radius 1 with isotropic homogeneous
growth (γr = γθ = 1 + β R). A. Opening a as a function of β after the critical value βcrit . B. Radial
stress as a function of initial radius R for β = 0.88 and β = 0.98; dashed curve: base solution, solid
curves, the cavitated solution with zero radial tension at the boundaries.
[1290] on the neo-Hookean shell. Later on, Hill [575], Haughton and Ogden [551],
Ogden [975], and Chen and Healey [198] obtained results for general strain-energy
functions and thin pressurized shells (see also [465] for the role of strain-stiffening
on the bifurcation condition). Work on asymptotic limits, the relationship with other
stability criterion and the eversion problem include studies by Fu [397], Haughton
and Kirkinis [547, 550], and Haughton and Chen [549].
We consider the possible bifurcations of a growing shell following the method
outlined in Section 14.8. We start with the base solution obtained for the radially
symmetric deformation of a growing shell under pressure. Once the change in radius
due to growth and applied stress has been established, we study the existence of
nearby non-radial deformation with respect to changes in the control parameters.
We first look at the case of an incompressible elastic shell subject to constant
anisotropic growth and uniform hydrostatic pressure P. The base solution is given
in Section 15.2.1 and we recall that
[G] = diag(γr , γθ , γθ ), [A(0) ] = diag(α−2 , α, α), (15.73)
so that
r = γr /α2 , r/R0 = αγθ , (15.74)
and
A3 a3
α3 = γ 1 − 03 + 3 3 , (15.75)
R0 γθ R0
where γ = γr /γθ . The elastic stretch at the outer boundary is

3
A3 A0
αb3 = γ 1 − 03 + αa3 . (15.76)
R0 B0
The stretch α satisfies the equation
dα α α4
= − . (15.77)
dr r rγ
To lowest order, the base solution for the radial stress is

α
∂α W
tr(0) (α) = γ
dα, (15.78)
αa γ − α3
= W (α−2 , α, α). We further restrict our attention to the case of

where, as before, W
a growing neo-Hookean material for which we have
= μ (α−4 + 2α2 − 3).

W (15.79)
2
For a given set of fixed parameters A, B, γr , γθ , the base solution is completely

specified by the pressure P. The problem is now to establish the existence of solu-
tions that are not radially symmetric for some values of these parameters.
Step 1: Kinematics of incremental deformations. The general form for the first-
order deformation χ(1) describing the bifurcation of a spherical shell is
χ(1) = u(r, θ, ϕ)er + v(r, θ, ϕ)eθ + w(r, θ, ϕ)eφ . (15.80)
The standard procedure is to expand the functions (u, v, w) in spherical harmonics

and solve Equation (14.116) for each mode. However, in the determination of the
parameters where a bifurcation occurs, it has been established in the classic case
[1290, 1313], and verified in the case of a growing shell, that the equation for w
decouples so that w does not play a role in the bifurcation. Furthermore, the equations
for all spherical harmonics are identical. Therefore, the analysis to zeroth order,
corresponding to Legendre polynomials, is sufficient to identify the bifurcation point.
We conclude that the bifurcation analysis of the spherical shell reduces to the
analysis of axisymmetric deformations, that is,
χ(1) = u(r, θ)er + v(r, θ)eθ . (15.81)
To compute the deformation gradient of χ(1) , we recall that if f = f (r, θ, ϕ), then
1 1
grad( f ) = (∂r f )er + (∂θ f )eθ + (∂ϕ f )eϕ . (15.82)
r r sin θ
Using this identity to compute the gradient of χ(1) leads to

⎡ ⎤
(1) ∂r u (∂θ u − v) /r 0
F = grad χ(1) = ⎣ ∂r v (∂θ v + u) /r 0 ⎦. (15.83)
0 0 (u + v cot θ) /r
The condition (14.105) gives
r ∂r u + 2u + ∂θ v + v cot θ = 0. (15.84)
Step 2: Linearized equations. In order to compute the components of div T(1) = 0

given explicitly by (14.116), we also need the expression for the divergence of an
arbitrary tensor A in spherical coordinates:
1 2
1
[div A]r = ∂r r Arr + ∂θ ( Aθr sin θ)
r 2 r sin θ
1 Aθθ + Aϕϕ
+ ∂ϕ Aϕr − , (15.85)
r sin θ r
1
1
[div A]θ = 2 ∂r r 2 Ar θ + ∂θ (Aθθ sin θ))
r r sin θ
1 Aθr − cot θ Aϕϕ
+ ∂ϕ Aϕθ + , (15.86)
r sin θ r
1
1
[div A]ϕ = 2 ∂r r 2 Ar ϕ + ∂θ Aθϕ sin θ

r r sin θ
1 Aϕr + cot θ Aϕθ
+ ∂ϕ Aϕϕ + . (15.87)
r sin θ r
Then, the third component of div T(1) = 0, written in spherical coordinates, vanishes
identically.
Step 3: Linearized constitutive equations. We adapt the general expressions of

the connections (14.122–14.127) relating the linearized Cauchy stress tensor to the
instantaneous elastic moduli to express the two remaining equations as
∂2u ∂u ∂2u ∂u ∂ (1)

c1 + c2 + c3 + c4 + c5 u = q , (15.88)
∂r 2 ∂r ∂θ2 ∂θ ∂r

∂2v ∂v ∂2u ∂u 1 ∂ (1)
c6 2 + c7 + c8 + c9 −v = q , (15.89)
∂r ∂r ∂r ∂θ ∂θ r ∂θ
with
c1 = L1111 − L1122 − L1212 + α1 W1 , (15.90)

1 −1 ∂ 2
c2 = r r L1111 −r 2 L1122 +L2222 −2L1122 −2L1212

r ∂r

−1 ∂
2
+L2233 +2α1 W1 −α2 W2 +r r α1 W1 , (15.91)

∂r
c3 = r −2 L2121 , c4 = r −2 cot(θ)L2121 , c5 = 2r −2 L2121 , (15.92)

∂ 2
c6 = L1212 , c7 = r −2 r L1212 , (15.93)

∂r
1
c8 = (L1122 −L2222 +L1212 −α1 W1 ) , (15.94)
r
∂ 2
c9 = r −2 r −1 r L1212 − L2222 + L2233 − α2 W2 . (15.95)

∂r
Equations (15.84) and (15.88–15.89) form a system of three linear partial differential
equations for (u, v, q (1) ). The standard method to solve (15.88–15.89) is to express all
variables of r and θ into products of a function of r alone and a Legendre polynomial
in cos(θ). We write
u(r, θ) = Un (r )Pn (cos θ), (15.96)

d
v(r, θ) = Vn (r ) Pn (cos θ), (15.97)
dθ
q (1) (r, θ) = Q n (r )Pn (cos θ), (15.98)
where Pn (cos θ) are the Legendre polynomials for which the following standard
equality [2, Chap. 8] holds:
d2 d
Pn (cos θ) + cot θ Pn (cos θ) + n(n + 1)Pn (cos θ) = 0. (15.99)
dθ2 dθ
Step 4: The boundary-value problem. By substituting (15.97–15.98) into (15.84–
15.89), together with identity (15.99), one can express Vn and Q n as a function of Un
and its derivatives to obtain a closed fourth-order linear ordinary differential equation
for Un :
d 4 Un d 3 Un d 2 Un dUn
C4 + C 3 + C 2 + C1 + C0 Un = 0. (15.100)
dr 4 dr 3 dr 2 dr
In the case of constant anisotropic growth, these coefficients are
C4 = r 4 γ 2 , (15.101)
C3 = 8α r γ,
3 3
(15.102)

6 2

C2 = r 4 − γ (n +n − 2) α + 16γα −γ n + n + 10 , (15.103)
2 2 2 3 2

C1 = r 2 2γ(n 2 +n−2)α9 −4 γ 2 (n 2 +n− 2)−2 α6

−4γ n 2 +n+8 α3 +2γ 2 n 2 +n+10 , (15.104)
2 2
6
C0 = (n +n−2) 4γα + γ (n +n−2) + 4 α − 16γα + 10 γ .
2 9 3 2
(15.105)
The boundary conditions are
T(1) n = 0 on r = a, (15.106)
T(1) n = −P (1) n on r = b. (15.107)
With the help of (14.115), the boundary conditions for the incremental deformation
take the form
T(1) n = L : F(1) + F(1) · A(0) · WA − q (1) 1 n. (15.108)

For the particular homogeneous isotropic case under consideration, the radial and
hoop components of the traction provide two boundary conditions on U (r ) at r =
a, b:
d 3 Un 2
3 d Un
r 3γ + 2r 2
(γ + 2α )
dr 3 dr 2
3 dUn
+r 8α − 2γ(2 + n + n 2 ) − γα6 (n 2 +n−2)
dr
+(−2γα6 + 4α3 − 2γ)(n 2 +n−2)Un = 0, (15.109)
d 2 Un 2 dUn (n 2 +n−2)
+ + U = 0. (15.110)
dr 2 r dr r2
The integration of (15.100) takes place between r = a and r = b and the central
problem of solving this equation lies in finding, for a given initial thickness A0 /B0 ,
the value of a such that the boundary conditions are satisfied (the outer radius b being
a function of a).
In the case of a growing shell, the outer radius b depends not only on the initial
thickness A0 /B0 , but also on the parameters γr and γθ that describe growth through
Equation (15.75).
The mode n = 0 corresponds to a symmetric increase in shell radius and is
not a true axisymmetric bifurcation but a limit-point instability as discussed in
Section 15.3. The stability of a pressurized shell and the critical mode (the first
unstable mode n > 0) depends on whether the pressure is applied internally (P < 0)
or externally (P > 0), its thickness, and the choice of strain-energy functions. If
the pressure is applied internally, a bifurcation analysis [975] shows that the unsta-
ble mode depends on the choice of strain-energy function and the thickness of the
shell. However, neo-Hookean shells under internal pressure are always stable under
axisymmetric perturbations. Therefore, this case will not be further discussed and,
instead, we focus our attention on the case of a growing material subject to external
pressure (P ≥ 0).
15.6.1 A Numerical Scheme
The first step in the analysis of the linearized equation (15.100) is to develop a
numerical scheme that will help us identify the critical values of the parameters
where a bifurcation takes place.
The general problem is to find the values of a parameter (say, λ) for which
there is a solution of a fourth-order equation with mixed boundary values. Let
y = [U, U , U , U ]T be the vector of variables and derivatives. The fourth-order
problem can be written as a system of four first-order equations:
y = f(y; r ), (15.111)
where r is the independent variable. The boundary-value conditions are given by

a set of four linear functions {ci (y(r ); r ), i = 1, 2, 3, 4}, that vanish at one of the
boundaries, namely
c1,2 (y(a); a) = 0, c3,4 (y(b); b) = 0. (15.112)
In order to find values of the parameter for which a solution of the equation with the
proper boundary conditions exists, the general strategy of the determinant method
[80, 553, 554] is used. We consider two copies of system (15.111), that is,
(y(1) ) = f(y(1) ; r ), (y(2) ) = f(y(2) ; r ). (15.113)
The parameter λ in the system is given by the relationship between the endpoints
a and b through relations such as (15.76). We integrate (15.113) with two linearly
independent conditions y(1) (a), y(2) (a) that satisfy both conditions at r = a (that is,
c1,2 (y(1,2) (a); a) = 0). Then, we carry the integration up to a point r = b, at which
the determinant
c3 (y(1) (r ); r ) c3 (y(2) (r ); r )

(r ) = , (15.114)
c4 (y(1) (r ); r ) c4 (y(2) (r ); r )
vanishes identically. By the linear superposition principle, if (b) = 0, then there

exists a linear combination of y (1) and y (2) that satisfies System (15.111) with bound-
ary values (15.112).
Note that the value of b is not known a priori and the integration proceeds with
a stop condition on (b) = 0. However, once b is known, we deduce the value of
A0 /B0 from (15.76).
As an application of this method, we consider the case of an internally pressurized
shell (P < 0) in the absence of growth. A typical plot of the strain αa , required to
excite different modes for various thicknesses, is shown in Figure 15.17 for a neo-
Hookean material. Once the critical strain αa is known for the nth mode, αb can be
computed from (15.27) and the critical pressure necessary to excite the mode n is
given by (15.44).
The following features appear:
• Thin shells are more prone to instabilities than thicker shells since the critical
strain increases with thickness;
• Unstable modes are not ordered. In particular, as the thickness decreases or
increases, higher modes can become unstable first;
• For thick shells, a detailed numerical analysis reveals that the limit A0 /B0 → 0
the n = 10 is the most unstable mode [465].
This computational analysis suggests that there are two regimes where further
analytical progress is possible, the thin-shell limit A/B → 1 and the thick-shell
limit A0 /B0 → 0. We consider these two limits separately and look at the effect of
the growth parameters on the stability of the shell.
(n)
αa
1 μ 8μ 10
μ
μ5 6
μ4
μ3
0.95
μ2
0.9
0.85
0.8
n=2 3 4 5 6
0.75 8 10
15 20
30
0.7 40
n=10 50
0.65
0.4 0.5 0.6 0.7 0.8 0.9 1
Figure 15.17 Critical strains αa for the nth mode of a neo-Hookean shell as a function of the
thickness A0 /B0 . Only the modes n = 2, 3, 4, 5, 6, 8, 10, 15, 20, 30, 40, 50 are shown. For a given
thickness A0 /B0 , as the external pressure is increased, the inner boundary of the shell moves inward
and the strain αa decreases until it reaches a critical curve. The critical mode is the first unstable
mode. For instance, at A0 /B0 = 6/10 the critical mode is n = 2, whereas for a thinner shell at
A0 /B0 = 3/4, the critical mode is n = 3 (both indicated by a black dot). Note that the amplitude
of the modes is arbitrary and has been chosen to show the structure of the solution.
15.6.2 Thin-Shell Limit
The numerical procedure for the calculation of the critical values at which bifurca-
tions occur can be adapted for the calculation of asymptotic solutions in the thin-shell
limit [397]. For thin shells, one can obtain analytic results by introducing the small
parameter = (b − a) and the stretched variable
r −a
ρ= . (15.115)

System (15.113) now reads

dz
= g(z; ρ; ), (15.116)
dρ
where z = [y(1) , y(2) ]T , g = [f(y(1) ), f(y(2) )]T , and the integration takes place
between ρ = 0 and ρ = 1. We seek solutions of this system of eight equations with
initial values z(ρ = 0; ) = [y0(1) , y0(2) ]T = z0 () where we choose y0(1) and y0(2) to be
linearly independent such that the conditions
c1,2 (y0(1) ; ρ = 0; ), c1,2 (y0(2) ; ρ = 0; ), (15.117)
given by (15.112) are identically satisfied. We look for solutions analytic in

k
z= zi i + O(k+1 ), (15.118)
i=0
where each zi satisfies an equation of the form
dzi
= gi (z1 , z2 , . . . , zi ), z(0)i = z0,i , (15.119)
dρ
where the linear vector fields giand the initial values

z0,i are, respectively, the coef-
ficients of the expansions g = i gi i and z0 = i z0,i i . This hierarchy of linear
systems can be integrated explicitly for increasing values of i and the solution z1,i at
ρ = 1 can be computed.
The solution at the outer boundary z1 = i z1,i i is a function of the unknown
parameter a (or, equivalently, the parameter αa ) and the condition for the existence
of a solution of the original problem is the vanishing of the determinant

c (y(1) (1)) c3 (y(3) (1))
= 3 (1) , (15.120)
c4 (y (1)) c4 (y(2) (1))
where [y(1) (1), y(2) (1)]T = z1 . We expand αa in powers of and solve = 0 to each
order. To third order, the solution reads
αa(n) = α0 (n) + α̃1 (γ, n) + α̃2 (γ, n)2 + O(3 ), (15.121)
where α0 is the first positive root of
(n + 2)(n − 1)α012 + 2(n 2 + n + 7)α06 − 3n(n + 1) = 0, (15.122)
and
α03 1
α̃1 = − , (15.123)
2γγθ3 2

1
α̃2 = −
72γθ 3 γ 2
−462 α0 6 γ 2 γθ 3 − 342 n 2 γα0 9 + 33 n 2 γ 2 γθ 9 + 98 nγ 2 γθ 9 + 6 n 4 α0 9 γ
−330γθ3 nα06 + 90 n 4 γα0 3 γθ 6 − 41 n 4 γ 2 γθ 9 + 672 γα0 9 + 24 n 5 γ 2 γθ 9
−180 γnα03 γθ6 − 348 nγα0 9 + 345 nγ 2 α0 6 γθ 3 − 57 n 4 α0 6 γθ 3 γ 2
+288 α0 6 γθ 3 n 2 γ 2 + 180 n 3 γα0 3 γθ 6 − 90 n 2 γα0 3 γθ 6 − 122 n 3 γ 2 γθ 9
+90 γθ 9 n 3 + 12 α0 9 n 3 γ + 45 γθ 9 n 4 + 315 γθ 9 n − 150 γθ 3 n 3 α0 6 −
−75 γθ 3 n 4 α0 6 + 8 n 6 γ 2 γθ 9 − 1470 γθ 3 α0 6 − 114 n 3 α0 6 γθ 3 γ 2

+360 γθ 9 n 2 + 405 γθ3 n 2 α0 6

× − 3γθ6 n 2 − α06 n 4 − 2α06 n 3 + 14α06 − 6n 2 α06
−1
−5nα06 − 6γθ6 n + 6n 3 γθ6 + 3n 4 γθ6 . (15.124)
Higher order coefficients can be computed in a similar manner. To obtain the critical
curves in terms of the parameter δ = (B0 − A0 )/A0 , we use the expression for the
volume of the deformed shell given by (15.16):
b3 − a 3 = γγθ3 (B03 − A30 ). (15.125)
To second order, this relation reads = γγθ3 δ/α02 and the expansion of the critical
curves in terms of δ becomes
αa(n) = α0 (n) + α1 (γ, n)δ + α2 (γ, n)δ 2 + O(δ 3 ), (15.126)
where i
γγθ3
αi = α̃i , for i > 0. (15.127)
α02
In Figure 15.18, the comparison between the numerical curves and the expansion
for thin shells in the absence of growth is given. In this case, we have
αa(n) = α0 (n) + α1 (1, n)δ + α2 (1, n)δ 2 + α3 (1, n)δ 3 + O(δ 4 ). (15.128)
The thin-shell expansion gives a satisfactory approximation of the numerical curves.

Moreover, the limiting value, corresponding to = δ = 0, is the exact critical strain.
We can also calculate an expression for the envelope of the critical curves
1
αa
0.95
0.9 Envelope n=2
0.85
0.8
0.75
3
4
0.7
5 6 8 10 15 20 30 40
50
Α/Β
0.65
0.6 0.7 0.8 0.9 1
Figure 15.18 Blow-up of the critical curves (solid lines) of Figure 15.17 together with the asymp-
totic curves (dotted lines) valid for thin shells and the critical envelope given by (15.129).
1 7 2 235 3
αa = 1 − δ − δ + δ + O δ3 . (15.129)
3 36 864
This envelope shown in Figure 15.18 is particularly important as it gives the critical
value of the bifurcation parameter at which the spherical shell becomes unstable.
15.6.3 Thick-Shell Limit
Equation (15.100) with boundary conditions (15.109) and (15.110) also admits a
WKB expansion [82] valid in the large n limit. Such solutions have been computed
in the classic case [397]. This expansion can be captured by setting
r
U (r ) = exp S(r )dr , (15.130)
a
with
∞

S=n Si (r )n −i . (15.131)
i=0
The functions Si (r ) can be obtained by substituting (15.130) in (15.100) and the

resulting systems of equations can be solved to each order in n. Since the problem
is of fourth order, the equation to order n −4 has four distinct solutions given by
1 α3
S0(1,2) = ± , S0(3,4) = ± . (15.132)
r r
Each of these branches gives rise to a separate solution
∞

r r
U (k) = exp S (k) dr = exp (n Si(k) n −i )dr . (15.133)
a a i=0
The equations for Si(k) with i > 0 are linear with coefficients depending only on
{S0(k) , . . . , Si−1
(k)
}.
We test the existence of a solution to the boundary-value problem by looking
for a linear combination of the four fundamental solutions satisfying the boundary
conditions. Such a solution exists when the determinant of the matrix built on the
four boundary conditions (15.109–15.110) evaluated on the four solutions vanishes
identically.
Let c1,2 (U ) and c3,4 (U ) be the boundary conditions (taken respectively at r = a
and r = b) and
Mi j = ci (U ( j) ), i, j = 1, . . . , 4. (15.134)
Then, the condition for the existence of a solution to the boundary-value problem in
the large n limit is simply given by the vanishing of
= det(Mi j ). (15.135)
However, the values of the parameters at which this determinant vanishes have to be
evaluated numerically.
Further theoretical progress can be achieved for thick shells, that is, for shells
such that (B0 − A0 )/A0 = O(1). In such a case, the expansion of the determinant
is itself an asymptotic series. An excellent approximation of the critical curves can
be obtained by finding the roots of the exponentially dominant term. This analysis
leads to
1 1 1 1
αa = α0 + α1 + α2 2 + α3 3 + O( 4 ), (15.136)
n n n n
where α0 is the first positive root of

1 √ 1/3 √ 2/3
α0 = 576 26 + 6 33 +234 26 + 6 33
12
1/3
√ 2/3 √
−576−54 26 + 6 33 33
≈ 0.6661423391, (15.137)
and
− (3γ + 11) α0 6 − (32 + 4γ) α0 3 + 9 + 5γ

α1 =
, (15.138)
12γα0 5 α0 3 + 2

α2 = (−70752γ −31349−3111γ 2 )α06 −(164574+73724γ −45474γ 2 )α03

−13163γ +27972γ +51387
2

3
3
2 −1
× 18 α0 3 +2 γ 2 α014 (α0 −1)3 (α0 +1)2 α0 2 +1+α0 α0 2 +1−α0 .
(15.139)
The limit from thick shell to the sphere can be obtained by taking the limit A0 → 0 at
fixed B0 or, alternatively, by taking A0 constant and B0 → ∞. This last limit corre-
sponds to a half-space under pressure and we recover the limiting value αa ≈0.66614
in the absence of growth obtained in Section 11.10.1 for such an instability.
15.6.4 Bifurcation of the Growing Shell
We now consider the effect of anisotropic growth on stability. We focus our attention
on radial growth and resorption. The effect of hoop growth is equivalent to radial
resorption up to an isotropic change of volume that can be taken into account by a
rescaling of the pressure (15.12).
15.6.4.1 Mechanical Versus Geometric Effects
In a growth process, there are two main effects that can influence the stability of a
structure. The first effect is due to geometry. When the volume of the spherical shell
changes, the relative thickness with respect to its radius also changes, affecting its
stability. For instance at a given pressure, a reduction in thickness after resorption
tends to destabilize the shell, since thinner shells are more unstable than thicker ones.
Likewise, an increase in the thickness after growth stabilizes the shell.
The second effect is related to mechanics through changes in residual stress. Radial
growth induces a compressive radial stress as observed in Figure 15.5A. This stress
tends to further destabilize the shell under pressure. Similarly, we see in Figure 15.5B
that radial resorption (or hoop growth) induces a tensile radial stress that stabilizes
the shell.
To illustrate these two opposing effects, we show the critical pressure Pcrit for the
mode n = 10 as a function of thickness and for small values of radial growth and
resorption in Figure 15.19.
Pcrit
7 γ=0.8 Fiber resorption
Thinner shells
Tensile residual stress
6
5
γ=1 n=10,
4
γ=1.2
3
Fiber growth
2 Thicker shells
Compressive residual stress
1
0
0.4 0.5 0.6 0.7 0.8 0.9 1
Figure 15.19 Critical pressure for mode n = 10 as a function of the initial thickness for radial
growth and resorption. Geometric effect of growth: For thin shells (A0 /B0 close to 1), radial
growth is stabilizing (larger critical pressure) and radial resorption is destabilizing (smaller critical
pressure). Mechanical effect of growth: For thick shells, radial growth is destabilizing and radial
resorption is stabilizing.
For thin shells (A0 /B0 close to 1), radial growth is stabilizing. A small change
in thickness has a direct effect on the critical strain that produces an instability (see
Figure 15.17 in the thin shell regime).
For thick shells, a change in thickness does not significantly change the critical
strain needed for the existence of an axisymmetric solution (see Figure 15.17 for
A0 /B0 < 1/2).
The effect of growth on thin shells can be explicitly understood by using the thin-
shell approximation and computing asymptotically the critical displacement for each
mode:
αa(n) = α0 (n) + α1 (γ, n)δ + α2 (γ, n)δ 2 + α3 (γ, n)δ 3 + O(δ 4 ), (15.140)
where δ = (B0 − A0 )/A0 and the coefficients αi (γ, n) are explicitly given
in (15.126). The critical mode number, n crit , is the unstable mode number n with
the largest value of αa(n) . It is found, as a first approximation, by the solution of
∂αa /∂n = 0. Once n crit is known, the envelope giving the critical value of the strain
for instability can be computed explicitly as αa (n crit ). In our case, it is given by

1 5 5 3γ 25 2 2
αa = 1 + − γ δ+ − + γ δ
2 6 24 4 72
γ
− 21 − 168 γ − 88 γ 2 δ 3 + O(δ 4 ). (15.141)

864
Pcrit
n=10
0.8
γ=3
0.6
0.4
0.2
0 0.5 0.6 0.7 0.8 0.9 1
-0.2
Figure 15.20 Growth-induced buckling: For larger values of γ, there exists a value of A0 /B0 such
that the mode n = 10 is unstable in the absence of external pressure
From this envelope, the critical pressure can be calculated explicitly by integrat-
ing (15.44) for a neo-Hookean energy:
2
Pcrit = 8(γδ)2 + (γδ)3 + O(δ 4 ), (15.142)
3
and, recalling that γ = γr /γθ , we have

d Pcrit
= 16δ 2 + 2δ 3 + O(δ 4 ), (15.143)
dγr (γr =1,γr =1)

d Pcrit
= −16δ 2 − 2δ 3 + O(δ 4 ). (15.144)
dγ θ (γr =1,γr =1)
We conclude that for thin shells, radial growth or hoop resorption has a stabilizing
effect, whereas radial resorption or hoop growth is destabilizing.
15.6.4.2 Spontaneous Instabilities
Another feature of interest can be seen in Figure 15.19. For thick shells and radial
growth, the pressure necessary to destabilize the shell decreases with the thickness.
This trend is due to the residual compressive stress associated with radial growth.
It suggests that, for sufficient growth, the shell might become unstable without any
external applied load. Indeed, for large enough values of γ, we find that there exists
γcrit
3 n=80
n=∞
n=40
2.5
1.5
Α0
/Β0
1
0 0.1 0.2 0.3 0.4 0.5
Figure 15.21 The critical value of radial growth necessary to induce a spontaneous instability in
a shell of initial thickness A0 /B0 for the modes n = 40, 80, ∞. Note that the WKB expansion is
not valid for low modes so the value of γcrit given by the mode n = ∞ is an upper bound for the
instability.
a value of A0 /B0 such that the mode n = 10 becomes unstable in the absence of an
external pressure, as shown in Figure 15.20.
Therefore, we can compute for a given A0 /B0 the value of γ = γcrit necessary to
induce an instability. To do so, we consider thick shells and high-mode numbers so
that the WKB expansion given in (15.136) is valid. In this regime, we compute γcrit
leading to an instability at zero pressure. In Figure 15.21, we plot the critical radial
growth γcrit as a function of the initial thickness for high-mode numbers (including
n = ∞ obtained by taking αa = α0 in the WKB expansion (15.136)). The exact
threshold for an instability is not known since the WKB expansion is only valid for
high-mode numbers and the first instability for thick shells occurs at mode n = 10.
Nevertheless, the value of γcrit provides an upper bound for the instability.
15.7 Instability Due to Differential Growth
Isotropic and inhomogeneous growth also produces residual stress. This stress field
may be sufficient to create an instability. Following the method presented in the
previous two sections, we consider the possibility of a bifurcation for the two cases
of differential growth discussed in Section 15.2.2.
15.7 Instability Due to Differential Growth 471
χ(0) χ(1)
Figure 15.22 The shell is first shrinking through differential growth until it builds enough residual
stress and becomes unstable (here, mode n = 15).
15.7.1 Instability in a Shrinking Shell
We first consider the case of a shell of initial radii A0 , B0 and current radii a, b that
shrinks according to the law
γr = γθ = g(r ) = 1 + ν(r − a), (15.145)
as shown in Figure 15.22. The constant ν is directly related to the change of volume
b3 − a 3
V = , (15.146)
B03 − A30
through
b
ρ2
B03 − A30 = 3 dρ. (15.147)
a g 3 (ρ)
The strain and stress for the radially symmetric solution are given by
r
α(r ) = , (15.148)
g(r )R0 (r )
r
α
tr(0) (r ) = ∂α Wsh (α) dρ, (15.149)
a ρ
where we choose W sh to be the auxiliary function for a neo-Hookean strain-energy

function.
We follow the same steps as before to obtain a linearized system of equations.
We use the spherical coordinates (r, θ, ϕ) and consider axisymmetric deformations
of the shell in the form
χ(1) = u(r, θ)er + v(r, θ)eθ . (15.150)
The nonvanishing components of div T(1) = 0 and the incompressibility condition

form a system of three coupled PDEs of second order for u, v and q (1) as a function
of (r, θ) with coefficients depending on the base finite-strain solution. Again, we
expand this solution in Legendre polynomials Pn (cos θ), that is, we write
u(r, θ) = Un (r )Pn (cos θ), (15.151)

d
v(r, θ) = Vn (r ) Pn (cos θ), (15.152)
dθ
q (1) (r, θ) = Q n (r )Pn (cos θ), (15.153)
and obtain, after simplification, a fourth-order differential equation describing the

amplitude of the field u with respect to the nth Legendre polynomial.
For each mode n, a solution of this boundary-value problem exists only for par-
ticular combinations of parameters a = a(ν, δ, n) where δ = (b − a) is the width of
the shell in the current configuration. These values of a are the critical values of the
inner radius where the nth mode first appears.
For small δ, these curves in parameter space can be obtained by a perturbation
expansion of the form
a = a0 + a1 δ + a2 δ 2 + O(δ 3 ), (15.154)
where a0 is the first positive solution of
(n + 2)(n − 1)a012 + 2(n 2 + n + 7)a06 = 3n(1 − n), (15.155)
and
1 3 ν 1
a1 = a + a0 − . (15.156)
2 0 2 2
The coefficient a2 is given by a similar expression in terms of ν and a0 . Once the
bifurcation of each individual mode is known, the critical value acrit where the shell
first bifurcates (independently of the mode number) is

1 1 1
acrit = 1 − − ν δ− 2 + 4ν + 5ν 2 δ 2 + O(δ 3 ). (15.157)

3 2 24
For a given δ, the value of acrit and bcrit = acrit + δ can be used in Equation (15.149)
to solve tr(0) (bcrit ) = −P with respect to ν. The value of νcrit = ν(acrit , bcrit ) can then
be used to compute the initial radii A0 and B0 , and hence, the current thickness.
The selected modes shown in Figure 15.23 are given by the largest integer n less
than

3 δ 8 + 3ν 29ν 2 + 220ν + 282 2
N= 1− + δ− δ . (15.158)
δ 12 24 384
15.7 Instability Due to Differential Growth 473
Current
thickness
(a/b)
1 P=0
n >9
9
0.95 8
7
n=6 P6
0.9
n=5 P5
0.85
n=4
P4
0.8
Initial
0.75
thickness
(A/B)
0.2 0.4 0.6 0.8 1
Figure 15.23 Instability of the shrinking shell. For an initial thickness A0 /B0 , the shell first
becomes unstable due to residual stress when the current thickness a/b reaches the critical value
given by the solid curves. The solid curves are obtained at constant pressure Pi given by the critical
value for instability of the ith mode without growth. The dashed lines denote the regions of the
mode of deformation n found at the bifurcation. The corresponding configurations n = 4, 5, 6 are
shown, but their amplitudes are not predicted by the theory.
The mode selected at the bifurcation depends on the initial thickness in such a
way that thinner shells become unstable with increasingly high-mode numbers, as
discussed in the previous section.
The basic physical process of the instability shown in Figure 15.23 is as follows:
A shell of thickness A0 /B0 becomes unstable when both its current thickness a/b
is small enough and sufficient compressive residual stress has built up in the shell.
Under a pressure Pn chosen to be the critical pressure for mode n, the shrinking shell
becomes unstable with possible mode 2 ≤ m ≤ n.
15.7.2 Instability of a Shell Growing Inside a Medium
We now turn our attention to the case of a growing shell inside an incompressible
elastic medium. The stability analysis in such interfacial problems is delicate due to
the matching of stresses at the interface in the deformed configuration. Nevertheless,
some insight can be gained by modeling the problem with the assumption that the
outside medium creates a hydrostatic compressive force equivalent to the one created
through growth and contact. In this context, we compute a lower bound for instability
Volume
increase (%)
150
n=4 n=5 6 7 8 9 n>9
n=3
140
Δμ=1/4
130
Δμ=1
120
110 Δμ=4
Initial
thickness
(A/B)
100
0.82 0.84 0.86 0.88 0.9 0.92 0.94 0.96 0.98 1
Figure 15.24 Instability of a growing embedded shell. For an initial thickness A0 /B0 , the shell
first becomes unstable due to contact stress when the volume increase (100v/V ) reaches the critical
value given by the solid curves (with different relative medium responses μ = 1/4, 1, 4). The
dashed lines represent the different instability modes.
in terms of the initial thickness and the volume increase for different relative medium
responses μ = μme /μsh .
The linearized equation for the instability threshold is the same one as in the
previous section and the same estimate for small current thicknesses can be used.
For each δ, the value of ν necessary for the radial problem to satisfy the boundary
values can be computed. Once this value is known, the initial thickness and volume
of the current configuration can be found for different values of μ as shown in
Figure 15.24.
This analysis shows the effect of differential growth on the stability of shells. In a
shrinking shell, two effects drive the instability: The shell thickness decreases and the
residual stress is compressive. Not surprisingly, in the absence of hydrostatic pressure,
the neo-Hookean shell is stable under moderate shrinking and only becomes unstable
when a large portion of the initial volume has been resorbed. Under compressive
loads, the shell can rapidly become unstable with lower or equal mode number than
the one given by the initial thickness.
In the case of a growing shell creating compressive stress through contact, the
instability for a thin shell is driven rapidly by the volume increases and the rigidity
of the external medium.
Chapter 16
Growing Cylinders
Among the many typical biological structures, tubular structures such as hollow
stems and blood vessels abound in nature. Tubes are typically used for transport, to
carry information, or for mechanical support. Their morphogenesis usually involves
complex genetic and biochemical processes [1018] mediated by mechanical forces.
Residual stress plays a particularly important role in tubular structures: It regulates
hoop stress gradients in blood vessels [407], it is used to improve the rigidity of
growing plant stems [1274], it mechanically stabilizes arteries [485], and it is known
to assist in the proper functioning of airways and oesophagus [534]. In this chapter,
we study different aspects of growing tubular structures.
First, we look at the problem of cavity opening that transforms cylinders into
tubes. Second, we consider several types of instability in growing structures and
from their applications in nature. Third, we study the surprising rotation of some
tubular fungi. We start with a general study of growing incompressible cylindrical
structures.

476 16 Growing Cylinders
Figure 16.1 A. Geometry

of the deformation under
axial force N and pressure N
P, the tube on the left is in
its initial configuration, the a b
tube on the right is in the
current configuration, due to
both growth and elasticity. A0 B0
A1 A2 ...An
H0 P h
16.1 Kinematics of the Growing Cylinder
We consider the general problem of a growing cylindrical structure consisting of N

different elastic, isotropic, and incompressible cylinders each defined by a diagonal
growth tensor and with, possibly different, but homogeneous, material properties. The
deformation and the interpretation of the growth tensor are sketched in Figure 16.1.
When the growth tensor is taken to be piecewise constant, the problem is a gener-
alization of the inclusion problem which consists in inserting pieces of elastic mate-
rials inside an elastic body by deforming both the inclusions and the body itself [348,
1354]. The cylindrical growth problem is the generalization of the circular shrink-fit
problem which consists in fitting a disk in a circular opening with a different radius
[35, 692]. It is highly relevant in biology where many cylindrical structures can be
described as multiple cylindrical layers with residual stresses. The overall properties
of these structures are inherited both from the elastic response of the individual layers
and from their relative geometric arrangement. The relative geometry of the different
layers is fully described by the growth of an initial stress-free multilayer cylinder.
16.1 Kinematics of the Growing Cylinder 477
In the initial configuration B0 , the structure has inner and outer radii A0 and
An ≡ B0 , and height H0 . This cylinder is further divided into n cylindrical shells
labeled by i from 1 to n so that the inner and outer radii of the ith cylindrical shell
is Ai−1 and Ai , respectively, as shown in Figure 16.1. Each shell is homogeneous,
isotropic, and incompressible.
We assume that the cylinder retains its integrity and cylindrical geometry in all
deformations. Following Section 11.8.3, the total deformation from the initial to
the current configuration, x = χ(X0 , t), in cylindrical coordinates, {r, θ, z} and
{R0 , 0 , Z 0 } reads
r = r (R0 ), θ = 0 , z = ζ Z 0 , (16.1)
where ζ is the constant axial stretch of the cylinder such that h 0 = ζ H0 , and the
continuous function r (R0 ) gives the location of circles of radius R0 after deformation.
The total deformation gradient is then
r
F = r er ⊗ er + eθ ⊗ eθ + ζez ⊗ ez , (16.2)
R0
where the prime denotes differentiation with respect to R0 . Since the cylindrical
geometry is respected in all configurations, we can use the same bases vectors:
E R0 = er and so on. We introduce the deformation gradient in terms of the principal
stretches
λi (R0 ) = (λi,r , λi,θ , λi,z ), (16.3)
where R0 ∈ [Ai−1 , Ai ), for i = 1, . . . , n − 1 and R0 ∈ [An−1 , An ], when i = n.

The deformation gradient reads:

n

[F(R0 )] = diag(λi ) H(R0 − Ai−1 ) − H(R0 − Ai ) , R0 ∈ [A0 , An ], (16.4)
i=1
where H(·) is the Heaviside function.

Using the multiplicative decomposition, we assume that F = AG where the
elastic deformation tensor is defined in terms of the elastic stretches αi (R0 ) =
(αi,r , αi,θ , αi,z ), so that

n

[A(R0 )] = diag(αi ) H(R0 − Ai−1 ) − H(R0 − Ai ) , R0 ∈ [A0 , An ].
i=1
(16.5)
Similarly, the growth tensor is defined in each shell by a growth vector γ i
γ i = (γi,r , γi,θ , γi,z ). (16.6)

z
Axial growth
γz >1
γr >1
Radial growth
Hoop growth
γθ>1
Figure 16.2 Local deformation of a cylindrical element due to the three growth functions con-
sidered. A cylindrical element in the initial configuration is transformed into another cylindrical
element in the virtual state by increasing its height, depth, or angular span.
The geometric interpretation of this special choice of growth tensor is depicted in

Figure 16.2. For the entire tube, we define the growth function

n

g(R0 ) = diag(γ i ) H(R0 − Ai−1 ) − H(R0 − Ai ) , R0 ∈ [A0 , An ]. (16.7)
i=1
Due to the incompressibility constraint, we have
det F = det(AG) = det G = g. (16.8)
That is,
dr 1
r = g(R0 )R0 , (16.9)
dR0 ζ
which can be solved to give

R0
2
r 2 = a2 + g(ρ)ρ dρ. (16.10)
ζ A0
16.1 Kinematics of the Growing Cylinder 479
For a given growth function g(R0 ), we define the position of the ith radius as ai =
r (Ai ) in the current configuration.
The deformation is diagonal in cylindrical coordinates and only depends on R0 .
Therefore, it follows from Equation (11.156) that the Cauchy stress tensor is also
diagonal in these coordinates:
[T] = diag(tr , tθ , tz ), ⇔ T = tr er ⊗ er + tθ eθ ⊗ eθ + tz ez ⊗ ez . (16.11)
The Cauchy equation, div T = 0, in cylindrical coordinates, reduces to
dtr 1
+ (tr − tθ ) = 0. (16.12)
dr r
It is then possible to solve for the radial stress using the expression for the single
cylinder given by Equation (13.54), that is

n

tr (R0 ) = τ (R0 ; Ai−1 , a, ζ)H(R0 − Ai−1 ) − τ (R0 ; Ai , a, ζ)H(R0 − Ai )
i=1
(16.13)
with
gi R0
τ (R0 ; Ai−1 , a, ζ) ≡ w(R0 )R0 dR0 , (16.14)
ζ Ai−1
and where
tθ − tr
w(R0 ) = . (16.15)
r2
For a strain-energy function W = W (αr , αθ , αz ), we have

1 ∂W ∂W
w(R0 ) = αθ − αr , (16.16)
r2 ∂αθ ∂αr
where
R0 r (R0 ) ζ
αr = γθ γz , αθ = , αz = , (16.17)
r (R0 )ζ R0 γθ γz
so that w(R0 ) is a piecewise function of R0 .

Once the radial component of the stress is known, the hoop stresses are given by
∂W ∂W
tθ = tr + αθ − αr , (16.18)
∂αθ ∂αr
∂W ∂W
tz = tr + αz − αr . (16.19)
∂αz ∂αr
Note that since tr (R0 ) is given by the integral of a piecewise continuous function, it is
also continuous. Continuity for the radial component of the stress tensor is expected
since we require, physically, continuity of the traction vector at the interface between
two cylindrical shells. However, in general, both the hoop and axial stress components
given by (16.18–16.19) will be discontinuous at the interfaces.
The boundary conditions have been discussed in Section 11.11.4. For given pres-
sure P and axial force N , they are given by
b
tθ − tr
C1 : dr = P, (16.20)
a r
b
C2 : π (2tz − tr − tθ )r dr = N − χPπa 2 , (16.21)
a
where χ = 1 for a capped cylinder, and χ = 0 for an open cylinder. These conditions
provide, for given P and N , the values of the current inner radius a = r (A0 ) and
the axial stretch ζ that fully determine the deformation and stress at all points in the
structure.
As an example, we consider the case where the growth tensor is piecewise constant
in a four-layer cylinder and plot the resulting stress components in Figure 16.3. We
observe the continuity of the radial stress and discontinuity of the hoop and axial
stresses.
16.2 Application: Cavitation in Plants
We first consider the possible role of mechanical stress in the opening of cavities
in elastic tissues. In Section 15.5.2, we observed that cavitation can be triggered
by either anisotropic or inhomogeneous growth in spherical geometry. However, for
typical materials, elastic cavitation is not possible in cylindrical geometry, since a
similar computation to the one performed in Section 15.5.2 reveals that cavitation is
only possible for cylinders if W ∼ αn with n < 2 which rules out the neo-Hookean
case. Nevertheless, the same conditions that create cavitation in spheres will create
large tensile stress at the core of a cylinder. These stresses may be sufficient to rupture
the material. We investigate this possibility in the context of cavity opening in the
world of plants following [469].
16.2.1 Background
Most people are familiar with the simple observation that many plants such as dande-
lions, chives, and bamboo have hollow stems. Similarly, many roots and water plants
have tissues with large airy tissues known as aerenchyma as shown in Figure 16.4.
These pockets of gas in the plant tissues fulfill many functions such as mechanical
economy and rigidity for hollow stems [583], buoyancy for hydrophytes, increased
16.2 Application: Cavitation in Plants 481
gas transport and sap flow, change in the scattering of light for chlorophyl production,
and seed projection [1064].
In the nineteenth century, Sachs [946, 1107] characterized the formation of
aerenchyma as being either lysigenous, that is, created through uniform growth and
cell death, or schizogenous, that is, air spaces are created through a process of dif-
ferential growth creating tension in tissues that leads to the mechanical tearing of
adjoining cells [352].
Schizogenous formation is associated with large and regular structures in plant
tissues such as hollow stems of which Sachs writes [1107]: “I will refer only to
the one obvious fact that while the pith is no longer able to grow in proportion, it
A. γ(R0)
1.4
1.3
1.2
1.1
1.0
0.9
R0
1.0 1.2 1.4 1.6 1.8 2.0
B.
1.0 tθ
0.5 tz
t R0
r
1.2 1.4 1.6 1.8 2.0
- 0.5
- 1.0
- 1.5
- 2.0
Figure 16.3 Residual stress in a four-layer cylinder. Here, we choose N = P = 0, isotropic

growth in all layers (γi = γi,r = γi,z = γi,θ ) with γ1 = 1.2, γ2 = 1.4, γ3 = 1, γ4 = 1.1 and
μ = 1, ζ = 1, A0 = 1, A1 = 1.2, A2 = 1.3, A3 = 1.7, A4 = 2.
A. C.
B. D.
E.
F.
Figure 16.4 A. Aerenchyma tissue in the roots of corn (Zea mays L., from Kramer [726]),
B. Diaphragms in a sliced stem, C. Hollow stem in hollyhock Althaea rosea., D. Transverse section
of the flowering scape of chive (Allium Schoenoprasum, from Sachs [1107]), E. Aerenchyma in
water plant helps buoyancy (cat-tail: Sparganium emersum). F. Patterns in the stems of horsetail
(Equisetum, picture courtesy of Thomas Schoepke).
becomes ruptured while a cavity arises in the interior. This may easily be observed in
the flower stems of the Teazel and Dandelion.” While lysigenous aerenchyma forma-
tion has received considerable attention by plant physiologists [725], schizogenous
aerenchyma formation with its combined mechanical and developmental regulation
is not well understood [657].
In a beautiful study, Takano et al. [1219] analyzed the effect of mechanical stress
and gibberellins on stem hollowing in bean plants (see also [173]). Gibberellins are
a well-known class of plant growth hormones involved in stem elongation [930].
In Takano’s experiment, the application of gibberellin to bush bean plants, which
are naturally not hollow, increases the length of the stem, reduces its diameter, and
induces stem hollowing. Conversely, in bean pole plants, which are naturally hollow,
the inhibition of gibberellin or mechanical stresses induced by rubbing the stem
cause an increase in the thickening, a reduction in axial length, and prevent stem
hollowing [660, 1043].
While Takano’s experiment suggests a connection between stem hollowing and
differential growth, other authors have argued that cavity opening is the result of a
chemical lysing process acting on the cell wall [173]. It is therefore of interest to
study the possible role of mechanical stresses in this process.
16.2.2 The Model
We consider a simple model of stem growth where the stem is an incompressible

neo-Hookean cylinder with initial outer radius B0 subject to radial, angular, and axial
deformation and growth along the three cylindrical coordinates (r, θ, z). As discussed
above, cavitation is not possible for a neo-Hookean cylinder. Therefore, we explore
whether growth-induced stresses can be sufficient to induce rupture. Assuming that
the cylinder retains its symmetry during deformation, the deformation tensors given
in the previous section reduce to
[F] = diag(r , r/R, ζ), (16.22)

[G] = diag(γr , γθ , γz ), (16.23)
[A] = diag(1/(αζ), α, ζ), (16.24)
where α = αθ . The radial stress is then given by

R
γr γz wα
tr (R) = tr (0) + dρ, (16.25)
0 ζγθ αρ
where w(α) = W (1/(αζ), α, ζ). Added to this last equation are the usual boundary
conditions tr (B) = 0 and zero resultant load on the top and bottom, leading to the
condition
b
r tz (r, ζ) dr = 0. (16.26)
0
Together, these conditions fix the strains α, ζ and the tension at the origin, tr (0).
Following known patterns of growth in stems [1029], we assume that on a given
cross section, growth is isotropic and inhomogeneous using a linear dependence with
respect to the radius R0 . Further, we assume that axial growth creates tissue tension
by being larger in the pith than the epidermis [1030, 1274], that is,
γr = γθ = ν1 + ν2 R0 , γz = μ1 + μ2 (B0 − R0 ). (16.27)
Keeping the growth gradients ν2 and μ2 constant, an increase of the parameter μ1

corresponds to axial growth and an increase of ν1 induces stem thickening which
theoretically reproduce the changes in growth patterns corresponding to Takano’s
experiments.
16.2.3 Analysis
We compute the maximal radial stresses in the cylinder cross section for various
situations. The axial stress is always compressive due to the fact that the pith grows
faster than the outer layers [1030, 1274]. However, both hoop and radial stresses
at the origin (tr (0) and tθ (0)) are equal and positive (i.e., tensile) for most realistic
values of the growth parameters.
Due to the zero load boundary condition on the outer edge and the form of
the imposed growth laws, cross-sectional rings expand in the virtual state, but are
restricted by the solid cylinder geometry in the current configuration creating a tensile
radial stress. It follows that the hoop and radial stresses are maximal at the origin.
In Figure 16.6, we plot this maximal value as a function of the axial and sectional
growth parameters (μ1 and ν1 respectively). In addition, we plot the value of breaking
stress for pith tissue estimated from [955] in which the Young’s modulus for the pith
is around 1 Mpa (Figure 1 in [955]) and values of the breaking stresses are around 0.3
MPa (Figure 2 in [955]). Since we have rescaled the stresses by setting the Young’s
modulus E = 3, the breaking stress is around tbreak = 1 in rescaled variables.
16.2.4 Discussion
The analysis shows that the sign (compressive or tensile) and magnitude of the
mechanical stresses acting on the cross section of a stem are consistent with the
hypothesis that aerenchyma is the result of mechanical tearing. In the case of stem
elongation, a faster elongation promotes hollowing by increasing the stress at the
cylinder core.
A.
B.
B.
C.
Figure 16.5 Takano’s experiment on bean plants. A. In bush bean plants, addition of gibberellins
induces elongation and hollowing in a species with a naturally full stem. B. In pole bean plants,
a species that has naturally a hollow stem, the addition of uniconazole inhibits GA3 and prevents
hollowing. C. In pole bean plants, mechanical stress induced by moderate rubbing with the thumb
and forefinger back and forth 19 times, once daily for 5 days (adapted from [1219]).
A. B.
tr (0) 2 =1.0 tr (0)
1.4
1.1
2 =0.9
1.2
1.0
2 =0.8 tbreak
1.0
0.9
0.8
0.8 2=1.0
0.6
1 2=0.9
2=0.8
5 10 15 20 25 0.8 1.0 1.2 1.4 1.6 1
Stem elongation Stem thickening
Figure 16.6 Radial stress at the origin of a growing cylinder with constant inhomogeneous and
anisotropic growth (gr = gθ = ν1 + ν2 R0 , gz = μ1 + μ2 (B0 − R0 )) with parameter μ2 = 2.
A. The growth parameter ν1 = 1 is constant and the axial growth parameter μ1 is varied. B. The
axial growth parameter μ1 = 10 is constant and the constant sectional growth parameter is varied.
The dashed line corresponds to an estimate of the breaking strength. The material is assumed to be
elastic with a strain-energy function Wnh .
Indeed, in Figure 16.6A, an increase of μ1 implies an increases of the stress

at the origin. This increase is consistent with the observation of Takano shown in
Figure 16.5A.
Similarly, the inhibition of gibberellins corresponds to a decrease of μ1 consistent
with the non-hollowing of pole plants observed in Figure 16.5B.
Conversely, in the second case shown in Figure 16.6B, an increase of the stem
thickness reduces the stress at the origin below the critical rupture stress. This trend
is consistent with the observation of Takano shown in Figure 16.5C where thickening
induced by mechanical stresses prevents cavity formation.
The key effect that allows for tissue separation is cross-sectional differential
growth, that is the coefficient ν2 describing the relative rate of growth of the outer
layers versus the inner layers. However, in general, the effect of differential growth
may not be sufficient to initiate tearing and a local decrease of the breaking stress
due to cell wall degradation is also necessary. Indeed, based on similar experiments
on celery and tomato pithiness, Jaffe and coworkers have argued that chemical lysis
play a dominant role in aerenchyma [666]. Yet another possible mechanism for the
opening of cavity is not through tearing, but through cell–cell separation by introduc-
ing intercellular space at tricellular junctions [665]. This separation requires either
weakening of the pectin in the middle lamina or large compressive stresses. How-
ever, the mechanical compressive axial stresses generated by differential growth are
not sufficient by themselves to initiate cell separation and we conclude that a full
understanding of cavity opening in plants requires the coupling of both chemical and
mechanical effects.
16.3 Bifurcation of Growing Cylinders 487
16.3 Bifurcation of Growing Cylinders
We now consider the possibility of buckling in a growing elastic cylinder. Before

doing so, it is of interest to recall the basic mechanisms through which a cylindrical
structure can become unstable in the absence of growth.
16.3.1 Buckling Versus Barreling
It is well known from daily experience that under a large enough compressive axial
load an elastic beam will buckle. This phenomenon known as elastic buckling or
Euler buckling is a paradigmatic instability of mechanics. The critical load for buck-
ling was first derived by Euler in 1744 [350, 351, 984] and further refined for higher
modes by Lagrange in 1770 [748, 1234]. Both authors obtained their result on the
basis of simple beam equations first derived by Bernoulli [1240] (see Figure 16.7).
Since then, Euler buckling has played a central role in the stability and mechanical
properties of slender structures from nano- to macro-structures in physics, engineer-
ing, biochemistry, and biology [954, 1236].
A. B.
Figure 16.7 Axial buckling. A. Illustrations from Euler manuscript [350]. B. Lagrange solutions
for modes n =1, 2, and 3 [748].
We can use the theory of the planar elastica developed in Section 5.8 to look at the
possibility of a bifurcation for an elastic rod of length L. Recalling that θ = θ(s) is the
angle between the tangent to the rod’s centerline and the direction of a compressive
load N = N ex , the equation for the transverse deflection y = y(s) and for the angle
are
y = sin θ, (16.28)

θ + α sin θ = 0, (16.29)
where the prime denotes the derivative with respect to the arc length, and α = N /E I
is positive for compressive forces, i.e., N > 0. Here, E I is the beam’s bending
stiffness, where E is the Young’s modulus, and I is the second moment of area.
Following Section 5.5, the second moment of area, in the case of a cylindrical shell
of inner radius A and outer radius B, is
B 4 − A4
I =π . (16.30)
4
For a hinged-hinged beam as shown in Lagrange’s drawings in Figure 16.7B, the
boundary conditions are
θ (0) = θ (L) = 0, y(0) = y(L) = 0. (16.31)
Linearizing the equations around the base solution, y(s) = θ(0) = 0, leads to
y = θ, (16.32)
θ + α θ = 0. (16.33)
This boundary-value problem has nontrivial solutions of the form

nπs
θ = cos , (16.34)
L
nπs
y = sin , (16.35)
L
with αL/π = n 2 ∈ N \ {0}. The smallest load N for which there exists a nontrivial
solution corresponds to n = 1 and is the celebrated Euler buckling load
π2 E I
NEuler = , (16.36)
L2
where, according to Euler, “π is the circumference of a circle whose diameter is one”
[351].
A. B.
Figure 16.8 Navier and Saint-Venant description of a barreling instability [938] and computations
of a distorted cross-section of beams under flexure [1239].
Many other approaches in the theory of one-dimensional structures lead to the

same conclusion. For instance, one can analyze the exact solutions of the planar
elastica, as given in Section 5.8, or use an energy argument [1236].
The one-dimensional theory can be used with a variety of boundary conditions.
It is particularly easy to generalize and it can be used for large geometric deflections
of the axis [33]. However, since material cross sections initially perpendicular to
the axis remain undeformed and perpendicular to the tangent vector, no information
on the elastic deformation around the central curve can be obtained. In particular,
other modes of instability such as the barreling instability shown in Figure 16.8A,
cannot be obtained. Barreling is an axisymmetric deformation mode of a cylinder or
a cylindrical shell. These modes will typically occur for sufficiently short structures,
where the assumptions used in deriving the Kirchhoff equations fail.
The two-dimensional theory of shells can be used when the thickness of the
cylindrical shell is small enough. Then the stability analysis of shell equations such
as the Donnell-von Kármán equations leads to detailed information on symmetric
instability modes, their localization, and selection [642]. However, the theory cannot
be directly applied to obtain information on the buckling instability (asymmetric
buckling mode).
The three-dimensional theory of nonlinear elasticity provides a full description of
the different instability modes. In the case of long slender structures under loads, the
buckling instability can be captured by assuming that the object is either a rectangular
beam [102, 787, 967] or a cylindrical shell under axial load. Then, the buckling
instability can be recovered by a bifurcation argument following the general theory
given in Section 11.10.
This problem was first addressed in 1955 within the framework of nonlinear
elasticity by Wilkes [1325] who showed that the linearized system around a finite
axial strain can be solved exactly in terms of Bessel functions. While Wilkes only
analyzed the first axisymmetric mode (n = 0, see Figure 16.9), he noted in his
Euler validity
ζ
1.0
0.9
Increase in compressive loads
0.8
n=1
0.7
n=2
0.6
n=0
n=0
n=2
0.5
n=3
0.4
n=4
0.3 n=5
n=6
n=7
0.2 n=8
0.1 bπ
ζL
0.0
0 5 10 15 20 25 30
Slender tubes Stubby tubes
Figure 16.9 Bifurcation curves (stretch as a function of stubbiness) of an homogeneous neo-

Hookean (W = (I1 −3)/2) cylindrical tube for different circumferential number n = 0, 1, 2, . . . , 9.
with b/a = B/A = 2. For a given geometry, a bifurcation point is reached when the axial stretch
ζ reaches a bifurcation curves. The selected mode is the mode with the largest value of ζ (adapted
from [486]).
conclusion that the asymmetric mode (n = 1) corresponds to the Euler strut and
doing so, opened the door to further investigation by Fosdick and Shield [374] who
recovered Euler’s criterion asymptotically from Wilkes’ solution.
These initial results constitute the basis for much of the modern theory of elastic
stability of cylinders within the framework of three-dimensional nonlinear elasticity
[96, 295, 311, 552, 553, 1008, 1166]. The experimental verification of Euler’s
criterion was considered by Southwell [1180] and by Beatty and Hook [73].
A complete bifurcation analysis of cylindrical shells and cylinders is given by
Destrade et al. [267, 486] and is summarized here. First, an example of different
types of instabilities found in an incompressible neo-Hookean cylindrical shell of
different length is shown in Figure 16.9. The known standard features of the stability
problem for the cylindrical shell are recovered, namely: For slender tubes, the Euler
buckling (n = 1) is dominant and becomes unavoidable as slenderness increases;
there is a critical slenderness value at which the first barreling mode, n = 0, is the first
unstable mode (in a thought experiment where the axial strain would be incrementally
increased until the tube becomes unstable); and for very large bπ/(ζ L), the critical
compression ratio tends asymptotically to the value ζ ≈ 0.44375, which corresponds
to the surface instability of a half-space under uniaxial compression discussed in
Section 11.10.1.
We observe in Figure 16.9 that the Euler buckling criterion is only valid for slender
structures and we wish to make that statement more precise. Therefore, we focus on
the Euler buckling instability by taking the limit ζ → 1 for a cylindrical shell of
initial radii A and B. We consider a Mooney–Rivlin strain-energy function, which
for λ close to 1, corresponds to the most general form of third-order incompressible
elasticity:

μ 1 1
Wmr = ( + α)(I1 − 3) + ( − α)(I2 − 3) , (16.37)
2 2 2
where α ∈ [−1/2, 1/2]. In the limit ζ → 1, an asymptotic expression in ν = B/L

can be obtained:

3 π 3 B 2 μ ρ4 − 1 ν 2
Ncrit = −
4 ρ4

1 π 5 B 2 μ ρ2 − 1 18α(ρ2 + 1)2 − 17(ρ2 − 1)2 ν 4
+ O(ν 6 ), (16.38)
48 ρ6
where ρ = B/A and μ = E/3. We recover to order ν 2 the classic Euler buckling
formula (16.36) and we note that to order O(ν 4 ), higher order elasticity effects,
controlled by the parameter α, first appear.
The largest correction to the Euler criterion is obtained when α = −1/2 and in
the limit of a solid circular cylinder ( A → 0 at fixed B). In this case, the ratio of
the fourth- to second-order corrections is 13π 2 ν 2 /12 ≈ 10.7ν 2 and amount of about
10% for a rod of length L = 10B.
For a neo-Hookean material (α = 1/2), the maximal deviation from Euler’s
criterion is found in the limit of thin shells B → A in which the ratio of the fourth-
to second-order corrections is −3π 2 ν 2 /4 ≈ −7.4ν 2 and amount of about -7% for a
rod of length L = 10B. In general, for any third-order elastic materials we have, up
to corrections of order O(ν 6 ),

3 13
NEuler 1 − π 2 ν 2 ≤ Ncrit ≤ NEuler 1 + π 2 ν 2 . (16.39)
4 12
We see that the maximal error drops below 1% for rods of length L 33B.
It is also of interest to understand which bifurcation will first appear depending
on the geometry of the initial cylindrical shell. Here, we restrict our attention to
the neo-Hookean case. We want to identify the geometric values at which there is
a transition between buckling and barreling modes. We refer to this transition as a
dimensional transition, in the sense that the material mostly behaves as a slender
one-dimensional structure when it buckles according to mode n = 0 and mostly
3.0
Thick tubes
B/A
2.8
2.6
A. n=0
2.4
2.2 B A
2.0
n=1
1.8
1.6
n=2 L
B.
Thin tubes
1.4
1.2
B/L
1.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Slender tubes Stubby tubes
Figure 16.10 Dimensional transition for a neo-Hookean cylindrical tube of initial length L and
initial radii A and B. All tubes in the n = 1 region become unstable through Euler buckling. As
the tube gets stubbier or thinner (arrows), it will not buckle, but it will be subject to a barreling
instability. Note that only the transition curve from mode n = 0 are shown. Tubes in the barreling
regions may be susceptible to other unstable modes. Adapted from [486].
as a two-dimensional structure when it barrels with mode n = 2. Both modes of

instability can be captured by, respectively, a one- or a two-dimensional theory. The
results in Figure 16.10 show that for sufficiently slender and thin tubes (B/L small,
B/A large), the Euler buckling mode is the dominant instability mode. At a given
thickness, the Euler buckling mode disappears in favor of a barreling mode when the
tube is stubby enough (shorter tube at fixed radius) as indicated by the arrow A in
Figure 16.10.
For short tubes, at a fixed stubbiness, the Euler buckling mode disappears at the
expense of the n = 2 barreling mode as indicated by the arrow B in Figure 16.10.
This analysis guarantees that for slender enough tubes (ν = B/L small enough), the
Euler buckling mode is always dominant as long as [486]
B 3 17 4 4 161 6 6
> ρ∗ = 1 + π 2 ν 2 − π ν + π ν + O ν8 . (16.40)
A 4 32 384
Therefore, as long as this inequality is satisfied, we can focus further bifurcation

analyses on the Euler buckling mode only. Finally, we note that a local bifurca-
tion analysis does not prevent the cylinder from becoming unstable through global
dynamical bifurcation as found in thin stubby cylinders through the formation of
diamond patterns [492, 764].
16.3.2 Bifurcation and Buckling in Growing Cylinders
We apply the bifurcation scheme given in Section 14.8 to the problem of computing
bifurcations in growing cylinders. The computations are tedious, but the main steps
are conceptually simple. Following Section 16.1, we assume that, for a given diag-
onal growth tensor G = G(R0 ), there exists a family of cylindrical solutions. We
denote by χ(0) this deformation and by A(0) and F(0) the corresponding base solution
for the elastic and geometric deformation tensors. The problem is then to write the
boundary-value problem for the linearized equations around that base solution.
Step 1: Kinematics of incremental deformations. We expand the deformation

χ = χ(0) + χ(1) , where the χ(0) is the finite deformation, and χ(1) is given by the
vector
χ(1) = [u(r, θ, z), v(r, θ, z), w(r, θ, z)]. (16.41)
Let F(0) = Grad(χ(0) ), then the incremental deformation tensor F(1) , defined through
F = (1 + F(1) ) · F(0) , is given by
⎡ ⎤
u r (u θ − v)/r u z
[F(1) ] = ⎣ vr (u + vθ )/r vz ⎦ , (16.42)
wr wθ /r wz
where the (r, θ, z) subscripts denote partial derivatives. The elastic stress tensors are
given by
A(0) = F(0) G−1 , A(1) = F(1) . (16.43)
The incompressibility constraint is

tr F(1) ≡ u r + (u + vθ )/r + wz = 0. (16.44)
Step 2-3: Linearized equations. The Cauchy stress tensor is expanded around the
base solution to give
T = T(0) + T(1) + O( 2 ), (16.45)
and the constitutive relationship to leading and first order reads
T(0) = A(0) WA(0) − p (0) 1, (16.46)

and
T(1) = L : F(1) + F(1) A(0) WA(0) − p (1) 1, (16.47)
where p = p (0) + p (1) , the derivatives of W with respect to A are evaluated at A(0) ,
and L is the fourth-order tensor of instantaneous elastic moduli defined by
(0)
L : F(1) = A(0) WAA : F(1) A(0) . (16.48)
The explicit form for an isotropic material is given in Section 14.8. The condition
div T(1) = 0 leads to
T
div (M) + F(1) grad p (0) − grad p (1) = 0, (16.49)
where M ≡ L : F(1) . In cylindrical coordinates, these three equations are given by

(0)
∂M11 1 ∂M21 ∂M31 1 (1) d p ∂ p (1)
+ + + (M11 − M22 ) + F11 − = 0, (16.50)
∂r r ∂θ ∂z r dr ∂r
(0)
∂M12 1 ∂M22 ∂M32 1 (1) d p ∂ p (1)
+ + + (M21 + M12 ) + F12 − = 0, (16.51)
∂r r ∂θ ∂z r dr ∂θ
(0)
∂M13 1 ∂M23 ∂M33 1 (1) d p ∂ p (1)
+ + + M13 + F13 − = 0. (16.52)
∂r r ∂θ ∂z r dr ∂z
Substituting (16.42) into the incremental equations given in (16.49) results in

r (L2211 − L3311 ) − L2222 + L2233 − L2121 + L1331 vθ /r 2
+(L1111 − L3311 − L1331 )u rr

r (L1111 + ( p (0) ) − L3311 ) + L1111 − 2L3311 + L2233 − L1331 u r /r
+(L2112 + L2211 − L1331 − L3311 )vr θ /r + L2121 u θθ /r 2 + L3131

+ r (L2211 − L3311 ) − L2222 + L2233 + L1331 u/r 2 − ( p (1) )r = 0,
(16.53)
(r L1212 + L1212 + L2222 − L2233 − L3223 )u θ /r − (r L1212 + L1212 )v/r

+(L2222 − L2233 − L3223 )vθθ /r + L3232 r vzz + L1212 r vrr
+(L2211 − L2233 − L3223 + L1221 )u r θ + (r L1212 + L1212 )vr − ( p (1) )θ = 0,
(16.54)

L2323 wθθ /r 2 + r L1313 + r ( p (0) ) + L1331 − L2332 u z /r
+L1313 wrr + (L3333 − L2332 − L2233 )wzz + (r L1313 + L1313 )wr /r
+(L3311 + L1331 − L2332 − L2233 )u r z − ( p (1) )z = 0, (16.55)
where the prime denotes differentiation with respect to r .

Step 4: The boundary-value problem. The set of partial differential equations for
the variables u, v, w, p (1) can be further simplified to a boundary-value problem for
a set of differential equations in the variable r by Fourier expanding the dependence
in θ and z, that is
u = f (r ) cos nθ cos ψz, v = g(r ) sin nθ cos ψz, (16.56)

(1)
w = h(r ) cos nθ sin ψz, p = k(r ) cos nθ cos ψz. (16.57)
In order that the incremental displacements be single-valued, we take the mode

number n to be an integer n ≥ 0. Since the incremental end displacement w must
vanish on the ends, we have
m m
ψ= π= π, m = 1, 2, 3, ... (16.58)
l λL
where m is set to unity, without loss of generality, so that ψ is inversely proportional to

the length of the cylinder. The expressions in (16.56–16.57) are substituted in (16.53–
16.55) and h(r ) can be eliminated by means of the incompressibility constraint
(16.44). The resulting incremental equilibrium equations are
k =

r (L2211 − L3311 ) − L2222 + L2233 − L2121 + L1331 ng/r 2
+(L2112 + L2211 − L1331 − L3311 )ng /r + (L1111 − L3311 − L1331 ) f

+ r (L1111 + ( p (0) ) − L3311 ) + L1111 − 2L3311 + L2233 − L1331 f /r

+ r (L2211 − L3311 ) − L2222 + L2233 + L1331 f /r 2
−n 2 L2121 f /r 2 − ψ 2 L3131 f, (16.59)
r L1212 g = −nk + (r L1212 + L1212 + L2222 − L2233 − L3223 )n f /r

+(r L1212 + L1212 )g/r + (L2222 − L2233 − L3223 )n 2 g/r
+r ψ 2 L3232 g + (L2211 − L2233 − L3223 + L1221 )n f − (r L1212 + L1212 )g ,
(16.60)
L1313 f =

n 2 f + (ng + f )/r L2323 /r 2 − r L1313 + r ( p (0) ) + L1331 − L2332 ψ 2 f /r

−L1313 ( f + ng )/r − 2(r + ng )/r 2 + 2( f + ng)/r 3
+(L3333 − L2332 − L2233 )ψ 2 ( f + ng)/r + (L3333 + L3311 − L1331 )ψ 2 f

−(r L1313 + L1313 ) f + ( f + ng )/r − ( f + ng)/r 2 /r − ψ 2 k. (16.61)
The corresponding incremental boundary conditions on the inner and outer wall
r = a, b is T(1) n = P (1) n which takes the form
(L : F(1) + F(1) A(0) WA(0) − p (1) 1)n = P (1) n, (16.62)
where the incremental internal pressure, P (1) , is taken to be P (1) = 0.

The three boundary condition equations on r = a, b are given by
(L1122 −L1133 )( f +ng)+(L1111 −L1133 +αr W1 )r f −r k = P (1) , (16.63)

r g − g − n f = 0, (16.64)
r 2 f + r f + (ψ 2 r 2 + n 2 − 1) f = 0. (16.65)
The growth dependence comes about through the coefficients of the above system of
equations. The position vector depends on G through (16.10) and the instantaneous
elastic moduli depend on α = αθ and its derivatives through the relation
dα α ζα3 γθ2 αγθ

= − − . (16.66)
dr r rg γθ
For a given growth tensor G and strain-energy function W , the bifurcation points
corresponds to the critical value of P = Pcrit such that a nontrivial solution for this
boundary-value exists. This critical value can be solved by the determinant method
as explained in Section 15.6.1.
16.3.3 The Effective Rigidity of a Growing Cylinder
During growth, the effective property of a structure may change. Due to the residual
stress field, it may support larger loadings before buckling, or, on the contrary, be
prone to buckling under its own weight. Here, we are interested in quantifying the
effective rigidity of a growing cylinder in the presence of residual stress. We are
considering a grown hyperelastic incompressible cylinder through a growth process
G. Since the cylinder is in large deformation and supports residual stress, the classic
notion of Young’s modulus does not apply directly and an explicit computation of
the rigidity is not possible.
We can perform the following thought experiment to obtain an effective Young’s
modulus: Consider the grown cylinder and measure all its geometric parameters
(inner and outer radii a0 and b0 and height l0 being the geometric parameters for zero
load) and subject it to a normal load N until it buckles at Ncrit . The critical buckling
pressure
Ncrit
Pcrit = , (16.67)
π(b02 − a02 )
can be used to compute the effective Young’s modulus, defined as the Young’s mod-
ulus of an equivalent homogeneous elastic cylinder with the same geometry, but with
no residual stress, and buckling at the same pressure, that is
0
Pcrit Eul
er bu
ckl
ing
−5
−10
−15
σ2
0 0.1 0.2 0.3 0.4
Figure 16.11 Bifurcation modes of a cylindrical tube (A = 0.0001, B = 1). The buckling mode
n = 1 is shown. The curve gives the value of the parameters where buckling occurs and the line
indicates the prediction given by the Euler buckling formula (where μ = 1 = E/3).
4Pcrit
E eff = , (16.68)
π2 σ2

where σ = b02 + a02 /l0 is the stubbiness ratio (the inverse of the slenderness ratio).
We saw in Section 16.3.1 that the Euler buckling formula is valid for cylinders
that are sufficiently slender. Therefore, in order to obtain an estimate of the cylinder
rigidity through an effective Young’s modulus, we compute the critical pressure Pcrit
as a function of σ and obtain the effective Young’s modulus, E eff , from the slope k
of the graph
π2
Pcrit (σ 2 ) = kσ 2 , k= E eff . (16.69)
4
This procedure is validated and illustrated in the absence of growth in Figure 16.11,
where the critical pressure for the buckling mode (n = 1) of a neo-Hookean cylinder
is plotted as a function of σ 2 together with the Euler buckling criterion. Therefore,
we conclude that the bifurcation analysis provides a method to define an effective
Young’s modulus.
A. B.
Eeff
tz
0.6
Tensile
0.4
3
0.2
0
κ=0.1
−0.2 2.5
κ=0.2
−0.4
κ=0.3
Compressive
−0.6
κ=0.4 2
−0.8
κ=0.5
−1
r κ
1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6
Figure 16.12 A. Axial stress profiles for neo-Hookean cylinders (μ = 1) undergoing axial growth,
γz = 1 + κ(b − r ). B. The effective modulus as a function of the growth parameter κ, where
γz = 1 + κ(b − r ). For κ = 0, we recover the Young’s modulus E = 3.
16.3.4 An Example
We apply this method to the case of an isotropic neo-Hookean cylinder with radial-
dependent growth. The growth tensor G is a function of the radial position r in the
current configuration. We assume a simple, linear growth law in the axial direction,
γr = γθ = 1, γz = 1 + κ(r − a), (16.70)
where κ is a parameter characterizing growth in the system. It is interesting to consider

the residual stresses created in the grown material in the absence of loads, N = 0,
for various values of the growth parameter. As axial growth increases, a large com-
pressive zone develops along the longitudinal direction as shown in Figure 16.12A.
We can now determine the overall rigidity of the grown cylinder. We assume that
the cylindrical shell has grown axially by a given amount and we test the stability of
the resulting residually stressed cylindrical shell. In Figure 16.12B we observe that
the effect of axial growth is to reduce the stability of the structure. This effect can
be understood as follows Differential growth creates a large zone of compression in
the cylinder. The cylinder is pre-compressed due to growth and will therefore buckle
under smaller loads.
16.4 Application: Tissue Tension in Plants 499
A.
B.
Figure 16.13 Growth-induced longitudinal tissue tension in a stalk of rhubarb. When partially
peeled, the epidermis bends backward demonstrating the presence of residual stress and surface
tension in the original configuration. When completely peeled (last picture on the right), the pith
increases in length, as shown by the dashed lines. A. from Gager’s “Fundamental of Botany” [412].
B. Picture courtesy of Ellen Kuhl (adapted from [591]).
16.4 Application: Tissue Tension in Plants
16.4.1 Background
We briefly recall the problem of tissue tension first discussed in Section 2.2.1 and
illustrated in Figure 16.13.
When the stem of a flower, such as a dandelion, is cut axially, it bends backwards as
shown in Figure 16.14. This change of shape is due to the presence of residual stress
Figure 16.14 Left: When the stem of a dandelion is cut axially, it bends backward. Right: When a
section is completely cut, it relieves stress by curving (held with suspicion by Zakkai, the author’s
second son).
in the original cylindrical configuration. This residual stress, called tissue tension by
plant physiologists from the nineteenth century, is induced by a differential resting
length of the outer tissue, the epidermis, with respect to the inner softer tissue. For
instance, when the epidermis of a rhubarb stalk is completely peeled, it is markedly
shorter than the pith, its inner tissue, as shown in Figure 2.14.
We conclude that, in their original configuration, the epidermis is in tension and the
pith in compression. Many authors [562, 739, 741] have argued that this differential
length originates from the differential extensional growth of the cell walls in outer and
inner layers, creating an irreversible change in the resting lengths of both tissues.
Note, however, that when a tissue is immersed in a bath with solutes of different
concentration, it may also change its resting length, indicating that there is a hydraulic
component to the relative extension of the tissues.
Mechanically, each separate rhubarb tissue is soft and cannot support its own
weight without buckling. However, the combined configuration is rigid and allows
the plant to shoot upwards. Experiments have established that the outer tissue plays a
substantial role in the overall rigidity, accounting for as much as 70% of the rigidity
for less than 10% of the cross-sectional area [954, p. 300]. Therefore, the overall
flexural rigidity of the residually stressed structure is much larger than the rigidity
of each component. The problem is then to understand how axial differential growth
and/or difference in tissue stiffness change the overall flexural rigidity. We note that
in order to understand the mechanics of these structures, the biological origin of the
residual stresses is not relevant as we take their existence as a starting point.
16.4.2 The Model
We use a simple model of a stem viewed as a cylindrical shell composed of two

material layers with different growth and elastic properties. We are interested in
understanding the specific role of axial growth, therefore, we neglect the possible
effect of anisotropic reinforcement by assuming that the tissue is made out of a
material that is hyperelastic, incompressible, homogeneous, isotropic and subject to
growth solely along the axial direction (taken to be the z-axis).
Since typical ranges for elongation or compression can be as much as 6 to 40%
of the original length [659], the material should be considered as being in large
deformations. We take growth as a fait accompli by postulating that each cylindrical
shell has grown axially by a given amount and study the grown structure at that
time. Following the discussion in the previous section, we analyze the stability of the
grown residually stressed cylindrical shell to obtain an effective Young’s modulus.
We use the general setup of Section 16.1 and assume that in its initial configuration,
the cylindrical shells have height L 0 and radii A0 , A1 , A2 . Since growth is purely
axial, we assume that the growth tensor has the form

diag(1, 1, γ1 ) R0 ∈ [A0 , A1 ],
[G] = (16.71)
diag(1, 1, γ2 ) R0 ∈ [A1 , A2 ],
with γ1 and γ2 constant.

Next, we model the two tissues separately based on experiments [564] in which
tests were performed on the inner and outer tissues of the hypocotyl of the sunflower
(Helianthus annuus). The data for the inner tissue (data points in Figure 16.15A)
shows that the stress–strain curve for the inner tissue is nearly linear within the range
of forces applied (0–0.4 N). Therefore, we model the inner tissue as a neo-Hookean
material with strain-energy function
μ1
W1 = (I1 − 3). (16.72)
2
We consider elastic deformations of the form [A] =diag(α, α, 1/α2 ). Following the
discussion in Section 11.8.1, the axial stress for a strip of neo-Hookean material is
given by (taking λ = 1/α2 in Equation (11.190))

1
tz = μ1 − α 2
. (16.73)
α4
A nonlinear regression can be performed to fit this relation to the experimental data
while solving for the parameter μ1 . Doing so, we obtain μ1 =1.25 MPa with an R-
squared value of 0.991. Figure 16.15A compares the approximate experimental data
points [564] and the corresponding stress-strain fit.
The outer tissue is a complex tissue containing both the epidermis and one or two
layers of cortical parenchyma. Uniaxial extension data, shown in Figure 16.15B,
shows a very strong strain-stiffening effect. Therefore, we model the outer layer as
a Fung material with
μ2 ν(I1 −3)
W2 = e −1 , (16.74)
2ν
where μ2 is the elastic material constant and ν controls the strain-stiffening property
of the outer layer. The data points were fitted to the axial stress
∂W2
tz = αz − p. (16.75)
∂αz
In a uniaxial test on a rectangular strip of tissue along the z-direction, the elastic tensor
is again given by [A] = diag(α, α, 1/α2 ) and the boundary condition tx = t y = 0
implies p = α∂x W2 . Substituting the strain-energy function W2 into the stress–strain
relationship, the axial stress is given by

1 1
tz = μ2 − α exp ν(2α + 4 − 3) .
2 2
(16.76)
α4 α
A nonlinear regression analysis produces values μ2 = 6.7 MPa and ν = 195 with
an R-squared value of 0.992. This large value of ν is needed to recover the observed
exponential behavior at small strains. Figure 16.15 compares the approximate exper-
imental data points and the corresponding stress–strain fit. Both models are shown
to be very good fits to the data.
The relevant part of this analysis is not the precise values of the constants, but
the overall qualitatively different properties of the two tissues in the elastic regime.
Therefore, for the rest of our analysis, we adopt a neo-Hookean model for the inner
tissue and a Fung model for the other tissue and perform a general analysis of the
response of the overall structures for arbitrary values of the material and geometric
constants.
A. Inner layer B. Outer layer

stress (MPa)
stress (MPa)
fit (neo-Hookean model) fit (Fung model)
0.3 experimental data experimental data
3
0.25 2.5
0.2 2
0.15 1.5
0.1 1
0.05 0.5
ε ε
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Figure 16.15 The experimental data from a uniaxial test on the tissue of a sunflower hypocotyl
[564]. A. The data points for the inner layer are fitted to the constitutive relationship using a neo-
Hookean strain-energy function (R-squared value is 0.991). B. The outer layer data and a fit to the
constitutive relationship using a Fung-type strain-energy function (R-squared value is 0.992).
16.4.3 Analysis
16.4.3.1 The Base Solution
First, we consider the semi-inverse problem consisting in computing the deformation

of the two-layer cylinder into another two-layer cylinder for given material, geomet-
ric, load, and growth parameters. The general solution is given in Section 16.1 with
the particular choice P = 0, n = 2 and the form of the strain-energy functions given
in the previous section. It is particularly interesting to consider the residual stresses
created in the grown material in the absence of loads, that is by choosing N = 0. In
Figure 16.16, we show the axial stress profile as we vary the ratio of the inner layer
stiffness to the outer layer stiffness and the ratio of the inner layer width to the outer
layer width (w1 /w2 ) where wi = (A1 − A0 ) and w2 = (A2 − A1 ). As the thickness
of the inner layer decreases, large stress gradients are created in the outer layer.
16.4.3.2 Estimates
Before performing a stability analysis of the residually stressed cylinders, we derive

estimates based on simple considerations from linear elasticity.
First, we consider a cylinder made out of two cylinders of different materials in
the absence of growth with respective radii A0 , A1 and A2 and length L as before.
Then, a standard way to estimate the Young’s modulus [1236] of the assembly is to
consider the sum of the flexural rigidities, that is
E est I = E 1 I1 + E 2 I2 , (16.77)
tz a2 outer layer
a1 in tension
a0
5 w / w = 2, μ = 10
2 1 1
w1/ w2 = 8, μ1= 10
w1/ w2 = 2, μ1= 1
4
w1/ w2 = 8, μ1= 1
1 inner layer
in compression
r
−1
0 0.2 0.4 0.6 0.8 1
Figure 16.16 Tissue tension in two-layer cylinders with varying thickness and shear moduli (A1 =
0.9, A2 = 1, ν = 20, γ2 = 1, γ1 = 1.2).
where E est is an estimate for the effective Young’s modulus provided by the linear
theory of elasticity. Explicitly, since the moments of area for each cylindrical shells
are known, we have
μ1 (A41 − A40 ) + μ2 (A42 − A41 )

E est = 3 . (16.78)
A42 − A40
This estimate provides a linear relationship with respect to changes in the rigidity of
the outer layer. However, it fails to take into account the effect of strain-stiffening
since the parameter ν does not appear explicitly.
Second, we consider a grown two-layer cylinder with a neo-Hookean material
inside and a Fung material outside. We assume that the cylinder in the current con-
figuration has height l0 and radii a0 , a1 , a2 where a0 1 so that a02 , a04 can be
neglected. Before we consider an estimate of the flexural rigidity, we need to model
the Fung material under extension. To do so, we consider the first approximation that
captures both the exponential behavior and the small deformation limit, that is we
express the axial tension as
2
T2 = 3μ2 2 e3ν 2 , (16.79)
where 2 is the axial strain, computed with respect to the unstressed grown config-
uration of the outer cylinder. The apparent Young’s modulus of the outer cylinder
is obtained as the slope of the tangent to this curve as a function of 2 , that is, we
estimate the Young’s modulus of the outer cylinder as
∂T2 2
E2 = = 3μ2 (6ν 22 + 1)e3ν 2 . (16.80)
∂ 2
Similarly, we have, for the inner cylinder T1 = 3μ1 1 and E 1 = 3μ1 , where 1 is
the strain with respect to the unstressed configuration of the inner cylinder. Next, we
compute the rest shape of the grown cylinders by considering the case where no net
traction is applied at the two caps, that is, we have
N =0

= π T1 a12 + T2 (a22 − a12 )

= 3π μ1 1 a12 + μ2 e3ν 2 2 (a22 − a12 ) ,
2
(16.81)
where i = l0 /γi − 1 and the two cylinders are assumed, without loss of generality,
to have height 1 before they start growing. The equation N = 0 can be solved for
l0 to obtain the residual strain. Since the equation is transcendental in l0 , no useful
estimates of l0 can be obtained, except in the case ν = 0 which is an upper bound,
lest for the length l0 :
μ1 a12 + μ2 (a22 − a12 )

lest = γ1 γ2 . (16.82)
μ1 γ2 a12 + μ2 γ1 (a22 − a12 )
We can use again Equation (16.77) to obtain an estimate

2
μ1 a14 + μ2 (6ν 22 + 1)e3ν 2 (a24 − a14 )
E est = 3 , (16.83)
a24
where i = lest /γi − 1 and a1 , a1 are obtained using the assumption a0 1 and
from the conservation of volume of each cylindrical layer, namely A21 γ1 = b02 lest
and (A22 − A21 )γ2 = (a22 − a12 )lest . This estimate is compared with the results of the
numerical analysis of the stability equation in the next section.
16.4.3.3 Bifurcation Analysis and Effective Rigidity
We consider a grown two-layer cylinder with a neo-Hookean material inside and a

Fung material outside and compute the effective Young’s modulus as a function of
the parameters by performing a full stability analysis and extracting the critical value
of the axial load leading to buckling when parameters are varied. Without loss of
Eeff
ν=0
35 ν=1
ν=5
30 ν=10
ν=20
ν=50
25 ν=100
Eest
20
Eest
15
10
5
μ2 μ
0
/ 1
0 5 10 15 20 25 30
Figure 16.17 The effect of outer wall stiffness and strain-stiffening properties on the stability of a
two-layer cylinder in the absence of residual stress. A = 0.0001, B = 0.9, C = 1. The solid line
corresponds to the estimate (16.78).
generality, we take γ2 = μ1 = 1. For comparison, we scale the effective Young’s

modulus by the Young’s modulus of a neo-Hookean one-layer cylinder E nh .
First, we study the effect of differential stiffness between layers in the absence of
growth. In Figure 16.17, we observe that the stiffness of the outer layer μ2 provides a
substantial improvement on the overall rigidity of the structure, which is essentially
linear in the ratio of moduli and well approximated by the estimate (16.78). Note
that the correction due to the nonlinearity of the Fung model cannot be captured
since, in the regime considered, the estimate (16.78) does not depend explicitly on
the strain-stiffening parameter ν. However, the strain-stiffening property of the outer
wall has very little, if any, effect. Essentially, in this regime, the outer layer is in a
regime where it behaves as a neo-Hookean material.
Second, we fix the value of μ2 and consider the effect of growth by varying the
parameter γ1 . In Figure 16.18, we observe that in the absence of strain-stiffening, the
effect of growth is to reduce the stability of the structure. Indeed, growth creates a
large zone of compression in the inner tissue, as observed in Figure 16.16, so that the
cylinder is pre-compressed due to growth and therefore buckles for a smaller load.
Therefore, there is no gain in stiffness due to growth in a homogeneous, isotropic,
incompressible, two-layer neo-Hookean cylinder. For a strain-stiffening outer layer,
the situation is radically different and even modest differential axial growth has
a dramatic effect on the rigidity of the structure. After growth, large stresses are
needed for the small changes in strains that would occur on the concave side of the
E
eff
40
ν =0
ν =1
ν =5
35
ν =10 Eest (ν=100)
ν =20
ν =100
Eest (ν=20)
30
25
Eest (ν=10)
20
15 Eest (ν=5)
10
γ
1
5
1 1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.4 1.45 1.5
Figure 16.18 The effect of axial differential growth on the rigidity of a two-layer cylinder for
different strain-stiffening properties (initial size A = 0.0001, B = 0.9, C = 1.0 μ2 = 12). The
estimate given by (16.83) captures most of the exponential increase of the rigidity as the relative
growth between outer and inner layer increases.
cylinder during bending. The exponential increase of the effective Young’s modulus
is qualitatively captured by the estimate (16.83) that predicts an exponential term of
the form exp[κν(γ1 − 1)2 ], where κ depends on the moduli μ1 and the geometric
parameters.
We conclude that the main effect of differential growth is to bring the outside
layer in a regime where the nonlinear stiffening response can be fully utilized.
16.4.4 Discussion
The increase in flexural rigidity of cylindrical structures through differential axial

growth and nonlinear strain-stiffening appears to be a simple generic mechanism
through which structural properties is controlled in biological systems. In plants, this
differential axial extension is only a component of a complex mechanical and growth
environment [377–379]. In particular, for trees, apart from primary vertical growth,
a secondary cambial radial growth also takes place. Cambial growth allows branches
and trunks to gain in girth, but it also takes place differentially to allow the overall tree
structure to attain a complicated mechanical balance. The net mechanical effect of
1/γ =1.00
1
1/γ =0.98
1
1/γ =0.96
1
1/γ =0.94
1
Figure 16.19 Growth-induced longitudinal tissue tension in a stalk of rhubarb. The stalk is modeled
by assuming growing volume and surface elastic elements. As the relative shrinkage of the surface
increases, the surface tension creates higher curvatures for quarter slices [591] (figure courtesy of
Ellen Kuhl).
this combined growth is a complicated stress field with a combination of tension and
compression wood [910]. However, the cellular mechanism sensing residual stresses
during the differentiation process of the wood cells remains elusive [55].
The outer layer in tension is usually very thin and can be approximated by a surface
condition [591]. The idea is to consider a cylinder with an external surface. Both
volumes and surface can grow independently. The inner volume follows the usual
rule of volumetric growth whereas surface growth appears as a boundary condition.
This construction leads to an elegant formulation of the problem that can be used for
analytical and computational purposes. For instance, Figure 16.19 shows the result
of the computation of equilibrium solution of a shrinking skin on a quarter cylinder.
In order to isolate and understand the effect of differential axial growth, we have
neglected important effects necessary to obtain a precise picture of the material
properties of specific biological structures, notably, anisotropic response and inho-
mogeneity. However, the analysis presented here was performed using the general
framework of nonlinear elasticity which can easily be generalized to include these
effects. In particular, axial growth is also found in arteries [601], a system which
crucially depends on anisotropy as we show next.
16.5 Application: The Buckling of Arteries 509
Pressure (kpa)
0N (full) 0.31N 1.48N 3.94N
40 axial in situ 0.11N 0.70N 2.47N
5.90N
Prestretch
0N-Control
30 (dotted)
9.90N
selectedexp.data
20
Model (split SEF)
13.3
10
0
1 1.02 1.04 1.06 1.08 1.1 1.12 1.14 1.16
-5 Axial Stretch
Figure 16.20 Pressure-axial stretch plot for a human iliac artery. At a fixed axial tension, the
pressure is varied and the axial stretch measured [1133]. (picture courtesy of Gerhard Holzapfel. )
16.5 Application: The Buckling of Arteries
16.5.1 Background
As discussed in Section 2.1.3, the arterial system is a highly complicated and reg-
ulated mechanical system dedicated to carry blood from the heart to organs while
sustaining varying pressures, loads, and stresses. Starting in the 1960s many differ-
ent puzzling experimental observations revealed that large arteries have mechanical
responses that cannot be readily understood by elementary mechanics:
• As early as 1881, Roy demonstrated that the elastic response of arteries is nonlinear
[1101].
• During regular homeostatic functions, arteries are tethered to other tissues and
they operate under a large stretch of about 1.3 to 1.6 [773]. They will therefore
naturally retract to their preferred rest length when cut.
• Patel and Fry showed that during the cardiac cycle, and despite large variations of
pressure, the end-to-end distance of major dog arteries remains mostly unchanged
[248, 1019]. This property is also observed in human arteries. In Figure 16.20, we
show the axial stretch of a human iliac artery obtained by fixing the axial tension
and varying the pressure. It is observed that close to the homeostatic axial stretch,
no axial stretch is observed while the pressure is varied.
• Even after cutting, in their rest length and in the absence of pressure, Bergel noted
that arteries are still axially stressed [168] showing that the resting lengths of
different layers are different [605] as shown in Figure 16.21B.
• Fung and, independently, Vaishnav and Vossoughi in the 1980s showed that arteries
are also residually stressed radially. This observation is obtained by first slicing
a ring out of an artery and by observing how this ring opens when a radial cut is
made. The opening that results from the cut is measured by the so-called opening
angle, a kinematic measure of residual strains [408] as discussed in Section 12.4.
• Arteries respond to sustained increase in axial loading by lengthening [264, 659]
in order to attain a homeostatic state [448].
The stresses associated with the opening angle are known to play a fundamental
role in regulating transmural stress gradient and lowering hoop stress at the inner
walls [184, 635]. For instance, we showed, in Section 13.8.3.1, that radial differential
growth can be tuned to reduce hoop stress gradient. Most models of arteries [601,
635, 1054] include the effect of radial residual stress but fail to properly account for
differences in axial stresses.
Experimental observations [605, 607] show that arteries exhibit axial residual
stress and that these stresses may be important in the regulation of homeostatic
properties. For instance, Jackson et al. [658] showed that partial unloading of axial
tension can induce arterial tortuosity. They studied the relationship between the
development of tortuosity and mechanical forces imposed on arterial tissue. The
axial strain in rabbit carotid arteries was reduced from 62 ± 2% and maintained at
33 ± 2% for twelve weeks. After this time period, all vessels became tortuous due
to tissue growth and remodeling, as shown in Figure 16.22. They concluded that
substantial axial strains are necessary to maintain arterial stability.
Arterial tortuosity is a well-known condition, shown in Figure 16.23, that can
occur in response to abnormal growth stimuli or genetic defects [7, 146, 250, 260,
382, 420].
The stability of arteries under axial loads is essential to maintaining normal arte-
rial function. Instability can lead to serious complications including stroke, vertigo,
blackout, fainting, and persistent tinnitus [16, 1009, 1304], yet the biomechanical
A. B.
Figure 16.21 Differential resting length in arteries. The separation of the three tissue components
of fresh human aortic patches demonstrate the existence of different resting length in the hoop (A)
and axial (B) directions (picture courtesy of Gerhard Holzapfel [606]).
basis of why arterial buckling occurs is not well understood. Further experimental
studies on the instability of arteries were conducted by Han [535] who set porcine
carotid arteries at a fixed axial stretch and gradually pressurized the arteries until
significant mechanical buckling occurred. The results supported by a linear stability
analysis [536] showed that the critical pressure leading to buckling is directly related
to the axial stretch ratio and is very close to a linear relationship in the tested range.
Here, following [485], we study the mechanics and stability of arteries including
radial and axial residual stresses and we investigate the possibility of mechanical
buckling in large deformations. We use the theoretical framework of large deforma-
tion morphoelasticity together with a bifurcation analysis to answer the following
questions:
• What are the mechanical factors (anisotropy, residual stress, ...) responsible for
the constancy of artery length in normal physiological regime?
• Could axial off-loading of an artery result in mechanical buckling?
• What is the role of residual stresses in arterial stability?
Figure 16.22 A. Buckling of the carotid artery of a rabbit following imposed axial shortening.
B. Mean tortuosity index of unmanipulated left carotid arteries; of arteries five weeks after surgical
off-loading of axial strain without and with doxycycline (DOX) treatment; and of arteries five weeks
after sham manipulations [658].
Figure 16.23 Angiogram of the right vertebral artery showing marked vessel tortuosity [7].
16.5.2 The Model

16.5.2.1 Elastic Response
The arterial wall consists of three layers, each containing specific histological fea-
tures: the intima (the innermost layer), the media, and the adventitia (the outer layer).
The intima is known to have negligible (solid) mechanical contributions [602] and
therefore the arterial wall is approximated as a two-layer structure containing the
media and adventitia.
Much is known about arteries and the simple toy model proposed in
Section 11.8.3.1 is not sufficient to analyze the respective role of anisotropy and
differential growth. Indeed, numerous constitutive models have been proposed for
arteries (see review in [678]). Most models [208, 269, 1268] are formulated by fitting
a given constitutive equations to experimental data. Here, we start with the multi-
Figure 16.24 A two-layer

model representing the
media and adventitia with a
continuous family of fibers
representing the effect of
Φ
collagen fibers. 1
Φ
2
layer model for arteries introduced by Holzapfel, Gasser, and Ogden [601, 604] based
on the theory of anisotropic materials with two fibers described in Section 11.11.2.
This model has the advantage of being structural in the sense that it incorporates
layer-specific histological features. Each layer is reinforced with collagen fibers with
preferred directions that render the material anisotropic [66]. The model accounts
for the orientations of the collagen fibers in each layer. The contribution of collagen
fibers in the arterial wall is considered to be negligible at low pressures and therefore,
the mechanical response is assumed to be isotropic. However, at higher pressures,
the collagen fibers are stretched and the resulting mechanical response is anisotropic.
For each layer Holzapfel et al. [601] suggest a strain-energy function of the form
W (I1 , I4 , I6 ) = Wiso (I1 ) + Waniso (I4 , I6 ), (16.84)
where the invariants I1 , I4 , I6 of the right Cauchy–Green tensor C = FT F are given

by
I1 = tr C, I4 = C : M ⊗ M, I6 = C : M ⊗ M . (16.85)
As explained in Section 11.11.2, the invariants I4 and I6 are the squares of the stretches
in the respective directions of the two families of collagen fibers. The unit vectors
M and M’ describe the orientations of these fibers and these vectors are defined for
each layer in the reference configuration. In the usual cylindrical coordinate system,
the components of the direction vectors are
⎡ ⎤ ⎡ ⎤ ⎡ ( j) ⎤ ⎡ ⎤
( j)
Mr 0 Mr 0
⎢ ( j) ⎥ ⎣
⎣ Mθ ⎦ = cos j ⎦ , ⎣ M θ ⎦ = ⎣ cos j ⎦ ,
( j)
j = 1, 2, (16.86)
( j)
Mz sin j M (z j) − sin j
where the indices j = 1 and j = 2 refer to the inner layer (the media) and the outer
layer (the adventitia), respectively. The fibers are assumed to lie in the tangent plane.
The angle between the collagen fibers and the circumferential direction is denoted by
j as shown in Figure 16.24. The isotropic and anisotropic strain-energy functions
in (16.84) are then given by
μ
Wiso = (I1 − 3), (16.87)
2
k ν(In −1)2
Waniso (I4 , I6 ) = H(I4 − 1)H(I6 − 1) {e − 1}, (16.88)
2ν n=4,6
where the material parameter k has the dimension of stress and ν is a dimensionless
parameter. The Heaviside function, H(·), indicates that the anisotropic term con-
tributes to the mechanical response only when the fibers are extended, in other words
when either I4 > 1 or I6 > 1.
Typical material and geometric data used here are from a rabbit carotid artery
[598] (A0 = 0.71, A1 = 0.97, A2 = 1.10, μ1 = 3.0000 kPa, k1 = 2.3632 kPa,
ν1 = 0.8393, μ2 = 0.3000 kPa, k2 = 0.5620 kPa, ν2 = 0.7112, 1 = 29.0 ◦ , 2 =
62.0 ◦ ).
The cylinder is subjected to an internal pressure, P, and therefore the axial equi-
librium and the boundary conditions at the inner and outer surfaces yield
a2
2 2r tz dr = Pa 2 , (16.89)
a0
tr (a0 ) = −P, tr (a2 ) = 0, (16.90)
where tr and tz are the Cauchy stresses in the radial and axial directions, respectively.
Note that the cylindrical symmetry will be preserved in the finite deformation because
the two families of fibers are mechanically equivalent and are arranged symmetrically.
16.5.2.2 Growth Response
We assume that the growth tensor is constant and diagonal in each layer so that

diag(γr , 1, γz ), A0 ≤ R0 < A1 ,
[G] = (16.91)
diag(γr , 1, 1), A1 ≤ R0 ≤ A2 ,
where we have assumed without loss of generality that γz = 1 in the inner layer. We
also assume that the hoop residual stress in both layers results from the same radial
growth γr . To estimate these constants from experimental data, we use the method
of Taber and Humphrey [1213].
In our case, we use data on the opening angle from a radial cut on a thin ring
of artery to determine the value of the constants appearing in the growth tensor.
Within the framework of morphoelasticity, we assume that the current state of an
unloaded artery is the result of a local growth step characterized by a growth tensor
G followed by an elastic step characterized by an elastic tensor A. In addition to the
current configuration state B, we also consider a configuration B1 , which represents
a single opened ring due to a radial cut as shown in Figure 16.25. The deformation Aa
to configuration B1 will contain residual stress if the elements of V are geometrically
locally incompatible. Finally, the deformation Ab bends the ring closed and the ring
is subjected to an internal pressure P to produce the final configuration B.
The cylindrical polar coordinates in B1 are given by (ρ, ϑ, ζ). The elastic tensor
A can be decomposed into the product of two tensors, A = Aa Ab , where
[Aa ] = diag(αaρ , αaϑ , αaζ ), (16.92)
and
[Ab ] = diag(αbρ , αbϑ , αbζ ). (16.93)
The solution for the configuration B is given in Section 16.1 for n = 2 layers.
Once this solution is known, the geometry and corresponding stresses in the cut
ring can be determined as follows. The total deformation in terms of the material
coordinates in B1 is given by
r = r (ρ), θ = πϑ/ϕ0 , z = Λζ, (16.94)
where Λ is the axial stretch and ϕ0 is the angle related to the opening angle ϕ =
2(π − ϕ0 ) defined in Figure 16.26.
The kinematic relationships are given by
αaρ = λr (αbρ γr )−1 , αaϑ = λθ (αbϑ γθ )−1 , αaz = λz (αbζ γz )−1 , (16.95)
αbρ = ∂r/∂ρ, αbϑ = πr/ϕ0 ρ, αbζ = Λ, (16.96)
and the incompressibility conditions are given by
αaρ αaϑ αaz = αbρ αbϑ αbz = 1. (16.97)
The equilibrium equation is

∂tρ t ρ − tϑ
+ = 0, (16.98)
∂ρ ρ
where the Cauchy stress components in B1 are given by
∂W
ti = αai − p̂, i = ρ, θ, z, (16.99)
∂αai
Unstressed Intact pressurized

reference state artery after growth
F
Bf
B0
G A
Ab
Aa
Zero-stress state
after growth V Unloaded artery
with radial cut
B1
Figure 16.25 Decomposition into a single opened ring due to a radial cut. In addition to the standard
multiplicative decomposition F = AG, we consider an extra configuration where the cylinder is
open. We can then relate the geometry of opening to the growth tensor.
where p̂ is the Lagrange multiplier in B1 , and the strain-energy function is W =

W (αaρ , αaϑ , αζ ). In the unloaded configuration B1 , we assume zero net axial force
and bending moment [1054] and zero normal traction on the inner and outer bound-
aries,
ρ(a2 ) ρ(a2 )
tζ ρ dρ = 0, tϑ ρ dρ = 0, (16.100)
ρ(a0 ) ρ(a0 )
tρ [ρ(a0 )] = tρ [ρ(a2 )] = 0, (16.101)
where ρ(a0 ) and ρ(a2 ) are the inner and outer radii in B1 .
(a2) (a0)
Figure 16.26 Schematic of a radial cut in an artery and the corresponding opening angle ϕ.
200
Opening angle φ
180
160
140
120
100
80
60
40
r
20
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
Figure 16.27 The opening angle of an artery as a function of the radial growth. The axial growth
is assumed to be γz = 1.2.
Substituting the expressions from (16.95–16.96) into the incompressibility con-

dition (16.97) yields
πΛr ∂r
= 1, (16.102)
ϕ0 ρ ∂ρ
which can be integrated to obtain

1/2
πΛ 2
ρ(r ) = ρ(a0 ) +
2
(r − a0 )
2
. (16.103)
ϕ0
Equations (16.100) can be written in terms of the material coordinate r using the
relationship in (16.102),

πΛ a2
πΛ a2
r tζ (r ) dr = 0, r tϑ (r ) dr = 0. (16.104)
ϕ0 a0 ϕ0 a0
The equilibrium equation (16.98) is integrated to obtain an equation for the radial
stress
πΛ a2 r
tρ (r ) = (tϑ (r ) − tρ (r )) dr. (16.105)
ϕ0 a0 ρ(r )2
Substituting the boundary condition at the outer boundary (tρ (a2 ) = 0) into the
previous equation yields
20 P (kPa)
18
16
14
12
10
4 0.05 N
0N 0.03 N
0.01 N 0.1 N
0.005 N
2 0.001 N 0.02 N ζ
1
1 1.2 1.4 1.6 1.8 2 2.2
Figure 16.28 Pressure-axial stretch plot for a two-layer anisotropic cylinder without residual stress.
Parameter values given in the main text.

πΛ a2
r
(tϑ (r ) − tρ (r ) dr = 0. (16.106)
ϕ0 a0 ρ(r )2
Equations (16.104) and (16.106) are solved to obtain the unknown parameters ρ(a),
Λ, and ϕ0 and the solution for B1 is completely determined. For the set of parameter
values given above for a rabbit carotid artery, the opening angle was found to be 160◦ .
For an axial growth γz = 1.2, the opening angle is plotted with respect to varying
values of the radial growth. The radial growth corresponding to an opening angle of
160◦ can then be determined as demonstrated in Figure 16.27.
16.5.3 Analysis
16.5.3.1 The Role of Nonlinearities
The solution to the cylindrical deformation problem is given by the method outlined
in Section 16.1 for n = 2 layers with the parameters computed in the previous
section. In order to study the mechanical factors responsible for the constancy of
artery length in normal physiological regime, we reproduce the conditions leading
to the pressure-axial stretch plot in Figure 16.20 with parameters from [598] and no
residual stress. An interesting trend is revealed in Figure 16.28. As the external load
is increased, the length does not change as the pressure increases which agrees with
the experimental findings where the end-to-end distance of major arteries remained
mostly unchanged despite large variations of pressure [248, 1019]. This result is
consistent with the inversion lines found in [1133] and can be obtained without
residual stress. Inversion lines are regimes of axial stretch where the end-to-end
distance remains constant as arterial pressure is varied. What is the origin of these
inversion lines? Are they due to anisotropy, nonlinearity, or both?
In Figure 16.29, we compare an isotropic neo-Hookean model with an isotropic
Fung model. The neo-Hookean model completely fails to account for the qualita-
tive response of arteries as it does not exhibit inversion lines, but instead reveals
a local maximum which is associated with the well-known limit-point instability
discussed in Section 11.8.3.1. However, the isotropic Fung model is seen to capture
the inversion lines of Figure 16.20. We conclude that the inversion lines are not due
to the anisotropic property of the material or its residual stress. They are a direct
consequence of the strong strain-stiffening behaviot that can be modeled by a Fung
strain-energy function (see also [979] for a discussion of these effects).
16.5.3.2 The Role of Radial Growth
Next, we study the effect of the radial residual stress by comparing the radial stress
profile with and without residual stress. The computed stresses in the wall of both
a pressurized artery with no residual stress and a pressurized artery with residual
stress are shown in Figure 16.30. We observe that the transmural stress gradient is
greatly reduced when residual stress is included in the model. We recover here the
well-known effect of residual stress on the stress field across the arterial wall [208]
discussed in Section 13.8.3.1.
16.5.3.3 The Role of Axial Growth
A bifurcation analysis can be used to determine the critical buckling pressure and
to study the overall effect of axial stretch and residual stress on the critical pressure.
The Holzapfel-Gasser-Ogden model is used with the parameter values found in the
previous section. The grown cylinder with geometric parameters a0 , a2 , l0 is sub-
jected to an internal pressure and axial stretch, and the resulting cylinder has an inner
radius a, outer radius c, and length l. The axial stretch ζ refers to the ratio l/l0 .
The artery is first subjected to a constant axial stretch, and then the internal pres-
sure is increased until the artery buckles. The critical internal buckling pressure for a
given axial stretch is computed using a bifurcation analysis similar to the one devel-
oped in Section 16.3.2 with one major difference: The general relationships for the
instantaneous elastic moduli given in Section 14.8 is only valid for the isotropic part
of the strain-energy function. Here, the strain-energy function is of the form
A. Neo-Hookean model
P (kPa)
8
5 0N
0.1 N
0.2 N
4 0.3 N
1
ζ
0
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
B. Fung model
18 P (kPa)
16
14
12
10
0N 4N
4 0.4 N 2N
0.2 N 1N
0.1 N 0.8 N
0.6 N
2
ζ
0
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
Figure 16.29 Pressure-axial stretch plot for a one-layer cylinder, A0 = 1.75, A2 = 2.5 without
axial stress for A. A neo-Hookean material (W = μ2 (I1 − 3) where μ = 30). B. An isotropic Fung
material (μ = 26.950, b1 = 0.9925, b2 = 0.418, b3 = 0.0089, b4 = 0.0749, b5 = 0.0295, b6 =
0.0193) [602], where the Fung material is described by the strain-energy function W = μ2 (exp(Q)−
1) where Q = b1 E 2 + b E 2 + b E 2 + 2b E
2 ZZ 3 RR 4 E Z Z + 2b5 E Z Z E R R + 2b6 E R R E .
600 A. (no radial growth) 600 B. (with radial growth)

500 500
400 400
t
300 θ 300
200 200
tz
t
100 100
θ
t t
r z
0 0
r t r
r
1.2 1.22 1.24 1.26 1.28 1.3 1.32 1.34 1.36 1.15 1.2 1.25 1.3 1.35 1.4 1.45
Figure 16.30 Stresses in the arterial wall A. without residual stress B. with residual stress. In both
cases the artery is subjected to an internal pressure P = 13.33 kPa.
W (I1 , I4 , I6 ) = Wiso (I1 ) + Waniso (I4 , I6 ), (16.107)
and the instantaneous elastic moduli are defined by

(0)
L : F(1) = A(0) WAA : F(1) A(0) . (16.108)
In components, the instantaneous elastic moduli are:

(0)
Li jkl = Aim Am jkn A(0)
nl , (16.109)
where A is the elastic moduli defined by
∂2 W
A= (A(0) ). (16.110)
∂A∂A
Since the strain-energy function is the sum of an isotropic and an anisotropic term,
we decompose the elastic moduli in isotropic and anisotropic components
Li jkl = Liiso
jkl + Li jkl ,
aniso
(16.111)
where the nonzero components of Liso are given by the general relationships (14.122–
14.127) with W = Wiso . The coefficients of Lianiso
jkl must be given by other relationships
[1044]. For our particular problem, due to the particular dependence of the anisotropic
part on I4 and I6 only, we can directly compute the derivatives of Waniso with respect
to its invariants:

∂Waniso ∂Waniso
Lianiso
jkl = 2 m i m k δ jl + m i m k δ jl
∂ I4 ∂ I6

∂ 2 Waniso ∂ 2 Waniso
+2 m m m
i j k l m + 2 m i j k l ,
m m m
∂ I42 ∂ I62
(16.112)

where m i = αi Mi , m i = αi Mi so explicitly m 1 = m 1 = 0, m 2 = m 2 = α2 cos β,
m 3 = −m 3 = α3 sin β. Note that there is no cross-derivatives with respect to I4 and
I6 due to the fact that our particular strain-energy function does not contain terms
that involve both invariants. We also have
∂Waniso ∂Waniso
=
∂ I4 ∂ I6
2 2 2
= k1 eν (−1+cos β α2 +sin β α3 ) −1 + cos β 2 α2 + sin β 2 α2 ,
2 2
2 3
(16.113)
and
∂ 2 Waniso ∂ 2 Waniso
=
∂ I42 ∂ I62
2 2 2
= k1 eν (−1+cos β α2 +sin β α3 ) 1 + 2ν −1 + cos β 2 α2 + sin β 2 α2 2 ,
2 2
2 3
(16.114)
due to the fact that I4 = I6 = α22 cos β 2 + α32 sin β 2 . Once the instantaneous elastic
moduli are known, the procedure to identify bifurcations by solving the linearized
boundary-value problem follows the same steps as before.
For our problem, in order to reproduce the condition of the experiments in Jackson
et al. [658], we compute the critical pressure leading to buckling as a function of a
prescribed axial stretch. The results are shown in Figure 16.31. The stability curve
for a grown residually stressed cylinder is compared to a stress-free cylinder with
the same geometric parameters. Axial in situ stretches of 1.5 have been reported for
rabbit iliac arteries [410]. Note that in this regime the residually stressed cylinder can
withstand a much greater internal pressure before it buckles. Furthermore, an artery
under a typical physiological internal pressure P = 13.33 kPa is shown to withstand
a greater reduced axial stretch if the artery is residually stressed.
Jackson et al. [658] performed a study that assessed the effect of reducing axial
stretch in rabbit carotid arteries from 62 ± 2% to 33 ± 2% using interposition grafts.
The arteries were held at an axial stretch of 30% for 12 weeks. The arteries displayed
tissue growth and remodeling and all vessels became tortuous. Our analysis predicts
that an artery subjected to an internal pressure of P = 13.33 kPa will buckle when
the axial stretch is reduced to a value in the range between 1.25 and 1.33 for the
given growth parameters. We conclude that the onset of axial tortuosity in arteries is
consistent with a buckling instability triggered by variations of internal pressure.
70 with residual stress

Pcrit
60
50
40
without
30 residual stress
20
10
ζ
0
1 1.1 1.2 1.3 1.4 1.5 1.6
Figure 16.31 The critical internal pressure versus the axial stretch with and without axial residual
stress (solid line: γz = 1.2, γr = 1.6256, dashed line: γz = γr = 1). The arrow represents the
release of axial loading resulting in an axial stretch of about 1.25 for P=13.3 kPa. The artery
becomes unstable.
16.5.4 Discussion
Our analysis of a two-layer morphoelastic model for the mechanical response of

large arteries with differential axial and hoop growth reveals a number of interesting
features.
First, material nonlinearities play an important role in capturing the correct behav-
ior of arteries during a cardiac cycle. These nonlinearities are responsible for the
inversion lines observed for arteries under axial tension.
Second, the hoop residual stress reduces hoop stress gradients, a well-known
feature of differential growth in arteries that is recovered in the full model.
Third, the axial residual stress plays an important role in preventing buckling.
This observation is consistent with our analysis of rigidification of plant stems which
revealed the interplay between the nonlinear response of the material with the residual
stress induced by differential growth. In the case where arteries become unstable
under the internal pressure through a buckling instability, the mechanical analysis of
arteries presented here is consistent with the onset of such an instability. However,
this analysis cannot explain the remodeling of the structure after the instability that
will eventually lead to permanent tortuosity. Such an analysis would require a post-
buckling analysis for a model of arteries involving both growth and remodeling.
P
χ(0) χ(1)
R0 r
A2
r
A1
A0
Figure 16.32 Radial instability in a two-layer cylinder can be induced through differential or
incompatible growth.
16.6 Circumferential Buckling and Mucosal Folding
So far, we have considered axial instabilities, resulting in buckling or barreling.

There is yet another interesting bifurcation that occurs in growing tubes. Consider a
two-layer cylindrical shell and assume that each layer has constant radial and hoop
growth but no axial growth as shown in Figure 16.32.
Depending on the geometry and stiffness of each layer, a number of interesting
instabilities can arise with or without external pressure P. For instance, if the external
cylinder is rigid and the inner tube grows isotropically, it will develop compressive
stresses and the inner layer will buckle circumferentially. Similarly, if the inner layer
is rigid and the outer layer grows isotropically, it will develop flower-like patterns
as shown in Figure 16.33D. Depending on the relative stiffness of the two materials,
their nonlinear responses, the geometry, and growth constants, various patterns can
develop in the growth process.
We refer to these type of patterns as circumferential buckling or mucosal folding
as they are observed in the oesophagus [790, 1349], blood vessels [778], and gas-
trointestinal tract [808]. Various examples are shown in Figure 16.33 and a detailed
analysis of mucosal folding in airways is considered in the next section.
We can capture the onset of such patterns by looking for bifurcations around a
grown configuration as shown in Figure 16.32. In the axial direction, one can either
assume that no extension takes place by assuming no end displacement (ζ = 1,
in-plane strain) or, by assuming free-ends (tz = 0, in-plane stress). Here, we will
consider the simpler situation and restrict the deformation of cylinders to a two-
dimensional cross-section by assuming no end displacement, so that ζ = 1 is known
from the onset. The general methods developed in this chapter are directly applicable
to this setting. Before considering an application, we study the simple case of a single-
layer cylinder growing under pressure.
16.6 Circumferential Buckling and Mucosal Folding 525
A.
1 mm
Airways
B.
(b)
(a)
Oesophagus
C. D.
Arteries Gels
Figure 16.33 Examples of mucosal folding A. Airways from [512]. The top row shows tissue
specimens in no-load state the bottom row shows the zero-stress state after a radial cut for rats
aged 1–32 weeks. B. Oesophagus ring. (a) the mucosal folding in vitro and (b) the zero-stress state
clearly showing that the folding pattern is due to the mismatch in length [1349]. C. Creasing in the
inner wall of arteries [508]. D. Isotropic growth of a gel (black) attached to a hard core (adapted
from [276]).
16.6.1 Example: Circumferential Buckling in a Cylinder
We consider a single-layer cylindrical shell under growth and external pressure and
no end displacement. The cylinder has radii A0 , A1 and a0 , a1 in the initial and current
configuration, respectively. Therefore, we have
[A] = diag(α−1 , α, 1), [G] = diag(γ, 1, 1). (16.115)
From our previous analyses, we known that the equilibrium solution is obtained by
solving
a1
αŴ (α)
−P= dr, (16.116)
a0 r
where Ŵ (α) = W (α−1 , α, 1). Here, P is the applied load on the cylindrical tube,
such that P > 0 in the case of external pressure and P < 0 for internal pressure.
For a given growth constant γ and fixed pressure P, (16.116) is an equation for the
unknown parameter a0 , since a1 is related to a0 through
a12 = a02 + γ(A21 − A20 ). (16.117)
The stability analysis proceeds by considering incremental deformations restricted

in the plane
χ(1) = u(r, θ)er + v(r, θ)eθ . (16.118)
Then, in polar coordinates, we have

⎡ uθ − v ⎤
ur
⎢ r ⎥
[F(1) ] = [A(1) ] = ⎢
⎣
⎥.
⎦ (16.119)
u + vθ
vr
r
The incompressibility condition is
u + vθ
tr(A(1) ) = u r + = 0. (16.120)
r
Inserting (16.119) into the incremental equilibrium equation, div T(1) = 0, yields two
differential equations involving the unknown functions u, v, and p (1) . To simplify,
these functions are assumed to be of the form
u(r, θ) = f (r ) sin(nθ),
v(r, θ) = g(r ) cos(nθ), (16.121)
(1)
p (r, θ) = h(r ) sin(nθ).
The buckling mode n corresponds to the number of folds in the buckled state. Making
these substitutions and using the incompressibility condition (16.120) to solve for
g(r ) in terms of f (r ), the system can be simplified to a single fourth-order differential
equation for f (r ):
B4 f (4) (r ) + B3 f (r ) + B2 f (r ) + B1 f (r ) + B0 f (r ) = 0, (16.122)
where

2 γ R0 2 + 2 r 2
B4 = R0 , B3 =
2
,
rγ
3 R0 4 γ − 8 R0 2 r 2 + n 2 R0 4 γ + r 4 n 2 γ
B2 =− ,
r 2 R0 2 γ
−3 r 4 n 2 γ R0 2 + 2 r 6 n 2 + 3 R0 6 γ − 4 R0 4 r 2 + n 2 R0 6 γ − 2 R0 4 n 2 r 2
B1 = ,
r 3 γ R0 4
−3 R0 4 γ + 4 R0 2 r 2 + 3 n 2 R0 4 γ − 4R0 2 n 2 r 2 − r 4 n 2 γ + r 4 n 4 γ
B0 = .
r 4 R0 2 γ
Note that R0 = R0 (r ) is determined from the finite deformation via
r 2 = a02 + γ(R02 − A20 ). (16.123)
We solve this equation with the incremental boundary conditions by setting γ = 1,

and choose the external pressure P as the bifurcation parameter. In Figure 16.34, the
critical value of P is plotted as a function of the radial growth γ, for various modes
n and for fixed tube thickness A/B = 0.5.
For a given γ, the smallest value of Pcr over all modes is the critical external
pressure, denoted Pcr∗ , at which the tube becomes unstable and buckles, and the
corresponding mode n cr is the critical buckling mode. For P < Pcr∗ , the tube remains
circular.
The value of Pcr∗ for γ = 1 is the critical pressure without growth, which occurs
for mode n cr = 2. Growth can have both stabilizing and destabilizing effects. For
this thickness, radial resorption (γ < 1) is destabilizing: Pcr∗ is smaller for all values
of γ < 1 than for γ = 1. Conversely, radial growth (γ > 1) has a stabilizing effect
initially, but it is destabilizing for large γ. Observe that the strongest tube defined as
the tube with the highest value of Pcr∗ , is for γr ≈ 1.72, the point marked b in Figure
16.34.
Along with the change in critical pressure, growth has a dramatic effect on the
shape after bifurcation as well. Note that in Figure 16.34, the lower most deformation
corresponds to buckling due solely to differential growth (since Pcr∗ = 0). It should
not be surprising that differential growth can cause instability without any external
loading as it was demonstrated earlier that growth can induce sufficient residual stress
to destabilize a structure even in the absence of external loads. Similarly, differential
Pcr 1.5
9 0.5
1.6
7
1.2 6
4
0.8 3
2
0.4
0
0 1 2 3 γ
1.5 1.5
1
1
0.5
0.5
Figure 16.34 Critical buckling pressure as a function of radial growth γ for modes n =
2, 3, 4, 6, 7, 9, with the corresponding tube deformations. Note that the amplitude has been chosen
to illustrate the structure - it has no physical significance since the shapes are produced in a linear
stability analysis. Adapted from [912].
growth can have a stabilizing effect by changing the geometry of the body: A stubbier
body is less likely to suffer buckling than a thin one. A greater range of possibilities
opens up when considering a tube with two connected layers of material with different
stiffness properties [912]. These ideas could be used in the design of biomaterials in
which the fine tuning of critical pressure is desired.
16.6.2 Example: A Two-Layer Cylinder
Based on the previous analysis, it is now straightforward to look at the bifurcation of

a two-layer cylinder, each layer with homogeneous properties. Here, we fix A0 = 1,
Figure 16.35 Critical

pressure as a function of the
growth rate ratio between
inner and outer layers for
isotropic growth in each
layer. The incremental
boundary conditions are zero
traction on both surfaces.
The critical mode is n cr = 2
at each point (A0 = 1, A1 =
1.1, A2 = 1.5, μ1 = 10,
μ2 = 1). Adapted from
[912].
A2 = 1.5 and the elastic moduli of the inner and outer neo-Hookean layers to be
μ1 = 10, μ2 = 1. Under this fixed geometry, we investigate buckling for different
isotropic growth in each layer. The critical pressure (smallest value over all modes)
is plotted as a function of the ratio of growth rates γ1 /γ2 in Figure 16.35. The critical
pressure only depends on the ratio. Deformations with different values of the growth
parameters with equal ratios are equivalent, up to an isotropic and equal growth of
both layers, so that γ1 /γ2 = 1 is equivalent to the case of no growth. Hence, higher
growth in the inner layer is destabilizing, creating a compressive stress on the inner
layer. Conversely, smaller growth in the inner layer creates a tensile, stabilizing effect
on the inner layer.
The boundary conditions imply that both edges are unconstrained in the buckling
deformation. If we forced the outer edges to remain circular during deformation by
A. 1.5
B.
1
Pcr 0.5
Pcr 2.5
6
2
2.5 1.5
2 1
3
5
1.5
3 0.5
3 4 1
0.5
4 3
2
5 3
1 4 6
6 2 6
7 7 9 γ1 / γ 4
5 5
6 γ1 / γ
8 8
0 2 1 2
0 0.5 1 1.5 0 0.5 1 1.5
Figure 16.36 Critical pressure as a function of growth rate ratio between inner and outer layers
for isotropic growth in each layer, for two different inner layers A. B = 1.1 and B. B = 1.2. The
critical mode is labeled at each point, and select deformation patterns are included ( A0 = 1, A1 =
1.1, A2 = 1.5, μ1 = 10, μ2 = 1). Adapted from [912].
Figure 16.37 Cartilages

structure of the lower
airways. From Gray’s
Anatomy of the Human Body
[498, plate 961].
requiring that the perturbation vanishes at the outer boundary, we obtain different
buckling modes. Instead of exciting mode n cr = 2, modes n cr = 3 to n cr = 9 are
excited first as the growth ratio increases. We also notice that the buckling pressure is
significantly higher, so that the tube is essentially stronger when the outer edge is kept
circular. Furthermore, increasing the thickness of the inner layer results in an overall
increase in the buckling pressure. It also has the effect of reducing the buckling mode.
If the inner layer is thinner, there exists a growth ratio such that isotropic growth in
each layer induces buckling without any external pressure (Figures 16.36 and 16.37).
16.7 Application: Asthma and Airway Remodeling
16.7.1 Background
Asthma is a complex disease characterized by a narrowing of the airway and

reduced lung function. Chronic asthma is often accompanied by irreversible struc-
tural changes to the airway wall, collectively referred to as airway remodeling [1068].
16.7 Application: Asthma and Airway Remodeling 531
Figure 16.38 Asthma is typically associated with an increase in wall thickness leading to a reduc-
tion of the air flow. Picture courtesy of Imtiaz Ibne Alam.
Airway remodeling is a complex process occurring at multiple time and length scales
and involving many different chemical, biochemical, and physical stimuli. Despite
a wealth of research, it is still not clear how each of the different structural changes
individually affects airway function [1281], nor is it known whether these different
changes are beneficial or detrimental to asthmatic patients [854]. A well-documented
key feature of airway remodeling is an increase in airway wall thickness as shown in
Figure 16.38, detected at all levels of the bronchial tree and all layers of the airway
wall [631, 662, 745, 953].
In response to certain stimuli, the smooth muscle surrounding the airway wall
contracts, and the luminal boundary folds or buckles. In normal physiological con-
ditions, such an event is marked by only modest narrowing of the airway [903]. In
asthmatic patients, however, this airway narrowing tends to occur for lower stimuli,
and also leads to exaggerated narrowing [696] as seen in Figure 16.39. Viewed as a
mechanical problem, this folding indicates an instability of the inner mucosal edge
in response to an external pressure provided by the contraction of smooth-muscle
cells.
Some of the key questions from a modeling perspective are
• What is the critical buckling pressure?
• What are the critical buckling modes (i.e., the number of folds in the buckled
state)?
• What is the role of airway wall thickness on the instability?
Several authors have modeled mechanical and geometric aspects of mucosal fold-
ing. Lambert [753] modeled the basement membrane as a single-layer elastic tube
and showed that the buckling mode can have a dramatic impact on airway narrowing;
in particular an airway with fewer folds will have a greater degree of occlusion if the
folds extend until epithelial cells come into contact.
This basic model was expanded upon in [754], in which a geometric constraint
was included as a mechanism for selecting the number of folds, and was further
expanded to include a thin layer of fluid on the inner edge of the airway allowing
for possible surface tension effects [576]. Other geometry-based models have been
proposed to take into account the inextensibility of the basement membrane: Seow et
al. [1136] showed that the number of folds was directly connected to tethers between
the airway and smooth muscle; Donovan and Tawhai [293] studied effective airway
radius given fixed folding geometry. A more sophisticated model for airways as a
two-layer elastic tube was proposed by Wiggs et al. [1323]. By solving the buckling
problem in a finite-elements analysis, these authors found that wall thickness has
a significant effect on the buckling mode and degree of narrowing. Similar finite-
elements models include the inertial effects of the basement membrane [56] and a
comparison with buckling experiments on rubber tubes [626].
16.7.2 The Model
Here, following [913] we consider a model of asthma that includes both nonlinear
material responses to large deformations, which are common in airway narrowing
[961], and the relative change of geometry and residual stress due to growth [661].
We address whether growth significantly impacts the mechanical response of the
airway to smooth-muscle contraction beyond the change in geometry, and thus be
an important factor in airway remodeling.
We model the airway as a two-layer cylindrical morphoelastic structure. We only
consider tissue interior to the smooth muscle. Following the model given by Wiggs et
al. [1323], a stiff and thin inner layer corresponds to the mucosal region, consisting of
the basement membrane, the lamina propria, and the epithelium [50]. Surrounding
this region is a portion of the submucosa region, consisting of loose connective
tissue. The inner layer is much thinner and stiffer than the outer layer as shown in
Figure 16.40.
The airway smooth muscle surrounds the outer layer, with smooth muscle con-
traction providing a force that deforms and eventually induces buckling in the tube.
Figure 16.39 Cross section

of a peripheral bronchus
showing small folds at
extremities of large folds
[754]. Arrow points to
submucosa. Smooth muscle
(sm), lumen (L). Bar is 50
μm.
Figure 16.40 Bilayer model Submucosa: Mucosa:

of airway wall, consisting of Soft, thick outer layer,
a stiff and thin inner layer, shear modulus μ  shear modulus μ
the mucosa, and a soft and
thick outer layer, the R0
submucosa. Surrounding the
submucosa is smooth A2
muscle, which applies a
normal pressure when it
contracts. A1
A0
Smooth muscle
Smooth-muscle contraction is a fairly complex process, and several models have

been formulated to couple the underlying chemical processes to mechanics [132,
1041, 1292]. The net mechanical effect of smooth muscle contraction is a normal
force applied at the airway wall–smooth-muscle interface [1041].
Since our purpose here is to focus on growth and buckling, smooth-muscle con-
traction is taken into account as an applied normal pressure boundary condition at
the outer edge. Also, since mucosal folding occurs at the mucosal region while the
smooth muscle remains roughly circular [754], we impose the boundary condition
that the outer edge remains circular in the deformed, buckled state.
Material dimensions for the airway in the initial configuration are the inner radius
A0 , the thickness of the inner layer, A1 − A0 , and the thickness of the outer layer
A2 − A1 .
We assume that the tissue is an isotropic, incompressible hyperelastic material
[119]. In the absence of detailed testing experiments on the different tissues, we
model the two layers as either neo-Hookean with
μi
Wi = (I1 − 3), i = 1, 2, (16.124)
2
or a Fung model, to explore the possible role of strain-stiffening
μi
Wi = [exp β(I1 − 3) − 1], i = 1, 2, (16.125)
2β
where β is a measure of strain-stiffening with typical values between 3 and 20.

The model is three dimensional, but it assumes only plane-strain deformations,
and the deformation is uniform along the tube axis and ζ = 1 in all deformations.
We can therefore focus our analysis on the two-dimensional problem and, as before,
the elastic deformations are fully characterized by the hoop elastic stretch α = αθ ,
so that
Figure 16.41 A comparison

of pressure vs area for the Area (mm2)
symmetric deformation of a
two-layer tube for four
different forms of model. Fung, β=20
Parameters are A0 = 0.98,
A1 = 1, A2 = 1.5, μ1 = 40,
and μ2 = 4.
Fung, β=3
neo-Hookean
Linear
0.4 1.2 2.0 2.8 3.6

Pressure (kPa)
I1 = α−2 + α2 + 1. (16.126)
The analysis of deformations and buckling proceeds as in the previous sections and
details will be omitted.
16.7.2.1 The Role of Nonlinear Strain-Stiffening
In Figure 16.41, we plot pressure against area for the symmetric deformation of
a two-layer tube for the neo-Hookean model, the Fung model with both β = 3
and β = 20, and also for a linear model. Each model gives qualitatively similar
pressure area relationships as expected, but a significant quantitative difference exists
between each curve. From this observation, we conclude that nonlinear effects may
be important for quantitative predictions. However, most of the qualitative trends
observed below remain unchanged when the neo-Hookean model is replaced by a
Fung model. With the Fung model, effects are amplified, i.e., a more drastic change
in buckling pressure and mode is observed. Therefore, the analysis will be restricted
to neo-Hookean materials.
16.7.3 Analysis
The bilayer model for airways contains two growth parameters for each layer, two
shear moduli, and three reference radii which makes for a very large parameter
space. Our goal is to investigate the effect of growth on the buckling of a normal
versus a remodeled airway. Therefore, in our study we keep reference dimensions

of the airway fixed and change the thickness of the airway by varying the growth
parameters. Motivated by physiological measurements [626, 662, 1323], we use as
reference for a normal airway, the base values A0 = 0.98, A1 = 1, A2 = 1.5 (all
measured in cm), and stiffness ratio μ1 /μ2 = 10.
We assume that when the smooth muscle contracts, it creates a pressure normal to
the circular cross section and the tube deforms in a symmetric fashion, maintaining
its circular shape, until a bifurcation point is reached at a critical pressure, at which
point the cylinder buckles to an asymmetric state. Physiologically, we are interested
in the critical pressure at which folds form as it can be linked to the magnitude of the
contractile force of the airway smooth muscle necessary to induce buckling of the
airway. We assume here that the relationship between normal pressure and contractile
force is monotonic. Thus, an increase in critical pressure corresponds to a greater
contractile force for buckling. In terms of airway hyper-responsiveness, buckling
pressure (or contractile force of airway smooth muscle for buckling) provides a
measure of the “strength” of an airway.
16.7.3.1 The Role of Differential Growth
The first effect we consider is isotropic growth, but with different growth rates in each
layer, that is, we assume γr = γθ = γ in each region, but γ1 = γ2 . As an example,
we fix γ2 = 1.2 and vary γ1 > 1. We show in Figure 16.42 the critical pressure and
critical buckling mode as a function of γ1 . In Figure 16.43 the same plots are shown
with γ1 = 1.2 fixed and γ2 varying. The critical buckling pressure only depends on
the ratio γ1 /γ2 . Different values with the same ratio are equivalent up to an isotropic
and equal growth multiple in each layer, with no change in the critical pressure (note
that the airway dimensions will vary by the same scalar multiple).
Comparing the critical pressure to the reference case of no growth, it is interesting
to note that when the ratio γ1 /γ2 > 1, the airway is less stable, i.e., it buckles at a
lower pressure. On the other hand, a greater pressure is required when this ratio is
less than one. The airway is weaker with respect to buckling when the inner layer
grows at faster rate than the outer layer, and likewise stronger when the outer layer
grows faster. Observe that this effect is significant as the change in critical pressure
increases by a factor of about 8 as γ2 changes from 1 to 1.5.
We compare the size of the nonsymmetric deformation for different parameters.
In this way we determine whether buckling at one set of parameters should result
in more or less narrowing initially than buckling at another set of parameter values.
At the lower value of γ1 , the buckling pressure is much higher, but the airway nar-
rows significantly more. At higher values, the airway narrowing is reduced, but the
buckling pressure decreases. There is essentially a trade-off between the two growth
rates. This trade-off does not occur in the case of fixed γ1 as seen in Figure 16.43.
The larger value of γ2 has a higher resistance to buckling and comparable narrowing
to the smaller value of γ2 .
1.5
0.5
2
1.5
A. 0.5 B.
Pcr (kPa) ncr
30
3.2
28
2.4
26
1.6
0.8 crit. pressure, no growth 24
0 22
1 1.2 1.4 γ1 1 1.2 1.4 γ1
Figure 16.42 A. Critical buckling pressure and B. buckling mode as a function of isotropic growth
of the inner layer, for fixed isotropic growth in the outer layer, γ2 = 1.2.
It may seem counterintuitive that extra growth in the outer layer is required to
strengthen the airway wall, since the inner layer is stiffer. In fact, a primary result of
Wiggs et al. [1323] was that increasing the thickness in the inner layer has a greater
impact on buckling than increasing the thickness of the outer layer. Indeed, differen-
tial growth creates residual stress, which induces a competition between mechanics
2 2
1.5 1.5
1 1
0.5 0.5
A. B.
Pcr (kPa) ncr
29
3.2
28
2.4
27
1.6
26
crit. pressure, no growth
0.8 25
γ2 γ2
0 24
1 1.2 1.4 1 1.1 1.2 1.3 1.4 1.5
Figure 16.43 A. Critical buckling pressure and B. buckling mode as a function of isotropic growth
of the outer layer, for fixed isotropic growth in the inner layer, γ1 = 1.2.
Pcr (kPa)
1.5
1.6 1
n=14
1.5
1.2
1
n=25
γr
0.8
1 1.5 2
Figure 16.44 Critical pressure as a function of anisotropic growth of the inner layer. All other
growth parameters are set to one.
and geometry. In particular, when the inner layer grows, it pushes against the outer
layer, creating a compressive residual stress in each layer. This mechanical effect is
destabilizing against external pressure. Conversely, when the outer layer grows faster,
it pulls the inner layer, creating a stabilizing tensile stress. This counterintuitive result
highlights the important mechanical role of differential growth.
16.7.3.2 Anisotropic Growth
Next, we consider the role of anisotropic growth. For simplicity, we assume that the
outer layer does not grow and explore buckling as a function of anisotropic growth in
the inner layer by varying the ratio γr /γθ . Thickening of the inner layer only occurs
with radial growth.
In Figure 16.44, the critical pressure is plotted for γθ = 1 fixed and γr varying
from 1 to 2. As γr increases, the inner layer doubles in size from γ = 1 to 2.
Correspondingly, the buckling pressure increases in a linear fashion. Again, there
is a competition between mechanics and geometry: Radial growth causes the inner
layer to be relatively thicker—a stabilizing geometric effect—but at the same time it
creates a radial compressive stress, which is destabilizing mechanically. In this case,
since the inner layer was very thin to begin with, the geometric effect is stronger,
and the newly grown airway increases its stability.
In terms of the degree of narrowing, with increasing growth the buckling mode
decreases monotonically and significantly, from n = 27 at γr = 1 to n = 14 at
γr = 2. In Figure 16.44, the deformations at γr = 1.2 and γr = 2 show the effect of
the buckling mode. In both cases, the inner radius at the point of bifurcation is about
0.87. Both plots are produced using the same-sized perturbation in the incremental
deformation, but the airway narrowing is exaggerated at the point where the lower
mode, γr = 2.
16.7.4 Discussion
The morphoelastic model shows that differential growth can have a significant effect
on airway buckling. Therefore, it may be an important contributing factor in under-
standing the mechanical behavior of airways which have undergone remodeling.
Isotropic growth with differing rates in each layer leads to the somewhat surprising
conclusion that if the stiff mucosa grows at a faster rate than the soft submucosa, the
airway actually becomes less stable. Anisotropic growth of the inner layer leads to a
strengthened airway, but it significantly decreases the buckling mode.
Differential growth provides a mechanism to explain otherwise contradictory
structural changes, for instance an airway wall becoming thicker and at the same
time less stable. However, airway remodeling involves more complex changes than
just an increase of the mucosal and submucosal layers [200, 960]. The airway smooth-
muscle layer also gains mass, and there is uncertainty as to whether the contractile
potential changes [1015]. Thus, whether the airways are actually less stable after
remodeling is not yet established.
The two-dimensional nature of the model and the plane-strain assumption are
supported by the fact that folds in airways are observed as longitudinal ridges [1346].
The linear stability analysis has the drawback that it only provides information on the
deformation up to the point of the buckling. Nevertheless, knowledge of the buckling
pressure is a good measure of the “strength” of an airway, and previous studies have
shown that a tube becomes much more compliant after buckling [1323], suggesting
that narrowing occurs more rapidly after buckling. Thus, buckling pressure seems
to be a valid measure for investigating the mechanical impact of growth on airway
narrowing.
The study presented here is mostly qualitative as many physiological values are
not available. Nevertheless, an estimate of the buckling pressure can be obtained. It
relates the normal force induced at the interface due to smooth-muscle contraction.
From a study on human airway walls [973], we estimate the shear modulus in the
stiff mucosal layer, μ1 , at around 120 kPA. Taking P ∗ ≈ 0.4 as a typical value of
buckling pressure leads to an estimate of 1.6 kPA as the amount of stress necessary to
induce buckling, comparable to pressures that canine airways can generate of about
3 kPA [523].
16.8 Residual Stress Through Fiber Contraction 539
16.8 Residual Stress Through Fiber Contraction
So far, we have studied the creation and effect of residual stress in cylindrical struc-
tures through inhomogeneous or anisotropic growth. Another possible mechanism
for stress creation is the contraction or extension of fibers in an anisotropic material.
Cylindrical geometry is particularly well suited to illustrate this process.
We recall that an anisotropic material with one family of fibers can be characterized
by the theory of invariants as described in Section 11.11.1. A general anisotropic
material with one fiber with reference direction M is fully characterized by a strain-
energy function
W = W (I1 , I2 , I3 , I4 , I5 ), (16.127)
where Ik are the three isotropic and anisotropic pseudo-invariants
1 2
I1 = tr B, [I − tr(B2 )], I3 = det B = J 2 ,
I2 = (16.128)
2 1
I4 = M · (CM), I5 = M · (C2 M), (16.129)
and C = FT F and B = FFT .

Fiber contraction can be modeled by assuming that in the absence of an isotropic
deformation, the strain-energy function does not vanish identically, that is, W (3, 3,
1, 1, 1) = 0.
Rather than developing a general theory of fiber contraction, we illustrate the main
idea by generalizing the standard fiber-reinforced model to include fiber contraction
[272, 1025], that is, we consider materials described by the strain-energy function
μ1 μ4
W (I1 , I4 ) = (I1 − 3) + (I4 − ν 2 )2 , (16.130)
2 4
The parameters ν describes the effect of fiber compression in the matrix. Fibers can
be inserted in the matrix while the matrix is under stress and, as we relax the matrix,
the fibers may become compressed or stretched in the reference configuration. Thus
the parameter ν is the stretch the fiber needs in order to attain its natural length in
the reference configuration. For example, if an unstressed fiber is added to an elastic
matrix in a state of tension, the fibers in the corresponding reference configuration
will be compressed. Hence, the fiber needs to be stretched by a factor ν > 1 to recover
its natural length in that reference configuration. Note that for ν = 1, W (3, 1) = 0
and the material is residually stressed; even in the absence of external loads, neither
the matrix nor the fibers are stress-free.
Figure 16.45 The twist τ as

a function of the fiber angle 0.04 τ
for a capped cylinder ν(2)
under pressure with a single
0.02
fiber in pre-compression.
(2)
ν
ν<
The particular case
Φ
P = 0.2, μ4 = 10 is used. In (1) <
this case ν (1) ≈ 1.002082 0.5 ν 1.0 π/2
and ν (2) ≈ 1.05641.
-0.02
-0.04 ν(1)
16.8.1 Rotation of a Pressurized Anisotropic Cylinder
As a example of the effect of fiber contraction, we consider the case of a caped

pressurized thin tube with a single pre-compressed fiber helically wrapping around
the tube so that, in the cylindrical basis (E R , E , E Z ), we have
M = cos E + sin E Z . (16.131)
Under pressure, the tube deforms to another tube so that the deformation reads, in
the usual cylindrical coordinates {r, θ, z} and {R, , Z },

R 2 − A2
r= a2 + , θ = + τζ Z, z = ζ Z, (16.132)
ζ
where, the axial stretch ζ, and the twist τ are constant, and as before, A and a are
the inner radii of the tube in the reference and current configurations, respectively.
In the thin tube limit, the axial stretch is simply λ = a/A.
We take, without loss of generality, μ1 = 1 in (16.130) and we are left with
three material parameters (, ν, μ4 ) and one load P. The problem is solved for
the radial stretch λ, the axial stretch ζ and the twist τ by the method described in
Section 11.11.4.
We now ask the following question: Does a capped tube with a right-handed
single helical fiber ∈ [0, π/2] under positive pressure P rotate clockwise (τ < 0,
left-handed) or counterclockwise (τ > 0, right-handed)?
We start by considering the simple problem of twist for very small values of pre-
compression as shown in Figure 16.45 (curve ν (1) ). As expected, a tube with right-
handed helical fiber under pressure creates a left-handed rotation. In the process, the
tube unwinds. Note that the direction of rotation is opposite to the rotation of a spring
in extension, discussed in Section 5.7.3.
Next, we consider the same system for increased values of the parameter ν. There,
we observe that the rotation changes direction. Ultimately for larger values of ν, the
rotation is completely inverted, i.e., entirely right-handed rotation for all values of
the fiber angle (curve ν (2) ).
16.8 Residual Stress Through Fiber Contraction 541
Figure 16.46 Generic Φ

change of twist as a function
of the angle and the π/2
pre-compression. For
ν < ν (1) , the rotation is
RH
left-handed (negative τ ) and
right-handed for ν > ν (2) .
For ν (2) < ν < ν (1) , the
rotation can be left-handed LH
or right-handed depending
on the fiber angle.
ν
0
1 ν ν(2)
(1)
Based on these observations, we are interested in identifying the precise values

of P, ν and for which such an inversion of rotation occurs. To do so, we use
the equations (11.345–11.347), obtained for the membrane limit of a pressurized
cylinder, with J6 = 0, J4 replaced by I4 − ν 2 , and an initial state with no load or
rotation M = F = τ = 0. These equations then read
ζ 2 + λ2 − ν 2
cos2 = , (16.133)
ζ 2 − λ2
1 − 2ζ 4 λ2 + ζ 2 λ4 = 0, (16.134)
ζ 2 λ4 − 1 = Pζ 3 λ4 . (16.135)
We can easily find the values of P, ν for which the inversion of rotation occurs at
= π/2 and = 0, corresponding to the vales ν (1) and ν (2) in Figure 16.45.
As shown in Figure 16.46, for ν < ν (1) , the rotation is left-handed and right-
handed for ν > ν (2) . For a given P, the first critical value ν (1) is the first real root
larger than 1 of
1/2 √

ν 5/2
ν + 8+ν
3 6 P ν + 8+ν
3 6 − 2 ν + 8 = 0. (16.136)
Similarly, for a given P, the second critical value ν (2) is the first real root larger than
1 of
ν 2 P 2 + 4(ν 6 − 1)(ν P − 1) = 0. (16.137)
Note that inversion only occurs for pressure values such that
√ √
6 2√
13+3 21
2 57 + 12 21 − 3 1 + 21
P≤ ≈ 0.749868,
55/6
which corresponds to a maximal axial stretch of ζ ≈ 1.17819. Note also that none
of these values depend on the fiber strength μ4 .
The possibility of an inversion is a rather surprising result. Indeed it demonstrates
that the chirality of the fibers does not necessarily determine the chirality of the cylin-
drical structure and its rotation, even for small values of pre-compression (less than
1%). In this range, the structure will exhibit macroscopic left-handed or right-handed
rotation depending on the fiber angle, and small changes in the pre-compression and
angle will change the rotation as observed in the rotation of Phycomyces described
in the next section.
16.9 Application: The Spiral Growth of Phycomyces
16.9.1 Background
Phycomyces blakesleeanus is the most studied Phycomyces, a genus of fungi first

reported by the Belgian biologist and cleric Jean-Baptiste Carnoy in 1870 [170].
After a complex reproductive cycle this fungus emerges as a strikingly large, single-
celled, aerial hypha—-known as the sporangiophore—reaching sizes of up to 10 cm
long as shown in Figure 16.47. The organism, its rotating growth phases, and its
remarkable tropic responses to various external stimuli has long been a source of
fascination to biologists and biophysicists alike [90, 219, 375, 416, 670].
Helical growth is the rotation of a structure as it extends axially. It is observed in
many different tubular structures across the plant kingdom [42]. However, it is not
known whether this type of growth serves a biological function or provides some
mechanical advantage, as hypothesized in the case of spiral grain of trees [66, 1132].
Nevertheless, elucidating the physical processes that lead to helical growth provides
insights into the fundamental problem of relating cell wall structure to growth and
form in plants. The physical explanation for helical growth is based on the intuitive
notion that a cylinder with a helically structured (cell) wall, under axial stress due to
turgor pressure, should unwind in the absence of an external axial moment. Hence,
helical growth can be viewed as the macroscopic mechanical consequence of cell
wall anisotropy. However, as we saw in the previous section, the direction of rotation
of a fiber-reinforced pressurized cylinder is not only related to the orientation of the
wall fibers, but also to its mechanical environment. We refer to right-handed growth
(RH) when a point on top of the cylinder, viewed from above, turns counterclockwise
during extension, and left-handed growth (LH) when the same point turns clockwise.
Similar to other systems exhibiting helical growth, the cell wall of the cylindrically
structured sporangiophore of Phycomyces blakesleeanus is anisotropic, essentially
composed of chitin microfibrils embedded in an elastic matrix of amorphous material
made out of chitosan and chitin [886, 1095]. As illustrated in Figure 16.48, after an
initial aerial growth phase (Stage I) and the development of the sporangium (Stages
II and III), axial growth resumes in Stage IV. A remarkable behavior is then observed
16.9 Application: The Spiral Growth of Phycomyces 543
A. B.
active growth zone
2r
Figure 16.47 A. The fungus Phycomyces blakesleeanus as drawn by Carnoy [170]. The spherical
dome on top of the tube is the sporangium that contains the spores that are disseminated on com-
pletion of the life cycle. B. Optical microscopy of a colony of Phycomyces (courtesy of Michael
Tabor and Robert Reinking).
[178]: in Stage IVa, RH growth takes place for approximately 1 hour; followed by
Stage IVb in which LH growth is observed that can lasts up to 48 hours. A further
rotational inversion (Stage IVc) has also been reported by some authors [993, 996,
1333].
The length and time scales in Stages IVa and IVb are as follows: A typical radius
of the sporangiophore is l = 50 μm and we will use this radius (assumed to be
constant throughout these phases) as the length scale. The wall thickness is about
l/83. The growth zone, which is the zone immediately below the sporangium where
new cell wall material is added, extends to a size h ≈ 60 l. A typical sporangiophore
length is 3 − 4 × 103 l.
LH
RH
growth (mm)
LH
time (hours)
Figure 16.48 The four stages of growth for the aerial hypha. In Stage IV helical growth reversal
is observed: Stage IVa exhibits right-handed growth and Stage IVb presents left-handed growth.
Stage IVb continues for up to two days when the organisms can reach up to 10 cm in height. LH
denotes left-handed (clockwise) rotation and RH denotes right-handed (counterclockwise) rotation
In Stage IVb, the typical axial velocity is a constant vz = 60 l/hr [1333]. This
axial extension is accompanied by an angular rotation in Stage IVb of approximately
4π/hr for a duration of 24–48 hours. In contrast, Stage IVa lasts about 1 hour. A
typical internal turgor pressure is about 0.4 MPa [234].
16.9.2 The Model
Here following [483], we present a mechanical model for the rotational reversals
observed in Stages IVa, IVb, and IVc. We view the sporangiophore as a growing,
fiber-reinforced, elastic structure, and include a dynamical remodeling and deposition
of the fibers. Our basic assumption is that new fibers are deposited in a stress-free
state along the direction of existing fibers.
We model the sporangiophore as an anisotropic, elastic, incompressible tube with
two families of embedded fibers following the description of Section 11.11.4.
The first family of fibers is in the hoop direction and provides the cylinder with
sufficiently strong radial reinforcement such that under pressure the radius can be
considered to be constant and expansion is confined to a pure extension along the
axis.
The second family of fibers winds helically (right-handed [1094]) around the axis
and induces a rotation of the cylinder under extension. The presence of the hoop
fibers together with the relative thinness of the wall compared to the radius are used
to justify the membrane assumption, that is, we assume that there is no variation of
strain or stress in the radial direction and that the tube radius is constant. Also, since
the growth zone is much larger than the radius, we neglect gradient effects in the
axial direction.
We consider finite deformations in which the cylinder is allowed to grow, rotate
around its axis, and elongate axially while remaining cylindrical. Thus, using the
notation of the previous section, in cylindrical coordinates with the z-axis corre-
sponding to the vertical axis of growth, the deformation is
r = R, θ = + τ ζ Z , z = ζ Z , (16.138)
where τ is the twist, and ζ is the axial stretch.

The material response is specified by the standard fiber-reinforcing model [272]
with the simplest possible dependence for its isotropic and anisotropic parts
μ1 μ4
W (I1 , I4 ) = (I1 − 3) + (I4 − ν 2 )2 , (16.139)
2 4
where the Cauchy–Green deformation tensor invariants are given explicitly by
I1 = ζ 2 τ 2 + ζ 2 + 1/ζ 2 + 1, (16.140)

I4 = ζ sin ζ τ 2 + 1 sin + 2τ cos + cos2 . (16.141)
The angle denotes the orientation of the fibers with respect to the horizontal
plane in the reference configuration and ν is the pre-compression of the fibers in the
initial configuration. The anisotropic contribution to the hoop stress is proportional
to (I4 − ν 2 ). Since I4 = 1 in an unstressed configuration, the case ν > 1 results in a
negative, i.e., compressive stress. We recall from (11.323) that the fiber orientation,
φ, in the current configuration is

ζ sin
φ = arctan , (16.142)
cos + ζτ sin
where we have assumed a thin-walled cylinder of constant (unit) radius.

τ
τmax
0.1
Φc
π/2
0
0.5
ν= 1.0 1.5
Φ
ν= 1.1
-0.1 ν= 1.0
5
1.0
-0.2
τmin
-0.3
Figure 16.49 Effect of fiber pre-compression on the twist τ as a function of the fiber angle
(μ = 5, ζ = 1.4).
16.9.3 Analysis
We first consider a static model and assume that our tube is uniform with height
H and a fixed fiber angle with a right-handed helical configuration 0 < < π/2.
We then subject it to an internal pressure leading to a given extension ζ that we
use to parameterize the internal load. Note the difference with the example from the
previous section where the pressure in a single fiber material was controlled rather
than the extension. Here, the presence of hoop fibers guarantees that the axial stretch
ζ is monotonic with the internal pressure.
16.9.3.1 Rotation of a Pressurized Sporangiophore
First, we consider the rotation of the tubular structure with uniform fiber direction and
in the absence of remodeling and growth processes. From the strain-energy function,
we compute the Cauchy stress tensor T. Since no axial moment is applied on the
surface of the tube, the boundary condition is simply Tθz = 0 where
Tθz = μ1 ζ 2 τ + ζμ4 sin (ζτ sin + cos ) ×

ζ sin ζ τ 2 + 1 sin + 2τ cos + cos2 − ν 2 , (16.143)
which leads to a cubic equation for τ for given values of , ν, ζ and ratio μ = μ4 /μ1 .
The first positive root is the twist needed to relax the torsional stress in the tube
induced by the anisotropy. Some of these solutions are shown in Figure 16.49.
In the absence of pre-compressed fibers (ν = 1) we see that the twist τ is always
negative, i.e., the top of tube will rotate clockwise by an angle = τ h = τ ζ H
Figure 16.50 Fiber φ,τ

orientation, φ, in the current
configuration as a function of 1.5
, and the corresponding

plot of τ ()). The dashed
diagonal line corresponds to 1.0
φ = . φ(Φ)
0.5
counter-clockwise rotation
π/2
0.5 1.0 Φ
clockwise rotation τ(Φ)
leading to left-handed growth. The effect of pre-compression in the fibers is quite

remarkable and nonintuitive. Even for small levels of pre-compression (e.g., ν =
1.05), at small fiber angles, the cylinder will rotate counterclockwise as shown in
Figure 16.49. The critical fiber angle c at which there is no rotation is given by
⎛ ⎞
ζ2− ⎠ ν2
c = arccos ⎝ , (16.144)
ζ2 − 1
and clearly requires ζ > ν. The inversion occurs when fibers go from compression
to tension, since the critical angle corresponds to the point I4 = ν 2 . In the absence
of pre-compression, the cylinder rotates clockwise. We also note that despite the
torsional inversion, the fiber orientation in the current configuration is always greater
than the orientation in the reference configuration (see Figure 16.50).
From the expression for Tθz it is possible to obtain a lower bound, τmin , and an
upper bound, τmax , for the twist in the limit of stiff fibers μ 1, namely

ζ2 − ν2 ν √
τmin = − , τmax = (ν − 1) μ. (16.145)
νζ ζ
16.9.3.2 Growth and Remodeling of a Pressurized Sporangiophore
We have shown that a fiber-reinforced elastic tube with right-handed helical fibers can
rotate clockwise or counterclockwise depending on the initial angle, axial extension,
and initial stress in the fiber. We use this key result to develop a model of growing
Phycomyces in which new fibers are continuously laid down in an evolving growth
zone at the top of the sporangiophore.
Our key assumptions are:
z=0 z=0 z=0 z=0
growth zone
z=h
Figure 16.51 At the beginning of Stage IV, we assume that the growing zone extends linearly to
a constant size h (drawing on the right adapted from [996]).
• New fibers are laid down stress-free along the direction of fibers in the current
configuration;
• The growth zone extends linearly in time to a size h within the first two hours as
A crucial concept is that the current configuration at a given time step becomes the
reference configuration for the next time step. The overall growth process can now
be explained as follows.
Consider a material point in the cell wall. Initially fibers are laid down at some
small angle 0 with ν0 = 1 at the top of the growth zone. The extension (due to the
turgor pressure) and the rotation (due to the relaxation of the torsional stress) of the
cylinder result in the fibers re-orienting to an angle φ1 > 0 . As growth proceeds,
the material point moves downward with respect to the top of the growth zone and
new fibers at the same material point are laid down with angle 1 = φ1 and zero
stress. This material point now has a mixture of fibers with angles 0 and 1 . The
effective stress in the fibers in this new reference configuration (at this material point)
is thus a combination of fibers with pre-compressions ν0 and ν1 .
A continuous model of this process is given by
∂
= κ1 (φ − ), (16.146)
∂t
∂ν
= κ2 (ζ − ν) sin , (16.147)
∂t
Φ
ν 1.5
1.35 Φ
1.30
ν
1.25 1.0
1.20
1.15
0.5
0.05 τ 1.10
t
1.05
0.2 0.4 0.6 0.8 1.0
0.2 0.4 0.6 0.8

t (hr)
0.05
-0.10
Figure 16.52 Dynamics of τ , ν, and for a given material point as a function of time (κ1 =
16 hr−1 , κ2 = 4 hr−1 ). While this process takes place, the material point at t = 0 starts on top of
the growth zone and is convected downward with respect to a reference frame moving with the top
of the growth zone.
where , φ, ν, ζ are all evaluated at Z , the distance from the top of the sporangiophore
of a material point in the reference configuration.
Following the property of the angle φ = φ() given above, we have
lim (Z , t) = π/2, and lim ν(t) = ζ ∀Z . (16.148)

t→∞ t→∞
The time constant 1/κ1 is fitted so that when a material point leaves the growth zone
the associated fiber is oriented along the axis, i.e., (1) ≈ π/2 and similarly 1/κ2 is
fitted to match the time for rotational inversion. As an example of the dynamics gen-
erated by this system (coupled to the equations for φ and τ ), we show in Figure 16.52
the evolution of τ , , and ν as a function of time.
Each time the tube extends and twists as an elastic response to the extensional
and torsional stresses, there is an evolution of the reference configuration due to the
deposition of new fibers and other wall building materials. The net result is a contin-
uous remodeling of the cell wall leading to an irreversible growth of the tube. The
rotation of a small disk of height Z at point Z measured from the top of the growth
ω
stage IVb
0.5 1.0 z=0 -1.0

t (hr)
growth zone
-0.5
z=h
-1.0
stage IVa stage IVc
Figure 16.53 Rotational velocity (in units of radians per unit time, tr ) of the sporangiophore as a
function of time. The time -1 on the right-side of the time axis refers to one hour before the end of
growth, i.e., the time in this simulation at which the growth zone starts to retract.
zone with fiber angle and pre-compression ν (in the reference configuration) is
θ = τ ((Z ), ν(Z ))z = τ ζZ . Thus the total rotational velocity, ω, can be
estimated as H
ω = tr−1 τ ((Z ), ν(Z ))ζ dZ , (16.149)
0
where tr is a characteristic time scale corresponding to the remodeling time of the

reference configuration.
We can now use our model for the evolution of the fiber angle and the fiber pre-
compression to compute the rotational dynamics of the growing cylinder. We assume
that the growth zone extends linearly in time with a constant velocity vz up to a length
h = H ζ (corresponding to the end of Stage IVa). Subsequently, the length of the
growth zone remains stationary (Stage IVb) and the tube extends, while rotating at
a constant rate, until the growth zone retracts resulting in another inversion (Stage
IVc) as shown in Figure 16.53.
Initially before the growth zone is fully established, only the fibers in the top part of
the zone play a role and the net rotation is counterclockwise leading to right-handed
growth (Stage IVa).
Once the growth zone is fully established, the observed rotation of the sporangium
is determined by the integral over the entire zone, hence leading to a left-handed
spiral (Stage IVb). Interestingly, the model predicts that at any given time there are
individual material points in the growth zone turning both counterclockwise (top
of the growth zone) and clockwise (bottom). The integrated effect of these rotations
depends on the extent of the growth zone. This mechanism is entirely consistent with
the detailed observations of Cohen and Delbrück [218] on the rotation of individual
points in the growth zone of which they write: “The twist reaches negative values
during the First few tenths of a millimeter. It then turns positive. The curious negative
values of the twist in the uppermost portion of the growth zone are undoubtedly real”.
Estimates of the rotational velocity depend on the choice of the time scale tr . From
Figure 5 we see that in Stage IVb |ω| ∼ 1.5, and for the range of tr , 1/κ1 ≤ tr ≤ 1/κ2 ,
one finds the corresponding range of 1 to 4 turns/hr. This result is qualitatively
consistent with the observed 2 turns/hr.
16.9.4 Discussion
The model discussed here relies on two simple fundamental assumptions: First, new
fibers are laid down along the direction of existing fibers; second, new fibers are
laid down in a stress-free state. The rotational inversion is then a direct consequence
of the response of a tubular, fiber-reinforced, morphoelastic material. The beauty of
the inversion phenomenon is that it relies on the nonlinear anisotropic response of
the system and leads to two apparent nonintuitive behaviors: First, the fibers tend to
align with the axis, independently of the rotation; second, right-handed helical fibers
can lead to both clockwise and counterclockwise rotation depending on the angle
and level of pre-compression.
At a smaller scale, a fundamental problem of plant physiology is to understand the
regulation of expansive growth in organisms with walled cells (plant, algal, and fungal
cells). The sporangiophore of Phycomyces blakesleeanus has been used as a model
organism for investigating the relationship between the rate of growth regulation and
the mechanical properties of the cell walls [995, 997].
Part V
Conclusion: Where It Does Not End
Last scene of all,

That ends this strange eventful history,
Is second childishness and mere oblivion,
Sans teeth, sans eyes, sans taste, sans everything.
Chapter 17
Ten Challenges
The theory of growth presented so far is still in infancy. A number of key challenges
need to be addressed thoroughly before it stands as a complete mechanical theory
that can be readily accepted and used universally. Below, I outline briefly some of
these challenges in the hope that they will motivate further studies.
17.1 The Rheology of Growth
We discussed in Chapter 13.1, the problem of varying time scales appearing in the
growth and remodeling of materials. In standard rheology, it is appreciated that
depending on the time scale at which a system is being observed, it may be described
by different theories. The classic example is the evolution of a mountain: A mountain
is a solid on human time scales but would appear to flow on geological time scales. In
1928, this example was used by Markus Reiner and Eugene Bingham to introduce the
notion of a Deborah number, a term inspired by a biblical passage. Reiner famously
said: “In her famous song after the victory over the Philistines, [the prophetess
Deborah] sang The mountains flowed before the lord” [268, 1072]. The Deborah
number De is defined as

556 17 Ten Challenges
Time of stress relaxation

De = . (17.1)
Time of observation
It combines a material parameter and an observation parameter. At low value of
the Deborah number, corresponding to high stress relaxation or long observation
time, the material behaves mostly as a Newtonian viscous fluid. For high Deborah
numbers, corresponding to a very long time of relaxation or a very short observation
time, the material is dominated by a solid-like behavior.
Closely associated with the Deborah number is the Weissenberg number W i
[1309, 1322] describing the relative contribution of viscous forces with respect to
elastic forces
Viscous forces
Wi = . (17.2)
Elastic forces
Depending on the relative values of these two numbers, different regimes can be
obtained, as classified by Pipkin [1039]. Materials with small Weissenberg and Deb-
orah numbers can be studied within the framework of linear viscoelasticity. Very
large Deborah numbers correspond to either linear elastic-like materials (small W i)
or nonlinear elastic materials (larger W i) as the Weissenberg number in that regime
gives a relative measure of the strain amplitude. Intermediate values of W i and De
correspond to nonlinear viscoelastic materials, which are still poorly understood
[651].
The problem with growing biological systems is that, in addition to the mate-
rial and observation time scales used to characterize the rheology of polymers,
it involves new time scales associated with mass addition, remodeling, and other
active processes. It also involves new length scales associated with cellular processes.
Therefore, it cannot be simply modeled as an inert viscoelastic material. New dimen-
sionless numbers must be introduced to keep track of these new time and length scales
and to classify possible material behaviors associated with development, growth,
remodeling, and homeostasis.
Nevertheless the study of complex fluids gives us the understanding that the type
of theories to be used to model a particular phenomenon depends crucially on the
period of observation. For instance, in the study of axon growth in Section 4.4.4,
we observed that the elastic properties of axons can be obtained on very short time
scales by plucking the axon and observing its deformation. On longer time scales,
the growth of the axon can be described either as a viscous fluid [1002] or by the
evolution of the reference configuration [1067].
More generally, the evolution of the reference configuration can be used to cap-
ture many processes such as plasticity and viscosity. Therefore, the central idea of
an evolving reference configuration offers a unifying framework to understand the
rheology of growing biological systems. Within this framework, the description of
particular asymptotic regimes based on the type of biological, materials, and obser-
vational time scales should provide systematic methods to obtain simple models for
specific systems and guide new experiments.
17.2 The Regulation of Growth and Growth Size 557
• Challenge #1: Within a general theory of evolving reference configurations for

anelastic systems, classify different rheological regimes corresponding to different
realizations of growth in biological systems.
17.2 The Regulation of Growth and Growth Size
During development and regular life time in many plants and animals, the body and
most of the organs are actively regulated to maintain size and shape. This regulation
mechanism occurs through growth and appropriate shrinking. One of the fundamental
questions of both development biology and of physiology is to uncover how the size
of an organism or an organ is determined and controlled [1289]: Why do arms have
the same size? Why do elephants grow larger than mice? How do eye balls fit snugly
in the eye socket and match the focal length of their optics? How do clownfish change
their body size according to their social status? How does the size of a snake heart
change by a factor two to three before and after a meal is digested?
Figure 17.1 In two

salamander species of
different sizes, the
amputation and reattachment
of a limb (arrows) early in
development lead, in mature
animals, to the original limb
size (adapted from [1264]).
Ambystoma
punctatum
Ambystoma
tigrinum
The size of an organism or an organ depends both on the number and size of cells
and on fluid and extracellular material. As a first approximation, it is the number
of cells that decide the overall size of an organism. For instance, a 70 kg human
contains about 1013 cells whereas a 25 g mouse (Mus musculus) has about 3 × 109
cells. The weight ratio 70,000/25 = 2,800 is roughly equivalent to the cell number
ratio of 3,333 [223]. It is therefore natural to consider cell proliferation and death as
the main driver for size control. However, a closer examination and the experimental
manipulation of either cell proliferation or cell size in various organs have revealed
that some organs manage to retain their size by modulating cell size. These organs
may have fewer but larger cells, or more cells of smaller size [265, 1308].
A central dogma of developmental biology is that apart from an overall effect of
extrinsic factors on the entire organism, such as nutrition, organ size, proportion, and
patterning are mostly controlled intrinsically. This dogma is illustrated by the cele-
brated experiments of Twitty and Schwind [1264] who exchanged limbs and eyes in
species of salamanders of different sizes. As shown in Figure 17.1, the transplanted
organ reached the size dictated by its own species, suggesting that growth follows
internal positional and size information. The proposed theoretical mechanism for
such an internal programming is that growth is controlled by morphogens. Mor-
phogens are molecules produced by cells at discrete sources of an organ that create
extracellular concentration gradients over some distance from the source. Therefore,
morphogens provide cues for the position of different cell types, limiting cell prolif-
eration, or initiating the production of secondary morphogens [1338]. These elusive
morphogens were eventually identified in the wing imaginal disk of Drosophilia (fruit
fly), a model organ of a model organism of developmental biology. Proteins encoded
by the decapentaplegic (Dpp) and wingless (Wg) genes were shown to pattern the
fly’s disk wing [776, 1364] and control cell proliferation as illustrated in Figure 17.2.
The wing blade develops over 4 days starting from approximately 50 cells to about
50,000 in 10 cycles [1017]. Remarkably, all cells divide homogeneously in space
until proliferation stops.
A. B. C.
Wild type Dpp expression Dpp inhibition
Figure 17.2 The role of morphogens in the regulation of size is revealed in experiments on
Drosophilia’s disk wing. Compared to the wild type (A), an over-expression of Dpp (shown by
the arrow in B) induces proliferation in surrounding cells and leads to an enlarged wing. C. Con-
versely, inhibiting Dpp creates a smaller wing (adapted from [1017, 1363]).
17.2 The Regulation of Growth and Growth Size 559
Despite the well-accepted fact that morphogens play a crucial part in organizing
growth, patterning, and controlling size, the mechanisms by which cells interpret
these gradients and possibly other spatial and physical clues remain elusive despite
growing evidence of the role of mechanotransduction [355, 361]. In particular, wing
size cannot be simply explained by the length scale associated with Dpp gradient
[10]. A possible mechanism for size control is that there exist extra morphogens
produced at the differentiation stage that are diluted as mass increases [1299]. How-
ever, these chemicals have not yet been identified. Another possible mechanism is
that physical forces acting on the cells are used to carry global information [11,
632, 952]. These models assume that proliferation depends on the concentration of
morphogens. Stronger gradients in the central region of the disc create differential
growth and, accordingly, regions of tensile hoop stress in the periphery. At the center
of the disc, compressive forces increase with disk size up to a critical value, where
growth stops. More sophisticated models [27] rely on anisotropic growth response
and active actin contraction to explain the scaling of the wing disc.
Another striking example of global organization is found in plants. Wild type
leaves of snapdragon (Antirrhinum) are naturally flat [937]. The cin mutant of Antir-
rhinum has excess growth in the marginal regions and has a crinkly and folded shape.
Flatness is not only a local notion associated with zero curvature, it is a global prop-
erty that is carried through the entire organ through local regulation. If some parts of
the leaf grow much faster, it will develop stresses that will naturally lead to buckling.
The fine balance between stresses, geometry, genetics, and diffusion seems to be key
in many other plant systems to control size and shape [217, 502, 1097].
• Challenge #2: Determine the respective roles of mechanics and genetics in deter-
mining the size of animals and organs. Understand how morphogen gradients, chem-
ical clues and pathways, physical inputs, and mechanical transduction combine to
obtain organism size information and conspire to control global size through local
growth.
17.3 The Elusive Growth Law
In the theory of continuum mechanics, the fundamental equations are obtained from a
consideration of kinematics for the type of deformations allowed, physical arguments
for the balance of forces acting in a material; and constitutive laws describing the
material properties of a material. For the case of elastic materials, the proper choice
of a constitutive law, based either on fundamental statistical and thermodynamical
principles [40] or purely phenomenological considerations, is still an active field of
research. Nevertheless, there is a large body of the literature describing these laws and
how they should be used when modeling a given system [119, 597, 978]. There is an
excellent agreement between the three approaches necessary to put a physical theory
on firm grounds: A theoretical physical description of the material at the micro-
and macro-scales, detailed experiments measuring the response of these systems
under loads, and reliable computer simulations. Each links theory to experiments
and allows for the exploration of more complex systems. This physical theory of
elastic materials is further established by various mathematical treatments providing
a rigorous foundation for the theory as well as insight into its mathematical structure.
The situation is not as pristine for the theory of growing solids. Indeed, the kine-
matics of growth for continuum solids introduces a new object, the growth tensor
G. Therefore, an appropriate constitutive equation, namely the growth law, must be
postulated for this new object. Following the discussion of Chapter 14, this consti-
tutive law takes the form of an evolution law describing the rate of change of the
growth tensor as a function of possible growth stimuli, be it chemical, genetics, or
physical. However, the precise form that these laws should take is still not correctly
understood and it is perhaps one of the most challenging problems in establishing a
mechanical theory of growth.
For the present discussion, we restrict our attention here to the coupling of growth
to mechanics by neglecting all other stimuli. These effects will be discussed as another
challenge. We know that the overall change of mass is given by (det G). Yet, even
if we assume that growth is isotropic, that is G = g1, the functional form of g is
still an unsolved issue. Worst, the choice of dependent variables, of stresses, strains,
tractions, or forces is not agreed upon. Whereas some of these formulations may
be mathematically equivalent, it is not clear whether growth processes in biological
tissues are primarily governed by stress or strain [634].
In the theory of elastic solids, the use of the second law of thermodynamics and
the absence of entropy sinks at the microscale can be used to find constraints on
possible constitutive laws. It is therefore natural to expect that it could be similarly
useful in giving us insight into the growth law. However, we have seen that growth
fulfills many different roles in living systems. In normal conditions, it is used in
the early development of an organism to reach an adult size as well as the regular
maintenance of an organism by homeostasis. It is also found in pathology, through
abnormal growth response and tumor growth. It is therefore unreasonable to expect
that all these different processes be described by a universal law for growth based
on simple physical principles. Indeed, we do not expect that pressure applied to
a growing embryo, a healthy artery, or a growing tumor would result in the same
growth process or even that it would result in a systematic mass addition or mass
resorption.
The different biological processes related to growth are governed by biological
pathways and typically require external sources of energy. The existence of these
pathways indicate that there is information at the cell level that is processed through
cellular mechanisms so that the cells are able to react to external stimuli. This cellular
information and energy input create at the tissue level possible sinks of entropy. In
the presence of these sinks, the thermodynamic argument breaks down since no strict
inequality can be established among the various fields appearing in the theory. Yet,
thermodynamics allows us to distinguish between purely physical passive processes
and biological active processes. It also helps us to identify particular combinations of
mechanical quantities, such as Mandel’s and Eshelby’s stress tensors, that are likely
to play a fundamental role in a growth law.
17.3 The Elusive Growth Law 561
In the absence of a well-established theory for a growth law, a phenomenological

approach can be used to formulate evolution laws based on both experiments and
general observations. For instance, the notion of homeostasis can be used to postulate
the existence of a special homeostatic stress that a given system maintains through
mass addition and removal. In such cases, it is assumed that the rate of change
of the growth tensor vanishes identically when the system reaches the homeostatic
stress and is such that the system tends to this stress for small enough perturbations.
However, even in this simple case, the specific form that the growth law must take to
ensure the stability of the homeostatic state is not well understood as observed in the
simple case of homogeneous deformations discussed in Section 14.6. Yet, the theory
of morphoelasticity constitutes a powerful framework in which new experiments can
be systematically formulated.
Another approach consists in using the theory of mixtures for which the exchange
between different components is easier to model based on fundamental principles
and out-of-equilibrium thermodynamics. The response of the different active con-
stituents to different mechanical stimuli would require specific modeling based on
their cellular properties. If the promises of these theories are fulfilled, it would then
be possible to coarse-grain these models to obtain the homogenized response at the
tissue level. We consider the challenges at the cellular level in the next Section.
• Challenge #3: Use the theory of morphoelasticity as a general constitutive frame-

work to identify possible functional forms of growth laws and to design new exper-
imental protocols. Test these protocols theoretically and computationally in order
to classify families of growth laws. Following the established protocols, run experi-
ments on different classes of biological systems to obtain specific constitutive laws
within these families.
17.4 Multiscale: From Discrete to Continuous and Back
Closely related to the previous challenge is the problem of connecting our accu-
mulated knowledge of growth at the tissue level to our understanding of cellular
mechanics. In particular, we know that mechanical signals at the tissue level are
transmitted into a cascade of cellular and molecular events [629, 664], as illustrated
in Figure 17.3. These mechanotransduction processes have been abundantly docu-
mented in various biophysical experiments where different cells lines were loaded
statically or cyclically while their proliferation or increase in various components
are recorded [1294].
These experiments demonstrate the great complexity and diversity of cellular
responses to mechanical stimuli and, in particular, how the extracellular matrix and
cell adhesion play an equally important role in dictating overall tissue properties.
So far, most of these studies are disconnected from similar considerations at the
tissue or organ levels and one of the key challenges of biophysics is to connect this
microscopic response to macroscopic scales [1198].
Figure 17.3 Illustration of

various known cellular
mechanotransduction
mechanisms (from [1294]).
The starting point to connect multiple scales is to reconcile the discrete cell-level
models described in Section 3 with the continuum theory of morphoelasticity. This
connection can be established by the methods of homogenization. Homogenization
denotes the set of mathematical techniques for the study of physical media with a
well-defined microstructure with typical length scales much smaller than the system
size [1115]. The basic idea behind homogenization is to average the properties of the
microstructure to obtain the effective properties of the body, seen as an homogeneous
system, at larger scales [86, 438, 884]. Homogenization is easier to achieve when the
heterogeneous material has a periodic, or nearly periodic, microstructure. Then, the
process amounts to approximate asymptotically, as ε → 0, the solutions of partial
differential equations with coefficients oscillating with a period ε [19].
In biomechanics, since many materials often have a regular structure [439] these
methods have been successful in modeling the response of bones [594], the effective
properties of soft-tissues [160, 359], the design of scaffold in tissue engineering
[593, 1149], and fluid transport in tumors [1148].
If a tissue does not have the required periodicity, the traditional methods of homog-
enization cannot be directly applied but bounds on the effective properties can be
obtained by comparing the material with another material that has well-defined struc-
ture, and that shares some of the statistical properties of the original system [801,
802, 892, 1042].
So far, applications of homogenization in biomechanics have mostly focused
on either obtaining effective strain-energy density functions for elastic materials or
averaged equations for fluid transport in poroelastic materials. Assuming that we have
suitable models to describe the mechanical response of a tissue under loads at any
17.4 Multiscale: From Discrete to Continuous and Back 563
Mechano- Cell Growth

Microscale Cell Deformations
ECM Growth
Cell level Local Forces transduction Fiber production
Localization Homogenization
Macroscale Stress and Strain Morpho- Growth Tensor

Tissue level F, T elasticity G
Figure 17.4 The required steps to include microscopic properties in a continuum model for growth
given time, we can try to use homogenization on longer time scales in order to obtain
a growth law based on a simple rule of evolution for a microscopic lattice [199]. Ideas
and methods from homogenization in plasticity theory may be particularly relevant
for this task. However, in plasticity problems there is a well-defined lattice structure
and the anelastic part of the deformation can be directly linked to slip lines and other
microscopic defects [1070, 1071]. These approaches seem particularly promising
in establishing a rigorous framework linking microstructure to tissue evolution. To
obtain the feedback between mechanics and growth, the deformation at the cell
level must be known. This is the problem of localization. If stresses and strains are
known at the macroscale, the forces or deformations experienced by the cell level
are not simply given by the pointwise value of the corresponding fields. Localization
provides a method to obtain microscale information from both the macroscale and the
microscopic structure. The different steps in a true multiscale approach are sketched
in Figure 17.4. We emphasize that this cartoon is a highly idealized picture of a
multiscale process and that much thought must be given in defining the different
concepts.
A real theory of growth will only emerge when there is a clear bridge between
microscopic response and macroscopic behavior. The development of the theory of
elastomers by Flory, Rivlin, and others in the 1940s relied on simple observations
of the microscopic structures of polymers that resulted in the derivation of the neo-
Hookean strain-energy density function using proper statistical averaging. This, in
turn, served as a basis for further refinement and development. It is hoped that a
similar development will take place in the mechanical theory of biological growth.
• Challenge #4: Starting from a simple representation of tissues, with given elastic
properties and rules for growth based on microscopic biophysical laws, systemati-
cally obtain average laws for growth in terms of macroscopic quantities such as the
growth and stress tensors. Localize these effective models to check independently
the local behavior of cells when the tissue is loaded.
17.5 Growth Versus Diffusion
Our discussion of a possible growth law has been restricted so far to mechanical stim-
uli. Clearly, the growth of most biological systems is not just driven by mechanics but
is typically the result of coupling with other physical or chemical fields. In particular,
the traditional view of morphogenesis is that patterns appearing during development
are driven by spatiotemporal variations of various chemical concentrations compet-
ing with each other and creating Turing-type instabilities [819, 927]. The equations
modeling these phenomena typically take the form of nonlinear reaction–diffusion
equations, which in the simplest cases read
∂ci
= Di ci + f i (c1 , . . . , cn ), i = 1, . . . , n, x ∈ ⊂ R N , (17.3)
∂t
where (c1 , . . . , cn ) is a vector of concentrations is the N -dimensional Laplacian

and (D1 , . . . , Dn ) are the diffusivities, assuming no cross-diffusion. The functions
( f 1 , . . . , f n ) describes the reaction kinetics between the different species.
In his celebrated article of 1952, Turing showed that by choosing n = 2 and
( f 1 , f 2 ) linear in (c1 , c2 ), these equations are sufficient to generate spontaneously
spatial patterns from uniform solutions [1259]. He further suggested that pattern
formation in morphogenesis could be modeled by the evolution of morphogens,
dictating the time and place at which cells would grow and remodel. These ideas are
the starting point of a large body of work on biological pattern formation [440, 820,
860].
Even in the absence of any mechanical feedback, there is an interesting coupling
between growth and Turing-type patterns [52, 252, 253]. If we assume that the
domain on which a pattern is established t grows slowly, it will affect the pattern
as the growth of the domain induces a flow in the current configuration so that, we
have now
∂ci
+ div(uci ) = Di ci + f i (c1 , . . . , cn ), i = 1, . . . , n, x ∈ t ⊂ R N . (17.4)
∂t
The term div(uci ) can be decomposed into an advection term u · grad(ci ) due to
the motion of material points during growth and a dilution term ci div(u) due to local
volume change. The flow u is determined by the deformation of the domain due to
growth. Since, we assume that we have no elastic response, we have simply
x = χ(X, t) X ∈ 0 , (17.5)
where 0 is the initial domain. This growth function is related to the growth tensor
G through its gradient: G = Gradχ. Then, the local flow is determined by
∂x ∂χ
u(x, t) = = . (17.6)
∂t ∂t
17.5 Growth Versus Diffusion 565
Activator concentration
space
time
Figure 17.5 Space–time evolution of activator concentration profile on an exponentially growing

domain. Light and dark shading represent high and low concentrations, respectively (adapted from
[252]).
The slow evolution of the domain creates an interesting variation of patterns. For
instance, in one dimension (N = 1) with two species (n = 2), a simple activator–
inhibitor system shows frequency-doubling of patterns as seen in Figures 17.5 and
17.6
The coupling between chemical fields and mechanical fields has long been sus-
pected to play an important role in biological pattern formation [671, 928, 929,
1244], but the relative importance of each effect and of their coupling remain poorly
understood. An interesting case in point is the problem of phyllotaxis, the arrange-
ment of leaves on a plant stem. These patterns also appear in the arrangement of seeds
on sunflower heads, bracts on pine cones, and spines on cacti, shown in Figure 17.7.
These initial elements, appearing originally as small bumps on leaves or homologues,
are collectively referred to as primordia. Understanding and modeling phyllotaxis
has been an active field of research for centuries, going back to Leonardo da Vinci and
Kepler (see [1026]). The geometry of such patterns can be modeled from a purely
geometric point of view as the sequential optimal packing of new primordia in a
given geometry [43] .
Phyllotaxis can also be explained as a developmental process, related to purely bio-
chemical effects, by following the transport and localization of the growth hormone
auxin. In this scenario, phyllotaxis emerges from the chemical interaction between
existing and incipient primordia in a growing structure, mediated by actively trans-
ported auxin. High auxin concentration triggers the formation of new primordia
Active pattern formation region
Wildtype
Mouse limb Wildtype

bud development
1D Turing model
on a growing domain
Doublefoot
Figure 17.6 Digit formation in the development of a mouse limb (Mus musculus). Top: tradi-
tional self-organization model where pattern formation only occurs at the distalmost part of the
limb bud. Bottom: 1 D-turing model in a growing domain reproducing both the wild type and the
supernumerary digit mutant when growth speed is increased (adapted from [893]).
which are separated from each other by the local depletion of auxin around them as
verified in computer simulations [674, 1178] (Figure 17.8A).
Yet another approach is to explain the apparition of new primordia as a mechanical
buckling instability due to differential growth [312, 315, 507, 1187]. Modeling the
soft outer layer where cell division and growth occurs as an overdamped, curved,
Figure 17.7 Phyllotaxis is the particular spiral arrangement of markers found in plants. A. Seeds
on sunflower heads. B. Bracts on pine cones with 8 counterclockwise and 13 clockwise spirals. C.
Spines on cacti [1026]. Pictures courtesy of Patrick Shipman.
17.5 Growth Versus Diffusion 567
Figure 17.8 Simulation of phyllotaxis for a biochemical model (A. courtesy of Richard Smith),
a mechanical model (B. from [1152] courtesy of Patrick Shipman) and a coupled mechani-
cal/biochemical model (C. courtesy of Matt Pennybacker [1026]).
elastic shell with circumferential and radial stresses leads to buckling patterns with
many phyllotactic features, such as the whorls and Fibonacci progressions [947,
1151, 1152], as shown in Figure 17.8B. At the cellular level, experimental evidence
and simulation further showed that mechanical stress plays an important role in the
organization of the microtubule cytoskeleton, which in turn leads to morphogenetic
patterns by setting anisotropic direction for cell division [528].
The biochemical and mechanical approaches can be combined by assuming that
there is a coupling between the local concentrations of key molecules, the local stress
field, and the growth field [948]. These different feedback mechanisms provide a
complete picture of the different patterns and allow to recover almost all features of
phyllotaxis [1026], as illustrated in Figure 17.8C.
Despite the fact that both chemical diffusion and mechanical stresses play an
important role in growth and remodeling [145], their coupling is usually weak in the
sense that the coefficients appearing in the equations governing either process only
depend weakly on the coupling. Therefore, it is mostly through changes in geometry
that the two processes are coupled.
• Challenge # 5: Determine the respective role of diffusion and mechanics in estab-

lishing biological patterns. In particular, identify systems dominated by one or the
other effect and systems where the coupling between both is necessary to obtain
generic and robust patterns.
17.6 Multiphysics: Coupling Growth with Other Fields
Biochemical diffusion and mechanics are not the only stimuli affecting growth. An
example of a physical stimulus is found in the growth of the eye and the process of
emmetropization which describes how, in health, the human eye grows during child-
hood to adopt the right shape and size for focussed vision. The goal of emmetropiza-
Figure 17.9 An illustration

of differing focal lengths for
blue light. In this example,
local growth of the sclera is Image blured
possible only where the growth
retina is in front of the best
focus point for blue light.
Top: There is growth in the
peripheral sclera and at the
posterior pole. Bottom:
growth ceases when the
retina is either at the best
focus point or behind it Image in focus
(adapted from [695]). no growth
no growth
tion is to position the retina at the correct axial distance behind the lens for the optical
power of the anterior eye. When this process fails, the individual is either myopic,
with excessive axial length, or hyperopic, with insufficient axial length. Animal stud-
ies have revealed that emmetropization is sensitive to the wavelength of light [858]
and that the application of negative powered lenses induces myopia [1289]. These
studies suggest that apart from possible genetic and environmental factors, the signal
controlling emmetropization is locally interpreted at the retina [1312] and is pro-
duced through the optic of the eye lens. A possible hypothesis, shown in Figure 17.9,
is that the retina can separately detect blur in red and blue light and that it uses the
differing amount of blur at these wavelengths to determine not only how blurred an
image is but also whether that blur is due to hyperopia or myopia. This signal is used
to drive growth according to the assumption that the posterior sclera is capable of
local growth to relieve the stress due to intraocular pressure whenever that region of
the sclera is in front of the position of best focus for blue light. As the eye elongates
to reach a focussed vision, it changes the overall geometry of the eye, hence the focus
of the lens. Hence, there is a two-way coupling between optics and growth.
Another example of physical coupling is found in the swelling in the brain follow-
ing trauma. Swelling is the volume increase due to fluid intake from the capillaries.
As a first step, it can therefore be modeled as a growth process where swelling is
directly induced by a change of mass. This approach can be used to understand the
stress created during swelling and the release of stresses following decompressive
craniectomy, a surgical procedure used to reduce stresses in swollen brains [1305,
1306]. An example of such computations, based on the theory of morphoelasticity,
is shown in Figure 17.10.
17.6 Multiphysics: Coupling Growth with Other Fields 569
Figure 17.10 Finite-element simulation for 10% brain swelling showing regions of high shear
stress close to the opening. Insert: Transverse section showing radial fiber stretch and possible
axonal damage (adapted from [487]).
However, at the tissue level, the water intake is tightly controlled by difference
in turgor pressure due to electrochemical imbalance [762]. In this case, the change
of volume can only be properly understood by coupling the solid deformations with
both fluid and ions transport. Again, the coupling acts both ways since the increase
in volume reduces the concentration of fixed charge densities. This coupling can be
achieved close to equilibrium using the methods of irreversible thermodynamics that
provide a systematic way to derive generic transport equations for coupled fields.
The theory of coupled fields is a well-developed theory of continuum mechan-
ics even though some fundamental physical and mathematical issues remain [340].
It combines aspects of solid mechanics, mixture theory, and the thermodynamics of
irreversible processes. However, this theory has not yet been adapted to the specificity
of biological systems and needs to be reformulated to include evolving microstruc-
tures and mass addition. The theory needs to be both generalized to include such
key aspects of biological systems but it also needs to be adapted and simplified, in a
pragmatic way, to address the reality of experimental biology which often does not
give the luxury of experimental control or access to microscopic parameters.
• Challenge #6: Understand the intrinsic coupling between different physical the-
ories in tissues, organs, and organisms with evolving microstructures. Develop a
general theory of out-of-equilibrium nonlinear thermodynamics coupling multiple
fields for growing biological systems. Couple this theory to gene regulatory networks
and biochemical pathways.
17.7 A Theory of Accretive Growth
Most of the work on growth focuses on volumetric, surface, and tip growth. In these
cases, growth is expressed as a change of the volume element attached to a material
point. In the case of surface growth or tip growth, only points close to the boundary
experience growth, but no new material points are added to the system. In accretive
growth, new layers of materials are added at the boundary through a process of
deposition, as found in the problem of crystal growth [154, 1319] or protoplanetary
disks fusion [105, 1318]. Examples of biological systems evolving by accretive
growth include seashells [202, 1265], horns [1173], coral [676, 875], and, to some
extend, bone growth.
Figure 17.11 Cranial skull

in the newborn. The calvaria
is composed of the bones
that can be seen from the top.
Large sutures and fontanelles
are present in the newborn
17.7 A Theory of Accretive Growth 571
An important example of accretive growth is found in the early growth and devel-
opment of the skull and in particular, the bones of the cranial vault composed of the
superior portions of the frontal bone, the occipital bone, and the parietal bones (see
Figure 17.11). At birth, these different bones are joined by soft fibrous cartilaginous
tissues referred to as sutures and fontanelles. Bones of the cranial vault form pri-
marily by intramembranous ossification, a mode of ossification where bone forms
directly from undifferentiated mesenchymal cells without first forming a cartilagi-
nous scaffold.
There are two principal modes of growth and remodeling to increase intracranial
volume and shape the skull after birth, both related to accretive growth: sutural growth
and plate growth. Sutural growth takes place at the suture and results in the deposition
of new bone at the margin of the bone plates. Sutural growth is the primary mode of
intracranial volumetric expansion for the skull as approximately 90% of intracranial
volume growth occurs within the first two years after birth in humans. The leading
theory is that the differentiation of osteoblasts and the ossification at the sutural
edge is regulated by signaling interactions between the mesenchyme, the osteogenic
fronts, and the dura mater. The main signaling is believed to be of mechanical origin:
As the brain grows, it generates tensile strain across the sutures that influences bone
deposition [333, 909] at the suture edge and inhibits ossification within the suture
itself [989] (Figure 17.12).
Plate growth takes place on both surfaces of the bone plates and results in remod-
eling and thickness increase. Typically, it happens by mass removal at the inner
surface and mass deposition at both the inner and outer surface. Mechanically, the
pressure exerted by the growing brain stimulates bone resorption by osteoclasts. On
the outer surface, a layer of osteoblasts continuously produces new bone growth and
is believed to be controlled by the tension in the periosteum [567].
While there seems to be broad agreement in the literature about the basic physical
mechanisms at play during growth, the detailed signaling processes that lead to this
overall behavior is extremely intricate as it relies on different cell lines as well as
on many different biochemical pathways [1081]. From a mechanical point of view,
A. suture B.
t
+ ++ + + + +
periostu
m
+ + + + +
+ - --- - - - -
- - -
+ +
bone + + h
er
dura mat
P
brain P
sutural growth plate growth
Figure 17.12 Cranial growth and remodeling occur primarily through A. sutural growth in the
suture and is remodeled by B. resorption and addition of material on the upper and lower surfaces
of the bone plates (middle sketch adapted from [909]).
explicit forms for the two growth processes have yet to be proposed. For instance,
assuming that edge growth depends on the mechanical forces applied at the suture
level, does it depend on force, stress, tension, or strains? These different mechani-
cal quantities are related but since the thickness of the calvarial bones change (by
a factor of about 3) and the stiffness of the suture changes (by a factor of about
50), the different measures of force change drastically during the growth process.
Furthermore, what reasonable law for this process could explain some of the basic
observations, both in healthy and pathological skull growth? In particular, detailed
simultaneous studies of skulls and brains in mice and humans indicate that the brain
and skull fit tightly together even in the case of premature closure or different geno-
types [951, 1079, 1080]. How does this synchrony take place? And how does it fail,
in craniosynostosis, when sutures fuse early [1326]?
This important example underlines some of the difficulties related to accretive
growth. From a theoretical point of view the main problem stems from the fact that
the body changes by adding new material points. This change of boundary can eas-
ily be described by an accretion field that describes the motion of the boundary in
time [914, 1173]. This approach was used in the modeling of seashells described in
Chapter 7. However, this description is purely geometric and the material properties
and mechanical state of these new points must also described. Therefore, unlike the
problem of volumetric growth where a reference material manifold is postulated such
that all mechanical quantities can be defined, in accretive growth, the boundaries of
reference manifold is also changing [579, 630]. This evolution is quite different to
the local expansion of a volume element and is particularly difficult to model from a
continuum perspective even for a rigid body [987]. For a deformable solid, tracking
the evolution of the boundary becomes a challenge. It can easily be performed as
a discrete iterative process but its continuous version remains difficult even in the
simplest cases [338, 1358].
• Challenge #7: Unify the theory of volumetric and accretive growth within a single
framework. Establish a constitutive theory for accreted materials. Develop compu-
tational and mathematical tools to properly handle boundary accretion. Apply these
ideas to explain the morphological patterns found in seashells and the pathological
shape of skulls observed in craniosynostosis.
17.8 Dynamics and Post-bifurcation Behavior
Volumetric growth creates residual stress. These stresses can be used to regulate a
physiological system, to maintain homeostasis, or to develop instabilities that will
evolve into morphological patterns. The theory presented so far can be used to detect
these instabilities and to understand how a system remains close to a dynamic equilib-
rium. After an instability sets in, the system enters a regime that is mostly controlled
by nonlinearities. The question is then to understand how the system behaves post-
bifurcation.
17.8 Dynamics and Post-bifurcation Behavior 573
C.
A. B.
D. E. F.
Figure 17.13 Post-bifurcation patterns for the growth of a thin layer on an elastic foundation
An interesting system to study post-bifurcation instabilities is the growth of a thin

elastic layer attached to an elastic foundation [436]. As the thin elastic layer grows, it
develops residual stresses that eventually trigger a wrinkling instability. Physically,
this pattern reduces the overall energy of the system by relaxing stresses in the stiffer
layer. Theoretically, this initial wrinkling instability can be easily captured by a linear
instability. However, the development of this wrinkling pattern beyond the onset of
instability is surprisingly intricate and depends on the various parameters of the
systems. Phenomenologically, the wrinkles can evolve into creases, ridges, buckles,
folds, or period-doubling wrinkles as sketched in Figure 17.13.
The pattern adopted by the system depends on the relative stiffnesses of the two
layers [139], the thickness of the thin layer, the growth of the top layer [789], the
curvature of the foundation [141], the adhesion energy between the layers [185,
1277], the imperfection of the substrate [166], the anisotropic response [592, 1194]
and the nonlinear elastic response of the materials [645]. For small ratios of layer
μl to foundation μs stiffnesses, μl /μs 10, the system localizes the deformation
and a fold or crease develops as observed in many biological systems. For example,
the deep folding patterns that are formed during the growth of brains are believed to
be partially caused by this instability [70, 142, 362]. The analysis of this instability
is particularly difficult due to the existence of multiple unstable linear modes and
possible contact. Surprisingly, a complete theoretical description is still lacking.
For large ratios of μl /μs 10 and sufficient growth, a period-doubling insta-
bility occurs due to nonlinearities in the substrate response [125, 167] as shown in
Figure 17.14. Whereas period-doubling is well understood in dynamical systems,
understanding the development of a spatial period-doubling pattern is more chal-
lenging even in the absence of growth [401].
The wrinkling of a growing layer on a substrate is usually studied as a quasi-static
problem in which the system relaxes between different growth steps. This is well
justified by the difference in time scales and the fact that growth is imposed as an
Wrinkling Period doubling Period tripling
Brain folding
Bilayer gel
Simulation
Figure 17.14 Period doubling and tripling in the mammalian cortex, swelling polymeric gels, and
numerical simulations of a growing thin layer on an elastic foundation (courtesy of Ellen Kuhl and
adapted from [138, 143]).
external field. However, for many biological systems the evolution of the growth
field depends intimately on the mechanical field itself. When an instability takes
place, this coupling and dynamics over the long time scales of growth should be
taken into account. A natural question is then to understand the dynamics of growing
systems. We saw in Section 14.6 that for homogeneous deformations, the coupling
between the growth dynamics and the nonlinear response of the material is such that
the resulting dynamics may be quite complex even in one and two dimensions.
When the deformation is nonhomogeneous, the problem becomes challenging.
Consider for instance a system with a growth law driven by stress, such that there
is no growth when the Cauchy stress reaches a given homeostatic value. A natural
question is then to determine under which conditions this state of stress is stable.
Surprisingly, this simple question is far from trivial in the presence of residual stress.
The analysis of a system of two growing cylindrical shells with constant growth
reveals that the dynamics in these systems depends not only on the growth law but
also on the geometry and nonlinear response of the material.
• Challenge #8: Develop theoretical and computational methods to analyze the

behavior of morphoelastic systems beyond initial bifurcations. Identify and classify
generic post-bifurcation patterns. Describe the long-term dynamics of growth and
determine generic criteria for the stability of homeostatic states based on the growth
law. Find conditions under which growth dynamics produces oscillatory growth.
17.9 Active Forces, Actives Stresses, and Active Strains
An important mechanical effect found in living systems is the presence of active

forces such as muscular fiber contraction [1060] or the traction exerted by neu-
ronal growth cones [1067]. These active forces are essential for all aspects of life
[656] from motility [493] and cell division [515] at the cellular scale to muscular
motion at the organ level [366]. They are developed actively within the material
by expanding energy. At the cellular level, the term active material mostly refers
17.9 Active Forces, Actives Stresses, and Active Strains 575
to filament networks containing molecular motors such as myosin [1144]. These

force-generating units are attached to filaments and consume adenosine triphosphate
(ATP) to change their molecular configurations which result in the relative sliding
and tensioning of these filaments. In an active fluid system, such as bacterial sus-
pension responding to chemotactic stimuli, these forces create a bath of interacting
self-propelled microorganisms which are known to undergo complex dynamics and
pattern formation as a result of the interplay between active force generation and
hydrodynamic interactions [1113].
Mechanically, these forces are not due to the action of external loads or external
body forces and are not present in a typical engineered material. From a modeling
perspective, they can be introduced in the governing equations at different levels.
Here, we consider the effect of active forces on an elastic solid. Assuming that active
forces do not generate torques, their presence does not alter the basic governing
equation expressing the balance of linear and angular momenta, that can be expressed
in the current configuration as
div T + ρb = ρv̇, TT = T, (17.7)
where, as detailed in Section 11.2.2, ρ is the density, v the velocity of a material

point, T is the Cauchy stress tensor and b a body force density.
Active forces: At the microscale, the forces developed by a cell due to division,
apoptosis, or motion act locally at the points of contact with the elastic medium
[1058]. Assuming that there are N such discrete points in the medium, these forces
can be represented as local force dipoles and included within the body force as
b = be + ba , (17.8)
where be represents the external body forces and the active body force term is
ba = div Q, (17.9)
where

N
Q= D(i) δ(x − x(i) ), (17.10)
i=1
represents the body force stress due to the force dipoles. Here, D(i) is a symmetric
tensor and δ(x − x(i) ) is Dirac’s distribution centred at the point of contact x(i) . By
analogy, the effects of these force dipoles can be treated in the same way as electric
charges in a polarizable medium, defects in solids, or interacting active molecules in
nematic liquid crystals [829].
In linear elasticity, a suitable Green’s function can be introduced to represent the
effects of highly localized forcing terms for the Cauchy equation (17.7). However,
in nonlinear elasticity, the presence of distributions as forcing terms cannot be easily
handled unless a weak form of the equations is used within a computational frame-
work [1297]. It may therefore be more suitable to directly introduce an active stress
tensor.
Active stresses: Rather than postulating the existence of active forces, the presence
of active mechanisms within a continuum tissue can be modeled by an active stress.
In this approach, the Cauchy stress tensor T is split into two components
T = Te + Ta , (17.11)
where Te represents the stress produced by the elastic response of the material in
the absence of active mechanisms. The choice of an additive decomposition between
active and elastic components is made out of convenience. Following the descrip-
tion of Section 11.3, for a hyperelastic incompressible material with a strain-energy
density W = W (F), the elastic stress is given constitutively by
∂W
Te = F − p1. (17.12)
∂F
The active stress Ta is a symmetric tensor postulated phenomenologically to represent

the effect of local contractions. Since contraction is typically localized and directed
it will be both inhomogeneous and anisotropic. The stress generated depends also
on the deformation itself, therefore, we expect the active stress to be a function of
the deformation tensor Ta = Ta (F). Along these lines, different models have been
proposed in connection with heart muscle contraction [112, 1012, 1177]. It follows
immediately from this description and (17.8) that this active stress tensor defines an
active body force through the relationship
ρba = div Ta . (17.13)
Active strains: Muscular contraction or relative sliding of actin filaments can be

interpreted as an internal remodeling of the continuum. This remodeling can be
modeled by the methods used in the theory of growth as a change in the reference
configuration. In this approach [201, 936], the deformation gradient is decomposed
multiplicatively similar to the one used in morphoelasticity
F = Fe Fa , (17.14)
where Fe and Fa are the elastic and active deformation tensors, respectively. Since,
typically we do not expect a change in volume during an active response, we restrict
the choice of active strains so that det Fa = 1. The advantage of this representation
is that it can be linked directly to the relative displacement of materials taking place
during contraction.
From a theoretical point of view, this decomposition is exactly the one used in
morphoelasticity and the entire machinery developed to understand growth problems
17.9 Active Forces, Actives Stresses, and Active Strains 577
Telsecopic extension of the tongue
III. Fire (am)

(ep)
(am)
(is)
I. Rest position II. Loaded position
(am)
(am)
(is) (is)
(ep) (ep) (ep)
(am)
(am) (ep): Entoglossal process (bone)

(am): accelerator muscle
(is): intralingual sheaths
Figure 17.15 The ballistic projection of the chameleon tongue relies on an active mechanism of
elastic energy build-up by muscular contraction in the tongue complex. Muscle fibers are wound
spirally around collagenous sheaths that expand telescopically (adapted from [918], photograph of
Chamaeleo calyptratus courtesy of Stephen Deban).
can be used directly to understand the active stresses generated by active strains with
the appropriate change of terminology. It has also been argued that it presents several
advantages from a theoretical and computational point of view [28, 1034].
Active fibers: Yet another way to model active growth is at the level of the strain-
energy density, either by using pseudo-elasticity [1004] or by assuming that the zero-
strain-energy along contracting fibers differs from the zero-strain state for the rest
of the tissue [918]. As an example, following the theory of invariants for anisotropic
materials from Section 11.11, we assume that the strain-energy density only depends
on the invariants I1 , I2 for the isotropic part and I4 for the anisotropic part, so that
W = W (I1 − 3, I2 − 3, I4 − ν). (17.15)
We recall that I4 represents the stretch along fibers. When ν = 1, the strain-energy
has a global minimum in the absence of deformations given by I1 = I2 = 3 and
I4 = 1. However, the elastic energy increases when ν=1 describing the rest state
of the fiber (with 0 < ν < 1 describing active fiber contraction). This approach
was used to model the active contraction of muscle fibers in the chameleon tongue
resulting in ballistic extension, as shown in Figure 17.15.
• Challenge #9: Understand the links between different methods to model active
processes within a unifying framework. Connect active processes at the tissue level
to specific mechanisms and energy expenditure at the microscale.
17.10 The Mathematical Foundations
The framework described in this monograph is a first step in building a general

mechanical theory of growing and evolving biological systems. In building such a
theory, one seeks to harmonize experiments, theory, and simulations so that each
discipline can inform the other. A general theoretical framework where ideas and
concepts are clearly discussed is of fundamental importance but is not sufficient.
Another crucial aspect of such a construction is to build rigorous mathematical foun-
dations for the theory so that it stands on solid ground. Very little is known about
the mathematical aspects of this theory and most of the outstanding mathematical
problems lingering in classical nonlinear elasticity [54] are completely open when
it comes to the more general theory of morphoelasticity. In particular there are two
important mathematical aspects of the theory that need further discussion.
Geometry: An important aspect of morphoelasticity is the distinction between

growth deformations and elastic deformations modeled through the decomposition
of the deformation gradient F = AG. Being able to separate the two processes allows
us to use a material constitutive law for the instantaneous response of the material
and an evolution law for the remodeling and growth dynamics. Both processes can
be clearly identified and the consequences of a change in geometry can be directly
obtained by the proper kinematics. The traditional view is that the growth process
is described by a local map, mapping vectors in the tangent space of the initial con-
figuration at a point to vectors in the tangent space of the virtual configuration. This
decomposition is perfectly suitable to define all kinematic and mechanical quanti-
ties as well as to obtain the governing equations of morphoelasticity. However, the
mathematical status of this so-called intermediate or virtual state is not clear. The
usual picture of a dislocated configuration viewed as a collection of sub-bodies is
unsatisfactory as there could be infinitely many of these pieces and their connections
must be somehow specified. What is the mathematical nature of such an object?
In Section 12.5.1, we showed that differential geometry provides a natural way
to justify the multiplicative decomposition. In this framework, the reference config-
uration is seen as a material manifold with vanishing curvature tensor and a torsion
tensor T directly determined by the growth tensor G through
T (G) = G−1 Skw(Grad G). (17.16)
Therefore, the virtual configuration can be formally defined as a Weitzenböck mani-

fold with torsion tensor T (G) and its tangent bundle is the natural space on which all
kinematic quantities appearing in morphoelasticity can be defined. This a posteriori
17.10 The Mathematical Foundations 579
justification of the morphoelasticity approach provides a rigorous way to answer

fundamental questions on the mathematical nature of growth processes. Borrowing
ideas from the geometric theory of defects in solids, the theory can then be further
generalized to introduce new effects associated with growth and remodeling such as
localized point or line growth. These effects are known to require a generalization of
the Weitzenböck manifold to include nonvanishing curvature and nonmetricity [1352,
1353, 1355].
An interesting difficulty arises when considering the dynamic evolution of growth.
If we assume that the growth tensor depends on the stress, the torsion is determined
by a set of partial differential equations involving torsion itself. The evolution of the
geometry and topology of manifolds through differential equations, such as Ricci
flows [206], is an important topic of differential geometry that is of direct relevance
to the mathematical description of growth. The use of differential geometry also
opens the door to the development of new numerical schemes taking advantage of
the underlying geometric structure [31, 32].
Analysis: Independent of the geometric nature of the governing equations, the prob-
lem of growth can be formulated as either a variational problem with respect to the
modified strain-energy density function or as a set of nonlinear partial differential
equations. Viewed as a problem in partial differential equations, several results on
well-posedness and local existence of solution of morphoelasticity can be established
[419] giving hope that general global results will follow. A natural question is to see
how the classical problems of elasticity extend to morphoelasticity. For instance, as
discussed in Section 11.9, Ericksen’s problem states that there exist several classes
of universal solutions for isotropic materials [677, 833]. For simple enough growth
tensors, such as dilations, these universal solutions persist but the general conditions
on the growth tensor where the universal solutions persist are not known. Similarly,
the general conditions on both the strain-energy functions and the growth tensor
under which there exist local or global energy minimizers for problems in morphoe-
lastostatics are largely unknown. The smoothness of these minimizers if they exist
is also unknown. Once the existence of such solutions can be established, general
bifurcation and stability problems can be studied for problems that involved both
mixed boundary conditions and varying growth parameters.
• Challenge #10: Provide a full mathematical description of growth using the tools of
differential geometry. Describe the evolution of these geometric quantities and link
them, in a practical way to physically and biologically relevant objects. Study the
evolution of manifolds under generic flows relevant to growth processes. Generalize
Ericksen’s problem for systems with residual stress. Prove the existence of static
solutions for morphoelasticity under generic conditions on both the growth tensor
and the strain-energy functions.
References
1. Atlas of Actinomycetes. The Society for Actinomycetes (Japan), Asakura Publishing Co.
ltd. (1997) 65, 235, 237
2. Abramowitz, M., Stegun, I.A.: Handbook of mathematical functions: with formulas, graphs,
and mathematical tables. Courier Corporation (1964) 459
3. Acharya, A., Bassani, J.L.: Lattice incompatibility and a gradient theory of crystal plasticity.
J. Mech. Phys. Solids 48(8), 1565–1595 (2000) 369, 371
4. Ackerly, S.C.: Kinematics of accretionary shell growth, with examples from brachiopods
and molluscs. Paleobiology 15(2), 147–164 (1989) 7
5. Adam, J.A., Maggelakis, S.A.: Diffusion regulated growth characteristics of a spherical
prevascular carcinoma. Bull. Math. Biol. 52(4), 549–582 (1990) 38, 72
6. Adams, D.S., Keller, R., Koehl, M.A.: The mechanics of notochord elongation, straightening
and stiffening in the embryo of Xenopus laevis. Development 110(1), 115–130 (1990) 338
7. Ades, L.C., Knight, W.B., Byard, R.W., Bateman, J.F., Esquivel, J.A.D., Mee, R.B.B., Haan,
E.A., Milewicz, D.M.: Clinicopathologic findings in congenital aneurysms of the great ves-
sels. Am. J. Med. Genet. 66(3), 289–299 (1996) 512
8. Adkins, J.E., Rivlin, R.S.: Large elastic deformations of isotropic materials. IX. The defor-
mation of thin shells. Philos. Trans. Roy. Soc. A 244, 505–531 (1952) 304, 442
9. Adkins, J.E., Rivlin, R.S.: Large elastic deformations of isotropic materials X. Reinforcement
by inextensible cords. Philos. Trans. Roy. Soc. A 248(944), 201–223 (1955) 321
10. Aegerter-Wilmsen, T., Aegerter, C.M., Bisseling, T.: Model for the robust establishment of
precise proportions in the early Drosophila embryo. J. Theor. Biol. 234(1), 13–19 (2005)
559
11. Aegerter-Wilmsen, T., Aegerter, C.M., Hafen, E., Basler, K.: Model for the regulation of
size in the wing imaginal disc of Drosophila. Mechanisms of development 124(4), 318–326
(2007) 559
12. Aeschlimann, M., Tettoni, L.: Biophysical model of axonal pathfinding. Neurocomp. 38(40),
87–92 (2001) 85
13. Agoras, M., Lopez-Pamies, O., Ponte Castañeda, P.: A general hyperelastic model for in-
compressible fiber-reinforced elastomers. J. Mech. Phys. Solids 57(2), 268–286 (2009) 294
14. Alarcón, T., Byrne, H.M., Maini, P.K.: A cellular automaton model for tumour growth in
inhomogeneous environment. J. Theor. Biol. 225(2), 257–274 (2003) 52
15. Alberts, B., Johnson, A., Lewis, J., Raff, M., Roberts, K., Walter, P.: Molecular Biology of
the Cell. Taylor & Francis Group (2008) 320
16. Aleksic, M., Schütz, G., Gerth, S., Mulch, J.: Surgical approach to kinking and coiling of
the internal carotid artery. J. Cardiovasc. Surg. 45(1), 43–48 (2004) 510
Interdisciplinary Applied Mathematics 45, DOI 10.1007/978-0-387-87710-5
582 References
17. Alexander, H.: Tensile instability of initially spherical balloons. Int. J. Eng. Sci 9, 151–160
(1971) 304, 442
18. Alford, P.W., Humphrey, J.D., Taber, L.A.: Growth and remodeling in a thick-walled artery
model: effects of spatial variations in wall constituents. Biomech. Model. Mechanobiol. 7,
245–262 (2008) 63, 409
19. Allaire, G.: Homogenization and two-scale convergence. SIAM J. Math. Anal. 23(6), 1482–
1518 (1992) 562
20. Alley, T.R.: Head shape and the perception of cuteness. Developmental Psychology 17(5),
650–654 (1981) 17
21. Ambrosi, D., Ateshian, G.A., Arruda, E.M., Cowin, S.C., Dumais, J., Goriely, A., Holzapfel,
G.A., Humphrey, J.D., Kemkemer, R., Kuhl, E., Olberding, J.E., Taber, L.A., Garikipati, K.:
Perspectives on biological growth and remodeling. J. Mech. Phys. Solids 59, 863–883 (2011)
47, 49
22. Ambrosi, D., Guana, F.: Stress-modulated growth. Math. Mech. Solids 12(3), 319–342
(2007) 390, 407, 408
23. Ambrosi, D., Guillou, A.: Growth and dissipation in biological tissues. Continuum Mech.
Therm. 19(5), 245–251 (2007) 407
24. Ambrosi, D., Guillou, A., Di Martino, E.S.: Stress-modulated remodeling of a non-
homogeneous body. Biomech. Model. Mechanobiol. 7(1), 63–76 (2008) 407
25. Ambrosi, D., Mollica, F.: On the mechanics of a growing tumor. Int. J. Eng. Sci. 40, 1297–
1316 (2002) 76
26. Ambrosi, D., Mollica, F.: The role of stress in the growth of a multicell spheroid. J. Math.
Biol. 48, 477–499 (2004) 45, 48, 409
27. Ambrosi, D., Pettinati, V., Ciarletta, P.: Active stress as a local regulator of global size in
morphogenesis. Int. J. Non-Linear Mech. 75, 5–14 (2014) 559
28. Ambrosi, D., Pezzuto, S.: Active stress vs. active strain in mechanobiology: constitutive
issues. J. Elasticity 107(2), 199–212 (2012) 577
29. Ambrosi, D., Preziosi, L.: On the closure of mass balance models for tumor growth. Math.
Model. Meth. Appl. Sci. 12(5), 737 (2002) 76
30. Amrouche, C., Ciarlet, P.G., Gratie, L., Kesavan, S.: On Saint Venant’s compatibility con-
ditions and Poincaré’s lemma. C. R. Math. 342(11), 887–891 (2006) 362
31. Angoshtari, A., Yavari, A.: A geometric structure-preserving discretization scheme for in-
compressible linearized elasticity. Comput. Methods Appl. Mech. Eng. 259, 130–153 (2013)
579
32. Angoshtari, A., Yavari, A.: Differential complexes in continuum mechanics. Arch. Rat.
Mech. Anal. 216(1), 193–220 (2015) 372, 579
33. Antman, S.S.: Nonlinear problems of elasticity. Springer New York (2005) 66, 97, 101, 107,
257, 262, 489
34. Antman, S.S., Negrón-Marrero, P.V.: The remarkable nature of radially symmetric equilib-
rium states of aeolotropic nonlinearly elastic bodies. J. Elasticity 18(2), 131–164 (1987)
450
35. Antman, S.S., Shvartsman, M.M.: The shrink-fit problem for aeolotropic nonlinearly elastic
bodies. J. Elasticity 37(2), 157–166 (1994) 476
36. Arai, T., Kuroishi, T., Saito, Y., Kurita, Y., Naruke, T., Kaneko, M.: Tumor doubling time
and prognosis in lung cancer patients: evaluation from chest films and clinical follow-up
study. Jpn. J. Clin. Oncol. 24(4), 199–204 (1994) 376
37. Araujo, R.P., McElwain, D.L.S.: A history of the study of solid tumour growth: The contri-
bution of mathematical modelling. Bull. Math. Biol. 66(5), 1039–1091 (2004) 17
38. Arias, J.L., Fernandez, M.S.: Biomimetic processes through the study of mineralized shells.
Mater. Charact. 50(2-3), 189–195 (2003) 188
39. Arnett, T.: Regulation of bone cell function by acid–base balance. Proc. Nutrition Soc.
62(02), 511–520 (2003) 36
40. Arruda, E.M., Boyce, M.C.C.: A three-dimensional constitutive model for the large stretch
behavior of rubber elastic materials. J. Mech. Phys. Solids 41(2), 389–412 (1993) 320, 559
References 583
41. Ashby, M.F., Gibson, L.J., Wegst, U., Olive, R.: The mechanical properties of natural mate-
rials. I Material property charts. Proc. Roy. Soc. Lond. A 450, 123–140 (1995) 227
42. Astbury, W., Preston, R.: The structure of the cell wall in some species of the filamentous
green alga Cladophora. Proc. Roy. Soc. Lond. B 129(854), 54–76 (1940) 542
43. Atela, P., Golé, C., Hotton, S.: A dynamical system for plant pattern formation: a rigorous
analysis. J. Nonlinear Sci. 12(6), 641–676 (2002) 565
44. Ateshian, G.A., Humphrey, J.D.: Continuum mixture models of biological growth and re-
modeling: past successes and future opportunities. Annu. Rev. Biomed. Eng. 14, 97–111
(2012) 49
45. Atkinson, G.F.: Lessons in Botany. Henry Holt and Company. New York (1900) 40, 47
46. Audoly, B., Pomeau, Y.: Elasticity and geometry. Oxford Univ. Press (2010) 115
47. Avery, W.B., Herakovich, C.T.: Effect of fiber anisotropy on thermal stresses in fibrous
composites. J. Appl. Mech. 53(4), 751–756 (1986) 450
48. Avery Jr, G.S.: Structure and development of the tobacco leaf. Am. J. Bot. 20, 565–592
(1933) 26
49. Azeloglu, E.U., Albro, M.B., Thimmappa, V.A., Ateshian, G.A., Costa, K.D.: Heterogeneous
transmural proteoglycan distribution provides a mechanism for regulating residual stresses
in the aorta. Am. J. Physiol. -Heart C. 294(3), H1197–H1205 (2008) 45
50. Bai, A., Eidelman, D.H., Hogg, J.C., James, A.L., Lambert, R.K., Ludwig, M.S., Martin, J.,
McDonald, D.M., Mitzner, W.A., Okazawa, M., et. al.: Proposed nomenclature for quanti-
fying subdivisions of the bronchial wall. J. Appl. Physiol. 77, 1011–1014 (1994) 532
51. Baker, M., Ericksen, J.L.: Inequalities restricting the form of the stress-deformation relations
for isotropic elastic solids and Reiner-Rivlin fluids. J. Wash. Acad. Sci 44(3), 33–35 (1954)
291
52. Baker, R.E., Maini, P.K.: A mechanism for morphogen-controlled domain growth. J. Math.
Biol. 54(5), 597–622 (2007) 564
53. Ball, J.M.: Discontinuous equilibrium solutions and cavitation in nonlinear elasticity. Philos.
Trans. Roy. Soc. London A 306(1496), 557–611 (1982) 451
54. Ball, J.M.: Some open problems in elasticity. In: Geometry, mechanics, and dynamics, pp.
3–59. Springer (2002) 578
55. Bamber, R.K.: A general theory for the origin of growth stresses in reaction wood: how trees
stay upright. IAWA J. 22(3), 205–212 (2001) 42, 508
56. Bando, K., Yamashita, D., Ohba, K.: Numerical Simulation for Mechanism of Airway Nar-
rowing in Asthma. Jpn. Soc. Mech. Eng. C 45(4), 1020–1027 (2002) 532
57. Bard, L., Boscher, C., Lambert, M., Mege, R.M., Choquet, D., Thoumine, O.: A molecular
clutch between the actin flow and N-cadherin adhesions drives growth cone migration. J.
Neurosci. 28(23), 5879 (2008) 80
58. Barkley, M.D., Zimm, B.H.: Theory of twisting and bending of chain macromlecules; analy-
sis of the fluorescence depolarization of DNA. J. Chem. Phys. 70 (1979) 63
59. Barlow, P.W., Brain, P., Adam, J.S.: Differential growth and plant tropism: a study assisted
by computer simulation. Environmental Exp. Bot. 29 (1989) 39, 63
60. Baron, C.J.: The structural mechanics and morphogenesis of extant regular echinoids having
rigid tests. PhD, 270pp. Berkeley: University of California. (1991) 217
61. Barron, S., Foss, J.A., Riley, E.P.: The effect of prenatal cocaine exposure on umbilical cord
length in fetal rats. Neurotoxicol. Teratol. 13, 503–506 (1991) 154
62. Bartnicki-Garcia, S.: Hyphal tip growth: outstanding questions. In: edited by Heinz
D. Osiewacz (ed.) Molecular Biology of Fungal Development, vol. 15, pp. 29–58. CRC
Press, Inc, Boca Raton, Florida (2002) 238
63. Bartnicki-Garcia, S., Bracker, C.E., Gierz, G., Lopez-Franco, R., Lu, H.S.: Mapping the
growth of fungal hyphae: Orthogonal cell wall expansion during tip growth and the role of
turgor. Biophys. J. 79(5), 2382–2390 (2000) 234, 235, 238
64. Bartnicki-Garcia, S., Hergert, F., Gierz, G.: Computer simulation of fungal morphogenesis
and the mathematical basis for hyphal tip growth. Protoplasma 153, 46–57 (1989) 239
584 References
65. Bashforth, F., Adams, J.C.: An attempt to test the theories of capillary action: by comparing
the theoretical and measured forms of drops of fluid. University Press (1883) 216
66. Baskin, T.I.: Anisotropic expansion of the plant cell wall. Annu. Rev. Cell Dev. Biol. 21,
203–222 (2005) 513, 542
67. Baskin, T.I., Jensen, O.E.: On the role of stress anisotropy in the growth of stems. J. Exp.
Bot. 64, 4697–4707 (2013) 64, 96
68. Bastien, R., Bohr, T., Moulia, B., Douady, S.: Unifying model of shoot gravitropism reveals
proprioception as a central feature of posture control in plants. Proc. Natl. Acad. Sci. USA
110(2), 755–760 (2013) 65
69. Bastien, R., Douady, S., Moulia, B.: A unifying modeling of plant shoot gravitropism with
an explicit account of the effects of growth. Front. Plant Sci. 5, 136. (2014) 65
70. Bayly, P.V., Taber, L.A., Kroenke, C.D.: Mechanical forces in cerebral cortical folding: a
review of measurements and models. J. Mech. Behav. Biomed. Mater. 29, 568–581 (2014)
573
71. Beatty, M.F.: A class of universal relations in isotropic elasticity theory. J. Elasticity 17(2),
113–121 (1987) 305
72. Beatty, M.F.: Introduction to nonlinear elasticity. In: Nonlinear effects in fluids and solids,
pp. 13–112. Springer (1996) 304, 446
73. Beatty, M.F., Hook, D.E.: Some experiments on the stability of circular, rubber bars under end
thrust(Circular rubber bars stability under compressive end loads compared for experimental
data and analytical solutions). Int. J. Solids Struct. 4, 623–635 (1968) 490
74. Belgacem, H.B., Conti, S., DeSimone, A., Müller, S.: Rigorous bounds for the Föppl—von
Kármán theory of isotropically compressed plates. J. Nonlinear Sci. 10(6), 661–685 (2000)
256
75. Bell, P.R.: Twining of the hop (Humulus lupulus L. ). Nature 181, 1009–1010 (1958) 65, 167
76. Beloussov, L.V.: Mechanically based generative laws of morphogenesis. Phys. Biol. 5,
015,009 (2008) 410
77. Beloussov, L.V., Dorfman, J.G., Cherdantzev, V.G.: Mechanical stresses and morphological
patterns in amphibian embryos. J. Embryol. Exp. Morph. 34(3), 559–574 (1975) 45, 410
78. Beloussov, L.V., Grabovsky, V.I.: A geometro-mechanical model for pulsatile morphogen-
esis. Comput. Methods Biomech. Biomed. Engin. 6(1), 53–63 (2003) 12, 419
79. Beloussov, L.V., Grabovsky, V.I.: Morphomechanics: goals, basic experiments and models.
Int. J. Dev. Biol. 50(2), 81–92 (2006) 45, 410
80. Ben Amar, M., Goriely, A.: Growth and instability in elastic tissues. J. Mech. Phys. Solids
53, 2284–2319 (2005) 46, 127, 400, 461
81. Ben Amar, M., Pomeau, Y.: Theory of dendritic growth in a weakly undercooled melt.
Europhys. Lett. 2 (1986) 3
82. Bender, C.M., Orszag, S.A.: Advanced Mathematical Methods for Scientists and Engineers
I. Springer Science & Business Media (1999) 465
83. Benham, C.J.: An elastic model of the large structure of duplex DNA. Bioploymers 18,
609–623 (1979) 63, 136
84. Benham, C.J.: Geometry and mechanics of DNA superhelicity. Biopolymers 22, 2477–2495
(1983) 63, 136
85. Benirschke, K., Kaufmann, P.: Pathology of the human placenta. Fourth edition. Springer,
New York (2000) 155
86. Bensoussan, A., Lions, J.L., Papanicolaou, G.: Asymptotic analysis for periodic structures,
vol. 374. American Mathematical Soc. (2011) 562
87. Berg, E.E.: Chinese foot binding. Orthopaedic Nursing 14(5), 66–69 (1995) 28
88. Berg, H.C., Anderson, R.A.: Bacteria swin by rotating their flagellar filaments. Nature 245
(1973) 155
89. Bergel, D.H.: The visco-elastic properties of the arterial wall. Ph.D. thesis, University of
London (1960) 45
90. Bergman, K., Burke, P.V., Cerdá-Olmedo, E., David, C.N., Delbrück, M., Foster, K., Goodell,
E., Heisenberg, M., Meissner, G., Zalokar, M., et al.: Phycomyces. Bacteriol. Rev. 33(1), 99
(1969) 63, 542
References 585
91. Berry, C.L.: The growth and development of large arteries. Exp. Embryol. Teratology 1,
34–64 (1974) 32
92. Beveridge, D.L., Barreiro, G., Byun, K.S., Case, D.A., Cheatham, T.E., Dixit, S.B., Giudice,
E., Lankas, F., Lavery, R., Maddocks, J.H., et al.: Molecular dynamics simulations of the 136
unique tetranucleotide sequences of DNA oligonucleotides. i. research design and results on
d (c p g) steps. Biophys. J. 87(6), 3799–3813 (2004) 116
93. Bhattacharya, K., Lewicka, M., Schäffner, M.: Plates with incompatible prestrain. Arch. Rat.
Mech. Anal. pp. 1–39 (2015) 252
94. Biggs, N., Lloyd, E.K., Wilson, R.J.: Graph Theory. Clarendon, Oxford (1976) 150
95. Bigoni, D.: Nonlinear solid mechanics: bifurcation theory and material instability. Cam-
bridge University Press (2012) 311
96. Bigoni, D., Gei, M.: Bifurcations of a coated, elastic cylinder. Int. J. Solids Struct. 38(30-31),
5117–5148 (2001) 299, 490
97. Bilby, B.A., Bullough, R., Smith, E.: Continuous distributions of dislocations: a new appli-
cation of the methods of non-Riemannian geometry. Proc. Roy. Soc. Lond. A 231(1185),
263–273 (1955) 363
98. Bilby, B.A., Gardner, L.R.T., Stroh, A.N.: Continuous distributions of dislocations and the
theory of plasticity. In: Proceedings of the Ninth International Congress of Applied Mechan-
ics, Bruxelles, vol. 8, pp. 35–44 (1957) 353
99. Billington, E.W.: The Poynting effect. Acta Mech. 58(1), 19–31 (1986) 293
100. Biot, M.A.: Bending of an infinite beam on an elastic foundation. ZAMP 2(3), 165–184
(1922) 143
101. Biot, M.A.: General theory of three-dimensional consolidation. J. Appl. Phys. 12(2), 155–
164 (1941) 394
102. Biot, M.A.: Exact theory of buckling of a thick slab. Appl. Sci. Res 12, 183–198 (1962) 318,
489
103. Biot, M.A.: Mechanics of incremental deformation. Wiley, New York (1965) 46, 298, 311
104. Black, R., Turner, S.J., Johnson, M.S.: The early life history of Bembicium vittatum Philippi,
1846 (Gastropoda: Littorinidae). Veliger 37(4), 393–399 (1994) 194
105. Blum, J., Wurm, G.: The growth mechanisms of macroscopic bodies in protoplanetary disks.
Annu. Rev. Astron. Astrophys. 46, 21–56 (2008) 570
106. Bock, R.D., Wainer, H., Petersen, A., Thissen, D., Murray, J., Roche, A.: A parameterization
for individual human growth curves. Human Biology 45(1), 63–80 (1973) 15
107. Bodenstein, J.: A dynamic simulation model of tissue growth and cell patterning. Cell dif-
ferentiation 19(1), 19–33 (1986) 55
108. Boettiger, A., Ermentrout, B., Oster, G.: The neural origins of shell structure and pattern in
aquatic mollusks. Proc. Natl. Acad. Sci. 106(16), 6837 (2009) 189
109. Bogen, D.K., McMahon, T.A.: Do cardiac aneurysms blow out? Biophys. J. 27 (1979) 295,
304
110. Bogin, B.: Patterns of human growth, vol. 23. Cambridge University Press (1999) 12, 17
111. Bohn, H.F., Günther, F., Fink, S., Speck, T.: A Passionate Free Climber: Structural Devel-
opment and Functional Morphology of the Adhesive Tendrils in Passiflora discophora. Int.
J. Plant Sc. 176(3), 294–305 (2015) 150
112. Böl, M., Abilez, O.J., Assar, A.N., Zarins, C.K., Kuhl, E.: In vitro/in silico characterization
of active and passive stresses in cardiac muscle. Int. J. Multiscale Comp. Eng. 10, 171–188
(2012) 576
113. Bonner, J.: The action of the plant growth hormone. J. Gen. Physiol. 17(1), 63–76 (1933) 42
114. Boudaoud, A.: Growth of walled cells: From shells to vesicles. Phys. Rev. Lett. 91, #018,
104 (2003) 228
115. Boudaoud, A.: The buckling of a swollen thin gel layer bound to a compliant substrate. J.
Appl. Mech. 75, 051,002–1 (2008) 403
116. Boudaoud, A., Chaïeb, S.: Mechanical phase diagram of shrinking cylindrical gels. Phys.
Rev. E 68(2), 21,801 (2003) 3, 403
586 References
117. Bourgery, J.M., Jacob, N.H.: Traité complet de l’anatomie de l’homme, comprenant la
médecine opératoire. CA Delaunay (1832) 34, 35
118. Bowen, R.M.: Toward a thermodynamics and mechanics of mixtures. Arch. Rat. Mech. Anal.
24(5), 370–403 (1967) 394
119. Boyce, M.C., Arruda, E.M.: Constitutive models for rubber elasticity: a review. Rubber
Chem. Technol. 73, 504–523 (2000) 293, 533, 559
120. Boyce, M.C., Weber, G.G., Parks, D.M.: On the kinematics of finite strain plasticity. J. Mech.
Phys. Solids 37(5), 647–665 (1989) 401
121. Boyer, J., Wu, G.: Auxin increases the hydraulic conductivity of auxin-sensitive hypocotyl
tissue. Planta 139(3), 227–237 (1978) 96
122. Bracker, C.E., Murphy, D.J., Lopez-Franco, R.: Laser microbeam manipulation of cell mor-
phogenesis growing in fungal hyphae. Proc. SPIE 2983, 67–80 (1997) 238
123. Brand, A., Lee, K., Veses, V., Gow, N.A.R.: Calcium homeostasis is required for contact-
dependent helical and sinusoidal tip growth in Candida albicans hyphae. Molecular micro-
biology 71(5), 1155–1164 (2009) 238, 419
124. Brand, R., Claes, L.: The law of bone remodelling. J. Biomech. 22, 185–187 (1989) 34
125. Brau, F., Vandeparre, H., Sabbah, A., Poulard, C., Boudaoud, A., Damman, P.: Multiple-
length-scale elastic instability mimics parametric resonance of nonlinear oscillators. Nature
Physics 7(1), 56–60 (2011) 573
126. Bray, D.: Axonal growth in response to experimentally applied mechanical tension. . Dev.
Biol. 102(2), 379 (1984) 30, 80, 81
127. Broca, P.: Sur le volume et la forme du cerveau suivant les individus et suivant les races,
vol. 1. Hennuyer (1861) 20
128. Brodland, G.W.: The differential interfacial tension hypothesis (dith): a comprehensive the-
ory for the self-rearrangement of embryonic cells and tissues. J. Biomech. Eng. 124(2),
188–197 (2002) 51
129. Brodland, G.W.: Computational modeling of cell sorting, tissue engulfment, and related
phenomena: A review. Appl. Mech. Rev. 57(1), 47–76 (2004) 51
130. Brodland, G.W., Chen, H.H.: The mechanics of cell sorting and envelopment. J. Biomech.
33(7), 845–851 (2000) 51
131. Brodland, G.W., Veldhuis, J.H.: Computer simulations of mitosis and interdependencies
between mitosis orientation, cell shape and epithelia reshaping. J. Biomech. 35(5), 673–681
(2002) 58
132. Brook, B., Peel, S., Hall, I., Politi, A., Sneyd, J., Bai, Y., Sanderson, M., Jensen, O.: A bio-
mechanical model of agonist-initiated contraction in the asthmatic airway. Respir. Physiol.
Neurobiol. 170(1), 44–58 (2010) 533
133. Brooks, R.C.: Origins, usage and production of screws: An historical perspective. History
and Technology, an International Journal 8(1), 51–76 (1991) 116
134. Brown, C.L., Sommer, H.E., Pienaar, L.V.: The predominant role of the pith in the growth and
development of internodes in Liquidambar styraciflua (Hamamelidaceae). I. Histological
basis of compressive and tensile stresses in developing primary tissues. Am. J. Bot. 82(6),
769–776 (1995) 42
135. Brown, C.L., Sommer, H.E., Pienaar, L.V.: The predominant role of the pith in the growth
and development of internodes in Liquidambar styraciflua (Hamamelidaceae). II. Pattern
of tissue stress and response of different tissues to specific surgical procedures. Am. J. Bot.
82(6), 777–781 (1995) 42
136. Brown, J.H., West, G.B.: Scaling in biology. Oxford University Press (2000) 21
137. Bruck, H., Evans, J., Peterson, M.: The role of mechanics in biological and biologically
inspired materials. Exp. Mech. 42(4), 361–371 (2002) 328
138. Budday, S., Andres, S., Steinmann, P., Kuhl, E.: Primary and secondary instabilities in soft
bilayered systems. PAMM 15(1), 281–282 (2015) 574
139. Budday, S., Kuhl, E., Hutchinson, J.W.: Period-doubling and period-tripling in growing
bilayered systems. Philos. Mag. 95, 3208–3224 (2015) 573
References 587
140. Budday, S., Raybaud, C., Kuhl, E.: A mechanical model predicts morphological abnormal-
ities in the developing human brain. Scientific reports 4, 5644 (2014) 45
141. Budday, S., Steinmann, P., Goriely, A., Kuhl, E.: Size and curvature regulate pattern selection
in the mammalian brain. Extreme Mech. Lett. 4, 193–198 (2015) 20, 573
142. Budday, S., Steinmann, P., Kuhl, E.: The role of mechanics during brain development. J.
Mech. Phys. Solids 72, 75–92 (2014) 573
143. Budday, S., Steinmann, P., Kuhl, E.: Secondary instabilities modulate cortical complexity in
the mammalian brain. Philos. Mag. 95(28-30), 3244–3256 (2015) 574
144. Buganza, T.A., Gosain, A.K., Kuhl, E.: Stretching skin: The physiological limit and beyond.
Int. J. Non-Linear Mech. 47(8), 938–949 (2012) 320
145. Buganza Tepole, A., Kuhl, E.: Computational modeling of chemo-bio-mechanical coupling:
a systems-biology approach toward wound healing. Comput. Methods Biomech. Biomed.
Eng. 19(1), 13–30 (2016) 567
146. Bullitt, E., Ewend, M.G., Aylward, S., Lin, W., Gerig, G., Joshi, S., Jung, I., Muller, K.,
Smith, J.K.: Abnormal vessel tortuosity as a marker of treatment response of malignant
gliomas: preliminary report. Technol. Cancer Res. Treat. 3(6), 577–584 (2004) 510
147. Bunce, J.A.: Leaf elongation in relation to leaf water potential in soybean. J. Exp. Bot. 28(1),
156–161 (1977) 92
148. Burgert, I.: Exploring the micromechanical design of plant cell walls. Am. J. Bot. 93(10),
1391–1401 (2006) 320
149. Burgreen, D., Manitt, P.J.: Thermal buckling of a bimetallic beam. J. Eng. Mech. Div. -ASCE
95, 421–32 (1969) 65
150. Burgreen, D., Regal, D.: Higher mode buckling of bimetallic beam. J. Eng. Mech. Div.
-ASCE 97(4), 1045–1056 (1971) 65
151. Burridge, R., Keller, J.B.: Poroelasticity equations derived from microstructure. J. Acoust.
Soc. Am. 70, 1140–1146 (1981) 395
152. Burström, H.: Tissue tensions during cell elongation in wheat roots and a comparison with
contractile roots. Physiol. Plantarum 25(3), 509–513 (1971) 42
153. Burton, A.C.: The importance of the shape and size of the heart. American Heart J. 54(6),
801–810 (1957) 32, 33
154. Burton, W., Cabrera, N., Frank, F.C.: The growth of crystals and the equilibrium structure
of their surfaces. Philos. Trans. Roy. Soc. London A 243(866), 299–358 (1951) 570
155. Buschmann, H., Hauptmann, M., Niessing, D., Lloyd, C.W., Schäffner, A.R.: Helical growth
of the Arabidopsis mutant tortifolia2 does not depend on cell division patterns but involves
handed twisting of isolated cells. Plant Cell 21(7), 2090 (2009) 65, 121
156. Byrne, H., Drasdo, D.: Individual-based and continuum models of growing cell populations:
a comparison. J. Math. Biol 58(4-5), 657–687 (2009) 55
157. Byrne, H., Preziosi, L.: Modelling solid tumour growth using the theory of mixtures. Math.
Med. Biol. 20, 341–366 (2003) 396
158. Byrne, H.M., Chaplain, M.A.J.: Growth of nonnecrotic tumors in the presence and absence
of inhibitors. Math. Biosci. 130(2), 151–181 (1995) 38, 72
159. Cabana, T., Jolicoeur, P., Michaud, J.: Prenatal and postnatal growth and allometry of stature,
head circumference, and brain weight in Québec children. Am. J. Human Biol. 5(1), 93–99
(1993) 17
160. Caillerie, D., Mourad, A., Raoult, A.: Cell-to-muscle homogenization. application to a con-
stitutive law for the myocardium. ESAIM: Mathematical Modelling and Numerical Analysis-
Modélisation Mathématique et Analyse Numérique 37(4), 681–698 (2003) 562
161. Candolle, A.P.d.: Organographie végétale. Chez Deterville, Paris (1827) 150, 151
162. Candolle, A.P.d.: Physiologie végétale. Béchet Jeune, Paris (1832) 151
163. Canejo, J.P., Godinho, M.H.: Cellulose perversions. Materials 6(4), 1377–1390 (2013) 155
164. Canham, P.B., Finlay, H.M., Dixon, J.G., Boughner, D.R., Chen, A.: Measurements from
light and polarised light microscopy of human coronary arteries fixed at distending pressure.
Cardiovasc. Research 23(11), 973–982 (1989) 339
588 References
165. Cannon, W.B.: Physiological regulation of normal states: some tentative postulates concern-
ing biological homeostatics. In: A. Petit (ed.) Ses Amis, ses Collèges, ses Élèves, p. 91.
Paris: Les Éditions Médicales (1926) 409
166. Cao, Y., Hutchinson, J.W.: From wrinkles to creases in elastomers: the instability and
imperfection-sensitivity of wrinkling. Proc. Roy. Soc. Lond. A 468(2137), 94–115 (2012)
573
167. Cao, Y., Hutchinson, J.W.: Wrinkling phenomena in neo-Hookean film/substrate bilayers. J.
Appl. Mech. 79(3), 031,019 (2012) 573
168. Cardamone, L., Valentin, A., Eberth, J.F., Humphrey, J.D.: Origin of axial prestretch and
residual stress in arteries. Biomech. Model. Mechanobiol. 8(6), 431–446 (2009) 45, 509
169. Carmo, M.P.D.: Differential geometry of curves and surfaces. Prentice-Hall, New Jersey
(1976) 99, 209, 254
170. Carnoy, J.B.: Recherches anatomiques et physiologiques sur les champignons. Bull. Soc.
Roy. Bot. Belgique 9, 157–321 (1870) 542, 543
171. Carol, R.J., Dolan, L.: Building a hair: tip growth in Arabidopsis thaliana root hairs. Philos.
Trans. Roy. Soc. B 357(1422), 815–821 (2002) 238
172. Carpita, N.C., Gibeaut, D.M.: Structural models of primary cell walls in flowering plants:
consistency of molecular structure with the physical properties of the walls during growth.
Plant J. 3(1), 1–30 (1993) 89
173. Carr, S.M., Jaffe, M.J.: Pith autolysis in herbaceous, dicotyledonous plants: Experimental
manipulation of pith autolysis in several cultivated species. Ann. Bot-London 75(6), 587–592
(1995) 483
174. Case, T.J.: On the evolution and adaptive significance of postnatal growth rates in the ter-
restrial vertebrates. Q. Rev. Biol. pp. 243–282 (1978) 155, 376
175. Casey, J.: On Volterra dislocations of finitely deforming continua. Math. Mech. Solids 9(5),
473–492 (2004) 358
176. Casey, J., Naghdi, P.M.: A remark on the use of the decomposition F = Fe F p in plasticity.
Trans. ASME 47, 672–675 (1980) 48, 401
177. Caspar, D.L.D.: Bacterial flagellar coiling explained by slip-and-click strand switching.
Nature Struct. Biol. 5, 92–94 (1998) 155
178. Castle, E.S.: Spiral growth and reversal of spiraling in Phycomyces, and their bearing on
primary wall structure. Amer. J. Bot. 29(8), 664–672 (1942) 321, 543, 544
179. Cermelli, P., Fried, E., Sellers, S.: Configurational stress, yield and flow in rate–independent
plasticity. Proc. Roy. Soc. Lond. A 457(2010), 1447–1467 (2001) 48
180. Cermelli, P., Gurtin, M.E.: On the characterization of geometrically necessary dislocations
in finite plasticity. J. Mech. Phys. Solids 49(7), 1539–1568 (2001) 358, 371, 372, 373
181. Chada, S., Lamoureux, P., Buxbaum, R.E., Heidemann, S.R.: Cytomechanics of neurite
outgrowth from chick brain neurons. J. Cell Sci. 110 ( Pt 10), 1179–86 (1997) 81
182. Champy, C.: Sexualité et hormones. Paris: Gaston Doin (1924) 21
183. Chater, K.F., Hopwood, D.A.: Streptomyces. In: A.L. Sonenshein, J.A. Hoch, R. Losick
(eds.) Streptomyces, vol. Bacillus subtilis and other gram-positive bacteria: biochemistry
and Molecular genetics. American Society for Microbiology (1993) 235
184. Chaudhry, H.R., Bukiet, B., Davis, A., Ritter, A.B., Findley, T.: Residual stresses in oscillat-
ing thoracic arteries reduce circumferential stresses and stress gradients. J. Biomech. 30(1),
57–62 (1997) 43, 510
185. Chaudhury, M.K., Chakrabarti, A., Ghatak, A.: Adhesion-induced instabilities and pattern
formation in thin films of elastomers and gels. Eur. Phys. J. A 38(7), 1–26 (2015) 573
186. Chaurasia, B.D., Agarwal, B.M.: Helical structure of the human umbilical cord. Acta Anat.
103 (1979) 154
187. Chavarrıa-Krauser, A., Jager, W., Schurr, U.: Primary root growth: a biophysical model of
auxin-related control. Funct. Plant Biol. 32, 849–862 (2005) 25, 96
188. Chebli, Y., Geitmann, A.: Mechanical principles governing pollen tube growth. Funct. Plant
Sc. Biotechnol. 1(2), 232–245 (2007) 12, 419
References 589
189. Checa, A.: A model for the morphogenesis of ribs in ammonites inferred from associated
microsculptures. Palaeontology 37(4), 863–888 (1994) 197
190. Checa, A.G., Crampton, J.S.: Mechanics of sculpture formation in Magadiceramus? ran-
gatira rangatira (Inoceramidae, Bivalvia) from the Upper Cretaceous of New Zealand.
Lethaia 35(4), 279–290 (2002) 197
191. Checa, A.G., Jiménez-Jiménez, A.P.: Rib fabrication in Ostreoidea and Plicatuloidea (Bi-
valvia, Pteriomorphia) and its evolutionary significance. Zoomorphology 122(3), 145–159
(2003) 197
192. Chen, H.H., Brodland, G.W.: Cell-level finite element studies of viscous cells in planar
aggregates. J. Biomech. Eng. 122(4), 394–401 (2000) 57
193. Chen, J.C.W.: The kinetics of tip growth in the Nitella rhizoid. Plant Cell Physiol. 14(4),
631–640 (1973) 25
194. Chen, X., Brodland, G.W.: Multi-scale finite element modeling allows the mechanics of
amphibian neurulation to be elucidated. Phys. Biol. 5(1), 015,003 (2008) 58
195. Chen, Y.: Singularity theory and nonlinear bifurcation analysis. London Math. Soc. Lecture
Note Ser. pp. 305–344 (2001) 311
196. Chen, Y., Hoger, A.: Constitutive functions of elastic materials in finite growth and defor-
mation. J. Elasticity 59, 175–193 (2000) 400, 438
197. Chen, Y.C., Haughton, D.: Stability and bifurcation of inflation of elastic cylinders. Proc.
Roy. Soc. Lond. A Mat. 459(2029), 137–156 (2003) 299, 311
198. Chen, Y.C., Healey, T.J.: Bifurcation to pear-shaped equilibria of pressurized spherical mem-
branes. Int. J. Non-Linear Mech. 26, 279–291 (1991) 304, 442, 456
199. Chenchiah, I.V., Shipman, P.D.: An energy-deformation decomposition for morphoelasticity.
J. Mech. Phys. Solids 67, 15–39 (2014) 563
200. Chernyavsky, I.L., Croisier, H., Chapman, L.A., Kimpton, L.S., Hiorns, J.E., Brook, B.S.,
Jensen, O.E., Billington, C.K., Hall, I.P., Johnson, S.R.: The role of inflammation resolution
speed in airway smooth muscle mass accumulation in asthma: insight from a theoretical
model. PloS one 9(3), e90,162 (2014) 538
201. Cherubini, C., Filippi, S., Nardinocchi, P., Teresi, L.: An electromechanical model of cardiac
tissue: Constitutive issues and electrophysiological effects. Prog. Biophys. Mol. Biol. 97(2),
562–573 (2008) 576
202. Chirat, R., Moulton, D.E., Goriely, A.: Mechanical basis of morphogenesis and convergent
evolution of spiny seashells. Proc. Natl. Acad. Sci. USA 110(15), 6015–6020 (2013) 8, 197,
204, 205, 570
203. Chladni, E.F.F.: Die Akustik. Breitkopf & Härtel (1830) 251
204. Cho, K., Gent, A.N.: Cavitation in model elastomeric composites. J. Mat. Sc. 23(1), 141–144
(1988) 451
205. Chouaieb, N., Goriely, A., Maddocks, J.H.: Helices. Proc. Natl. Acad. Sci. USA 103(25),
9398–9403 (2006) 100, 116, 117
206. Chow, B., Lu, P., Ni, L.: Hamilton’s Ricci flow, vol. 77. American Mathematical Soc. (2006)
579
207. Chuong, C., Fung, Y.: Compressibility and constitutive equation of arterial wall in radial
compression experiments. J. Biomech. 17(1), 35–40 (1984) 302
208. Chuong, C.J., Fung, Y.C.: Three-dimensional stress distribution in arteries. J. Biomech. Eng.
105, 268–274 (1983) 390, 512, 519
209. Ciarlet, P.G.: Mathematical elasticity Vol 1: Three dimensional elasticity. North Holland,
Amsterdam (1996) 256
210. Ciarlet, P.G.: Theory of shells, vol. 3. Elsevier (2000) 256
211. Ciarlet, P.G.: An introduction to differential geometry with applications to elasticity. J.
Elasticity 78(1-3), 1–215 (2005) 361
212. Cilley, F.: Fundamental propositions in the theory of elasticity; a study of primary or self-
balancing stresses. Amer. J. Sci. 11(64), 269–290 (1901) 45
213. Cisternas, J., Holmes, P.: Buckling of extensible thermoelastic rods. Math. Comput. Mod-
elling 36(3), 233–243 (2002) 98
590 References
214. Clark, R.B., Cowey, J.B.: Factors controlling the change of shape of certain nemertean and
turbellarian worms. J. Exp. Biol. 35(4), 731–748 (1958) 336
215. Clark, W.E.L.G., Medawar, P.B.: Essays on Growth and Form presented to D’Arcy Went-
worth Thompson. Oxford University Press (1945) 28
216. Clayton, J.D., Bammann, D.J., McDowell, D.L.: A geometric framework for the kinematics
of crystals with defects. Philos. Mag. 33(35), 3983–4010 (2005) 359
217. Coen, E., Rolland-Lagan, A.G., Matthews, M., Bangham, J.A., Prusinkiewicz, P.: The ge-
netics of geometry. Proc. Natl. Acad. Sci. USA 101(14), 4728–4735 (2004) 559
218. Cohen, R., Delbrück, M.: Distribution of stretch and twist along the growing zone of the
sporangiophore of Phycomyces and the distribution of response to a periodic illumination
program. J. Cell. Comparative Physiol. 52(3), 361–388 (1958) 550
219. Cohen, R., Delbrück, M.: Photoreactions in Phycomyces: Growth and tropic responses to
the stimulation of narrow test areas. J. Gen. Physiol. 42(4), 677–695 (1959) 542
220. Cohn, F.: Studies on bacteria. Contributions to the biology of plants (in German) 1, 127–222
(1875) 234
221. Coleman, B.D., Dill, E.H., Lembo, M., Lu, Z., Tobias, I.: On the dynamics of rods in the
theory of Kirchhoff and Clebsch. Arch. Rat. Mech. Anal. 121, 339–359 (1993) 66, 97
222. Coleman, B.D., Noll, W.: The thermodynamics of elastic materials with heat conduction and
viscosity. Arch. Rat. Mech. Anal. 13(1), 167–178 (1963) 383, 405
223. Conlon, I., Raff, M.: Size control in animal development. Cell 96(2), 235–244 (1999) 558
224. Conway, J.: The game of life. Scientific American 223(4), 4 (1970) 51
225. Cook, J.: Mathematical models for dermal wound healing: wound contraction and scar
formation. Ph.D. thesis, University of Washington (1995) 48
226. Cook, T.A.: The curves of life. Reprinted (DOVER-1979), London (1914) 116, 187
227. Cooper, S.J.: From Claude Bernard to Walter Cannon. Emergence of the concept of home-
ostasis. Appetite 51(3), 419–427 (2008) 409
228. Cortes, D.H., Lake, S.P., Kadlowec, J.A., Soslowsky, L.J., Elliott, D.M.: Characterizing the
mechanical contribution of fiber angular distribution in connective tissue: comparison of two
modeling approaches. Biomech. Model. Mechanobiol. 9(5), 651–658 (2010) 340, 342
229. Cosgrove, D.J.: Cell wall yield properties of growing tissue. evaluation by in vivo stress
relaxation. Plant Physiol. 78, 347–356 (1985) 93
230. Cosgrove, D.J.: Mechanism of rapid suppression of cell expansion in cucumber hypocotyls
after blue-light irradiation. Planta 176(1), 109–116 (1988) 92
231. Cosgrove, D.J.: Wall extensibility: its nature, measurement and relationship to plant cell
growth. New Phytol. 124(1), 1–23 (1993) 89, 91, 92
232. Cosgrove, D.J.: Growth of the plant cell wall. Nature Rev. Mol. Cell Biol. 6(11), 850–861
(2005) 90
233. Cosgrove, D.J., Green, P.B.: Rapid suppression of growth by blue light. Biophysical mech-
anism of action. Plant Physiol. 68(6), 1447–1453 (1981) 96
234. Cosgrove, D.J., Ortega, J.K.E., Shropshire Jr, W.: Pressure probe study of the water relations
of Phycomyces blakesleeanus sporangiophores. Biophys. J. 51(3), 413–423 (1987) 544
235. Cosserat, E., Cosserat, F.: Théories des Corps Déformables. Hermann, Paris (1909) 251
236. Coutand, C., Fournier, M., Moulia, B.: The gravitropic response of poplar trunks: key roles
of prestressed wood regulation and the relative kinetics of cambial growth versus wood
maturation. Plant Physiol. 144(2), 1166–1180 (2007) 25
237. Coutts, M., Grace, J.: Wind and trees. Cambridge University Press (1995) 28
238. Cowan, R.S.: A monograph of the genus Eperua (Leguminosae-Caesalpinioideae). Smith-
sonian Institution Press Washington, DC (1975) 38
239. Cowin, S., Holzapfel, G.A.: On the modeling of growth and adaptation. Holzapfel G. A.,
Odgen RW. Mechanics of biological tissue. New York: Springer pp. 29–46 (2006) 47
240. Cowin, S.C.: Strain or deformation rate dependent finite growth in soft tissues. J. Biomech.
29, 647–649 (1996) 48
241. Cowin, S.C.: How is a tissue built? J. Biomech. Eng. 122, 553–569 (2000) 49
References 591
242. Cowin, S.C.: Tissue growth and remodeling. Annu. Rev. Biomed. Eng. 6, 77–107 (2004) 9,
48, 49
243. Cowin, S.C.: On the Modeling of Growth and Adaptation. In: G.A. Holzapfel, R.W. Ogden
(eds.) Mechanics of Biological Tissue, pp. 29–46. Springer Verlag, New York (2006) 49
244. Cowin, S.C.: The specific growth rates of tissues: A review and a re-evaluation. J. Biomech.
Eng. 133(4), 041, 001–041, 021 (2011) 16, 49
245. Cowin, S.C., Doty, S.B.: Tissue Mechanics. Springer (2009) 395
246. Cowin, S.C., Hegedus, D.M.: Bone remodeling I: A theory of adaptive elasticity. J. Elasticity
6, 313–326 (1976) 47
247. Cowin, S.C., Van Buskirk, W.: Surface bone remodeling induced by a medullary pin. J.
Biomech. 12(4), 269–276 (1979) 47
248. Cox, R.H.: Anisotropic properties of the canine carotid artery in vitro. J. Biomech. 8(5),
293–300 (1975) 509, 519
249. Cox, R.W., Peacock, M.A.: The velocity field of growing ear cartilage. J. Anat. 126, 555–566
(1978) 25
250. Coyle, P.: Diameter and length changes in cerebral collaterals after middle cerebral artery
occlusion in the young rat. Anat. Rec. 210(2), 357 – 364 (1984) 510
251. Cramer, G.R., Bowman, D.C.: Kinetics of maize leaf elongation. I. Increased yield threshold
limits short-term, steady-state elongation rates after exposure to salinity. J. Exp. Bot. 42(244),
1417–1426 (1991) 31
252. Crampin, E.J., Gaffney, E.A., Maini, P.K.: Reaction and diffusion on growing domains:
scenarios for robust pattern formation. Bull. Math. Biol. 61(6), 1093–1120 (1999) 564, 565
253. Crampin, E.J., Hackborn, W.W., Maini, P.K.: Pattern formation in reaction-diffusion models
with nonuniform domain growth. Bull. Math. Biol. 64(4), 747–769 (2002) 564
254. Culmann, K.: Die Graphische Statik. Verlag Von Meyer & Zeller (1866) 34, 35
255. Cyron, C.J., Aydin, R.C., Humphrey, J.D.: A homogenized constrained mixture (and mechan-
ical analog) model for growth and remodeling of soft tissue. Biomech. Model. Mechanobiol.
pp. 1–15 (2016) 396
256. Daerden, F., Lefeber, D.: Pneumatic artificial muscles: actuators for robotics and automation.
Euro. J. Mech. Environ. Eng. 47(1), 11–21 (2002) 338
257. Dafermos, C.M.: The second law of thermodynamics and stability. Arch. Rat. Mech. Anal.
70(2), 167–179 (1979) 311
258. Dallon, J.C., Othmer, H.G.: How cellular movement determines the collective force generated
by the Dictyostelium discoideum slug. J. Theor. Biol. 231(2), 203–222 (2004) 56
259. Damuth, J.: Scaling of growth: plants and animals are not so different. Proc. Natl. Acad. Sci.
USA 98(5), 2113–2114 (2001) 358
260. Danks, D.M., Campbell, P.E., Stevens, B.J., Mayne, V., Cartwright, E.: Menkes’s kinky
hair syndrome; An inherited defect in copper absorption with widespread effects. Pediatrics
50(2), 188–201 (1972) 510
261. Darwin, C.: The Movements and Habits of Climbing Plants. Appleton, New York (1888)
47, 151, 152, 163
262. David, G., Humphrey, J.D.: Further evidence for the dynamic stability of intracranial saccular
aneurysms. J. Biomech. 36, 1143–1150 (2003) 304, 447
263. Davis, H.G.: Conservative surgery. D. Appleton & company (1867) 36, 37
264. Davis, N.P., Han, H.C., Wayman, B., Vito, R.: Sustained axial loading lengthens arteries in
organ culture. Ann. Biomed. Eng. 33(7), 867–877 (2005) 32, 510
265. Day, S.J., Lawrence, P.A.: Measuring dimensions: the regulation of size and shape. Devel-
opment 127(14), 2977–2987 (2000) 558
266. De Gennes, P.G., Brochard-Wyart, F., Quéré, D.: Capillarity and wetting phenomena: drops,
bubbles, pearls, waves. Springer Verlag (2004) 212
267. De Pascalis, R., Destrade, M., Goriely, A.: Nonlinear correction to the euler buckling formula
for compressed cylinders with guided-guided end conditions. J. Elasticity 102(2), 191–200
(2011) 294, 490
592 References
268. Dealy, J.M.: Weissenberg and Deborah numbers—their definition and use. Rheol. Bull.
79(2), 14–18 (2010) 555
269. Delfino, A., Stergiopulos, N., Moore, J.E., Meister, J.J.: Residual strain effects on the stress
field in a thick wall finite element model of the human carotid bifurcation. J. Biomech. 30(8),
777–786 (1997) 295, 512
270. Demiray, H.: A note on the elasticity of soft biological tissues. J. Biomech. 5(3), 309–311
(1972) 295
271. Demiray, H., Weizsäcker, H.W., Pascale, K., Erbay, H.A.: A stress-strain relation for a rat
abdominal aorta. J. Biomech. 21(5), 369–374 (1988) 295
272. Demirkoparan, H., Pence, T.J.: Torsional swelling of a hyperelastic tube with helically wound
reinforcement. J. Elasticity 92(1), 61–90 (2008) 324, 539, 545
273. Dennerll, T.J., Lamoureux, P., Buxbaum, R.E., Heidemann, S.R.: The cytomechanics of
axonal elongation and retraction. J. Cell Biol. 109(6 Pt 1), 3073–83 (1989) 82
274. Dera, G., Eble, G.J., Neige, P., David, B.: The flourishing diversity of models in theoreti-
cal morphology: from current practices to future macroevolutionary and bioenvironmental
challenges. Paleobiology 34(3), 301–317 (2008) 189
275. Dervaux, J., Ben Amar, M.: Morphogenesis of growing soft tissues. Phys. Rev. Lett. 101(6),
068,101 (2008) 46, 252, 256
276. Dervaux, J., Ben Amar, M.: Buckling condensation in constrained growth. J. Mech. Phys.
Solids 59(3), 538–560 (2011) 525
277. Dervaux, J., Ciarletta, P., Ben Amar, M.: Morphogenesis of thin hyperelastic plates: A
constitutive theory of biological growth in the Föppl-von Kármán limit. J. Mech. Phys.
Solids 57(3), 458–471 (2009) 46, 252, 258
278. Deseri, L., Piccioni, M.D., Zurlo, G.: Derivation of a new free energy for biological mem-
branes. Continuum Mech. Therm. 20(5), 255–273 (2008) 208
279. Deseri, L., Zurlo, G.: The stretching elasticity of biomembranes determines their line tension
and bending rigidity. Biomech. Model. Mechanobiol. 12(6), 1233–1242 (2013) 208
280. Destrade, M.: Rayleigh waves and surface stability for Bell materials in compression: com-
parison with rubber. Q. J. Mech. Appl. Math. 56, 593–604 (2003) 316, 317
281. Destrade, M., Ogden, R.W.: On the third-and fourth-order constants of incompressible
isotropic elasticity. J. Acoust. Soc. Am. 128, 3334–3343 (2010) 294
282. Détienne, P., Thiel, J.: Monographie des Wapas de Guyane Française. Bois For. Trop. 216,
43–68 (1988) 39
283. Devlin, K.: Do you believe in fairies, unicorns, or the BMI? Mathematical Association of
America (2009) 12
284. DiCarlo, A., Quiligotti, S.: Growth and balance. Mech. Res. Commun. 29(6), 449–456 (2002)
48
285. Dichmann, D.J., Li, Y., Maddocks, J.H.: Hamiltonian formulations and symmetries in rod
mechanics. IMA Vol. Math. Appl. 82, 71–114 (1996) 66, 97
286. Dinwoodie, J.: Growth stresses in timber–a review of literature. Forestry 39(2), 162–170
(1966) 42
287. Dodds, P.S., Rothman, D.H., Weitz, J.S.: Re-examination of the “3/4-law” of metabolism.
J. Theor. Biol. 209(1), 9–27 (2001) 15
288. Dolbow, J., Fried, E., Ji, H.: Chemically induced swelling of hydrogels. J. Mech. Phys. Solids
52(1), 51–84 (2004) 403
289. Doliwa, A., Santini, P.: The integrable dynamics of a discrete curve. In: V. L., P. Winternitz
(eds.) Symmetries and Integrability of Difference Equations (CRM Proc. and Lecture Notes
vol 9), pp. 91–102 (1996) 102
290. Dollhofer, J., Chiche, A., Muralidharan, V., Creton, C., Hui, C.: Surface energy effects for
cavity growth and nucleation in an incompressible neo-Hookean material—-modeling and
experiment. Int. J. Solids Struct. 41(22-23), 6111–6127 (2004) 451
291. Domokos, G., Healey, T.J.: Multiple helical perversions of finite, intristically curved rods.
Int. J. Bifurcat. Chaos 15(03), 871–890 (2005) 158
References 593
292. Donaldson, H.H.: A comparison of the albino rat with man in respect to the growth of the
brain and of the spinal cord. J. Comp. Neurol. Psych. 18(4), 345–392 (1908) 12
293. Donovan, G.M., Tawhai, M.H.: A simplified model of airway narrowing due to bronchial
mucosal folding. Respir. Physiol. Neurobiol. 171(2), 144–150 (2010) 532
294. Dorfmann, A., Fuller, K.N.G., Ogden, R.W.: Shear, compressive and dilatational response
of rubberlike solids subject to cavitation damage. Int. J. Solids and Struct. 39(7), 1845–1861
(2002) 451
295. Dorfmann, A., Haughton, D.M.: Stability and bifurcation of compressed elastic cylindrical
tubes. Int. J. Eng. Sci. 44, 1353–1365 (2006) 490
296. Dowaikh, M.A., Ogden, R.W.: On surface waves and deformations in a pre-stressed incom-
pressible elastic solid. IMA J. Appl. Math. 44, 261–284 (1990) 316
297. Drapaca, C.S., Fritz, J.S.: A mechano-electrochemical model of brain neuro-mechanics:
application to normal pressure hydrocephalus. Int. J. Num. Anal. Mod. Ser. B 1, 82–93
(2012) 395
298. Drasdo, D.: Buckling instabilities of one-layered growing tissues. Phys. Rev. Lett. 84(18),
4244 (2000) 55
299. Drasdo, D., Hoehme, S.: A single-cell-based model of tumor growth in vitro: monolayers
and spheroids. Phys. Biol. 2(3), 133–147 (2005) 55
300. Drasdo, D., Hoehme, S., Block, M.: On the role of physics in the growth and pattern formation
of multi-cellular systems: what can we learn from individual-cell based models? J. Stat. Phys.
128(1-2), 287–345 (2007) 55
301. Drasdo, D., Höhme, S.: Individual-based approaches to birth and death in avascular tumors.
Math. Comput. Modelling 37(11), 1163–1175 (2003) 55
302. Driessen, N.J.B., Boerboom, R.A., Huyghe, J.M., Bouten, C.V.C., Baaijens, F.P.T.: Com-
putational analyses of mechanically induced collagen fiber remodeling in the aortic heart
valve. J. Biomech. Eng. 125, 549 (2003) 419, 420
303. Driessen, N.J.B., Cox, M.A.J., Bouten, C.V.C., Baaijens, F.P.T.: Remodelling of the angular
collagen fiber distribution in cardiovascular tissues. Biomech. Model. Mechanobiol. 7(2),
93–103 (2008) 321, 328, 420
304. Driessen, N.J.B., Peters, G.W.M., Huyghe, J.M., Bouten, C.V.C., Baaijens, F.P.T.: Remod-
elling of continuously distributed collagen fibres in soft connective tissues. J. Biomech.
36(8), 1151–1158 (2003) 419, 420
305. Driessen, N.J.B., Wilson, W., Bouten, C.V.C., Baaijens, F.P.T.: A computational model for
collagen fibre remodelling in the arterial wall. J. Theor. Biol. 226(1), 53–64 (2004) 419, 420
306. Drozdov, A., Khanina, H.: A model for the volumetric growth of a soft tissue. Mathematical
and Computer Modelling 25(2), 11–29 (1997) 48
307. Drozdov, A.D.: Volumetric growth of viscoleastic solids. Mech. Solids 25, 99–106 (1990)
48
308. Dubois, E.: Sur le rapport du poids de l’encéphale avec la grandeur du corps chez les
mammifères. Bulletins de la Socit́é d’Anthropologie de Paris 8(1), 337–376 (1897) 20
309. Dubois, E.: Sur le rapport du poids de l’encéphale avec la grandeur du corps chez les
mammifères. L’année psychologique 5(1), 593–599 (1898) 20
310. Düchting, W., Vogelsaenger, T.: Recent progress in modelling and simulation of three-
dimensional tumor growth and treatment. Biosystems 18(1), 79–91 (1985) 52
311. Duka, E.D., England, A.H., Spencer, A.J.M.: Bifurcation of a solid circular elastic cylinder
under finite extension and torsion. Acta Mech. 98(1), 107–121 (1993) 299, 490
312. Dumais, J.: Plant Morphogenesis: A Role for Mechanical Signals. Curr. Biol. 19(5), R207–
R208 (2009) 566
313. Dumais, J., Long, S.R., Shaw, S.L.: The mechanics of surface expansion anisotropy in
Medicago truncatula root hairs. Plant Physiol. 136, 3266–3275 (2004) 234
314. Dumais, J., Shaw, S.L., Steele, C.R., Long, S.R., Ray, P.M.: An anisotropic-viscoplastic
model of plant cell morphogenesis by tip growth. Int. J. Develop. Biol. 50(2-3), 209–222
(2006) 240
594 References
315. Dumais, J., Steele, C.R.: New evidence for the role of mechanical forces in the shoot apical
meristem. J. Plant Growth Regul. 19(1), 7–18 (2000) 566
316. Dutrochet, M.: Recherche sur la volubilité des tiges de certains végétaux et sur la cause de
ce phénomène. C. R. Acad. Sci. 19, 295–303 (1844) 151
317. Dyson, R.J., Jensen, O.E.: A fibre-reinforced fluid model of anisotropic plant cell growth. J.
Fluid Mech. 655, 472–503 (2010) 96, 321
318. Dyson, R.J., Vizcay-Barrena, G., Band, L.R., Fernandes, A.N., French, A.P., Fozard, J.A.,
Hodgman, T.C., Kenobi, K., Pridmore, T.P., Stout, M., et al.: Mechanical modelling quantifies
the functional importance of outer tissue layers during root elongation and bending. New
Phytologist 202(4), 1212–1222 (2014) 96
319. Eastham, J., Oosterhuis, D.M., Walker, S.: Leaf water and turgor potential threshold values
for leaf growth of wheat. Agron. J. 76(5), 841–847 (1984) 91
320. Eastwood, M., Mudera, V.C., McGrouther, D.A., Brown, R.A.: Effect of precise mechan-
ical loading on fibroblast populated collagen lattices: morphological changes. Cell Motil.
Cytoskeleton 40(1), 13–21 (1998) 419
321. Eckart, C.: The thermodynamics of irreversible processes. IV. The theory of elasticity and
anelasticity. Phys. Rev. 73(4), 373 (1948) 86, 362, 363
322. Edmonds, H.W.: The spiral twist of the normal umbilical cord in twins and singletons. Am.
J. Obst. & Gynec. 67, 102–120 (1954) 154
323. Efrati, E., Klein, Y., Aharoni, H., Sharon, E.: Spontaneous buckling of elastic sheets with a
prescribed non-Euclidean metric. Physica D 235(1-2), 29–32 (2007) 252, 256
324. Efrati, E., Sharon, E., Kupferman, R.: Elastic theory of unconstrained non-Euclidean plates.
J. Mech. Phys. Solids 57(4), 762–775 (2009) 256
325. Eichenbaum, G.M., Kiser, P.F., Dobrynin, A.V., Simon, S.A., Needham, D.: Investigation of
the swelling response and loading of ionic microgels with drugs and proteins: The depen-
dence on cross-link density. Macromolecules 32(15), 4867–4878 (1999) 403
326. Eisenberger, M., Clastornik, J.: Vibrations and buckling of a beam on a variable winkler
elastic foundation. J. Sound Vib. 115(2), 233–241 (1987) 143
327. Elkin, B.S., Shaik, M.A., Morrison 3rd, B.: Fixed negative charge and the donnan effect: a
description of the driving forces associated with brain tissue swelling and oedema. Philos.
Trans. Roy. Soc. A 368(1912), 585–603 (2010) 395
328. Ellers, O.: A mechanical model of growth in regular sea urchins: predictions of shape and a
developmental morphospace. Proc. Roy. Soc. Lond. B 254(1340), 123–129 (1993) 217
329. Ellers, O., Telford, M.: Causes and consequences of fluctuating coelomic pressure in sea
urchins. Biol. B. 182(3), 424–434 (1992) 217
330. Elsgaard, L., Prieur, D.: Hydrothermal vents in Lake Tanganyika harbor spore-forming ther-
mophiles with extremely rapid growth. J. Great Lakes Re. 37(1), 203–206 (2011) 376
331. Elżanowski, M., Preston, S.: A model of the self-driven evolution of a defective continuum.
Math. Mech. Solids 12(4), 450–465 (2007) 361
332. England, J.L.: Statistical physics of self-replication. J. Chem. Phys. 139(12), 121,923 (2013)
376
333. Enlow, D.H.: Facial growth. WB Saunders Company (1990) 571
334. Ente, G., Penzer, P.H.: The umbilical cord: normal parameters. J. R. Soc. Health August
(1991) 154
335. Entov, V.M.: Mechanical model of scoliosis. Mech. Solids 18, 199–206 (1983) 48, 63
336. Epstein, M.: A question of constant strain. J. Elasticity 17(1), 23–34 (1987) 361
337. Epstein, M.: The geometrical language of continuum mechanics. Cambridge University
Press (2010) 361
338. Epstein, M.: Kinetics of boundary growth. Mech. Res. Commun. 37(5), 453–457 (2010)
175, 572
339. Epstein, M.: The elements of continuum biomechanics. Wiley. com (2012) 49, 377, 380
340. Epstein, M., Goriely, A.: Self-diffusion in remodeling and growth. ZAMP 63(2), 339–355
(2012) 378, 394, 569
References 595
341. Epstein, M., Maugin, G.A.: Thermomechanics of volumetric growth in uniform bodies. Int.
J. Plasticity 16(7-8), 951–978 (2000) 48, 377, 408
342. Ericksen, J.L.: Deformations possible in every isotropic, incompressible, perfectly elastic
body. ZAMP 5(6), 466–489 (1954) 305
343. Ericksen, J.L.: Deformations possible in every compressible, isotropic, perfectly elastic
material. J. Math. Phys 34, 126–128 (1955) 305
344. Erickson, R.O., Sax, K.B.: Elemental growth rate of the primary root of Zea mays. Proc.
Am. Philos. Soc. 100(5), 487–498 (1956) 25
345. Erickson, R.O., Sax, K.B.: Rates of cell division and cell elongation in the growth of the
primary root of Zea mays. Proc. Am. Philos. Soc. 100(5), 499–514 (1956) 25
346. Eringen, A.C., Ingram, J.D.: A continuum theory of chemically reacting media—I. Int. J.
Eng. Sci 3(2), 197–212 (1965) 394
347. Ermentrout, G.B., Edelstein-Keshet, L.: Cellular automata approaches to biological model-
ing. J. Theor. Biol. 160(1), 97–133 (1993) 51
348. Eshelby, J.D.: The determination of the elastic field of an ellipsoidal inclusion, and related
problems. Proc. Roy. Soc. Lond. A 241(1226), 376–396 (1957) 400, 476
349. Eskandari, M., Kuhl, E.: Systems biology and mechanics of growth. Wiley Interdisciplinary
Reviews: Systems Biology and Medicine 7(6), 401–412 (2015) 49
350. Euler, L.: Methodus inveniendi lineas curvas maximi minimive proprietate gaudentes sive
solutio problematis isoperimetrici latissimo sensu accepti. apud Marcum-Michaelem Bous-
quet & socios (1744) 123, 487
351. Euler, L.: Sur la force des colonnes. Mem. Acad. , Berlin 13 (1759) 487, 488
352. Evans, D.E.: Aerenchyma formation. New Phytol. 161, 35–49 (2003) 481
353. Evans, E.A., Skalak, R.: Mechanics and thermodynamics of biomembranes. CRC Press, Inc,
Boca Raton, Florida (1980) 208, 214, 220, 222
354. Evans, N.H., McAinsh, M.R., Hetherington, A.M.: Calcium oscillations in higher plants.
Curr. Opin. Plant Biol. 4(5), 415–420 (2001) 12, 419
355. Farge, E.: Mechanical induction of twist in the Drosophila foregut/stomodeal primordium.
Curr. Biol. 13(16), 1365–1377 (2003) 559
356. Farhadifar, R., Röper, J.C., Aigouy, B., Eaton, S., Jülicher, F.: The influence of cell mechanics,
cell-cell interactions, and proliferation on epithelial packing. Current Biology 17(24), 2095–
2104 (2007) 57
357. Faruk Senan, N.A., O’Reilly, O.M., Tresierras, T.N.: Modeling the growth and branching of
plants: A simple rod-based model. J. Mech. Phys. Solids 56(10), 3021–3036 (2008) 97
358. Federico, S., Gasser, T.C.: Nonlinear elasticity of biological tissues with statistical fibre
orientation. J. R. Soc. Interface 7(47), 955–966 (2010) 340
359. Federico, S., Herzog, W.: Towards an analytical model of soft biological tissues. J. Biomech.
41(16), 3309–3313 (2008) 562
360. Feijo, J.A., Sainhas, J., Holdaway-Clarke, T., Cordeiro, M.S., Kunkel, J.G., Hepler, P.K.:
Cellular oscillations and the regulation of growth: the pollen tube paradigm. BioEssays
23(1), 86–94 (2001) 12, 419
361. Fernandez-Sanchez, M., Brunet, T., Röper, J., Farge, E.: Mechanotransduction’s impact on
animal development, evolution, and tumorigenesis. Annu. Rev. Cell Dev. Biol. 31, 373–397
(2015) 559
362. Filas, B.A., Xu, G., Taber, L.A.: Mechanisms of brain morphogenesis. In: Computer Models
in Biomechanics, pp. 337–349. Springer (2013) 573
363. Finn, R.: The sessile liquid drop. I. Symmetric case. Pacific J. Math 88, 541–587 (1980) 217
364. Finn, R.: Equilibrium capillary surfaces. Springer (1986) 217
365. Firn, R., Digby, J.: The role of the peripheral cell layers in the geotropic curvature of sunflower
hypocotyls: a new model of shoot geotropism. Aust. J. Plant Physiol. 4, 337–347 (1977) 42
366. Fletcher, D.A., Geissler, P.L.: Active biological materials. Annu. Rev. Phys. Chem. 60, 469–
486 (2009) 574
367. Flory, P.J.: Network structure and the elastic properties of vulcanized rubber. Chem. Rev.
35(1), 51–75 (1944) 293
596 References
368. Flory, P.J.: Statistical thermodynamics of semi-flexible chain molecules. Proc. Roy. Soc.
Lond. A 234(1196), 60–73 (1956) 293, 353
369. Flory, P.J.: Thermodynamic relations for high elastic materials. Trans. Faraday Soc. 57,
829–838 (1961) 403
370. Flory, P.J.: Network topology and the theory of rubber elasticity. Brit. Polym. J. 17(2), 96–102
(1985) 293
371. Fond, C.: Cavitation criterion for rubber materials: a review of void-growth models. J. Poly-
mer Sci. B: Polymer Phys. 39, 2081–2096 (2001) 451
372. Föppl, A.: Vorlesungen über technische Mechanik, vol. 5. BG Teubner (1907) 256
373. Fosdick, R.L., Schuler, K.W.: On Ericksen’s problem for plane deformations with uniform
transverse stretch. International Journal of Engineering Science 7(2), 217–233 (1969) 305
374. Fosdick, R.L., Shield, R.T.: Small bending of a circular bar superposed on finite extension
or compression. Arch. Ration. Mech. Anal. 12(1), 223–248 (1963) 490
375. Foster, K.W., Lipson, E.: The light growth response of Phycomyces. J. Gen. Microbiol.
62(5), 590–617 (1973) 542
376. Foty, R.A., Steinberg, M.S.: The differential adhesion hypothesis: a direct evaluation. De-
velopmental biology 278(1), 255–263 (2005) 51
377. Fournier, M., Bailleres, H., Chanson, B.: Tree biomechanics: growth, cumulative prestresses,
and reorientations. Biomimetics 2 (1994) 507
378. Fournier, M., Bordonne, P.A., Guitard, D., Okuyama, T.: Growth stress patterns in tree stems.
Wood Sci. Technol. 24(2), 131–142 (1990) 64, 507
379. Fournier, M., Chanson, B., Thibaut, B., Guitard, D.: Mechanics of standing trees: modelling
a growing structure submitted to continuous and fluctuating loads. 2. Tridimensional analy-
sis of maturation stresses. Case of standard hardwood [cambial growth]. In: Annales des
Sciences Forestières, vol. 48 (1991) 42, 507
380. Fournier, M., Stokes, A., Coutand, C., Fourcaud, T., Moulia, B., et al.: Tree biomechanics
and growth strategies in the context of forest functional ecology. Ecology and biomechanics:
a mechanical approach to the ecology of animals and plants. Boca Raton, Taylor & Francis
Group pp. 1–33 (2006) 64
381. Fowler, D.R., Meinhardt, H., Prusinkiewicz, P.: Modeling seashells. SIGGRAPH Comput.
Graph. 26(2), 379–387 (1992) 189, 192
382. Franceschini, P., Guala, A., Licata, D., Di Cara, G., Franceschini, D.: Arterial tortuosity
syndrome. Am. J. Med. Genet. 91(2), 141–143 (2000) 510
383. Frankel, T.: The geometry of physics: an introduction. Cambridge University Press (2011)
368
384. Franko, A.J., Freedman, H.I.: Model of diffusion of oxygen to spheroids grown in stationary
medium—I. Complete spherical symmetry. Bull. Math. Biol. 46(2), 205–217 (1984) 38, 72
385. Franko, A.J., Sutherland, R.M.: Oxygen diffusion distance and development of necrosis in
multicell spheroids. Radiat. Res. pp. 439–453 (1979) 72
386. Fratzl, P.: Collagen: structure and mechanics, an introduction. In: P. Fratzl (ed.) Collagen:
structure and mechanics, pp. 1–13. Springer Verlag (2008) 320
387. Fratzl, P., Burgert, I., Gupta, H.S.: On the role of interface polymers for the mechanics of
natural polymeric composites. Phys. Chem Ch. Ph. 6(24), 5575–5579 (2004) 321
388. Freed, A.D., Einstein, D.R., Vesely, I.: Invariant formulation for dispersed transverse isotropy
in aortic heart valves. Biomech. Model. Mechanobiol. 4(2), 100–117 (2005) 340
389. Frei, E., Preston, R.D.: Cell wall organization and wall growth in the filamentous green algae
Cladophora and Chaetomorpha. I. The basic structure and its formation. Proc. Roy. Soc.
Lond. B pp. 70–94 (1961) 321
390. Frenkel, D.: Order through entropy. Nature materials 14(1), 9–12 (2015) 382
391. Freyer, J.P., Fink, N.H., Schor, P.L., Coulter, J.R., Neeman, M., Sillerud, L.O.: A system for
viably maintaining a stirred suspension of multicellular spheroids during NMR spectroscopy.
NMR Biomed. 3, 195–205 (1990) 38, 72
392. Fried, E.: An elementary molecular-statistical basis for the Mooney and Rivlin–Saunders
theories of rubber elasticity. J. Mech. Phys. Solids 50(3), 571–582 (2002) 294
References 597
393. Friedman, A.: Cancer models and their mathematical analysis. Lecture Notes in Math. 1872,
223 (2006) 38, 72
394. Friesecke, G., James, R.D., Müller, S.: The Föppl–von Kármán plate theory as a low energy
-limit of nonlinear elasticity. Comptes Rendus Mathematique 335(2), 201–206 (2002) 256
395. Fryba, L.: History of Winkler foundation. Vehicle System Dynamics 24(sup1), 7–12 (1995)
145
396. Fu, J.: Chinese moso bamboo: its importance. Bamboo 22(5), 5–7 (2001) 376
397. Fu, Y.: Some asymptotic results concerning the buckling of a spherical shell of arbitrary
thickness. Int. J. Non-linear Mech. 33, 1111–1122 (1998) 456, 462, 465
398. Fu, Y., Pearce, S., Liu, K.K.: Post-bifurcation analysis of a thin-walled hyperelastic tube
under inflation. Int. J. Nonlinear Mech. 43(8), 697–706 (2008) 304
399. Fu, Y., Rogerson, G.A., Zhang, Y.: Initiation of aneurysms as a mechanical bifurcation
phenomenon. Int. J. Nonlinear Mech. 47(2), 179–184 (2012) 304
400. Fu, Y., Xie, Y.: Stability of localized bulging in inflated membrane tubes under volume
control. Int. J. Eng. Sci. 48(11), 1242–1252 (2010) 304
401. Fu, Y.B., Cai, Z.X.: An asymptotic analysis of the period-doubling secondary bifurcation in
a film/substrate bilayer. SIAM J. Appl. Math. 75(6), 2381–2395 (2015) 573
402. Fu, Y.B., Ogden, R.W.: Nonlinear Elasticity. Theory and Applications. Cambridge University
Press, Cambridge (2001) 311
403. Fuhrhop, J.H., Helfrich, W.: Fluid and solid fibers made of lipid molecular bilayers. Chem.
Rev. 93, 1565–1582 (1993) 210
404. Fung, Y.C.: Elasticity of soft tissues in simple elongation. Am. J. Physiol. 213(6), 1532–1544
(1967) 295
405. Fung, Y.C.: Biorheology of soft tissues. Biorheology 10(2), 139–155 (1973) 295
406. Fung, Y.C.: What principle governs the stress distribution in living organs? In: Y.C. Fung,
E. Fukada, J.J. Wang (eds.) Biomechanics in China, Japan and USA, pp. 1–13. Science Press,
Beijing (1983) 43, 390
407. Fung, Y.C.: What are the residual stresses doing in our blood vessels? Ann. Biomed. Eng.
19(3), 237–249 (1991) 43, 475
408. Fung, Y.C.: Biomechanics: mechanical properties of living tissues. Springer, New York
(1993) 43, 221, 320, 510
409. Fung, Y.C.: Stress, strain, growth, and remodeling of living organisms. ZAMP 46(special
issue) (1995) 392, 400
410. Fung, Y.C., Fronek, K., Patitucci, P.: Pseudoelasticity of arteries and the choice of its math-
ematical expression. Am. J. Physiol. -Heart C. 237(5), 620–631 (1979) 522
411. Fung, Y.C., Liu, S.Q.: Change of residual strains in arteries due to hypertrophy caused by
aortic constriction. Circulation Res. 65(5), 1340–1349 (1989) 409
412. Gager, C.S.: Fundamentals of Botany. P. Blakiston’s Son & Co. (1916) 39, 47, 499
413. Gal, D., Dvorkin, J.: Stresses in anisotropic cylinders. Mech. Res. Commun. 22(2), 109–113
(1995) 450
414. Galileo, G.: Discorsi e dimostrazioni matematiche intorno a due nuove scienze: attenenti
alla Mechanica & i Movimenti Locali. Reprinted (Leiden: Elsevier, 1914) (1638) 17, 19
415. Galland, P.: Tropisms of Avena coleoptiles: sine law for gravitropism, exponential law for
photogravitropic equilibrium. Planta 215(5), 779–784 (2002) 65
416. Galland, P., Finger, H., Wallacher, Y.: Gravitropism in Phycomyces: threshold determination
on a clinostat centrifuge. J. Plant Physiol. 161(6), 733–9 (2004) 542
417. Galloway, J.W.: Helical paradigm: Imperative of form and function. Encyclopedia of life
sciences (2001) 116
418. Ganghoffer, J.F., Haussy, B.: Mechanical modeling of growth considering domain variation.
part i: constitutive framework. Int. J. Solids Struct. 42(15), 4311–4337 (2005) 363
419. Ganghoffer, J.F., Plotnikov, P.I., Sokołowski, J.: Mathematical modeling of volumetric ma-
terial growth. Arch. Appl. Mech. 84(9-11), 1357–1371 (2014) 579
420. Gardella, R., Zoppi, N., Assanelli, D., Muiesan, M.L., Barlati, S., Colombi, M.: Exclusion of
candidate genes in a family with arterial tortuosity syndrome. Am. J. Med. Genet. 126A(3),
221–228 (2004) 510
598 References
421. Garikipati, K.: The kinematics of biological growth. Appl. Mech. Rev. 62, 030,801 (2009)
49, 175
422. Garikipati, K., Arruda, E., Grosh, K., Narayanan, H., Calve, S.: A continuum treatment of
growth in biological tissue: the coupling of mass transport and mechanics. J. Mech. Phys.
Solids 52(7), 1595–1625 (2004) 48, 408
423. Garikipati, K., Narayanan, H., Arruda, E.M., Grosh, K., Calve, S.: Material forces in the
context of biotissue remodelling. Adv. Mech. Math. pp. 77–84 (2005) 386
424. Garikipati, K., Olberding, J., Narayanan, H., Arruda, E., Grosh, K., Calve, S.: Biological
remodelling: Stationary energy, configurational change, internal variables and dissipation.
425. Gasser, T.C., Auer, M., Labruto, F., Swedenborg, J., Roy, J.: Biomechanical rupture risk
assessment of abdominal aortic aneurysms: model complexity versus predictability of finite
element simulations. Europ. J. Vasc. and Endovasc. Surg. 40(2), 176–185 (2010) 304
426. Gasser, T.C., Ogden, R.W., Holzapfel, G.A.: Hyperelastic modelling of arterial layers with
distributed collagen fibre orientations. J. R. Soc. Interface 3(6), 15–35 (2006) 45, 340, 342,
344
427. Gayon, J.: History of the concept of allometry. American Zoologist 40(5), 748–758 (2000)
21
428. Geitmann, A.: Plant and fungal cytomechanics: quantifying and modeling cellular architec-
ture. Can. J. Bot. 84, 581–593 (2006) 234
429. Geitmann, A., Emons, A.M.C.: The cytoskeleton in plant and fungal cell tip growth. J.
Microscopy 198(3), 218–245 (2000) 238
430. Geitmann, A., Ortega, J.K.E.: Mechanics and modeling of plant cell growth. Trends Plant
Sci. 14(9), 467–478 (2009) 234
431. Geitmann, A., Steer, M.: The architecture and properties of the pollen tube cell wall. In: The
Pollen Tube, pp. 177–200. Springer (2006) 5
432. Gent, A.: A new constitutive relation for rubber. Rubber Chem. and Technol. 69 (1996) 295,
318
433. Gent, A.: Elastic instabilities of inflated rubber shells. Rubber Chem. Technol. 72, 263–268
(1999) 304, 442, 447
434. Gent, A., Lindley, P.: Internal Rupture of Bonded Rubber Cylinders in Tension. Proc. Roy.
Soc. Lond. A. 249(1257), 195–205 (1959) 451
435. Gent, A.N.: Elastic instabilities in rubber. Int. J. Non-Linear Mech. 40, 165–175 (1995) 295
436. Genzer, J., Groenewold, J.: Soft matter with hard skin: From skin wrinkles to templating
and material characterization. Soft Matter 2(4), 310–323 (2006) 573
437. Germain, S.: Recherches sur la théorie des surfaces élastiques. Mme. Ve. Courcier, Paris
(1821) 251
438. Geymonat, G., Müller, S., Triantafyllidis, N.: Homogenization of nonlinearly elastic materi-
als, microscopic bifurcation and macroscopic loss of rank-one convexity. Arch. Rat. Mech.
Anal 122(3), 231–290 (1993) 562
439. Gibson, L.J.: Biomechanics of cellular solids. J. Biomech. 38(3), 377–399 (2005) 562
440. Gierer, A., Meinhardt, H.: A theory of biological pattern formation. Kybernetik 12(1), 30–39
(1972) 564
441. Gillespie, P.G., Walker, R.G., et al.: Molecular basis of mechanosensory transduction. Nature
pp. 194–202 (2001) 36
442. Giri, B., Fujisaki, K., Todoh, M., Tadano, S.: Residual stress estimation of bone tissue using
X-ray imaging technique. J. Biomech. 40(2), 371 (2007) 44
443. Giri, B., Tadano, S., Fujisaki, K., Todoh, M.: Understanding site-specific residual strain and
architecture in bovine cortical bone. J. Biomech. 41(15), 3107–3115 (2008) 45
444. Girousse, C., Moulia, B., Silk, W., Bonnemain, J.L.: Aphid infestation causes different
changes in carbon and nitrogen allocation in alfalfa stems as well as different inhibitions of
longitudinal and radial expansion. Plant Physiol. 137(4), 1474–1484 (2005) 25
445. Glaser, O.C.: On the mechanism of morphological differentiation in the nervous system. I.
the transformation of a neural plate into a neural tube. Anat. Rec. 8(12), 525–551 (1914) 29
References 599
446. Glazier, D.S.: A unifying explanation for diverse metabolic scaling in animals and plants.
Biol. Rev. 85(1), 111–138 (2010) 21
447. Gleason, R.L., Humphrey, J.D.: A mixture model of arterial growth and remodeling in
hypertension: altered muscle tone and tissue turnover. J. Vasc. Res. 41, 352–363 (2004) 48,
396
448. Gleason, R.L., Humphrey, J.D.: Effects of a sustained extension on arterial growth and
remodeling: a theoretical study. J. Biomech. 38(6), 1255–1261 (2005) 32, 396, 510
449. Göktepe, S., Abilez, O.J., Kuhl, E.: A generic approach towards finite growth with examples
of athlete’s heart, cardiac dilation, and cardiac wall thickening. J. Mech. Phys. Solids 58(10),
1661–1680 (2010) 34, 402
450. Göktepe, S., Abilez, O.J., Parker, K.K., Kuhl, E.: A multiscale model for eccentric and
concentric cardiac growth through sarcomerogenesis. J. Theor. Biol. 265(3), 433–442 (2010)
34, 402
451. Goldstein, R.E., Goriely, A.: Dynamic buckling of morphoelastic filaments. Phys. Rev. E
74, 010,901 (2006) 97, 100, 125, 130, 131
452. Goldstein, R.E., Goriely, A., Hubber, G., Wolgemuth, C.: Bistable helices. Phys. Rev. Lett.
84 (2000) 136, 155
453. Goldstein, R.E., Petrich, D.M.: The Korteweg-de Vries hierarchy as dynamics of closed
curves in the plane. Phys. Rev. Lett. 67 (1991) 102
454. Goldstein, S.F., Charon, N.W.: Motility of the spirochete Leptospira. Cell Motil. Cytoskele-
ton 9(2), 101–10 (1988) 116, 156
455. Gompertz, B.: On the nature of the function expressive of the law of human mortality, and
on a new mode of determining the value of life contingencies. Phil. Trans. R. Soc. Lond.
115, 513–583 (1825) 16
456. Gonzalez, O., Stuart, A.M.: A First Course in Continuum Mechanics. Cambridge University
Press (2007) 350
457. Gooday, G.W., Trinci, A.P.J.: The eukaryotic microbial cell. In: G.W. Gooday, D. Lloyd,
A.P.J. Trinci (eds.) Wall structure and biosynthesis in fungi, vol. 30, pp. 207–251. Cambridge
University Press, London (1980) 5, 237
458. Gordeeva, I.A., Panzhenskii, V.I., Stepanov, S.E.: Riemann–Cartan manifolds. J. Math. Sc.
169(3), 342–361 (2010) 367
459. Gore, J., Bryant, Z., Nöllmann, M., Le, M.U., Cozzarelli, N.R., Bustamante, C.: DNA over-
winds when stretched. Nature 442(7104), 836–839 (2006) 119, 120
460. Goriely, A.: Integrability and Nonintegrability of Dynamical Systems. World Scientific Pub-
lishing Company (2001) 13
461. Goriely, A.: Knotted umbilical cords. In: J.A. Calvo, A. Stasiak, E. Rawdon (eds.) Physical
and Numerical Models in Knot Theory Including Their Application to the Life Sciences, pp.
109–126. World Scientific, Singapore (2004) 63, 154
462. Goriely, A.: Twisted elastic rings and the rediscoveries of Michell’s instability. J. Elasticity
84(3), 281–299 (2006) 133, 138
463. Goriely, A., Ben Amar, M.: Differential growth and instability in elastic shells. Phys. Rev.
Lett. 94(19), 198,103 (2005) 46, 400
464. Goriely, A., Ben Amar, M.: On the definition and modeling of incremental, cumulative, and
continuous growth laws in morphoelasticity. Biomech. Model. Mechanobiol. 6(5), 289–296
(2007). 10. 1007/s10237-006-0065-7 46, 400
465. Goriely, A., Destrade, M., Ben Amar, M.: Instabilities in elastomers and in soft tissues. Q.
J. Mech. Appl. Math. 59(4), 615–630 (2006) 311, 443, 456, 461
466. Goriely, A., Geers, M.G.D., Holzapfel, G.A., Jayamohan, J., Jérusalem, A., Sivaloganathan,
S., Squier, W., van Dommelen, J.A.W., Waters, S.L., Kuhl, E.: Mechanics of the brain:
perspectives, challenges, and opportunities. Biomech. Model. Mechanobiol. 14, 931 (2015)
22
467. Goriely, A., Karolyi, G., Tabor, M.: Growth induced curve dynamics for filamentary micro-
organisms. J. Math. Biol. 51(3), 355–366 (2005). 10. 1007/s00285-005-0328-4 239
600 References
468. Goriely, A., Moulton, D.E.: Morphoelasticity - a theory of elastic growth. In: O.U. Press
(ed.) New Trends in the Physics and Mechanics of Biological Systems (2010) 46, 127
469. Goriely, A., Moulton, D.E., Vandiver, R.: Elastic cavitation, tube hollowing, and differential
growth in plants and biological tissues. Europhys. Lett. 91, 18,001 (2010) 452, 480
470. Goriely, A., Neukirch, S.: Mechanics of climbing and attachment in twining plants. Phys.
Rev. Lett. 97(18), 184,302 (2006) 65, 97, 116, 125, 164, 165, 167, 168
471. Goriely, A., Nizette, M.: Kovalevskaya rods and Kovalevskaya waves. Regul. Chaotic Dyn.
5, 95–106 (2000) 115
472. Goriely, A., Nizette, M., Tabor, M.: On the dynamics of elastic strips. J. Nonlinear Sci. 11(1),
3–45 (2001) 136
473. Goriely, A., Robertson-Tessi, M., Tabor, M., Vandiver, R.: Elastic growth models. In:
R. Mondaini, P.M. Pardalos (eds.) Mathematical Modelling of Biosystems, Applied Op-
timization, vol. 12. Springer (2008) 63, 97
474. Goriely, A., Tabor, M.: Nonlinear dynamics of filaments I: Dynamical instabilities. Phys. D
105, 20–44 (1997) 136, 138
475. Goriely, A., Tabor, M.: Nonlinear dynamics of filaments II: Nonlinear analysis. Phys. D 105,
45–61 (1997) 136
476. Goriely, A., Tabor, M.: Nonlinear dynamics of filaments IV: The spontaneous looping of
twisted elastic rods. Proc. Roy. Soc. Lond. A 455, 3183–3202 (1998) 149
477. Goriely, A., Tabor, M.: Spontaneous helix-hand reversal and tendril perversion in climbing
plants. Phys. Rev. Lett. 80, 1564–1567 (1998) 63, 97, 136, 151, 159
478. Goriely, A., Tabor, M.: The nonlinear dynamics of filaments. Nonlinear Dynam. 21(1), 101–
133 (2000) 63, 66, 97, 134, 136
479. Goriely, A., Tabor, M.: Biomechanical models of hyphal growth in actinomycetes. J. Theor.
Biol. 222, 211–218 (2003) 5, 208, 242
480. Goriely, A., Tabor, M.: Self-similar tip growth in filamentary organisms. Phys. Rev. Lett.
90(10), 108,101 (2003) 5, 208, 234
481. Goriely, A., Tabor, M.: Estimates of biomechanical forces in Magnaporthe grisea. Mycol.
Res. 110(Pt 7), 755–759 (2006). 10. 1016/j. mycres. 2006. 03. 014 226
482. Goriely, A., Tabor, M.: Biomechanical modeling of growing filamentous cells. Fungal Biol.
Rev. 22, 77–83 (2008) 5
483. Goriely, A., Tabor, M.: Spontaneous rotational inversion in Phycomyces. Phys. Rev. Lett.
106(13), 138,103 (2011) 65, 121, 321, 544
484. Goriely, A., Tabor, M.: Rotation, inversion and perversion in anisotropic elastic cylindrical
tubes and membranes. Proc. Roy. Soc. Lond. A 469(2153) (2013) 299, 302, 338
485. Goriely, A., Vandiver, R.: On the mechanical stability of growing arteries. IMA J. Appl.
Math. 75(4), 549–570 (2010) 46, 475, 511
486. Goriely, A., Vandiver, R., Destrade, M.: Nonlinear Euler buckling. Proc. Roy. Soc. Lond. A
464(2099), 3003–3019 (2008) 46, 294, 490, 492
487. Goriely, A., Weickenmeier, J., Kuhl, E.: Stress singularities in swelling soft solids. Phys.
Rev. Lett. (2016) 569
488. Gould, S.J.: Allometry and size in ontogeny and phylogeny. Biol. Rev. 41(4), 587–638 (1966)
20
489. Gould, S.J.: Geometric similarity in allometric growth: a contribution to the problem of
scaling in the evolution of size. Am. Nat. pp. 113–136 (1971) 20
490. Gould, S.J.: An allometric interpretation of species-area curves: the meaning of the coeffi-
cient. Am. Nat. pp. 335–343 (1979) 20
491. Gould, S.J.: Covariance sets and ordered geographic variation in Cerion from Aruba, Bonaire
and Curaçao: a way of studying nonadaptation. Syst. Zool. 33, 217–237 (1984) 150
492. Grabovsky, Y., Harutyunyan, D.: Rigorous derivation of the formula for the buckling load
in axially compressed circular cylindrical shells. J. Elasticity 120(2), 1–28 (2015) 493
493. Gracheva, M.E., Othmer, H.G.: A continuum model of motility in ameboid cells. Bull. Math.
Biol. 66(1), 167–193 (2004) 574
References 601
494. Graham, B., van Ooyen, A.: Mathematical modelling and numerical simulation of the mor-
phological development of neurons. BMC Neurosci. 7, S9 (2006) 85
495. Gray, A.: Note on the coiling of tendrils. Proc. Am. Acad. Arts Sci. 4 (1858) 152
496. Gray, A.: Modern differential geometry of curves and surfaces with Mathematica, 2nd edn.
CRC Press, Boca Raton (1998) 209, 254
497. Gray, D.I., Gooday, G.W., Prosser, J.I.: Apical hyphal extension in Streptomyces coelicolor
A3(2). J. Gen. Microbiol 136, 1077 (1990) 237, 241
498. Gray, H.: Anatomy of the human body. Lea & Febiger (1918) 530
499. Green, A., Naghdi, P.: Non-isothermal theory of rods, plates and shells. Int. J. Solids Struct.
6(2), 209–244 (1970) 98
500. Green, A., Naghdi, P.: On thermal effects in the theory of rods. Int. J. Solids Struct. 15(11),
829–853 (1979) 98
501. Green, A., Zerna, W.: The equilibrium of thin elastic shells. Q. J. Mech. Appl. Math. 3(1),
9–22 (1950) 251
502. Green, A.A., Kennaway, J.R., Hanna, A.I., Bangham, J.A., Coen, E.: Genetic control of
organ shape and tissue polarity. PLoS Bio. 8(11), e1000,537 (2010) 559
503. Green, A.E., Naghdi, P.M.: On basic equations for mixtures. Q. J. Mech. Appl. Math. 22(4),
427–438 (1969) 394
504. Green, A.E., Zerna, W.: Theoretical Elasticity. Dover, New York (1992) 298, 318
505. Green, P., Erickson, R., Buggy, J.: Metabolic and Physical Control of Cell Elongation Rate
In Vivo Studies in Nitella 1. Plant Physiol. 47(3), 423–430 (1971) 91, 92
506. Green, P.B.: Growth Physics in Nitella: a Method for Continuous in Vivo Analysis of Exten-
sibility Based on a Micro-manometer Technique for Turgor Pressure 1. Plant Physiol. 43(8),
1169–1184 (1968) 91
507. Green, P.B., Steele, C., Rennich, S.: Phyllotactic patterns: a biophysical mechanism for their
origin. Ann. Bot. -London 77(5), 515–528 (1996) 566
508. Greensmith, J.E., Duling, B.R.: Morphology of the constricted arteriolar wall: physiological
implications. Am. J. Physiol. -Heart C. 247(5), H687 (1984) 525
509. Greenspan, H.: Models for the growth of a solid tumor by diffusion. Stud. Appl. Math. 51(4),
317–340 (1972) 75
510. Gregersen, H., Kassab, G., Fung, Y.: Review: The zero-stress state of the gastrointestinal
tract. Digest. Dis. Sci. 45(12), 2271–2281 (2000) 44, 45
511. Gregersen, H., Kassab, G., Pallencaoe, E., Lee, C., Chien, S., Skalak, R., Fung, Y.: Mor-
phometry and strain distribution in guinea pig duodenum with reference to the zero-stress
state. Am. J. Physiol. -Gastr. L. 273(4), G865–G874 (1997) 45
512. Gregersen, H., Lu, X., Zhao, J.: Physiological growth is associated with esophageal mor-
phometric and biomechanical changes in rats. Neurogastroent. Motil. 16(4), 403–412 (2004)
525
513. Gregersen, H., Weis, S., McCulloch, A.: Oesophageal morphometry and residual strain in a
mouse model of osteogenesis imperfecta. Neurogastroent. Motil. 13(5), 457–464 (2001) 45
514. Grenetz, P., List Jr, A.: A model for predicting growth responses in plants to changes in
external water potential: Zea mays primary roots. J. Theor. Biol. 39(1), 29–45 (1973) 85
515. Grill, S.W., Howard, J., Schäffer, E., Stelzer, E.H.K., Hyman, A.A.: The distribution of active
force generators controls mitotic spindle position. Science 301(5632), 518–521 (2003) 574
516. Grillo, A., Zingali, G., Federico, S., Herzog, W., Giaquinta, G.: The role of material inho-
mogeneities in biological growth. Theor. Appl. Mech 32, 21–38 (2005) 407
517. Grove, S.N., Bracker, C.E.: Protoplasmic organization of hyphal tips among fungi: Vesicles
and Spitzenkörper. J. Bacteriol. 104, 989 – 1009 (1970) 238
518. Grytz, R., Meschke, G.: A computational remodeling approach to predict the physiolog-
ical architecture of the collagen fibril network in corneo-scleral shells. Biomech. Model
Mechanobiol. 9(2), 225–235 (2010) 340
519. Gu, W., Lai, W., Mow, V.: Transport of multi-electrolytes in charged hydrated biological soft
tissues. Transport Porous Med. 34(1-3), 143–157 (1999) 395
602 References
520. Guckenheimer, J., Holmes, P.: Nonlinear oscillations, dynamical systems and bifurcations
of vector fields. Springer-Verlag, New York (1983) 412
521. Guillon, T., Dumont, Y., Fourcaud, T.: A new mathematical framework for modelling the
biomechanics of growing trees with rod theory. Math. Comput. Model. 55, 2061–2077 (2011)
97
522. Guillou, A., Ogden, R.W.: Growth in soft biological tissue and residual stress development.
Mechanics of biological tissue. Springer, Heidelberg pp. 47–62 (2006) 349
523. Gunst, S.J., Stropp, J.Q.: Pressure-volume and length-stress relationships in canine bronchi
in vitro. J. Appl. Physiology 64(6), 2522–2531 (1988) 538
524. Gupta, A., Steigmann, D.J., Stölken, J.S.: On the evolution of plasticity and incompatibility.
Math. Mech. Solids 12(6), 583–610 (2007) 352, 358, 371
525. Gurtin, M.E.: An introduction to continuum mechanics. Academic Press (1981) 262, 279,
290, 352
526. Gurtin, M.E., Fried, E., Anand, L.: The mechanics and thermodynamics of continua. Cam-
bridge University Press (2010) 262, 371, 382, 405
527. Haijun, Z., O. Y., Z.c.: Spontaneous curvature-induced dynamical instability of Kirchhoff
filaments: Application to DNA kink deformations. J. Chem. Phys. 110, 1247–1251 (1999)
138
528. Hamant, O., Heisler, M.G., Jönsson, H., Krupinski, P., Uyttewaal, M., Bokov, P., Corson, F.,
Sahlin, P., Boudaoud, A., Meyerowitz, E.M., Couder, Y., Traas, J.: Developmental patterning
by mechanical signals in Arabidopsis. Science 322(5908), 1650–1655 (2008) 567
529. Hamill, O.P., Martinac, B.: Molecular basis of mechanotransduction in living cells. Physiol.
Rev. 81(2), 685–740 (2001) 36
530. Hamilton, G.: Multicellular spheroids as an in vitro tumor model. Cancer Lett. 131(1), 29–34
(1998) 38, 76
531. Hammer, Ø.: A theory for the formation of commarginal ribs in mollusc shells by regulative
oscillation. J. Molluscan Stud. 66(3), 383–392 (2000) 197
532. Hammer, Ø., Bucher, H.: Reaction-diffusion processes: application to the morphogenesis of
ammonoid ornamentation. Geobios. 32(6), 841–852 (1999) 189
533. Hammer, Ø., Bucher, H.: Models for the morphogenesis of the molluscan shell. Lethaia
38(2), 111–122 (2005) 188
534. Han, H., Fung, Y.: Residual strains in porcine and canine trachea. J. Biomech. 24(5), 307–309
(1991) 44, 45, 475
535. Han, H.C.: A biomechanical model of artery buckling. J. Biomech. 40, 3672–3678 (2007)
511
536. Han, H.C.: Nonlinear buckling of blood vessels: A theoretical study. J. Biomech. 41(12),
2708–2713 (2008) 511
537. Han, W., Lindsay, S.M., Dlakic, M., Harrington, R.E.: Kinked DNA. Nature 386(6625), 563
(1997) 136
538. Hansen, A.C., Crane, R.L., Damson, M.H., Donovan, R.P., Horning, D.T., Walker, J.L.:
Some notes on a volume fraction mixture theory and a comparison with the kinetic theory
of gases. Int. J. Eng. Sci. 29(5), 561–573 (1991) 393
539. Hargittai, I., Pickover, C.A.: Spiral symmetry. World Scientific (1992) 188, 192
540. Hariton, I., DeBotton, G., Gasser, T.C., Holzapfel, G.A.: Stress-driven collagen fiber remod-
eling in arterial walls. Biomech. Model. Mechanobiol. 6(3), 163–175 (2007) 321, 328, 419,
420
541. Harold, F.M.: How hyphae grow: morphogenesis explained. Protoplasm 197, 137–147
(1997) 5, 237
542. Harold, F.M.: Force and compliance: rethinking morphogenesis in walled cells. Fungal
Genet. Biol. 37, 271–282 (2002) 237
543. Harris, J.E., Crofton, H.D.: Structure and function in the nematodes: internal pressure and
cuticular structure in ascaris. J. Exp. Biol. 34(1), 116–130 (1957) 336
544. Haslach, H.W., Humphrey, J.D.: Dynamics of biological soft tissue and rubber: internally
pressurized spherical membranes surrounded by a fluid. Int. J. Non-linear Mech. 39, 399–420
(2004) 447
References 603
545. Haughton, D.: Elastic membranes. In: Y.B. Fu, R.W. Ogden (eds.) Nonlinear Elasticity:
Theory and Applications, pp. 233–267. Cambridge University Press, Cambridge (2001) 210
546. Haughton, D.M.: On non-existence of cavitation in incompressible elastic membranes. Q. J.
Mech. Appl. Math. 39(2), 289–296 (1986) 451
547. Haughton, D.M.: On non-linear stability in unconstrained non-linear elasticity. Int. J. Non-
linear Mech. 39, 1181–1192 (2004) 456
548. Haughton, D.M.: A comparison of stability and bifurcation criteria for a compressible elastic
cube. J. Eng. Mech. 53(1), 79–98 (2005) 311
549. Haughton, D.M., Chen, Y.: On the eversion of incompressible elastic spherical shells. ZAMP
50, 312–326 (1999) 456
550. Haughton, D.M., Kirkinis, E.: A comparison of stability and bifurcation criteria for inflated
spherical elastic shells. Math. Mech. Solids 8, 561–572 (2003) 456
551. Haughton, D.M., Ogden, R.W.: On the incremental equations in non-linear elasticity-II.
Bifurcation of pressurized spherical shells. J. Mech. Phys. Solids 26, 111–138 (1978) 444,
456
552. Haughton, D.M., Ogden, R.W.: Bifurcation of inflated circular cylinders of elastic material
under axial loading–II. exact theory for thick-walled tubes. J. Mech. Phys. Solids 27, 489–
512 (1979) 299, 490
553. Haughton, D.M., Ogden, R.W.: Bifurcation of inflated circular cylinders of elastic material
under axial loading-I. Membrane theory for thin-walled tubes. J. Mech. Phys. Solids 27,
179–212 (1979) 299, 461, 490
554. Haughton, D.M., Orr, A.: On the eversion of compressible elastic cylinders. Int. J. Solids
Struct. 34, 1893–1914 (1997) 461
555. Healey, T.J., Montes-Pizarro, E.L.: Global bifurcation in nonlinear elasticity with an applica-
tion to barrelling states of cylindrical columns. In: The Rational Spirit in Modern Continuum
Mechanics, pp. 469–494. Springer (2004) 311
556. Hearle, J.W.S., Thwaites, J.J., (Eds.), J.A.: Mechanics of flexible fibre assemblies. Sijthoff
& Noordhoff, The Netherlands (1980) 152
557. Hecht, S.: Form and growth in fishes. J. Morphology 27(2), 379–400 (1916) 18
558. Hehl, F.W., Obukhov, Y.N.: Foundations of Classical Electrodynamics: Charge, flux, and
metric. Springer Science & Business Media (2003) 368
559. Heidemann, S.R., Buxbaum, R.E.: Tension as a regulator and integrator of axonal growth.
Cell Motil. Cytoskeleton 17(1), 6–10 (1990) 30, 80
560. Heidemann, S.R., Lamoureux, P., Buxbaum, R.E.: Cytomechanics of axonal development.
Cell Biochem. Biophys. 27(3), 135–155 (1997) 31, 80, 81
561. Van der Heijden, G.H.M., Champneys, A.R., Thompson, J.M.T.: Spatially complex local-
isation in twisted elastic rods constrained to a cylinder. Int. J. Solids and Struct. 39(7),
1863–1883 (2002) 63
562. Hejnowicz, Z.: Graviresponses in herbs and trees: a major role for the redistribution of tissue
and growth stresses. Planta 203(5), 136–146 (1997) 42, 65, 500
563. Hejnowicz, Z., Sievers, A.: Tissue stresses in organs of herbaceous plants I: Determination
in three dimensions in the hypocotyl of sunflower plants. J. Exp. Bot. 46(8), 1035–1043
(1995) 42
564. Hejnowicz, Z., Sievers, A.: Tissue stresses in organs of herbaceous plants III: Determination
in three dimensions in the hypocotyl of sunflower plants. J. Exp. Bot. 47(297), 519–528
(1996) 42, 501, 502
565. Helfrich, W.: Helical bilayer structures due to spontaneous torsion of the edges. J. Chem.
Phys. 85, 1085–1087 (1986) 116
566. Helmlinger, G., Netti, P.A., Lichtenbeld, H.C., Melder, R.J., Jain, R.K.: Solid stress inhibits
the growth of multicellular tumor spheroids. Nature Biotech. 15, 778–783 (1997) 38, 72, 76
567. Henderson, J.H., Longaker, M.T., Carter, D.R.: Sutural bone deposition rate and strain mag-
nitude during cranial development. Bone 34(2), 271–280 (2004) 571
568. Henderson, J.H., Nacamuli, R.P., Zhao, B., Longaker, M.T., Carter, D.R.: Age-dependent
residual tensile strains are present in the dura mater of rats. J. Roy. Soc. Interface 2(3),
159–167 (2005) 45
604 References
569. Hennel, J.W., Klinowski, J.: Magic-angle spinning: a historical perspective. In: New tech-
niques in solid-state NMR, pp. 1–14. Springer (2005) 338
570. Herbert, A., Rich, A.: The biology of left-handed Z-DNA. J. Biol. Chem. 271(20), 11,595–
11,598 (1996) 156
571. Hester, S.D., Belmonte, J.M., Gens, J.S., Clendenon, S.G., Glazier, J.A.: A multi-cell, multi-
scale model of vertebrate segmentation and somite formation. PLoS Comput Biol 7(10),
e1002,155 (2011) 53
572. Heusner, A.: Energy metabolism and body size I. Is the 0. 75 mass exponent of Kleiber’s
equation a statistical artifact? Respiration Physiol. 48(1), 1–12 (1982) 15
573. Heyn, A.N.J.: Further investigations on the mechanism of cell elongation and the properties
of the cell wall in connection with elongation. Protoplasma 25(1), 372–396 (1936) 93
574. Heyn, A.N.J.: The physiology of cell elongation. Bot. Rev. 6, 515–574 (1940) 93
575. Hill, J.M.: Closed form solutions for small deformations superimposed upon the symmetrical
expansion of a spherical shell. J. Elasticity 6, 113–123 (1976) 456
576. Hill, M.J., Wilson, T.A., Lambert, R.K.: Effects of surface tension and intraluminal fluid on
mechanics of small airways. J. Appl. Physiology 82(1), 233–239 (1997) 531
577. Himpel, G., Menzel, A., Kuhl, E., Steinmann, P.: Time-dependent fibre reorientation of
transversely isotropic continua - finite element formulation and consistent linearization.
Internat. J. Numer. Methods Eng. 73(10), 1413–1433 (2008) 420, 421
578. His, W.: On the principles of animal morphology. Proc. Roy. Soc. Edimburgh 15, 287–298
(1888) 28
579. Hodge, N., Papadopoulos, P.: A continuum theory of surface growth. Proc. Roy. Soc. Lond.
A 466(2123), 3135–3152 (2010) 175, 572
580. Hoehme, S., Drasdo, D.: A cell-based simulation software for multi-cellular systems. Bioin-
formatics 26(20), 2641–2642 (2010) 58
581. Hoffman, A.S.: “Intelligent” polymers in medicine and biotechnology. Macromol. Symp.
98(1), 645–664 (1995) 403
582. Hofhuis, H., Moulton, D., Lessinnes, T., Routier-Kierzkowska, A.L., Bomphrey, R.J., Mosca,
G., Reinhardt, H., Sarchet, P., Gan, X., Tsiantis, M., Ventikos, Y., Walker, S., A., Goriely,
R., Smith, Hay, A.: Morphomechanical innovation drives explosive seed dispersal. Cell 166,
222–233 (2016) 321
583. Hoga, C.J.J., Niklas, K.J.: On the economy and safety of hollow non-septate peduncles. Am.
J. Bot. 90(3), 356–363 (1990) 480
584. Hoger, A.: On the determination of residual stress in an elastic body. J. Elasticity 16, 303–324
(1986) 39
585. Hoger, A., Dyke, T.J.V., Lubarda, V.A.: Symmetrization of the growth deformation and
velocity gradients in residually stressed biomaterials. ZAMP 55, 848–860 (2004) 401
586. Hohl, M., Schopfer, P.: Water Relations of Growing Maize Coleoptiles 1 Comparison be-
tween Mannitol and Polyethylene Glycol 6000 as External Osmotica for Adjusting Turgor
Pressure. Plant Physiol. 95(3), 716–722 (1991) 91
587. Hohl, M., Schöpfer, P.: Growth at reduced turgor: irreversible and reversible cell-wall exten-
sion of maize coleoptiles and its implications for the theory of cell growth. Planta 187(2),
209–217 (1992) 92
588. Hohl, M., Schopfer, P.: Physical extensibility of maize coleoptile cell walls: apparent plastic
extensibility is due to elastic hysteresis. Planta 187(4), 498–504 (1992) 92, 93
589. Höhme, S., Hengstler, J.G., Brulport, M., Schäfer, M., Bauer, A., Gebhardt, R., Drasdo,
D.: Mathematical modelling of liver regeneration after intoxication with CCl 4. Chemico-
biological interactions 168(1), 74–93 (2007) 55
590. Holdaway-Clarke, T.L., Feijo, J.A., Hackett, G.R., Kunkel, J.G., Hepler, P.K.: Pollen tube
growth and the intracellular cytosolic calcium gradient oscillate in phase while extracellular
calcium influx is delayed. Plant Cell 9(11), 1999–2010 (1997) 12, 419
591. Holland, M.A., Kosmata, T., Goriely, A., Kuhl, E.: On the mechanics of thin films and
growing surfaces. Math. Mech. Solids 18(561-575) (2013) 402, 499, 508
References 605
592. Holland, M.A., Miller, K.E., Kuhl, E.: Emerging brain morphologies from axonal elongation.
Ann. Biomed. Eng. 43(7), 1640–1653 (2015) 85, 573
593. Hollister, S.J.: Porous scaffold design for tissue engineering. Nature materials 4(7), 518–524
(2005) 562
594. Hollister, S.J., Fyhrie, D.P., Jepsen, K.J., Goldstein, S.A.: Application of homogenization
theory to the study of trabecular bone mechanics. J. Biomech. 24(9), 825–839 (1991) 562
595. Holmes, M., Mow, V.: The nonlinear characteristics of soft gels and hydrated connec-
tive tissues in ultrafiltration. J. Biomech. 23(11), 1145–1156 (1990). doi:10.1016/0021-
9290(90)90007-P395
596. Holmes, M.H.: A theoretical analysis for determining the nonlinear hydraulic permeability
of a soft tissue from a permeation experiment. Bull. Math. Biol. 47(5), 669–683 (1985) 395
597. Holzapfel, G.A.: Biomechanics of soft tissue. The handbook of materials behavior models
3, 1049–1063 (2001) 261, 559
598. Holzapfel, G.A.: Computational biomechanics of soft biological tissue. Encyc. Comp. Mech.
2, 605–635 (2004) 514, 518
599. Holzapfel, G.A.: Collagen in arterial walls: biomechanical aspects. In: P. Fratzl (ed.) Colla-
gen: structure and mechanics, pp. 285–324. Springer Verlag (2008) 320
600. Holzapfel, G.A., Gasser, T.C.: A viscoelastic model for fiber-reinforced composites at finite
strains: Continuum basis, computational aspects and applications. Comput. Methods Appl.
Mech. Eng 190(34), 4379–4403 (2001) 338
601. Holzapfel, G.A., Gasser, T.C., Ogden, R.W.: A new constitutive framework for arterial wall
mechanics and a comparative study of material models. J. Elasticity 61, 1–48 (2000) 43,
295, 508, 510, 513
602. Holzapfel, G.A., Gasser, T.C., Ogden, R.W.: Comparison of a multi-layer structural model
for arterial walls with a Fung-type model, and issues of material stability. J. Biomech. Eng.
126, 264–275 (2004) 512, 520
603. Holzapfel, G.A., Ogden, R.W.: Constitutive modelling of passive myocardium: A structurally
based framework for material characterization. Philos. Trans. Roy. Soc. A 367(1902), 3445–
3475 (2009) 293
604. Holzapfel, G.A., Ogden, R.W.: Constitutive modelling of arteries. Proc. Roy. Soc. Lond. A
466(2118), 1551–1597 (2010) 293, 321, 326, 340, 342, 344, 513
605. Holzapfel, G.A., Ogden, R.W.: Modelling the layer-specific three-dimensional residual
stresses in arteries, with an application to the human aorta. J. R. Soc. Interface 7, 787–799
(2010) 43, 509, 510
606. Holzapfel, G.A., Sommer, G., Auer, M., Regitnig, P., Ogden, R.W.: Layer-specific 3D resid-
ual deformations of human aortas with non-atherosclerotic intimal thickening. Ann. Biomed.
Eng. 35(4), 530–545 (2007) 510
607. Holzapfel, G.A., Stadler, M., Schulze-Bauer, C.A.J.: A layer-specific three-dimensional
model for the simulation of balloon angioplasty using magnetic resonance imaging and
mechanical testing. Ann. Biomed. Eng. 30(6), 753–767 (2002) 43, 510
608. Honda, H., Tanemura, M., Nagai, T.: A three-dimensional vertex dynamics cell model of
space-filling polyhedra simulating cell behavior in a cell aggregate. J. Theor. Biol. 226(4),
439–453 (2004) 54
609. Honda, H., Yamanaka, H., Dan-Sohkawa, M.: A computer simulation of geometrical con-
figurations during cell division. J. Theor. Biol. 106(3), 423–435 (1984) 58
610. Hong, W., Liu, Z., Suo, Z.: Inhomogeneous swelling of a gel in equilibrium with a solvent
and mechanical load. Int. J. Solids and Struct. 46(17), 3282–3289 (2009) 403
611. Hooke, R.: de Potentia Restitutiva, Or of Spring Explaining the Power of Springing Bodies.
J. Martyn (1678) 121
612. Hopwood, N.: Producing development: The anatomy of human embryos and the norms of
Wilhelm His. Bulletin of the History of Medicine 74(1), 29–79 (2000) 28
613. Horgan, C.O.: Equilibrium solutions for compressible nonlinear elasticity. In: Nonlinear
Elasticity: Theory and Applications, London Mathematical Society Lecture Note Series, vol.
283, pp. 135–159. Cambridge University Press (2001) 451
606 References
614. Horgan, C.O., Murphy, J.G.: Constitutive modeling for moderate deformations of slightly
compressible rubber. J. Rheol. 53, 153–168 (2009) 296
615. Horgan, C.O., Murphy, J.G.: Torsion of incompressible fiber-reinforced nonlinearly elastic
circular cylinders. J. Elasticity 103(2), 235–246 (2011) 324
616. Horgan, C.O., Polignone, D.A.: Cavitation in nonlinearly elastic solids: A review. Appl.
Mech. Rev. 48, 471–485 (1995) 451
617. Horgan, C.O., Saccomandi, G.: Constitutive modeling of rubber-like and biological materials
with limited chain extensibility. Math. Mech. Solids 7, 353–371 (2002) 293, 295
618. Horgan, C.O., Saccomandi, G.: A molecular-statistical basis for the Gent constitutive model
of rubber elasticity. J. Elasticity 68, 167–176 (2002) 295
619. Horgan, C.O., Saccomandi, G.: A description of arterial wall mechanics using limiting chain
extensibility constitutive models. Biomech. Model. Mechanobiol. 1 (2003) 295, 447
620. Horgan, C.O., Saccomandi, G.: Constitutive models for compressible nonlinearly elastic
materials with limiting chain extensibility. J. Elasticity 77, 123–138 (2004) 293, 295
621. Horny, L., Kronek, J., Chlup, H., Gultova, E., Heller, L., Zitny, R., Vokoun, D.: Inflation-
extension test of silicon rubber-nitinol composite tube. In: 5th European Conference of the
International Federation for Medical and Biological Engineering, pp. 1027–1030. Springer
(2012) 299
622. Hotani, H.: Light microscope study of mixed helices in reconstituted Salmonella flagella. J.
Mol. Biol. 106, 151–166 (1976) 155
623. Howard, R.J., Ferrari, M.A.: Role of melanin in appressorium function. Exp. Mycol. 13(4),
403–418 (1989) 226
624. Howard, R.J., Ferrari, M.A., Roach, D.H., Money, N.P.: Penetration of hard substrates by a
fungus employing enormous turgor pressures. Proc. Natl. Acad. Sci. USA 88, 11,281–11,284
(1991) 226
625. Howard, R.J., Valent, B.: Breaking and entering: host penetration by the fungal rice blast
pathogen, Magnaporthe grisea. Annu. Rev. Microbiol. 50, 491–512 (1996) 5, 225, 321
626. Hrousis, C.A., Wiggs, B.R., Drazen, J.M., Parks, D.M., Kamm, R.D.: Mucosal folding in
biologic vessels. J. Biomech. Eng. 124(4), 334–341 (2002) 532, 535
627. Hsu, F.H.: The influences of mechanical loads on the form of a growing elastic body. J.
Biomech. 1(4), 303–311 (1968) 10, 47, 48, 400
628. Hu, T., Desai, J.P.: Characterization of soft-tissue material properties: large deformation
analysis. In: Medical Simulation, pp. 28–37. Springer (2004) 295
629. Huang, H., Kamm, R.D., Lee, R.T.: Cell mechanics and mechanotransduction: pathways,
probes, and physiology. American Journal of Physiology-Cell Physiology 287(1), C1–C11
(2004) 561
630. Huang, Z.: The equilibrium equations and constitutive equations of the growing deformable
body in the framework of continuum theory. Int. J. Non-Linear Mech. 39(6), 951–962 (2004)
572
631. Huber, H.L., Koessler, K.K.: The pathology of bronchial asthma. Arch. Intern. Med. 30,
689–760 (1922) 531
632. Hufnagel, L., Teleman, A.A., Rouault, H., Cohen, S.M., Shraiman, B.I.: On the mechanism
of wing size determination in fly development. Proc. Natl. Acad. Sci. USA 104(10), 3835–
3840 (2007) 559
633. Huiskes, R.: If bone is the answer, then what is the question? Journal of Anatomy 197(02),
145–156 (2000) 35
634. Humphrey, J.D.: Stress, strain, and mechanotransduction in cells. J. Biomech . Eng. 123(6),
638–641 (2001) 560
635. Humphrey, J.D.: Cardiovascular solid mechanics. Cells, tissues, and organs. Springer Verlag,
New York (2002) 4, 43, 48, 49, 261, 304, 320, 328, 447, 510
636. Humphrey, J.D.: Continuum biomechanics of soft biological tissues. Proc. Roy. Soc. Lond.
A 459, 3–46 (2003) 9, 43, 48, 49
637. Humphrey, J.D.: Towards a theory of vascular growth and remodeling. Mechanics of Bi-
ological Tissue (G. A. Holzapfel, RW Ogden, eds), Springer-Verlag, Heidelberg pp. 3–15
(2006) 48
References 607
638. Humphrey, J.D.: Need for a continuum biochemomechanical theory of soft tissue and cellular
growth and remodeling. In: Biomechanical Modelling at the Molecular, Cellular and Tissue
Levels, pp. 1–82. Springer (2009) 49, 394, 396
639. Humphrey, J.D., Rajagopal, K.R.: A constrained mixture model for growth and remodeling
of soft tissues. Math. Models. Meth. Appl. Sci. 12, 407–430 (2002) 48, 396
640. Humphrey, J.D., Rajagopal, K.R.: A constrained mixture model for arterial adaptations to a
sustained step change in blood flow. Biomech. Model. Mechanobiol. 2(2), 109–126 (2003)
48, 396
641. Humphrey, J.D., Rao, I.J., Rajagopal, K.R.: A mixture model for growth and remodeling in
soft tissues. Bioeng. Div. ASME 50, 645–646 (2001) 48, 396
642. Hunt, G.W., Lord, G.J., Peletier, M.A.: Cylindrical shell buckling: a characterization of
localization and periodicity. Discret. Contin. Dyn. S. 3(4), 505–518 (2003) 489
643. Hutchinson, J.: Control of gastropod shell shape; the role of the preceding whorl. J. Theor.
Bio. 140(4), 431–444 (1989) 192
644. Hutchinson, J.M.C.: Control of gastropod shell form via apertural growth rates. J. Morphol.
206(3), 259–264 (1990) 192
645. Hutchinson, J.W.: The role of nonlinear substrate elasticity in the wrinkling of thin films.
Philos. Trans. Roy. Soc. Lond. A 371(1993), 20120, 422 (2013) 573
646. Hutson, M.S., Brodland, G.W., Yang, J., Viens, D.: Cell sorting in three dimensions: topology,
fluctuations, and fluidlike instabilities. Phys. Rev. Lett. 101(14), 148,105 (2008) 58
647. Huxley, J.: Problems of relative growth. Methuen (1932) 20, 22
648. Huxley, J.S., Needham, J., Lerner, I.M.: Terminology of relative growth-rates. Nature 148,
225 (1941) 17
649. Huxley, J.S., Teissier, G.: Terminology of relative growth. Nature 137(3471), 780–781 (1936)
12, 17
650. Huyghe, J.M., Janssen, J.D.: Quadriphasic mechanics of swelling incompressible porous
media. Int. J. Eng. Sci. 35(8), 793–802 (1997) 395
651. Hyun, K., Wilhelm, M., Klein, C.O., Cho, K.S., Nam, J.G., Ahn, K.H., Lee, S.J., Ewoldt,
R.H., McKinley, G.H.: A review of nonlinear oscillatory shear tests: Analysis and application
of large amplitude oscillatory shear (LAOS). Progress in Polymer Science 36(12), 1697–
1753 (2011) 556
652. Iijima, A.: Growth of the intertidal snail, Monodonta labio (Gastropoda, Prosobranchia) on
the Pacific coast of central Japan. Bulletin of marine science 68(1), 27–36 (2001) 194
653. Illert, C.: Formulation and solution of the classical seashell problem. Il Nuovo Cimento D
11(5), 761–780 (1989) 192
654. Imatani, S., Maugin, G.A.: A constitutive model for material growth and its application to
three-dimensional finite element analysis. Mech. Res. Comm. 29, 477–483 (2002) 407
655. Isnard, S., Cobb, A.R., Holbrook, N.M., Zwieniecki, M., Dumais, J.: Tensioning the helix:
a mechanism for force generation in twining plants. Proc. Roy. Soc. Lond. B 276(1667),
2643–2650 (2009) 168
656. Iwasa, K.H., Chadwick, R.S.: Elasticity and active force generation of cochlear outer hair
cells. J. Acoust. Soc. Am. 92(6), 3169–3173 (1992) 574
657. Jackson, M.B., Armstrong, W.: Formation of aerenchyma and the processes of plant venti-
lation in relation to soil flooding and submergence. Plant Biol. 1, 274–287 (1999) 483
658. Jackson, Z.S., Dajnowiec, D., Gotlieb, A.I., Langille, B.L.: Partial off-loading of longitudinal
tension induces arterial tortuosity. Arterioscler. Thromb. Vasc. Biol. 25(5), 957–962 (2005)
510, 522
659. Jackson, Z.S., Gotlieb, A.I., Langille, B.L.: Wall Tissue Remodeling Regulates Longitudinal
Tension in Arteries. Circ. Res. 90(8), 918–925 (2002) 32, 501, 510
660. Jaffe, M., Forbes, S.: Thigmomorphogenesis: the effect of mechanical perturbation on plants.
Plant Growth Regul. 12(3), 313–324 (1993) 483
661. James, A.L.: Airway remodeling in asthma. Curr. Opin. Pulm. Med. 11, 1–6 (2004) 532
662. James, A.L., Paré, P.D., Hogg, J.C.: The mechanics of airway narrowing in asthma. Am.
Rev. Respir. Dis. 139(1), 242–246 (1989) 535
608 References
663. Janmey, P.A., McCormick, M.E., Rammensee, S., Leight, J.L., Georges, P.C., MacKintosh,
F.C.: Negative normal stress in semiflexible biopolymer gels. Nature Mat. 6(1), 48–51 (2006)
3, 293, 403
664. Janmey, P.A., McCulloch, C.A.: Cell mechanics: integrating cell responses to mechanical
stimuli. Annu. Rev. Biomed. Eng. 9, 1–34 (2007) 561
665. Jarvis, M.C.: Intercellular separation forces generated by intracellular pressure. Plant Cell
Environ. 21(12), 1307–1310 (1998) 486
666. Jarvis, M.C., Briggs, S.P.H., Knox, J.P.: Intercellular adhesion and cell separation in plants.
Plant Cell Environ. 26(7), 977–989 (2003) 486
667. John, K.L.: Comparative cardiac mechanics: Laplace’s law. J. Theor. Biol. 118(3), 339–343
(1986) 34
668. Johnson, A.S., Ellers, O., Lemire, J., Minor, M., Leddy, H.A.: Sutural loosening and skeletal
flexibility during growth: determination of drop-like shapes in sea urchins. Proc. Roy. Soc.
Lond. B 269(1488), 215–220 (2002) 217
669. Johnson, B.E., Hoger, A.: The dependence of the elasticity tensor on residual stress. J.
Elasticity 33, 145–165 (1993) 349, 350
670. Johnson, D.L., Gamow, R.I.: Phycomyces: growth responses of the sporangium. Plant Phys-
iol. 49(6), 898–903 (1972) 542
671. Jones, A.F., Byrne, H.M., Gibson, J.S., Dold, J.W.: A mathematical model of the stress
induced during avasculor tumor growth. J. Math. Biol. 40, 473–499 (2000) 565
672. Jones, G.W., Chapman, S.J.: Modeling growth in biological materials. SIAM Review 54(1),
52–118 (2012) 49, 54, 55, 563
673. de Jong, J.C., McCormack, B.J., Smirnoff, N., Talbot, N.J.: Glycerol generates turgor in rice
blast. Nature 389(6648), 244–244 (1997) 225
674. Jönsson, H., Heisler, M.G., Shapiro, B.E., Meyerowitz, E.M., Mjolsness, E.: An auxin-driven
polarized transport model for phyllotaxis. Proc. Natl. Acad. Sci. USA 103(5), 1633–1638
(2006) 566
675. Jülicher, F.: Supercoiling transition of closed DNA. Phys. Rev. E 49 (1994) 116
676. Kaandorp, J.A.: Analysis and synthesis of radiate accretive growth in three dimensions. J.
Theor. Biol. 175(1), 39–55 (1995) 570
677. Kafadar, C.B.: On Ericksen’s problem. Arch. Rat. Mech. Anal 47(1), 15–27 (1972) 579
678. Kalita, P., Schaefer, R.: Mechanical models of artery walls. Arch. Comput. Method. Eng.
15(1), 1–36 (2008) 512
679. Kamiya, R., Asakura, S.: Helical transformations of Salmonella flagella in vitro. J. Mol.
Biol. 106, 167–186 (1976) 116, 155
680. Kamiya, R., Asakura, S., Wakabayashi, K., Namba, K.: Transition of bacterial flagella from
helical to straight forms with different subunits arrangements. J. Mol. Biol. 131, 725–742
(1979) 155
681. Kardong, K.V.: Vertebrates: comparative anatomy, function, evolution. McGraw-Hill Boston
(2006) 21
682. Keener, J.P.: Knotted vortex filament in an ideal fluid. J. Fluid Mech. 211, 629–651 (1990)
63
683. Keith, A.: The Engines of the Human Body. London (1919) 35
684. Keller, J.B.: Tendril shape and lichen growth. Lecture on Mathematics in the Life Science
30, 257–274 (1980) 63
685. Kelvin, L., Tait, P.G.: Element of Natural Philosophy (Part one). P. F. Collier and Son, New
York (1901) 115
686. Ker, R.F.: Damage and fatigue. In: P. Fratzl (ed.) Collagen: structure and mechanics, pp.
111–131. Springer Verlag (2008) 320
687. Kessler, D.A., Koplik, J., Levine, H.: Pattern selection in fingered growth phenomena. Adv.
Phys. 37, 255–339 (1988) 3
688. Kiddie, G., McLean, D., Van Ooyen, A., Graham, B.: Biologically plausible models of
neurite outgrowth. Prog. Brain Res. 147, 67–80 (2004) 85
689. Kiener, L.C.: Species général et iconographique des coquilles vivantes. Baillière (1880) 189
References 609
690. Kier, W.M.: The diversity of hydrostatic skeletons. J. Exp. Biol. 215(8), 1247–1257 (2012)
336
691. Kier, W.M., Smith, K.K.: Tongues, tentacles and trunks: the biomechanics of movement in
muscular-hydrostats. Zool. J. Linn. Soc. 83(4), 307–324 (1985) 336
692. Kim, C.I., Schiavone, P.: A circular inhomogeneity subjected to non-uniform remote loading
in finite plane elastostatics. Int. J. Non-Linear Mech. 42(8), 989–999 (2007) 476
693. Kim, G.T., Cho, K.H.: Recent advances in the genetic regulation of the shape of simple
leaves. Physiologia Plantarum 126(4), 494–502 (2006) 28
694. Kim, Y., Stolarska, M.A., Othmer, H.G., Bellomo, N., Maini, P.: A hybrid model for tumor
spheroid growth in vitro I: Theoretical development and early results. Math. Model. Meth.
Appl. Sci. 17, 1773 (2007) 75
695. Kimpton, L.S., Hall, C.L., Bintu, B., Crosby, D., Byrne, H.M., Goriely, A.: A shell model
of eye growth and elasticity. Preprint (2015) 220, 568
696. King, G.G., Paré, P.D., Seow, C.Y.: The mechanics of exaggerated airway narrowing in
asthma: the role of smooth muscle. Resp. Physiol. 118, 1–13 (1999) 531
697. Kircher, A.: Mundus Subterraneus, 3rd edn. Joannem Janssonium, Amsterdam (1678) 19
698. Kirchhoff, G.: über das gleichgewicht und die bewegung einer elastischen scheibe. Journal
für die reine und angewandte Mathematik 40, 51–88 (1850) 251
699. Kirkinis, E., Ogden, R.W.: On extension and torsion of a compressible elastic circular cylin-
der. Math. Mech. Solids 7(4), 373–392 (2002) 299
700. Klapper, I.: Biological applications of the dynamics of twisted elastic rods. J. Comp. Phys.
125, 325–337 (1996) 171, 172
701. Klapper, I., Tabor, M.: Dynamics of twist and writhe and the modeling of bacterial fibers. In:
J. Mesirov, K. Schuiten, D.W. Summers (eds.) Mathematical Approaches to Biomolecular
Structure and Dynamics. Springer (1996) 171, 172
702. Klarbring, A., Olsson, T.: On compatible strain with reference to biomechanics of soft tissues.
ZAMM 85(6), 440–448 (2005) 363
703. Klarbring, A., Olsson, T., Stalhand, J.: Theory of residual stresses with application to an
arterial geometry. Arch. Mech. 59(4-5), 341–364 (2007) 363
704. Kleiber, M.: Body size and metabolic rate. Physiol. Rev 27(4), 511–541 (1947) 15
705. Klein, Y., Efrati, E., Sharon, E.: Shaping of elastic sheets by prescription of non-Euclidean
metrics. Science 315(5815), 1116–1120 (2007) 252, 256
706. Klisch, S.M.: Continuum models of growth with emphasis on articular cartilage. Mechanics
of biological tissues pp. 119–133 (2006) 404
707. Klisch, S.M., Asanbaeva, A., Oungoulian, S.R., Masuda, K., Eugene, J.M.A.T., Davol, A.,
Sah, R.L.: A cartilage growth mixture model with collagen remodeling: validation protocols.
J. Biomech. Eng. 130, 031,006 (2008) 395
708. Klisch, S.M., Chen, S.S., Sah, R.S., Hoger, A.: A growth mixture theory for cartilage with
application to growth-related experiments on cartilage explants. J. Biomech. Eng. 125, 169–
179 (2003) 395
709. Klisch, S.M., Dyke, T.J.V., Hoger, A.: A theory of volumetric growth for compressible elastic
biological materials. Math. Mech. Solids 6, 551–575 (2001) 48, 400, 434
710. Klisch, S.M., Hoger, A.: Volumetric growth of thermoelastic materials and mixtures. Math.
Mech. Solids 8, 377 – 402 (2003) 395
711. Klisch, S.M., Sah, R.L., Hoger, A.: A cartilage growth mixture model for infinitesimal strains:
solutions of boundary-value problems related to in vitro growth experiments. Biomech.
Model. Mechanobiol. 3(4), 209–223 (2005) 395
712. Knoche, S., Vella, D., Aumaitre, E., Degen, P., Rehage, H., Cicuta, P., Kierfeld, J.: Elastom-
etry of deflated capsules: Elastic moduli from shape and wrinkle analysis. Langmuir 29(40),
12,463–12,471 (2013) 222
713. Knops, R.J.: Elements of elastic stability theory. In: M. Hayes, G. Saccomandi (eds.) Topics
in Finite Elasticity, vol. 424, pp. 169–230. Springer Science & Business Media (2001) 311
714. Koch, A.L.: The problem of hyphal growth in streptomycetes and fungi. J. Theor. Biol.
171(2), 137–150 (1994) 5
610 References
715. Koch, A.L.: The problem of hyphal growth in streptomycetes and fungi. In: A.L. Koch (ed.)
Bacterial Growth and Form. Chapman & Hall, New York (1995) 5, 239
716. Koehl, M.A.R.: Physical modelling in biomechanics. Philos. Trans. Roy. Soc. Lond. B
358(1437), 1589–1596 (2003) 338
717. Koehl, M.A.R., Quillin, K.J., Pell, C.A.: Mechanical design of fiber-wound hydraulic skele-
tons: the stiffening and straightening of embryonic notochords. Amer. Zool. 40(1), 28–41
(2000) 338
718. Koga, H., Nakayachi, O.: Morphological studies on attachment of spores of Magnaporthe
grisea to the leaf surface of rice. J. Gen. Plant Pathol. 70, 11–15 (2004) 226
719. Koiter, W.T.: On the nonlinear theory of thin elastic shells. I- Introductory sections. II-
Basic shell equations. III- Simplified shell equations. Koninklijke Nederlandse Akademie
van Wetenschappen, Proceedings, Series B 69(1), 1–54 (1966) 251
720. Koiter, W.T., Van Der Heijden, A.: WT Koiter’s elastic stability of solids and structures.
Cambridge Univ Press (2009) 149
721. Kondaurov, V.I., Nikitin, L.V.: Finite strains of viscoelastic muscle tissue. J. Appl. Math.
Mech. 51(3), 346–353 (1987) 353
722. Kondo, K.: A proposal of a new theory concerning the yielding of materials based on Rie-
mannian geometry, I. J. Soc. Appl. Mech. Jpn 2(11), 123–128 (1949) 363, 400
723. Kondo, K.: Non-Riemannian geometry of imperfect crystals from a macroscopic viewpoint.
Memoirs of the unifying study of the basic problems in engineering science by means of
geometry 1, 6–17 (1955) 363
724. Kowalski, S.J., Rybicki, A.: Residual stresses in dried bodies. Dry. Technol. 25(4), 629–637
(2007) 45
725. Kozela, C., Regan, S.: How plants make tubes. Trends Plant Sci. 8(4), 159–164 (2003) 483
726. Kramer, P.J.: Water Relations of Plants. Academic Press, Inc. Orlando, FL. (1983) 482
727. Kringelbach, M.L., Stark, E.A., Alexander, C., Bornstein, M.H., Stein, A.: On cuteness:
Unlocking the parental brain and beyond. Trends in cognitive sciences 20, 545–558 (2016)
17
728. Kristie, D.N., Joliffe, P.A.: High-resolution studies of growth oscillations during stem elon-
gation. Can. J. Bot. 64(11), 2399–2405 (1986) 12, 419
729. Kröner, E.: Kontinuumstheorie der versetzungen und eigenspannungen, vol. 5. Springer
(1958) 353
730. Kröner, E.: Alleganeine kontinuumstheorie der verzetsmugen and eigenspannuugen. Arch.
Ration. Mech. Anal. 4, 273–334 (1960) 48, 353, 363
731. Kröner, E.: Continuum theory of defects. Les Houches, Session XXXV pp. 215–315 (1980)
363
732. Kröner, E., Seeger, A.: Nicht-lineare elastizitätstheorie der versetzungen und eigenspannun-
gen. Arch. Rat. Mech. Anal. 3(1), 97–119 (1959) 353
733. Kuhl, E.: Growing matter: a review of growth in living systems. J. Mech. Behavior Biomed.
Mat. 29, 529–543 (2014) 49, 401
734. Kuhl, E., Balle, F.: Computational modeling of hip replacement surgery: Total hip replace-
ment vs. hip resurfacing. Technische Mechanik 25(2), 107–114 (2005) 420
735. Kuhl, E., Garikipati, K., Arruda, E.M., Grosh, K.: Remodeling of biological tissue: mechan-
ically induced reorientation of a transversely isotropic chain network. J. Mech. Phys. Solids
53(7), 1552–1573 (2005) 420, 422
736. Kuhl, E., Holzapfel, G.A.: A continuum model for remodeling in living structures. J. Mater.
Sci. 42(21), 8811–8823 (2007) 419, 420
737. Kuhl, E., Steinmann, P.: On spatial and material settings of thermo-hyperelastodynamics for
open systems. Acta Mechanica 160(3-4), 179–217 (2003) 382
738. Kumar, V., Abbas, A.K., Fausto, N., Aster, J.C.: Robbins & Cotran pathologic basis of
disease. Elsevier Health Sciences (2009) 34
739. Kutschera, U.: Tissue stresses in growing plant organs. Physiol. Plantarum 77, 157–163
(1989) 42, 500
References 611
740. Kutschera, U.: Gravitropism of axial organs in multicellular plants. Adv. Space Res. 27(5),
851–860 (2001) 65
741. Kutschera, U.: The growing outer epidermal wall: Design and physiological role of a com-
posite structure. Ann. Bot-London 101(5), 615 (2008) 42, 89, 500
742. Kutschera, U., Niklas, K.J.: The epidermal-growth-control theory of stem elongation: An
old and a new perspective. J. Plant Physiol. 164(11), 1395–1409 (2007) 43
743. Kutschera, U., Schopfer, P.: Effect of auxin and abscisic acid on cell wall extensibility in
maize coleoptiles. Planta 167(4), 527–535 (1986) 93
744. Kutschera, U., Schopfer, P.: In-vivo measurement of cell-wall extensibility in maize coleop-
tiles: Effects of auxin and abscisic acid. Planta 169(3), 437–442 (1986) 93, 94, 96
745. Kuwano, K., Bosken, C.H., Paré, P.D., Bai, T.R., Wiggs, B.R., Hogg, J.C.: Small airways
dimensions in asthma and in chronic obstructive pulmonary disease. Am. Rev. Respir. Dis.
148(5), 1220–1225 (1993) 531
746. Kyriacou, S., Humphrey, J.D.: Influence of size, shape and properties on the mechanics of
axisymmetric saccular aneurysms. J. Biomech. 29(8), 1015–1022 (1996) 304
747. Lacro, R.V., Jones, K.L., Benirschke, K.: The umbilical cord twist: origin, direction, and
relevance. Am. J. Obstet. Gynecol. 157 (1987) 154
748. Lagrange, J.L.: Sur la figure des colonnes. Miscellanea Taurinensia 5, 123–66 (1770) 487
749. Lai, W.M., Hou, J.S., Mow, V.C.: A triphasic theory for the swelling and deformation be-
haviors of articular cartilage. J. Biomech. Eng. 113(3), 245–258 (1991) 395
750. Lakshmanan, M.: Rigid body motions, space curves, prolongation structures, fiber bundles,
and solitons. J. Math. Phys 20, 1667–1672 (1979) 102
751. Lally, C., Reid, A., Prendergast, P.: Elastic behavior of porcine coronary artery tissue under
uniaxial and equibiaxial tension. Ann. Biomed. Eng. 32(10), 1355–1364 (2004) 293
752. Lamb, G.L.J.: Solitons on moving space curves. J. Math. Phys. 18, 1654–1661 (1977) 102
753. Lambert, R.K.: Role of bronchial basement membrane in airway collapse. J. Appl. Physiol.
71, 666–673 (1991) 531
754. Lambert, R.K., Codd, S.L., Alley, M.R., Pack, R.J.: Physical determinants of bronchial
mucosal folding. J. Appl. Physiol. 77, 1206–1216 (1994) 531, 532, 533
755. Lamoureux, P., Buxbaum, R.E., Heidemann, S.R.: Direct evidence that growth cones pull.
Nature 340(6229), 159–162 (1989) 80, 81
756. Lamoureux, P., Heidemann, S.R., Martzke, N.R., Miller, K.E.: Growth and elongation within
and along the axon. Dev. Neurobiol. 70(3), 135–49 (2010). 10. 1002/dneu. 20764 64, 82
757. Lamoureux, P., Ruthel, G., Buxbaum, R.E., Heidemann, S.R.: Mechanical tension can spec-
ify axonal fate in hippocampal neurons. J. Cell Biol. 159(3), 499–508 (2002) 31, 80, 81
758. Lamoureux, P., Zheng, J., Buxbaum, R.E., Heidemann, S.R.: A cytomechanical investigation
of neurite growth on different culture surfaces. J. Cell Biol. 118(3), 655–661 (1992) 63
759. Lampl, M., Ashizawa, K., Kawabata, M., Johnson, M.L.: An example of variation and pattern
in saltation and stasis growth dynamics. Ann. Hum. Biol. 25(3), 203–219 (1998) 12
760. Lampl, M., Veldhuis, J.D., Johnson, M.L.: Saltation and stasis: a model of human growth.
Science 258(5083), 801–803 (1992) 12, 13
761. Landau, L.D., Lifshitz, E.M.: Theory of Elasticity. Pergamon Press, Oxford (1959) 111, 227,
261
762. Lang, G.E., Stewart, P.S., Vella, D., Waters, S.L., Goriely, A.: Is the Donnan effect sufficient
to explain swelling in brain tissue slices? J. R. Soc. Interface 11(96), 20140,123 (2014) 395,
569
763. Lang, G.E., Vella, D., Waters, S.L., Goriely, A.: Propagation of damage in brain tissue:
Coupling the mechanics of edema and oxygen delivery. Biomech. Model Mechanobiol. 14,
1197–1216 (2015) 395
764. Lange, C.G., Newell, A.C.: The post-buckling problem for thin elastic shells. SIAM J. Appl.
Math. 21 (1971) 149, 493
765. Langer, J.S.: Instabilities and pattern-formation in crystal-growth. Rev. Mod. Phys. 52 (1980)
3
612 References
766. Langewouters, G.J., Wesseling, K.H., Goedhard, W.J.A.: The static elastic properties of 45
human thoracic and 20 abdominal aortas in vitro and the parameters of a new model. J.
biomechanics 17(6), 425–435 (1984) 295
767. Lanir, Y.: A structural theory for the homogeneous biaxial stress-strain relationships in flat
collagenous tissues. J. Biomech. 12(6), 423–436 (1979) 321, 339
768. Lanir, Y.: Constitutive equations for fibrous connective tissues. J. Biomech. 16(1), 1–12
(1983) 293, 321
769. Lanir, Y., Fung, Y.C.: Two-dimensional mechanical properties of rabbit skin-II. Experimental
results. J. Biomech. 7(2), 171–182 (1974) 320
770. Lapicque, L.: Sur la relation du poids de l’encéphale aux poids du corps. Comptes Rendus
Séances Soc. Biol. Fil. ,(Sér 10) 5, 62–63 (1898) 20
771. de Laplace, P.S.: Traité de mécanique céleste: Théorie de l’action capillaire. Suppl. Courcier
(1806) 33, 216
772. Larson, K.: Circumnutation behavior of an exotic honeysuckle vine and its native congener:
influence on clonal mobility. Am. J. Bot. 87, 533–538 (2000) 150
773. Learoyd, B.M., Taylor, M.G.: Alterations with age in the viscoelastic properties of human
arterial walls. Circ. Res. 18(3), 278–292 (1966) 509
774. Lebon, G., Jou, D., Casas-Vázquez, J.: Understanding non-equilibrium thermodynamics.
Springer (2008) 406
775. Lechevalier, H.A., Lechevalier, M.P.: Biology of actinomycetes. Annu. Rev. Microbiol.
21(1), 71–100 (1967) 234
776. Lecuit, T., Brook, W.J., Ng, M., Calleja, M., Sun, H., Cohen, S.M.: Two distinct mechanisms
for long-range patterning by Decapentaplegic in the Drosophila wing. Nature 381(6581),
387–393 (1996) 558
777. Lee, E.H.: Elastic-plastic deformation at finite strains. Trans. ASME J. Appl. Mech. 54, 1–6
(1969) 48, 353
778. Lee, M.M.L., Chien, S.: Morphologic effects of pressure changes on canine carotid artery
endothelium as observed by scanning electron microscopy. Anatom. Res. 194, 1–14 (1978)
524
779. Lehoux, S., Castier, Y., Tedgui, A.: Molecular mechanisms of the vascular responses to
haemodynamic forces. J. Intern. Med. 259(4), 381–392 (2006) 32
780. Lekhnitskii, S.G.: Anisotropic plates. Tech. rep., Translated from the Second Russian Edition
by S. W. Tsai and T. Cheron Gordon and Breach, New York. (1957) 450
781. Lekszycki, T., dell’Isola, F.: A mixture model with evolving mass densities for describing
synthesis and resorption phenomena in bones reconstructed with bio-resorbable materials.
ZAMM 92(6), 426–444 (2012) 396
782. Léon, I.: Recherche nouvelles sur la cause du mouvement spiral des tiges volubiles. B. Soc.
Bot. Fr. 5, 624–629 (1858) 151
783. Léon, I.: Recherche nouvelles sur la cause du mouvement spiral des tiges volubiles (deuxième
partie). B. Soc. Bot. Fr. 5, 630 (1858) 151
784. Lepetit, J., Salé, P., Ouali, A.: Post-mortem evolution of rheological properties of the my-
ofibrillar structure. Meat Sci. 16(3), 161–174 (1986) 45
785. Lessinnes, T., Goriely, A.: Design and stability of a family of deployable structures. SIAM
J. Appl. Math. (2016) 311
786. Lessinnes, T., Moulton, D.E., Goriely, A.: Morphoelastic rods. Part II: Growing birods. J.
Mech. Phys. Solids. 100, 147–196 (2017) 98, 126
787. Levinson, M.: Stability of a compressed neo-Hookean rectangular parallelepiped. J. Mech.
Phys. Solids 16, 403–415 (1968) 489
788. Lewicka, M., Pakzad, M.R.: The infinite hierarchy of elastic shell models: some recent results
and a conjecture. In: Infinite Dimensional Dynamical Systems, pp. 407–420. Springer (2012)
256
789. Li, B., Jia, F., Cao, Y., Feng, X., Gao, H.: Surface wrinkling patterns on a core-shell soft
sphere. Phys. Rev. Lett. 106(23), 234,301 (2011) 573
References 613
790. Liao, D., Zhao, J., Yang, J., Gregersen, H.: The oesophageal zero-stress state and mucosal
folding from a GIOME perspective. World J. Gastroenterol 13(9), 1347–1351 (2007) 524
791. Linnaei, C.: Philosophia Botanica. Stockholm (1751) 151, 153
792. Lionnet, T., Joubaud, S., Lavery, R., Bensimon, D., Croquette, V.: Wringing out DNA. Phys.
Rev. Lett. 96(17), 178,102 (2006) 119
793. Listing, J.B.: Vorstudien über topologie. Göttinger Studien I (1847) 150
794. Liu, J., Huang, J., Su, T., Bertoldi, K., Clarke, D.R., Buehler, M.J.: Structural transition from
helices to hemihelices. PloS one 9(4), e93,183 (2014) 158
795. Liu, S.Q., Fung, Y.C.: Relationship between hypertension, hypertrophy, and opening angle
of zero-stress state of arteries following aortic constriction. J. Biomech Eng 111(4), 325–335
(1989) 409
796. Liu, W., Rahn, C.R.: Fiber-reinforced membrane models of mckibben actuators. J. Appl.
Mech. 70(6), 853–859 (2003) 338
797. Lloyd, C., Chan, J.: Helical microtubule arrays and spiral growth. Plant Cell Online 14(10),
2319–2324 (2002) 121, 321
798. Lockhart, J.A.: An analysis of irreversible plant cell elongation. J. Theor. Biol. 8, 264 – 275
(1965) 47, 88
799. López-Franco, R., Bartnicki-Garcia, S., Bracker, C.E.: Pulsed growth of fungal hyphal tips.
Proc. Natl. Acad. Sci. USA 91(25), 12,228 (1994) 12, 419
800. Lopez-Pamies, O.: Onset of cavitation in compressible, isotropic, hyperelastic solids. J.
Elasticity 94, 115–145 (2009) 294, 451
801. Lopez-Pamies, O., Ponte Castañed, P.: Homogenization-based constitutive models for porous
elastomers and implications for macroscopic instabilities: I—analysis. J. Mech. Phys. Solids
55(8), 1677–1701 (2007) 296, 562
802. Lopez-Pamies, O., Ponte Castañeda, P.: On the overall behavior, microstructure evolution,
and macroscopic stability in reinforced rubbers at large deformations: I—theory. J. Mech.
Phys. Solids 54(4), 807–830 (2006) 562
803. Lorenz, K.: Part and parcel in animal and human societies. Studies in animal and human
behavior 2, 115–195 (1971) 17
804. Loret, B., Simões, F.M.F.: A framework for deformation, generalized diffusion, mass transfer
and growth in multi-species multi-phase biological tissues. Eur. J. Mech. A-Solid. 24(5),
757–781 (2005) 408
805. Love, A.E.H.: The small free vibrations and deformation of a thin elastic shell. Phil. Trans.
R. Soc. Lond. A 179, 491–546 (1888) 251
806. Love, A.E.H.: A treatise on the mathematical theory of elasticity. Cambridge University
Press, Cambridge (1892) 111, 119
807. Lowery, L.A., Van Vactor, D.: The trip of the tip: understanding the growth cone machinery.
Nat. Rev. Mol. Cell Biol. 10(5), 332–343 (2009) 81
808. Lu, X., Zhao, J., Gregersen, H.: Small intestinal morphometric and biomechanical changes
during physiological growth in rats. J. Biomech. 38, 417–426 (2005) 524
809. Lubarda, V.A.: Elastoplasticity theory. CRC press (2001) 377
810. Lubarda, V.A.: Constitutive theories based on the multiplicative decomposition of deforma-
tion gradient: Thermoelasticity, elastoplasticity, and biomechanics. Appl. Mech. Rev. 57,
95–108 (2004) 48
811. Lubarda, V.A., Hoger, A.: On the mechanics of solids with a growing mass. Int. J. Solids
Struct. 39, 4627–4664 (2002) 48, 377, 401
812. Lubkin, S.: Unidirectional waves on rings: models for chiral preferences of circumnutating
plants. Bull. Math. Biol. 56, 795–810 (1994) 150
813. Lucca, G.: Representing seashells surface. In: Visual Mathematics. Mathematical Institute
SASA (2003) 192
814. Ma, H., Snook, L., Kaminskyj, S., Dahms, T.: Surface ultrastructure and elasticity in growing
tips and mature regions of Aspergillus hyphae describe wall maturation. Microbiology 151(Pt
11), 3679–3688 (2005) 116
614 References
815. Macklin, P., McDougall, S., Anderson, A.R.A., Chaplain, M.A.J., Cristini, V., Lowengrub,
J.: Multiscale modelling and nonlinear simulation of vascular tumour growth. J. Math. Biol.
58, 765–798 (2009) 76
816. MacMahon, T.A., Bonner, J.T.: On size and life. Scientific American Books (1983) 12
817. Macnab, R.M., Ornston, M.K.: Normal-to-curly flagellar transitions and their role in bacterial
tumbling. Stabilization of an alternative quaternary structure by mechanical force. J. Mol.
Biol. 112, 795–810 (1977) 116, 155
818. Maddocks, J.: Bifurcation theory, symmetry breaking and homogenization in continuum
mechanics descriptions of DNA. A Celebration of Mathematical Modeling: The Joseph B.
Keller Anniversary Volume pp. 113–136 (2004) 136
819. Maini, P.K., Painter, K.J., Chau, H.N.P.: Spatial pattern formation in chemical and biological
systems. J. Chem. Soc., Faraday Trans. 93(20), 3601–3610 (1997) 564
820. Maini, P.K., Woolley, T.E., Baker, R.E., Gaffney, E.A., Lee, S.S.: Turing’s model for biolog-
ical pattern formation and the robustness problem. Interface focus p. rsfs20110113 (2012)
150, 564
821. Mak, A., Lai, W., Mow, V.: Biphasic indentation of articular cartilage—1. Theoretical analy-
sis. J. Biomech. 20(7), 703–714 (1987) 395
822. Mallock, A.: Note on the instability of india-rubber tubes and balloons when distended by
fluid pressure. Proc. Roy. Soc. Lond. 49(296-301), 458–463 (1890) 261, 262
823. Malpas, P., Symonds, A.M.: Observations on the structure of the human umbilical cord.
Surgery Gynec. Obstet. 123, 746 (1966) 154
824. Mann, S.: Biomineralization: principles and concepts in bio-inorganic materials chemistry,
vol. 5. Oxford University Press (2001) 188
825. Manning, R.S., Hoffman, K.A.: Stability of n-covered circles for elastic rods with constant
planar intrinsic curvature. J. Elasticity 62, 1–23 (2001) 138
826. Manning, R.S., Maddocks, J.H., Kahn, J.D.: A continuum rod model of sequence-dependent
DNA structure. J. Chem. Phys. 105, 5626–5646 (1996) 63, 116, 136
827. Manning, R.S., Rogers, K.A., Maddocks, J.H.: Isoperimetric conjugate points with appli-
cation to the stability of DNA minicircles. Proc. Roy. Soc. Lond. A 454(1980), 3047–3074
(1998) 138
828. Mao, J.M., Delos, J.B.: Hamiltonian bifurcation theory of closed orbits in the diamagnetic
kepler problem. Phys. Rev. A 45(3), 1746–1761 (1992) 159
829. Marchetti, M.C., Joanny, J.F., Ramaswamy, S., Liverpool, T.B., Prost, J., Rao, M., Simha,
R.A.: Hydrodynamics of soft active matter. Rev. Mod. Phys 85(3), 1143 (2013) 576
830. Marder, M., Sharon, E., Smith, S., Roman, B.: Theory of edges of leaves. Europhys. Lett.
62(4), 498–504 (2003) 256
831. Marini, D.M., Hwang, W., Lauffenburger, D.A., Zhang, S., Kamm, R.: Left-handed helical
ribbon intermediates in the self-assembly of a β-sheet peptide. Nano Letters 2, 295–299
(2002) 116
832. Markus, M., Hess, B.: Isotropic cellular automaton for modelling excitable media. Nature
347(6288), 56–58 (1990) 52
833. Marris, A.W.: Two new theorems on Ericksen’s problem. Arch. Rat. Mech. Anal 79(2),
131–173 (1982) 579
834. Martin, R.R., Haines, H.: Application of Laplace’s law to mammalian hearts. Comp.
Biochem. Physiol. 34(4), 959–962 (1970) 34
835. Marty, G., Tsapis, N.: Monitoring the buckling threshold of drying colloidal droplets using
water-ethanol mixtures. Europ. Phys. J. E 27(2), 213–219 (2008) 432
836. Marzano, S.: An interpretation of Baker-Ericksen inequalities in uniaxial deformation and
stress. Meccanica 18(4), 233–235 (1983) 291
837. Masuda, Y.: Auxin-induced cell elongation and cell wall changes. J. Plant Res. 103(3),
345–370 (1990) 42
838. Matista, A.A., Silk, W.K.: An electronic device for continuous, in vivo measurement of
forces exerted by twining vines. Am. J. Bot. 84, 1164–1164 (1997) 149
839. Matsuo, E.S., Tanaka, T.: Patterns in shrinking gels. Nature pp. 482–485 (1992) 403
References 615
840. Matthews, M.A., van Volkenburgh, E., Boyer, J.S.: Acclimation of leaf growth to low water
potentials in sunflower. . Plant Cell Environ. 7(3), 199–206 (1984) 96
841. Mauck, R.L., Hung, C.T., Ateshian, G.A.: Modeling of neutral solute transport in a dy-
namically loaded porous permeable gel: implications for articular cartilage biosynthesis and
tissue engineering. J. Biomech. 125(5), 602 (2003) 395
842. Maugin, G.A.: The thermomechanics of nonlinear irreversible behaviors. World Scientific
(1999) 406
843. Maugin, G.A.: Pseudo-plasticity and pseudo-inhomogeneity effects in material mechanics.
J. Elasticity 71, 81–103 (2003) 86
844. Maxwell, J.C.: A treatise on electricity and magnetism. Clarendon, Oxford (1892) 116, 150
845. McDonald, A., Davies, W.: Keeping in touch: Responses of the whole plant to deficits in
water and nitrogen supply. Adv. Bot. Res. 22, 228–300 (1996) 96
846. McIlroy, D.N., Zhang, D., Kranov, Y., Norton, M.G.: Nanosprings. Appl. Phys. Lett. 79,
1540–1542 (2001) 116
847. McLean, D.R., Graham, B.P.: Stability in a mathematical model of neurite elongation. Math.
Med. Biol. 23(2), 101–117 (2006) 85
848. McLean, D.R., van Ooyen, A., Graham, B.P.: Continuum model for tubulin-driven neurite
elongation. Neurocomp. 58, 511–516 (2004) 85
849. McLellan, W., Koopman, H., Rommel, S., Read, A., Potter, C., Nicolas, J., Westgate, A.,
Pabst, D.: Ontogenetic allometry and body composition of harbour porpoises (phocoena
phocoena, l.) from the western north atlantic. J. Zool. 257(4), 457–471 (2002), 20
850. McMahon, J., Goriely, A., Tabor, M.: Spontaneous cavitation in growing elastic membranes.
Math. Mech. Solids 15, 57–77 (2010) 452
851. McMahon, J., Goriely, A., Tabor, M.: Nonlinear morphoelastic plates I: the genesis of residual
stress. Math. Mech. Solids 16, 833–871 (2011) 46, 252, 257
852. McMahon, J., Goriely, A., Tabor, M.: Nonlinear morphoelastic plates II: Exodus to buckled
states. Math. Mech. Solids 16, 833–871 (2011) 46, 252, 257
853. McMillen, T., Goriely, A.: Tendril perversion in intrinsically curved rods. J. Nonlinear Sci.
12(3), 241–281 (2002) 125, 149, 158, 159, 160
854. McParland, B.E., Macklem, P.T., Pare, P.D.: Airway hyperresponsiveness: From molecules
to bedside. J. Appl. Physiol. 95, 426–434 (2003) 531
855. Meade, J., Cowin, S., Klawitter, J., Van Buskirk, W., Skinner, H.: Bone remodeling due to
continuously applied loads. Calcified Tissue Int. 36, 25–30 (1984) 47
856. Medawar, P.B.: The laws of biological growth. Nature 148(3665), 772–774 (1941) 16, 22
857. Medawar, P.B.: The shape of the human being as a function of time. Proc. Roy. Soc. Lond.
B 132(867), 133–141 (1944) 17
858. Mehdizadeh, M., Nowroozzadeh, M.H.: Outdoor activity and myopia. Ophthalmology 116,
1229–1230 (2009). Author reply 568
859. Meineke, F.A., Potten, C.S., Loeffler, M.: Cell migration and organization in the intestinal
crypt using a lattice-free model. Cell proliferation 34(4), 253–266 (2001) 55
860. Meinhardt, H.: Models of biological pattern formation, vol. 6. Academic Press London
(1982) 564
861. Meinhardt, H.: The algorithmic beauty of sea shells. springer. Heidelberg, New York (with
PC-software) (2003) 176
862. Meinhardt, H., Klingler, M.: A model for pattern formation on the shells of molluscs. J.
Theor. Bio. 126(1), 63–89 (1987) 189
863. Meinhardt, H., Meinhardt, H.: Models of biological pattern formation, vol. 6. Academic
Press London (1982) 189
864. Melnik, A.V., Da Rocha, H.B., Goriely, A.: On the modeling of fiber dispersion in fiber-
reinforced elastic materials. Int. J. Non Linear Mech. (2015) 328, 340, 344
865. Melnik, A.V., Goriely, A.: Dynamic fiber reorientation in a fiber-reinforced hyperelastic
material. Math. Mech. Solids 18(6), 634–648 (2013) 321, 421, 423
866. Mendelson, N.H.: Helical Bacillus subtilis macrofibers: morphogenesis of a bacterial mul-
ticellular macroorganism. Proc. Natl. Acad. Sci. USA 75, 2478–2482 (1978) 64, 65, 170
616 References
867. Mendelson, N.H.: Bacterial macrofibers: the morphogenesis of complex multicellular bac-
terial forms. Sci. Progress Oxford 74, 425–441 (1990) 170
868. Mendelson, N.H.: Bacillus subtilis macrofibres, colonies and bioconvection patterns use
different strategies to achieve multicellular organization. Environ. Microbiol. 1, 471–477
(1999) 170
869. Menninger, E.A.: Flowering vines of the world: An encyclopedia of climbing plants. Hearth-
side Press Incorporated, New York (1970) 149
870. Menzel, A.: Modelling of anisotropic growth in biological tissues. Biomech. Model.
Mechanobiol. 3(3), 147–171 (2005) 321, 420, 421
871. Menzel, A.: Anisotropic remodelling of biological tissues. Mechanics of Biological Tissue
pp. 91–104 (2006) 328
872. Menzel, A.: Frontiers in inelastic continuum mechanics. Habilitationsschrift, University of
Kaiserslautern (2007) 419
873. Menzel, A., Kuhl, E.: Frontiers in growth and remodeling. Mech. Res. Commun. 42, 1–14
(2012) 49, 401, 419
874. Menzel, D. (ed.): A Mathematical Model of Cytoskeletal Dynamics and Morphogenesis in
Acetabularia, vol. The Cytoskeleton of the Algae. CRC Press, Boca Raton (1992) 96
875. Merks, R., Hoekstra, A., Kaandorp, J., Sloot, P.: Models of coral growth: spontaneous branch-
ing, compactification and the Laplacian growth assumption. J. Theor. Biol. 224(2), 153–166
(2003) 570
876. Merks, R.M.H., Glazier, J.A.: A cell-centered approach to developmental biology. Physica
A 352(1), 113–130 (2005) 53
877. Merodio, J., Ogden, R.W.: Instabilities and loss of ellipticity in fiber-reinforced compressible
non-linearly elastic solids under plane deformation. Int. J. Solids and Struct. 40(18), 4707–
4727 (2003) 324
878. Merodio, J., Ogden, R.W.: Mechanical response of fiber-reinforced incompressible non-
linearly elastic solids. Int. J. Nonlinear Mech. 40(2), 213–227 (2005) 324
879. Merodio, J., Ogden, R.W.: The influence of the invariant I8 on the stress–deformation and el-
lipticity characteristics of doubly fiber-reinforced non-linearly elastic solids. Int. J. Nonlinear
Mech. 41(4), 556–563 (2006) 324
880. Merodio, J., Ogden, R.W.: Extension, inflation and torsion of a residually stressed circular
cylindrical tube. Continuum Mech. Therm. pp. 1–18 (2015) 349
881. Merodio, J., Ogden, R.W., Rodriguez, J.: The influence of residual stress on finite deformation
elastic response. Int. J. Nonlinear Mech. 56, 43–49 (2013) 350
882. Mertens, J.J.: The shape and stabilization of the crimp in false-twist thermoplastic yarn. In:
Bulk, stretch and texture: Papers of the 61st annual conference of the textile institute (1966)
152
883. van der Meulen, M.C.H., Huiskes, R.: Why mechanobiology? A survey article. J. Biomech.
35(4), 401–414 (2002) 419
884. Michel, J.C., Moulinec, H., Suquet, P.: Effective properties of composite materials with
periodic microstructure: a computational approach. Comput. Methods Appl. Mech. Eng.
172(1), 109–143 (1999) 562
885. Michell, J.H.: On the stability of a bent and twisted wire. Messenger of Math. 11, 181–184
(1889) 133
886. Middlebrook, M.J., Preston, R.D.: Spiral growth and spiral structure: III. wall structure in
the growth zone of Phycomyces. Biochim. Biophys. Acta. 9, 32–37 (1952) 321, 542
887. Mihai, L.A., Chin, L., Janmey, P.A., Goriely, A.: A comparison of hyperelastic constitutive
models applicable to brain and fat tissues. J. R. Soc. Interface 12(110), 20150,486 (2015)
295
888. Mihai, L.A., Goriely, A.: Positive or negative Poynting effect? the role of adscititious in-
equalities in hyperelastic materials. Proc. Roy. Soc. Lond. A 467(2136), 3633–3646 (2011)
293, 403
889. Mihai, L.A., Goriely, A.: Numerical simulation of shear and the Poynting effects by the
finite element method: An application of the generalised empirical inequalities in non-linear
elasticity. Int. J. Nonlinear Mech. 49, 1–14 (2012) 293
References 617
890. Miller, A.E., Jones, M.C., Smith, D.W.: Tension: the basis of umbilical cord growth. J.
Pediatr. 101, 844 (1982) 63
891. Miller Jr, J.W.: The elastic properties of helical springs. Phys. Rev. 14(3), 129 (1902) 121
892. Milton, G.W.: The theory of composites. Cambridge University Press, Cambridge (2002)
562
893. Miura, T., Shiota, K., Morriss-Kay, G., Maini, P.K.: Mixed-mode pattern in doublefoot
mutant mouse limb—Turing reaction–diffusion model on a growing domain during limb
development. J. Theor. Biol. 240(4), 562–573 (2006) 566
894. Moessinger, A.C., Blanc, W.A., Marone, P.A., Polsen, D.C.: Umbilical cord length as an
index of fetal activity: experimental study and clinical implications. Pediatr. Res. 16, 109–
112 (1982) 154
895. Mogilner, A., Allard, J., Wollman, R.: Cell polarity: quantitative modeling as a tool in cell
biology. Science 336(6078), 175–179 (2012) 51, 59
896. Mogilner, A., Odde, D.: Modeling cellular processes in 3D. Trends Cell Biol. 21(12), 692–
700 (2011) 58
897. Mohl, H.v.: Principles of the anatomy and physiology of the vegetable cell. John van voorst,
paternoster row. , London (1852) 151
898. Duhamel du Monceau, H.R.: La physique des arbres (The physics of trees). L. Guérin et L.
-F. Delatour, Paris, France (1758) 5
899. Money, N.P.: Wishful thinking of turgor revisited: The mechanics of fungal growth. Fungal
Genet. Biol. 21, 173–187 (1997) 5, 237
900. Money, N.P., J., H.R.: Confirmation of a link between fungal pigmentation, turgor pressure,
and pathogenicity using a new method of turgor measurement. Fungal Genet. Biol. 20,
217–227 (1996) 226
901. Moon, H., Truesdell, C.: Interpretation of adscititious inequalities through the effects pure
shear stress produces upon an isotropic elastic solid. Arch. Rat. Mech. Anal. 55(1), 1–17
(1974) 292, 293
902. Moore, A., Burt, A.S.: On the locus and nature of the forces causing gastrulation in the
embryos of Dendraster excentricus. J. Exp. Zool. 82(1), 159–171 (1939) 45
903. Moore, B.J., Hilliam, C.C., Verburgt, L.M., Wiggs, B.R., Vedal, S., Paré, P.D.: Shape and
position of the complete dose-response curve for inhaled methacholine in normal subjects.
Am. J. Respir. Crit. Care. Med. 154, 642–648 (1996) 531
904. Mora, T., Boudaoud, A.: Buckling of swelling gels. Eur. Phys. J. E Soft Matter 20(2), 119–
124 (2006) 3
905. Morita, R.: Finite element analysis of a double membrane tube (DMS-tube) and its implica-
tion for gastropod shell morphology. J. Morphol. 207(1), 81–92 (1991) 197
906. Morita, R.: Mechanical constraints on aperture form in gastropods. J. Morphol. 207(1),
93–102 (1991) 197
907. Morrill, J.B., Santos, L.L.: A scanning electron microscopical overview of cellular and extra-
cellular patterns during blastulation and gastrulation in the sea urchin, Lytechinus variegatus.
In: The Cellular and Molecular Biology of Invertebrate Development, pp. 3–33. Univ. South
Carolina Press Columbia, SC (1985) 432
908. Moseley, H.: On the geometrical forms of turbinated and discoid shells. Philos. Trans. Roy.
Soc. Lond. B 128, 351–370 (1838) 188
909. Moss, M.L., Young, R.W.: A functional approach to craniology. American journal of physical
anthropology 18(4), 281–292 (1960) 571
910. Moulia, B., Coutand, C., Lenne, C.: Posture control and skeletal mechanical acclimation in
terrestrial plants: implications for mechanical modeling of plant architecture. Am. J. Bot.
93(10), 1477 (2006) 63, 508
911. Moulia, B., Fournier, M.: The power and control of gravitropic movements in plants: a
biomechanical and systems biology view. J. Exp. Bot. 60(2), 461–486 (2009) 65
912. Moulton, D.E., Goriely, A.: Circumferential buckling instability of a growing cylindrical
tube. J. Mech. Phys. Solids 59, 525–537 (2011) 528, 529
618 References
913. Moulton, D.E., Goriely, A.: Possible role of differential growth in airway wall remodeling
in asthma. J. Appl. Physiol. 110(4), 1003–1012 (2011) 532
914. Moulton, D.E., Goriely, A.: Surface growth kinematics via local curve evolution. J. Math.
Biol. 68(1-2), 81–108 (2014) 7, 8, 65, 176, 182, 195, 197, 201, 204, 572
915. Moulton, D.E., Goriely, A., Chirat, R.: Mechanical growth and morphogenesis of seashells.
J. Theor. Biol. 311, 69–79 (2012) 7, 176, 183, 196, 197, 199, 200
916. Moulton, D.E., Goriely, A., Chirat, R.: The morpho-mechanical basis of ammonite form. J.
Theor. Biol. 364, 220–230 (2015) 7, 188, 197, 199, 200, 201
917. Moulton, D.E., Lessinnes, T., Goriely, A.: Morphoelastic rods Part I: A single growing elastic
rod. J. Mech. Phys. Solids 61(2), 398–427 (2012) 46, 98, 140, 141
918. Moulton, D.E., Lessinnes, T., O’Keeffe, S., Dorfmann, L.A., Goriely, A.: The elastic secrets
of the chameleon tongue. Proc. Roy. Soc. Lond. A 472(2188), 20160,030 (2016) 577
919. Mow, V.C., Kuei, S.C., Lai, W.M., Armstrong, C.G.: Biphasic creep and stress relaxation of
articular cartilage in compression: Theory and experiments. J. Biomech. Eng. 102, 73–84
(1980) 395
920. Mueller-Klieser, W., Schreiber-Klais, S., Walenta, S., Kreuter, M.: Bioactivity of well-
defined green tea extracts in multicellular tumor spheroids. Int. J. Oncol. 21(6), 1307–1315
(2002) 38, 72
921. Mullen, J.L., Turk, E., Johnson, K., Wolverton, C., Ishikawa, H., Simmons, C., Soll, D.,
Evans, M.L.: Root-growth behavior of the Arabidopsis mutant rgr1. Roles of gravitropism
and circumnutation in the waving/coiling phenomenon. Plant Physiol. 118(4), 1139–1145
(1998) 63
922. Müller, I., H., S.: Inflating a rubber balloon. Math. Mech. Solids 7, 569–577 (2002) 304, 442
923. Müller, M.M., Ben Amar, M., Guven, J.: Conical defects in growing sheets. Phys. Rev. Lett.
101(15), 156,104 (2008) 252
924. Müller, N.J.C.: Handbuck der botanik. Heidelberg (1880) 24, 26, 41, 47, 90
925. Mura, T.: Micromechanics of defects in solids. Springer Science & Business Media (2013)
400
926. Murray, J.D.: Mathematical Biology, vol. 2. springer (2002) 47
927. Murray, J.D.: Mathematical Biology: I. An Introduction. Springer (2007) 564
928. Murray, J.D., Cook, J., Tyson, R., Lubkin, S.R.: Spatial pattern formation in biology: I.
Dermal wound healing. II. Bacterial patterns. J. Franklin Inst. 335(2), 303–332 (1998) 565
929. Murray, J.D., Maini, P.K., Tranquillo, R.T.: Mechanochemical models for generating bio-
logical pattern and form in development. Physics Reports 171(2), 59–84 (1988) 47, 565
930. Muto, H., Yabe, N., Asami, T., Hasunuma, K., Yamamoto, K.T.: Overexpression of constitu-
tive differential growth 1 gene, which encodes a RLCKVII-subfamily protein kinase, causes
abnormal differential and elongation growth after organ differentiation in Arabidopsis. Plant
Physiol. 136, 3124–3133 (2004) 483
931. Nagashima, T., Shirakuni, T., Rapoport, S.I.: A two-dimensional, finite element analysis of
vasogenic brain edema. Neurol. Med. Chir. 30, 1–9 (1990). doi:10.2176/nmc.30.1395
932. Naghdi, P.: Foundations of elastic shell theory. In: I.N. Sneddon, K. Hill (eds.) Progress in
Solid Mechanics, vol. 4, pp. 1–90. North Hollande Publishing Cy (1963) 251
933. Naghdi, P.: A critical review of the state of finite plasticity. ZAMP 41(3), 315–394 (1990)
48, 401
934. Naghdi, P.M.: The theory of shells and plates. Springer (1973) 256
935. Nakayama, K., Segur, H., Wadati, M.: Integrability and the motion of curves. Phys. Rev.
Lett. 69, 2603 (1992) 102
936. Nardinocchi, P., Teresi, L.: On the active response of soft living tissues. Journal of Elasticity
88(1), 27–39 (2007) 576
937. Nath, U., Crawford, B.C.W., Carpenter, R., Coen, E.: Genetic control of surface curvature.
Science 299(5611), 1404–1407 (2003) 256, 559
938. Navier, C.L.M.H., de Saint-Venant, B.: Résumé des Leçons sur l’Application de la Mé-
canique, vol. 2. Dunod (1864) 489
References 619
939. Negrón-Marrero, P., Antman, S.: Singular global bifurcation problems for the buckling of
anisotropic plates. Proc. Roy. Soc. Lond. A 427(1872), 95–137 (1990) 450
940. Nelson, D.: Experimental methods for determining residual stresses and strains in various
biological structures. Exp. Mech. pp. 1–14 (2013) 45
941. Nelson, D.: Review of methods for determining residual stresses in biological materials. In:
Proceedings of the 2012 Annual Conference on Exp. Appl. Mech. ., Exp. Appl. Mech., vol. 4,
pp. 173–182. Springer (2013) 45
942. Neukirch, S., Frelat, J., Goriely, A., Maurini, C.: Vibrations of post-buckled rods: The sin-
gular inextensible limit. J. Sound Vib. 331(3), 704–720 (2012) 115, 139
943. Neukirch, S., Goriely, A., Hausrath, A.C.: Chirality of coiled coils: Elasticity matters. Phys.
Rev. Lett. 100(3), 038,105 (2008) 116
944. Neukirch, S., Goriely, A., Hausrath, A.C.: Elastic coiled-coils act as energy buffers in the
ATP synthase. Int. J. Nonlinear Mech. 43, 1064–1073 (2008) 116
945. von Neumann, J., Burks, A.W.: Theory of self-reproducing automata. University of Illinois
Press (1966) 51
946. Newcombe, F.C.: The cause and conditions of lysigenous cavity-formation. Ann. Bot. Lond.
8, 403–421 (1894) 481
947. Newell, A.C., Shipman, P.D.: Plants and Fibonacci. J. Stat. Phys. 121(5-6), 937–968 (2005)
567
948. Newell, A.C., Shipman, P.D., Sun, Z.: Phyllotaxis: cooperation and competition between
mechanical and biochemical processes. J. Theor. Biol. 251(3), 421–439 (2008) 567
949. Newell, R.L.: The calcar femorale: a tale of historical neglect. Clinical Anatomy 10(1),
27–33 (1997) 34
950. Nielsen, S.L., Enriquez, S., Duarte, C., Sand-Jensen, K.: Scaling maximum growth rates
across photosynthetic organisms. Funct. Ecol. pp. 167–175 (1996) 376
951. Nieman, B.J., Blank, M.C., Roman, B.B., Henkelman, R.M., Millen, K.J.: If the skull fits:
magnetic resonance imaging and microcomputed tomography for combined analysis of brain
and skull phenotypes in the mouse. Physiological genomics 44(20), 992–1002 (2012) 572
952. Nienhaus, U., Aegerter-Wilmsen, T., Aegerter, C.M.: Determination of mechanical stress
distribution in Drosophila wing discs using photoelasticity. Mechanisms of development
126(11), 942–949 (2009) 559
953. Niimi, A., Matsumoto, H., Amitani, R., Nakano, Y., Mishima, M., Minakuchi, M., Nishimura,
K., Itoh, H., Izumi, T.: Airway wall thickness in asthma assessed by computed tomography.
Am. J. Respir. Crit. Care Med. 162, 1518–1523 (2000) 531
954. Niklas, K.J.: Plant Biomechanics: An Engineering Approach to Plant Form and Function.
The University of Chicago Press (1992) 487, 501
955. Niklas, K.J.: Influence of tissue density-specific mechanical properties on the scaling of
plant height. Ann. Bot. Lond. 72, 173–179 (1993) 484
956. Niklas, K.J.: Plant Allometry: the Scaling of Form and Process. University of Chicago Press
(1994) 22
957. Niklas, K.J.: Plant allometry: is there a grand unifying theory? Biol. Rev. 79(4), 871–889
(2004) 15, 21, 22
958. Niklas, K.J., Buchman, S.L.: The allometry of saguaro height. Am. J. Bot. pp. 1161–1168
(1994) 20
959. Nizette, M., Goriely, A.: Towards a classification of Euler-Kirchhoff filaments. J. Math.
Phys. 40, 2830–2866 (1999) 115, 123
960. Noble, P.B., Pascoe, C.D., Lan, B., Ito, S., Kistemaker, L.E.M., Tatler, A.L., Pera, T., Brook,
B.S., Gosens, R., West, A.R.: Airway smooth muscle in asthma: Linking contraction and
mechanotransduction to disease pathogenesis and remodelling. Pulm. Pharm. Therap. 29(2),
96–107 (2014) 538
961. Noble, P.B., Sharma, A., McFawn, P.K., Mitchell, H.W.: Airway narrowing in porcine
bronchi with and without lung parenchyma. Europ. Resp. J. 26(5), 804–811 (2005) 532
962. Nolan, D.R., Gower, A.L., Destrade, M., Ogden, R.W., McGarry, J.P.: A robust anisotropic
hyperelastic formulation for the modelling of soft tissue. J. Mech. Behav. Biomed. Mat. 39,
48–60 (2014) 350
620 References
963. Noll, W.: Materially uniform simple bodies with inhomogeneities. Arch. Rat. Mech. Anal
27(1), 1–32 (1967) 361
964. Nonami, H., Boyer, J.S.: Wall extensibility and cell hydraulic conductivity decrease in en-
larging stem tissues at low water potentials. Plant Physiol. 93(4), 1610–1619 (1990) 94
965. Norton, L.: A Gompertzian model of human breast cancer growth. Cancer research 48(24
Part 1), 7067–7071 (1988) 17
966. Novozhilov, V.: The theory of thin shells [translated by P. G. Lowe from the 1951 Russian
edition]. P. Noordhoff (1964) 251
967. Nowinski, J.L.: On the elastic stability of thick columns. Acta Mech. 7(4), 279–286 (1969)
489
968. Nutt, J.J.: Diseases and Deformities of the Foot. EB Treat (1913) 37
969. Nye, J.F.: Some geometrical relations in dislocated crystals. Acta metallurgica 1(2), 153–162
(1953) 372
970. Nyst, P.H.: Conchyliologie des terrains tertiaires de la Belgique. 1. partie. Terrain Pliocene
Scaldisien. Ann. d. Musee R. d’Histoire Nat. de Belg. Sir. Paleont (1878) 189
971. O’Dea, R.D., Byrne, H.M., Waters, S.L.: Continuum modelling of in vitro tissue engineering:
a review. In: Computational Modeling in Tissue Engineering, pp. 229–266. Springer (2012)
396
972. O’Dea, R.D., Waters, S.L., Byrne, H.M.: A multiphase model for tissue construct growth
in a perfusion bioreactor. Mathematical medicine and biology: a journal of the IMA 27(2),
95–127 (2010) 396
973. Ogawa, Y.: Study on the strength of human lower air-passages. J. Kyoto Prefect Univ. Med
66, 781–800 (1959) 538
974. Ogden, R.W.: Large deformation isotropic elasticity - On the correlation of theory and
experiment for incompressible rubberlike solids. Proc. Roy. Soc. Lond. A 326 (1972) 294,
304, 442
975. Ogden, R.W.: Non-linear Elastic Deformations. Dover, New york (1984) 262, 274, 281, 285,
287, 289, 304, 311, 313, 443, 456, 460
976. Ogden, R.W.: Recent advances in the phenomenological theory of rubber elasticity. Rubber
Chem. Technol. 59(3), 361–383 (1986) 294
977. Ogden, R.W.: Nonlinear elasticity with application to material modelling. Polish Academy
of Sciences (2003) 280
978. Ogden, R.W., Saccomandi, G., Sgura, I.: Fitting hyperelastic models to experimental data.
Comp. Mech. 34(484–502) (2004) 294, 559
979. Ogden, R.W., Schulze-Bauer, C.: Phenomenological and structural aspects of the mechanical
response of arteries. ASME Applied Mech. 242, 125–140 (2000) 519
980. Okada, K., Shimura, Y.: Reversible root tip rotation in Arabidopsis seedlings induced by
obstacle-touching stimulus. Science 250(4978), 274 (1990) 63
981. Okamoto, T.: Analysis of heteromorph ammonoids by differential geometry. Palaeontology
31, 35–52 (1988) 192
982. Okamoto, T.: Developmental regulation and morphological saltation in the heteromorph
ammonite Nipponites. Paleobiology 14(3), 272–286 (1988) 192
983. O’Keeffe, S.G., Moulton, D.E., Waters, S.L., Goriely, A.: Growth-induced axial buckling of
a slender elastic filament embedded in an isotropic elastic matrix. Int. J. Nonlinear Mech.
56, 94–104 (2013) 143, 145
984. Oldfather, W.A., Ellis, C.A., Brown, D.M.: Leonhard Euler’s Elastic Curves. Isis 20(1),
72–160 (1933) 487
985. Olson, W.K., Swigon, D., Coleman, B.D.: Implications of the dependence of the elastic
properties of DNA on nucleotide sequence. Philos. Trans. Roy. Soc. Lond. A 362(1820),
1403–1422 (2004) 136
986. Omens, J., McCulloch, A., Criscione, J.: Complex distributions of residual stress and
strain in the mouse left ventricle: experimental and theoretical models. Biomech. Model.
Mechanobiol. 1(4), 267–277 (2003) 44, 45
References 621
987. Ong, J.J., O’Reilly, O.M.: On the equations of motion for rigid bodies with surface growth.
Int. J. Eng. Sci. 42(19), 2159–2174 (2004) 572
988. van Ooyen, A., Graham, B.P., Ramakers, G.J.A.: Competition for tubulin between growing
neurites during development. Neurocomp. 38(40), 73–78 (2001) 85
989. Opperman, L.A.: Cranial sutures as intramembranous bone growth sites. Developmental
dynamics 219(4), 472–485 (2000) 571
990. O’Reilly, O.M.: Modeling Nonlinear Problems in the Mechanics of Strings and Rods: The
Role of the Balance Laws. Springer New York (2016) 107
991. O’Reilly, O.M., Tresierras, T.N.: On the evolution of intrinsic curvature in rod-based models
of growth in long slender plant stems. Int. J. Solids Struct. (2010) 97, 129
992. O’Reilly, O.M., Tresierras, T.N.: On the Static Equilibria of Branched Elastic Rods. Preprint
submitted to Int. J. Eng. Sci. (2010) 97
993. Ortega, J.K., Lesh-Laurie, G.E., Espinosa, M.A., Ortega, E.L., Manos, S.M., Cunning, M.D.,
Olson, J.E.: Helical growth of stage-IVb sporangiophores of Phycomyces blakesleeanus: the
relationship between rotation and elongation growth rates. Planta 216(4), 716–722 (2003)
543
994. Ortega, J.K.E.: Augmented growth equation for cell wall expansion. Plant Physiol. 79, 318–
320 (1985) 93, 96
995. Ortega, J.K.E.: Growth rate regulation of cells with walls: The sporangiophores of Phy-
comyces blakesleeanus used as a model system. Recent Res. Dev. Plant Physiol. 5, 1–19
(2012) 551
996. Ortega, J.K.E., Gamow, R.I.: The problem of handedness reversal during the spiral growth
of Phycomyces. J. Theor. Biol. 47(2), 317–332 (1974) 543, 548
997. Ortega, J.K.E., Munoz, C.M., Blakley, S.E., Truong, J.T., Ortega, E.L.: Stiff mutant genes
of Phycomyces affect turgor pressure and wall mechanical properties to regulate elongation
growth rate. Front. Plant Sci. 3 (2012) 551
998. Ortega, J.K.E., Zehr, E.G., Keanini, R.G.: In vivo creep and stress relaxation experiments to
determine the wall extensibility and yield threshold for the sporangiophores of Phycomyces.
Biophys. J. 56(3), 465 (1989) 92
999. Osada, T., Ross-Murphy, S.B.: Intelligent gels. Scientific American 268, 82–87 (1993) 403
1000. Osborne, W.A., Sutherland, W.: The elasticity of rubber balloons and hollow viscera. Proc.
Roy. Soc. Lond. B 81(551), 485–499 (1909) 261, 262, 446
1001. Osserman, R.: A survey of minimal surfaces. Courier Corporation (2002) 254
1002. O’Toole, M., Lamoureux, P., Miller, K.E.: A physical model of axonal elongation: Force,
viscosity, and adhesions govern the mode of outgrowth. Biophys. J. 94(7), 2610–2620 (2008)
82, 556
1003. Ozakin, A., Yavari, A.: Affine development of closed curves in Weitzenböck manifolds and
the Burgers vector of dislocation mechanics. Math. Mech. Solids 19, 299–307 (2014) 372
1004. Paetsch, C., Trimmer, B.A., Dorfmann, L.A.: A constitutive model for active–passive tran-
sition of muscle fibers. Int. J. Non-Linear Mech. 47(2), 377–387 (2012) 577
1005. Palsson, E.: A three-dimensional model of cell movement in multicellular systems. Future
Generation Computer Systems 17(7), 835–852 (2001) 56
1006. Palsson, E., Othmer, H.G.: A model for individual and collective cell movement in Dic-
tyostelium discoideum. Proc. Nat. Acad. Sci. USA 97(19), 10,448–10,453 (2000) 55
1007. Pampaloni, F., Reynaud, E.G., Stelzer, E.H.K.: The third dimension bridges the gap between
cell culture and live tissue. Nature Rev. Mol. Cell Biol. 8(10), 839–845 (2007) 432
1008. Pan, F., Beatty, M.F.: Instability of a Bell constrained cylindrical tube under end thrust-part
1: Theoretical development. Math. Mech. Solids 2(3), 243–273 (1997) 490
1009. Pancera, P., Ribul, M., Presciuttini, B., Lechi, A.: Prevalence of carotid artery kinking in
590 consecutive subjects evaluated by Echocolordoppler. Is there a correlation with arterial
hypertension? J. Intern. Med. 248(1), 7 (2000) 510
1010. Pandolfi, A., Holzapfel, G.A.: Three-dimensional modeling and computational analysis of
the human cornea considering distributed collagen fibril orientations. J. Biomech. Engineer-
ing 130(6), 061,006 (2008) 340
622 References
1011. Pandolfi, A., Vasta, M.: Fiber distributed hyperelastic modeling of biological tissues. Mech.
Mater. 44, 151–162 (2012) 340
1012. Panfilov, A.V., Keldermann, R.H., Nash, M.P.: Self-organized pacemakers in a coupled
reaction-diffusion-mechanics system. Phys. Rev. Lett. 95(25), 258,104 (2005) 576
1013. Panyukov, S., Rabin, Y.: Fluctuating elastic rings: statics and dynamics. Phys. Rev. E. 64,
011,909 (2001) 136
1014. Papastavrou, A., Steinmann, P., Kuhl, E.: On the mechanics of continua with boundary
energies and growing surfaces. J. Mech. Phys. Solids (2013) 402
1015. Paré, P.D., McParland, B.E., Seow, C.Y.: Structural basis for exaggerated airway narrowing.
Can. J. Physiol. Pharmacol. 85(7), 635–638 (2007) 538
1016. Park, Y.B., Cosgrove, D.J.: A revised architecture of primary cell walls based on biome-
chanical changes induced by substrate-specific endoglucanases. Plant Physiology 158(4),
1933–1943 (2012) 89
1017. Parker, J.: Morphogens, nutrients, and the basis of organ scaling. Evolution & development
13(3), 304–314 (2011) 558
1018. Patan, S.: Vasculogenesis and angiogenesis as mechanisms of vascular network formation,
growth and remodeling. J. Neuro-Oncol. 50, 1–15 (2000) 475
1019. Patel, D.J., Fry, D.L.: Longitudinal tethering of arteries in dogs. Circ. Res. 19(6), 1011–1021
(1966) 509, 519
1020. Pathmanathan, P., Cooper, J., Fletcher, A., Mirams, G., Murray, P., Osborne, J., Pitt-Francis,
J., Walter, A., Chapman, S.J.: A computational study of discrete mechanical tissue models.
Physical biol. 6(3), 036,001 (2009) 56
1021. Pearce, S.: Effect of strain-energy function and axial prestretch on the bulges, necks and
kinks forming in elastic membrane tubes. Math. Mech. Solids 17(8), 860–875 (2012) 304
1022. Pearce, S., Fu, Y.: Characterization and stability of localized bulging/necking in inflated
membrane tubes. IMA J. Appl. Math. 75(4), 581–602 (2010) 304
1023. Pence, T.J., Tsai, H.: On the cavitation of a swollen compressible sphere in finite elasticity.
Int. J. Nonlinear Mech. 40(2-3), 307–321 (2005) 403, 451
1024. Pence, T.J., Tsai, H.: Swelling induced cavitation of elastic spheres. Math. Mech. Solids 11,
527–551 (2006) 403, 452
1025. Pence, T.J., Wineman, A.: On some connections between equivalent single material and
mixture theory models for fiber reinforced hyperelastic materials. Int. J. Nonlinear Mech.
47(2), 285–292 (2012) 539
1026. Pennybacker, M.F., Shipman, P.D., Newell, A.C.: Phyllotaxis: Some progress, but a story
far from over. Physica D 306, 48–81 (2015) 565, 566, 567
1027. Penta, R., Ambrosi, D., Shipley, R.: Effective governing equations for poroelastic growing
media. Q. J. Mech. Appl. Math. 67(1), 69–91 (2014) 395
1028. Pericak-Spector, K.A., Sivaloganathan, J., Spector, S.J.: An explicit radial cavitation solution
in nonlinear elasticity. Math. Mech. Solids 7(1), 87–93 (2002) 451
1029. Peters, W.S., Hagemann, W., Tomos, D.A.: What makes plants different? principles of ex-
tracellular matrix function in ‘soft’ plant tissues. Comp. Biochem. Phys. A 125(2), 151–167
(2000) 484
1030. Peters, W.S., Tomos, A.D.: The history of tissue tension. Ann. Bot-London 77(6), 657–65
(1996) 42, 484
1031. Peters, W.S., Tomos, D.A.: The epidermis still in control. Acta Botanica 109, 264–267 (1996)
39, 42
1032. Pezzulla, M., Shillig, S.A., Nardinocchi, P., Holmes, D.P.: Morphing of geometric composites
via residual swelling. Soft Matter 11, 5812–5820 (2015) 256, 257
1033. Pezzulla, M., Smith, G.P., Nardinocchi, P., Holmes, D.P.: Geometry and mechanics of thin
growing bilayers. Soft Matter 12(4435–4442), 1–5 (2016) 256
1034. Pezzuto, S., Ambrosi, D.: Active contraction of the cardiac ventricle and distortion of the
microstructural architecture. International journal for numerical methods in biomedical en-
gineering 30(12), 1578–1596 (2014) 577
References 623
1035. Pfeffer, W.: Pflanzenphysiologie: ein Handbuch der Lehre vom Stoffwechsels und
Kraftwechsels in der Pflanze, vol. 3. W. Engelmann (1904) 64, 65
1036. Pfister, B.J., Iwata, A., Meaney, D.F., Smith, D.H.: Extreme stretch growth of integrated
axons. J. Neurosci. 24(36), 7978–7983 (2004) 80
1037. Pilbeam, D., Gould, S.J.: Size and scaling in human evolution. Science 186, 892–901 (1974)
20
1038. Pipkin, A.C.: Integration of an equation in membrane theory. ZAMP 19(5), 818–819 (1968)
212
1039. Pipkin, A.C.: Lectures on viscoelastic theory. Springer (1972) 556
1040. Polignone, D.A., Horgan, C.O.: Cavitation for incompressible anisotropic nonlinearly elastic
spheres. J. Elasticity 33(1), 27–65 (1993) 451
1041. Politi, A.Z., Donovan, G.M., Tawhai, M.H., Sanderson, M.J., Lauzon, A.M., Bates,
J.H.T., Sneyd, J.: A multiscale, spatially distributed model of asthmatic airway hyper-
responsiveness. J. Theor. Biol. 266, 614–624 (2010) 533
1042. Ponte Castañeda, P.: The effective mechanical properties of nonlinear isotropic composites.
1043. Pressman, E., Huberman, M., Alaoni, B., Jafe, M.J.: Thigmomorphogenesis: the effect of
mechanical perturbation and ethrel on stem pithiness in tomato [Lycopersicon esculentum
(Mill.) Plants]. Ann. Bot-London 52(1), 93–100 (1983) 483
1044. Prikazchikov, D.A., Rogerson, G.A.: On surface wave propagation in incompressible, trans-
versely isotropic, pre-stressed elastic half-spaces. Int. J. Eng. Sci. 42(10), 967–986 (2004)
521
1045. Proseus, T.A., Ortega, J.K.E., Boyer, S., J.: Separating growth from elastic deformation
during cell enlargement. Plant Physiol. 119, 775–784 (1999) 92, 93, 94
1046. Prosser, J.I., Tough, A.J.: Growth mechanisms and growth kinetics of filamentous microor-
ganisms. Crit. Rev. Biotechnol. 10, 253–274 (1991) 237, 246
1047. Pucci, E., Saccomandi, G.: On universal relations in continuum mechanics. Continuum
Mech. Therm. 9(2), 61–72 (1997) 305
1048. Putter, A.: Studien über physiologische Ähnlichkeit. vi. Wachstumsahmlichkeiten. Pflagers.
Arch. Gres. Physiol 180, 298–340 (1920) 12
1049. Putz, F.E., Mooney, H.A. (eds.): Biomechanical studies of vines, vol. The biology of climbing
plants. Cambridge University Press, Cambridge (1991) 162, 163, 165
1050. Qi, A., Zheng, X., Du, C., An, B.: A cellular automaton model of cancerous growth. J. Theor.
Biol. 161(1), 1–12 (1993) 52
1051. Qian, H., White, J.H.: Terminal twist induced continuous writhe of a circular rod with
intrinsic curvature. J. Biomol. Struct. Dyn. 16, 663–669 (1998) 136
1052. Qiu, G.Y., Pence, T.J.: Remarks on the behavior of simple directionally reinforced incom-
pressible nonlinearly elastic solids. J. Elasticity 49(1), 1–30 (1997) 324
1053. Quetelet, A.: Sur l’homme et le développement de ses facultés ou essai de physique sociale.
Bachelier (1835) 11
1054. Rachev, A.: Theoretical study of the effect of stress-dependent remodeling on arterial geom-
etry under hypertensive conditions. J. Biomech. 30, 819–827 (1997) 43, 400, 409, 510,
516
1055. Rajagopal, K., Srinivasa, A.: On the thermomechanics of materials that have multiple natural
configurations: Part II: Twinning and solid to solid phase transformation. ZAMP 55(6),
1074–1093 (2004) 396
1056. Rajagopal, K.R., Srinivasa, A.R.: On the thermomechanics of materials that have multiple
natural configurations Part I: Viscoelasticity and classical plasticity. ZAMP 55(5), 861–893
(2004) 396
1057. Ramasubramanian, A., Taber, L.A.: Computational modeling of morphogenesis regulated
by mechanical feedback. Biomech. Model. Mechanobiol. 7(2), 77–91 (2008) 410
1058. Ranft, J., Basan, M., Elgeti, J., Joanny, J.F., Prost, J., Jülicher, F.: Fluidization of tissues by
cell division and apoptosis. Proc. Natl. Acad. Sci. USA 107(49), 20,863–20,868 (2010) 55,
575
624 References
1059. Rao, I.J., Humphrey, J.D., Rajagopal, K.R.: Biological growth and remodeling: A uniaxial
example with possible application to tendons and ligaments. Comput. Model. Eng. Sci.
4(3/4), 439–456 (2003) 63
1060. Rassier, D.E., MacIntosh, B.R., Herzog, W.: Length dependence of active force production
in skeletal muscle. J. Appl. Physiol. 86(5), 1445–1457 (1999) 574
1061. Raup, D.M.: The geometry of coiling in gastropods. Proc. Natl. Acad. Sci. USA 47(4),
602–609 (1961) 189, 192
1062. Raup, D.M., Graus, R.R.: General equations for volume and surface area of a logarithmically
coiled shell. Math. Geol. 4(4), 307–316 (1972) 189, 192
1063. Rausch, M., Dam, A., Göktepe, S., Abilez, O., Kuhl, E.: Computational modeling of growth:
systemic and pulmonary hypertension in the heart. Biomech. Model. Mechanobiol. 10(6),
799–811 (2011) 34
1064. Raven, J.A.: Into the voids: The distribution, function, development and maintenance of gas
spaces in plants. Ann. Bot-London 78, 137–142 (1996) 481
1065. Ray, P.: Radioautographic study of cell wall deposition in growing plant cells. J. Cell Biol
35, 659–74 (1967) 89
1066. Ray, P.M., Green, P.B., Cleland, R.: Role of turgor in plant cell growth. Nature 239(5368),
163–164 (1972) 90
1067. Recho, P., Jerusalem, A., Goriely, A.: Growth, collapse, and stalling in a mechanical model
for neurite motility. Phys. Rev. E 93(3), 032,410 (2016) 556, 574
1068. Redington, A.E., Howarth, P.H.: Airway wall remodelling in asthma. Thorax 52, 310–312
(1997) 530
1069. Reeves, N.D., Maganaris, C.N., Ferretti, G., Narici, M.V.: Influence of 90-day simulated
microgravity on human tendon mechanical properties and the effect of resistive counter-
measures. J. Appl. Physiology 98(6), 2278–2286 (2005) 37
1070. Reina, C., Conti, S.: Kinematic description of crystal plasticity in the finite kinematic frame-
work: a micromechanical understanding of F = Fe Fp . J. Mech. Phys. Solids 67, 40–61
(2014) 563
1071. Reina, C., Schlömerkemper, A., Conti, S.: Derivation of Fe Fp as the continuum limit of
crystalline slip. J. Mech. Phys. Solids (2016) 563
1072. Reiner, M.: The Deborah number. Physics today 17(1), 62 (1964) 555
1073. Reinhardt, M.O.: Das wachsthum der pilzhyphen. Jahrbücher für wissenschaftliche botanik
23 (1892) 5, 6, 234
1074. Reissner, E.: Symmetric bending of shallow shells of revolutions. J. Math. Mech, 7(121-140)
(1958) 450
1075. Renardy, Y., Coward, A.V., Papageorgiou, D., Sun, S.M. (eds.): The nonlocal dynamics
of stretching, buckling filaments, vol. Advances in multi-fluid flows. SIAM, Philadelphia
(1996) 98
1076. Rice, J.R.: Inelastic constitutive relations for solids: an internal-variable theory and its ap-
plication to metal plasticity. J. Mech. Phys. Solids 19(6), 433–455 (1971) 408
1077. Rice, S.H.: The bio-geometry of mollusc shells. Paleobiology 24, 133–149 (1998) 176, 188,
189, 194
1078. Richards, F.: A flexible growth function for empirical use. J. Exp. Bot. 10(2), 290–301 (1959)
13
1079. Richtsmeier, J.T., Aldridge, K., DeLeon, V.B., Panchal, J., Kane, A.A., Marsh, J.L., Yan,
P., Cole, T.M.: Phenotypic integration of neurocranium and brain. J. Exp. Zool. B 306(4),
360–378 (2006) 572
1080. Richtsmeier, J.T., DeLeon, V.B.: Morphological integration of the skull in craniofacial anom-
alies. Orthodontics & craniofacial research 12(3), 149–158 (2009) 572
1081. Richtsmeier, J.T., Flaherty, K.: Hand in glove: brain and skull in development and dysmor-
phogenesis. Acta neuropathologica 125(4), 469–489 (2013) 571
1082. da Riva Ricci, D., Kendrick, B.: Computer modelling of hyphal tip growth in fungi. Can. J.
Bot. 20, 2455–2462 (1972) 239
References 625
1083. Riveros, F., Martufi, G., Gasser, T.C., Rodriguez-Matas, J.F.: On the impact of intraluminal
thrombus mechanical behavior in AAA passive mechanics. Ann. Biomed. Eng. pp. 1–12
(2015) 304
1084. Rivlin, R.S.: Large elastic deformations of isotropic materials. I. fundamental concepts.
Philos. Trans. Roy. Soc. A 240(822), 459–490 (1948) 261
1085. Rivlin, R.S.: Large elastic deformations of isotropic materials. III. some simple problems in
cylindrical polar co-ordinates. Philos. Trans. Roy. Soc. A 240(823), 509–525 (1948) 301
1086. Rivlin, R.S.: Large elastic deformations of isotropic materials. IV. further developments of
the general theory. Philos. Trans. Roy. Soc. A 241(835), 379–397 (1948) 301
1087. Rivlin, R.S., Saunders, D.W.: Large Elastic Deformations of Isotropic Materials. VII. Ex-
periments on the Deformation of Rubber. Philos. Trans. Roy. Soc. A 243(865), 251–288
(1951) 221
1088. Robertson, N.: The growth process in fungi. Annu. Rev. Phytopathol. 6(1), 115–136 (1968)
63
1089. Robertson, R.: Snail handedness-the coiling directions of gastropods. Natn. Geogr. Res.
Explor. 9, 104–119 (1993) 150
1090. Robertson-Tessi, M., El-Kareh, A., Goriely, A.: A mathematical model of tumor–immune
interactions. J. Theor. Biol. 294, 56–73 (2012) 17
1091. Rodriguez, E.K., Hoger, A., McCulloch, A.: Stress-dependent finite growth in soft elastic
tissue. J. Biomech. 27, 455–467 (1994) 48, 127, 353, 400, 404, 412
1092. Rodriguez, J., Merodio, J.: A new derivation of the bifurcation conditions of inflated cylin-
drical membranes of elastic material under axial loading. application to aneurysm formation.
Mech. Res. Commun. 38(3), 203–210 (2011) 304
1093. Roelofsen, P.A.: The origin of spiral growth in Phycomyces sporangiophores. Recueil Trav.
Bot. Neerl. 42, 72–110 (1950) 121
1094. Roelofsen, P.A.: Ultrastructure of the wall in growing cells. Adv. Bot. Res. 2, 69–149 (1965)
545
1095. Roelofsen, P.A., Houwink, A.: Architecture and growth of the primary cell wall in some
plant hairs and in the Phycomyces sporangiophore. Acta. Bot. Neerl. 2, 218–225 (1953) 63,
542
1096. Roesler, H.: The history of some fundamental concepts in bone biomechanics. J. Biomech.
20(11), 1025–1034 (1987) 35
1097. Rolland-Lagan, A.G., Bangham, J.A., Coen, E., et al.: Growth dynamics underlying petal
shape and asymmetry. Nature 422(6928), 161–163 (2003) 256, 559
1098. Roose, T., Chapman, S.J., Maini, P.K.: Mathematical models of avascular tumor growth.
SIAM Review 49(2), 179 (2007) 38, 72, 76
1099. Roose, T., Netti, P.A., Munn, L.L., Boucher, Y., Jain, R.K.: Solid stress generated by spheroid
growth estimated using a linear poroelasticity model. Microvascular R. 66, 204–212 (2003)
396
1100. Roux, W.: Der zuchtende Kampf der Teile, oder die Teilauslese im Organismus (Theorie
der’funktionellen Anpassung’). Wilhelm Engelmann, Leipzig (1881) 35
1101. Roy, C.S.: The elastic properties of the arterial wall. The Journal of physiology 3(2), 125–159
(1881) 509
1102. Rubin, J., Rubin, C., Jacobs, C.R.: Molecular pathways mediating mechanical signaling in
bone. Gene 367, 1–16 (2006) 36
1103. Ruhland, W. (ed.): Les mouvements d’exploration et enroulement des plantes volubiles, vol.
Handbuch des Pflanzenphysiologie (1962) 163
1104. S., M.R., Kier, W.M.: The functional morphology of starfish tube feet: the role of a crossed-
fiber helical array in movement. Biol. B. 188(2), 197–209 (1995) 336
1105. Saccomandi, G.: Universal results in finite elasticity. Nonlinear Elasticity: Theory and Ap-
plications 283, 97 (2001) 296
1106. Saccomandi, G.: Nonlinear elasticity for soft fibrous materials. In: Nonlinear Mechanics of
Soft Fibrous Materials, pp. 121–173. Springer (2015) 293
626 References
1107. Sachs, J.: Text-book of botany, morphological and physiological. Clarendon, Oxford (1875)
22, 31, 42, 47, 163, 481, 482
1108. Sacks, M.S.: Biaxial mechanical evaluation of planar biological materials. J. Elasticity 61,
199–246 (2000) 293
1109. Sacks, M.S., Sun, W.: Multiaxial mechanical behavior of biological materials. Annu. Rev.
Biomed. Eng. 5(1), 251–284 (2003) 293
1110. Sadik, S., Angoshtari, A., Goriely, A., Yavari, A.: A geometric theory of nonlinear morphoe-
lastic shells. J. Nonlinear Sci. pp. 1–50 (2016) 258
1111. Sadik, S., Yavari, A.: On the origins of the idea of the multiplicative decomposition of the
deformation gradient. Math. Mech. Solids p. 1081286515612280 (2015) 353
1112. Sadler, T.W.: Langman’s medical embryology. Lippincott Williams & Wilkins (2011) 570
1113. Saintillan, D., Shelley, M.J.: Active suspensions and their nonlinear models. Abbreviation
Title C.R. Phys. 14(6), 497–517 (2013) 575
1114. Saleuddin, A.S.M., Petit, H.P.: The mode of formation and the structure of the periostracum.
In: A.S.M. Saleuddin, K.M. Wilbur (eds.) The Mollusca, Physiology, pp. 199–233. Academic
Press, New York (1983) 188
1115. Sanchez-Palencia, E.: Homogenization in mechanics. a survey of solved and open problems.
Rend. Sem. Mat. Univ. Politec. Torino 44(1), 1–45 (1986) 562
1116. Santini, M.T., Rainaldi, G.: Three-dimensional spheroid model in tumor biology. Pathobi-
ology 67(3), 148–157 (2000) 38, 72
1117. Santulli, C., Patel, S.I., Jeronimidis, G., Davis, F.J., Mitchell, G.R.: Development of smart
variable stiffness actuators using polymer hydrogels. Smart Mater. Struct. 14, 434–440
(2005) 338
1118. Sarrus, F., Rameaux, J.: Application des sciences accessoires et principalement des math-
ématiques à la physiologie générale. Bulletin de l’Académie Royale Médecine, Paris 3,
1094–1100 (1839) 14
1119. Sass, F., Bouché, C.: Dubbels Taschenbuch für den Maschinenbau. Springer, Berlin (1953)
113
1120. Saul, R.A. (ed.): Correlations between umbilical cord structure–composition and normal
and abnormal fetal development, vol. Proceedings of the Grennwood Genetics Conference.
Jacobs Press, Clinton, SC (1988) 154
1121. Saunders, P.T., Trinci, A.P.J.: Determination of tip shape in fungal hyphae. J. Gener. Micro-
biol. 110, 469 – 473 (1970) 239
1122. Savazzi, E.: Shellgen: A basic program for the modeling of molluscan shell ontogeny and
morphogenesis. Comput. Geosci. 11(5), 521–530 (1985) 192
1123. Scammon, R.E.: The first seriatim study of human growth. Am. J. Phys. Anthrop. 10(3),
329–336 (1927) 10
1124. Scher, J.L., Holbrook, M.M., Silk, W.K.: Temporal and spatial patterns of twining force and
lignification in Ipomoea purpurea. Planta 213, 192–198 (2001) 149
1125. Schlik, T.: Modeling superhelical DNA: recent analytical and dynamical approaches. Curr.
Opin. Struct. Biol. 5, 245–262 (1995) 135
1126. Schmidt-Nielsen, K.: Scaling: why is animal size so important? Cambridge University Press
(1984) 12, 15
1127. Schmundt, D., Stitt, M., Jähne, B., Schurr, U.: Quantitative analysis of the local rates of
growth of dicot leaves at a high temporal and spatial resolution, using image sequence
analysis. Plant J. 16(4), 505–514 (1998) 25, 26
1128. Schneider, C.L.: On the nastic and traumatic responses in the pea test. Am. J. Bot. 29(3),
201–206 (1942) 42
1129. Schnur, J.M.: Lipid tubules: a paradigm for molecularly engineered structures. Science 262,
1669–1676 (1993) 116
1130. Schöne, B.R., Rodland, D.L., Wehrmann, A., Heidel, B., Oschmann, W., Zhang, Z., Fiebig,
J., Beck, L.: Combined sclerochronologic and oxygen isotope analysis of gastropod shells
(Gibbula cineraria, north sea): life-history traits and utility as a high-resolution environ-
mental archive for kelp forests. Marine Biology 150(6), 1237–1252 (2007) 194
References 627
1131. Schopfer, P.: Biomechanics of plant growth. Am. J. Bot. 93(10), 1415–1425 (2006) 92
1132. Schulgasser, K., Witztum, A.: The mechanism of spiral grain formation in trees. Wood Sci.
Technol. 41(2), 133–156 (2006) 542
1133. Schulze-Bauer, C.A.J., Mörth, C., Holzapfel, G.A.: Passive biaxial mechanical response of
aged human iliac arteries. J. Biomech. Eng. 125, 395–406 (2003) 509, 519
1134. Scianna, M., Preziosi, L.: Cellular Potts models: multiscale extensions and biological appli-
cations. CRC Press (2013) 53
1135. Secomb, T.W., Gross, J.F.: Flow of red blood cells in narrow capillaries: role of membrane
tension. Int. J. Microcirc: Clin. Exp. 2, 229 – 240 (1983) 208
1136. Seow, C.Y., Wang, L., Pare, P.D.: Airway narrowing and internal structural constraints. J.
Appl. Physiology 88(2), 527–533 (2000) 532
1137. Serikawa, K.A., Mandoli, D.F.: An analysis of morphogenesis of the reproductive whorl of
Acetabularia acetabulum. Planta 207(1), 96–104 (1998) 256
1138. Setton, L.A., Tohyama, H., Mow, V.C.: Swelling and curling behaviors of articular cartilage.
J. Biomech. Eng. 120(3), 355–361 (1998) 45
1139. Shapiro, J.A., Dworkin, M. (eds.): Mycelial life of Streptomyces coelicolor A3(2) and its
relatives, vol. Bacteria as multicellular organisms. Oxford University Press, Oxford (1997)
63, 64
1140. Sharon, E., Efrati, E.: The mechanics of non-Euclidean plates. Soft Matter 6(22), 5693–5704
(2010) 256
1141. Sharon, E., Marder, M., Swinney, H.L.: Leaves, flowers and garbage bags: Making waves.
Am. Sci. 92(3), 254–261 (2004) 256
1142. Sharon, E., Roman, B., Swinney, H.L.: Geometrically driven wrinkling observed in free
plastic sheets and leaves. Phys. Rev. E 75(4), 46,211 (2007) 256
1143. Shaw, S.L., Dumais, J., Long, S.R.: Cell Surface Expansion in Polarly Growing Root Hairs
of Medicago truncatula. Plant Physiol. 124, 959–969 (2000) 234, 239
1144. Shelley, M.J.: The dynamics of microtubule/motor-protein assemblies in biology and
physics. Annu. Rev. Fluid Mech. 48, 487–506 (2016) 575
1145. Shergold, O.A., Fleck, N.A., Radford, D.: The uniaxial stress versus strain response of pig
skin and silicone rubber at low and high strain rates. Int. J. Impact Eng. 32, 1384–1402
(2006) 295
1146. Sherman, J., Morrison, W.J.: Adjustment of an inverse matrix corresponding to changes in
the elements of a given column or a given row of the original matrix. Ann. Math. Stat. 20(4),
621–621 (1949) 402
1147. Shingleton, A.: Allometry: the study of biological scaling. Nature Education Knowledge
1(9), 2 (2010) 12
1148. Shipley, R.J., Chapman, S.J.: Multiscale modelling of fluid and drug transport in vascular
tumours. Bull. Math. Biol. 72(6), 1464–1491 (2010) 562
1149. Shipley, R.J., Jones, G.W., Dyson, R.J., Sengers, B.G., Bailey, C.L., Catt, C.J., Please, C.P.,
Malda, J.: Design criteria for a printed tissue engineering construct: a mathematical homog-
enization approach. J. Theor. Biol. 259(3), 489–502 (2009) 562
1150. Shipman, P., Goriely, A.: On the dynamics of helical strips. Phys. Rev. E 61, 4508 (2000)
138
1151. Shipman, P., Newell, A.: Phyllotactic patterns on plants. Phys. Rev. Lett. 92, 168,102 (2004)
567
1152. Shipman, P.D., Newell, A.C.: Polygonal planforms and phyllotaxis on plants. J. Theor. Biol.
236(2), 154–197 (2005) 567
1153. Shirinifard, A., Gens, J.S., Zaitlen, B.L., Popławski, N.J., Swat, M., Glazier, J.A.: 3D multi-
cell simulation of tumor growth and angiogenesis. PloS one 4(10), e7190 (2009) 58
1154. Shraiman, B.I.: Mechanical feedback as a possible regulator of tissue growth. Proc. Natl.
Acad. Sci. USA 102(9), 3318 (2005) 400
1155. Silk, W.K.: Growth rate patterns which maintain a helical tissue tube. J. Theor. Biol. 138(3),
311–327 (1989) 149
628 References
1156. Silk, W.K.: On the curving and twining of stems. Environ. Exper. Bot. 29, 95–109 (1989)
65, 136
1157. Silk, W.K.: Steady form from changing cells. Int. J. Plant Sci. 153(3), 49–58 (1992) 25
1158. Silk, W.K., Bogeat-Triboulot, M.B.: Deposition rates in growing tissue: Implications for
physiology, molecular biology, and response to environmental variation. Plant and Soil pp.
1–17 (2013) 25
1159. Silk, W.K., Haidar, A.: Growth of the stem of Pharbitis nil: Analysis of longitudinal and
radial components. Phisiol. Veg. 24 (1986) 25
1160. Silk, W.K., Holbrook, N.M.: The importance of frictional interactions in maintaining the
stability of the twining habit. Am. J. Bot. 92(11), 1820–1826 (2005) 149, 163, 165, 167
1161. Silk, W.K., Holbrook, N.M., Jessup, C.: The forceful nature of twining vines. Plant biome-
chanics pp. 638–646 (2000) 65
1162. Silk, W.K., Hubbard, M.: Axial forces and normal distributed loads in twining stems of
morning glory. J. Biomech. 24, 599–606 (1991) 149, 163, 164
1163. Silva, S.D., Chouduri, A.R.: A theoretical model for tilts of bipolar magnetic regions. Astron.
Astrophys. 272, 621–633 (1993) 63
1164. Simkiss, K., Wilbur, K.M.: Biomineralization. Academic Press, New York (1989) 188
1165. Simon, B.R.: Multiphase poroelastic finite element models for soft tissue structures. Appl.
Mech. Rev 45(6), 191–218 (1992) 395
1166. Simpson, H.C., Spector, S.J.: On barrelling instabilities in finite elasticity. J. Elasticity 14,
103–125 (1984) 490
1167. Simpson, H.D., Mortimer, D., Goodhill, G.J.: Theoretical models of neural circuit develop-
ment. Curr. Top. Dev. Biol. p. 1 (2009) 85
1168. Sivaloganathan, J.: On cavitation and degenerate cavitation under internal hydrostatic pres-
sure. Proc. Roy. Soc. London A 455(1990), 3645–3664 (1999) 451
1169. Sivaloganathan, J., Spector, S.J.: On cavitation, configurational forces and implications for
fracture in a nonlinearly elastic material. J. Elasticity 67(1), 25–49 (2002) 451
1170. Skacel, P., Bursa, J.: Numerical implementation of constitutive model for arterial layers with
distributed collagen fibre orientations. Comput. Methods Biomech. Biomed. Engin. 18(8),
816–828 (2015) 340
1171. Skalak, R.: Growth as a finite displacement field. In: D.E. Carlson, R.T. Shield (eds.) Pro-
ceedings of the IUTAM Symposium on Finite Elasticity. Martinus Nijhoff, The Hague (1981)
39, 48
1172. Skalak, R., Dasgupta, G., Moss, M., Otten, E., Dullemeijer, P., Vilmann, H.: Analytical
description of growth. J. Theor. Biol. 94(3), 555–577 (1982) 48
1173. Skalak, R., Hoger, A.: Kinematics of surface growth. J. Math. Biol. 35, 869–907 (1997) 7,
48, 176, 570, 572
1174. Skalak, R., Tozeren, A., Zarda, R.P., Chien, S.: Strain energy function of red blood cell
membranes. Biophys. J. 13, 245–264 (1973) 208, 214
1175. Skalak, R., Zargaryan, S., Jain, R.K., Netti, P.A., Hoger, A.: Compatibility and the genesis
of residual stress by volumetric growth. J. Math. Biology 34, 889–914 (1996) 39, 48, 358
1176. Smith, B., Schaffer, T., Viani, M., Thompson, J., Frederick, N., Kindt, J., Belcher, A., Stucky,
G., Morse, D., Hansma, P.: Molecular mechanistic origin of the toughness of natural adhe-
sives, fibres and composites. Nature 399(6738), 761–763 (1999) 188
1177. Smith, N.P., Nickerson, D.P., Crampin, E.J., Hunter, P.J.: Multiscale computational modelling
of the heart. Acta Numerica 13, 371–431 (2004) 576
1178. Smith, R.S., Guyomarc’h, S., Mandel, T., Reinhardt, D., Kuhlemeier, C., Prusinkiewicz, P.:
A plausible model of phyllotaxis. Proc. Natl. Acad. Sci. USA 103(5), 1301–1306 (2006)
566
1179. Sokolnikoff, I.S.: Mathematical Theory of Elasticity. McGraw-Hill, New York (1956) 111,
112
1180. Southwell, R.V.: On the analysis of experimental observations in problems of elastic stability.
Proc. Roy. Soc. London A 135(828), 601–616 (1932) 490
References 629
1181. Spalding, E.P., Miller, N.D.: Image analysis is driving a renaissance in growth measurement.
Curr. Opin. Plant. Biol. 16(1), 100–104 (2013) 25
1182. Speck, T., Burgert, I.: Plant stems: Functional design and mechanics. Annu. Rev. Mater. Res.
41, 169–193 (2011) 25
1183. Spencer, A.J.M.: Deformations of Fibre-Reinforced Materials. Oxford (1972) 321
1184. Spencer, A.J.M.: Continuum theory of the mechanics of fibre-reinforced composites, vol.
282. Springer New York: (1984) 321, 323
1185. Spruit, H.C.: Motion of magnetic flux tubes in the solar convection zone and chromosphere.
Astron. Astrophys. 98, 155–160 (1981) 63
1186. Steel, G.G.: Growth kinetics of tumours: cell population kinetics in relation to the growth
and treatment of cancer. Oxford: Clarendon Press (1977) 376
1187. Steele, C.R.: Shell stability related to pattern formation in plants. J. Appl. Mech. 67, 237–247
(2000) 566
1188. Steigmann, D.J.: Fluid films with curvature elasticity. Arch. Rat. Mech. Anal. 150(2), 127–
152 (1999) 208
1189. Steigmann, D.J.: Effects of fiber bending and twisting resistance on the mechanics of fiber-
reinforced elastomers. In: Nonlinear Mechanics of Soft Fibrous Materials, pp. 269–305.
Springer (2015) 326
1190. Stein, A.: The deformation of a rod of growing biological material under longitudinal com-
pression. J. Appl. Math. Mech. 59(1), 139–146 (1995) 48
1191. Steinberg, M.S.: Differential adhesion in morphogenesis: a modern view. Curr. Opin. Genet.
Dev. 17(4), 281–286 (2007) 51
1192. Stevens, C.F.: Darwin and Huxley revisited: the origin of allometry. J. Biol. 8, 14 (2009) 21
1193. Stevens, P.S.: Patterns in Nature. Little, Brown (1974) 197
1194. Stewart, P.S., Waters, S.L., El Sayed, T., Vella, D., Goriely, A.: Wrinkling, creasing, and
folding in fiber-reinforced soft tissues. Extreme Mechanics Letters (2016) 573
1195. Stewart, S.A., German, R.Z.: Sexual dimorphism and ontogenetic allometry of soft tissues
in rattus norvegicus. J. Morphology 242(1), 57–66 (1999) 17
1196. Stojanovic, R.: On the stress relation in non-linear thermoelasticity. Int. J. Non-Linear Mech.
4(3), 217–233 (1969) 353
1197. Stojanovic, R., Djuric, S., Vujosevic, L.: On finite thermal deformations. Arch. Mech. Stosow
16, 103–108 (1964) 353
1198. Stolarska, M.A., Kim, Y., Othmer, H.G.: Multi-scale models of cell and tissue dynamics.
Philos. T. R. Soc. A 367(1902), 3525–3553 (2009) 561
1199. Stone, J.R.: The evolution of ideas: a phylogeny of shell models. Am. Nat. pp. 904–929
(1996) 189
1200. Stratz, C.H.: Naturgeschichte des Menschen. F. Enke (1904) 11, 18
1201. Sun, D.N., Gu, W.Y., Guo, X.E., Lai, W.M., Mow, V.C.: A mixed finite element formulation
of triphasic mechano-electrochemical theory for charged, hydrated biological soft tissues.
Int. J. Numer. Meth. Eng. 45, 1375–1402 (1999) 395
1202. Sun, W., Sacks, M.S.: Finite element implementation of a generalized fung-elastic constitu-
tive model for planar soft tissues. Biomech. Model. Mechanobiol. 4(2-3), 190–199 (2005)
295
1203. Sutherland, R.M.: Cell and environment interactions in tumor microregions: the multicell
spheroid model. Science 240(4849), 177 (1988) 38, 72
1204. Synge, J.L., Chien, W.Z.: The intrinsic theory of elastic shells and plates. In: von Kármán
anniv. vol., pp. 103–120. Cal. Inst. Tech., Pasadena (1941) 251
1205. Szabó, A., Merks, R.M.H.: Cellular Potts modeling of tumor growth, tumor invasion, and
tumor evolution. Front. Oncol. 3(87.10), 3389 (2013) 53
1206. Taber, L.A.: Biomechanics of growth, remodeling and morphogenesis. Appl. Mech. Rev.
48, 487–545 (1995) 9, 34, 36, 49, 400, 409
1207. Taber, L.A.: An optimization principle for vascular radius including the effects of smooth
muscle tone. Biophys. J. 74(1), 109–114 (1998) 32
630 References
1208. Taber, L.A.: Biomechanical growth laws for muscle tissues. J. Theor. Biol. 193, 201–213
(1998) 48, 400, 409
1209. Taber, L.A.: Theoretical study of Beloussov’s hyper-restoration hypothesis for mechanical
regulation of morphogenesis. Biomech. Model. Mechanobiol. 7(6), 427–441 (2008) 78, 79,
410
1210. Taber, L.A.: Towards a unified theory for morphomechanics. Philos. Trans. Roy. Soc. Lond.
A 367(1902), 3555–3583 (2009) 410
1211. Taber, L.A., Chabert, S.: Theoretical and experimental study of growth and remodeling in
the developing heart. Biomech. Model. Mechanobiol. 1(1), 29–43 (2002) 409
1212. Taber, L.A., Eggers, D.W.: Theoretical study of stress-modulated growth in the aorta. J.
Theor. Biol. 180, 343–357 (1996) 400, 409
1213. Taber, L.A., Humphrey, J.D.: Stress-modulated growth, residual stress, and vascular hetero-
geneity. J. Biomech. Eng. 123, 528–535 (2001) 419, 514
1214. Tadano, S., Yamada, S.: Residual stress and structural anisotropy of cortical bone. In: Me-
chanics of Biological Systems and Materials, Volume 4, pp. 117–122. Springer (2014) 45
1215. Taiz, L., Zeiger, E.: Plant Physiol., 4th ed edn. Sinauer Associates, Sunderland, Mass. (2006)
47, 90
1216. Takamizawa, K.: Stress-free configuration of a thick-walled cylindrical model of the artery:
an application of Riemann geometry to the biomechanics of soft tissues. J. Appl. Mech.
58(3), 840–842 (1991) 353
1217. Takamizawa, K., Hayashi, K.: Strain energy density function and uniform strain hypothesis
for arterial mechanics. J. Biomech. 20(1), 7–17 (1987) 353
1218. Takamizawa, K., Matsuda, T.: Kinematics for bodies undergoing residual stress and its
applications to the left ventricle. J. Appl. Mech. 57(2), 321–329 (1990) 353, 363
1219. Takano, M., Takahashi, H., Suge, H.: Mechanical stress and gibberellin: Regulation of hol-
lowing induction in the stem of a bean plant, Phaseolus vulgaris L. Plant and Cell Physiol.
36(1), 101–108 (2001) 483, 485
1220. Tanaka, H., Tomita, H., Takasu, A., Hayashi, T., Nishi, T.: Morphological and kinetic evo-
lution of surface patterns in gels during the swelling process: Evidence of dynamic pattern
ordering. Phys. Rev. Lett. 68(18), 2794 (1992) 403
1221. Tanaka, T., Sun, S.T., Hirokawa, Y., Katayama, S., Kucera, J., Hirose, Y., Amiya, T.: Me-
chanical instability of gels at the phase transition. Nature 325(6107), 796–798 (1987) 403
1222. Taylor, R., Zheng, C., Jackson, R., Doll, J., Chen, J., Holzbaur, K., Besier, T., Kuhl, E.:
The phenomenon of twisted growth: humeral torsion in dominant arms of high performance
tennis players. Comp. Methods Biomech. Biomed. Eng. 12(1), 83–93 (2009) 28
1223. Tenney, S.M.: A tangled web: Young, Laplace, and the surface tension law. Physiology 8(4),
179–183 (1993) 33, 212
1224. Teodosiu, C.: Elastic models of crystal defects. Editura Academiei (1982) 372
1225. Tessier-Lavigne, M., Goodman, C.S.: The molecular biology of axon guidance. Science
274(5290), 1123 (1996) 85
1226. Thimann, K.V., Schneider, C.L.: Differential growth in plant tissues. Am. J. Bot. 25(8),
627–641 (1938) 42
1227. Thoma, R.: Untersuchungen über die Histogenese und Histomechanik des Gefäßsystems.
Enke (1893) 32
1228. Thompson, J.M.T., Heijden, G.H.M.v.d., Neukirch, S.: Supercoiling of DNA plasmids: me-
chanics of the generalized ply. Proc. Roy. Soc. Lond. A 458, 959–985 (2002) 63
1229. Thomson, W.T., Tait, P.G.: Treatise on Natural Philosophy. Cambridge (1867) 119, 133
1230. Thomspon, D.W.: On Growth and Form: The Complete Revised Edition. Dover, New York
(1992) 7, 8, 12, 20, 150, 188, 216
1231. Tiero, A., Tomassetti, G.: On morphoelastic rods. Math. Mech. Solids p. 10.
1177/1081286514546178 (2014) 408
1232. Tilby, M.J.: Helical shape and wall synthesis in a bacterium. Nature 266 (1977) 156
1233. Timoshenko, S.: Theory of Plates and Shells. McGraw-Hill, New York (1940) 217
1234. Timoshenko, S.: History of Strength of Materials. Dover, New York (1983) 487
References 631
1235. Timoshenko, S.P.: Analysis of bi-metal thermostats. J. Opt. Soc. Am. 11(3), 233–255 (1925)
64
1236. Timoshenko, S.P., Gere, J.M.: Theory of Elastic Stability. Mc Graw-Hill, New York (1961)
487, 489, 503
1237. Tobias, I., Coleman, B.D., Lembo, M.: A class of exact dynamical solutions in the elastic
rod model of DNA with implications for the theory of fluctuations in the torsional motion
of plasmids. J. Chem. Phys. 105, 2517–2526 (1996) 138
1238. Tobias, I., Olson, W.K.: The effect of intrinsic curvature on supercoiling–Predictions of
elasticity theory. Biopolymers 33, 639–646 (1993) 138
1239. Todhunter, I.: The Elastical Researches of Barré de Saint-Venant. University Press (1889)
489
1240. Todhunter, I.: A History of the Theory of Elasticity and of the Strength of Materials: From
Galilei to the Present Time. University press (1893) 487
1241. Tongen, A., Goriely, A., Tabor, M.: Biomechanical model for appressorial design in Mag-
naporthe grisea. J. Theor. Biol. 240(1), 1–8 (2006) 228
1242. Tozeren, A., Skalak, R.: Interaction of stress and growth in a fibrous tissue. J. Theor. Biol.
130(3), 337–350 (1988) 48
1243. Tranquillo, R.T., Murray, J.: Mechanistic model of wound contraction. J. Surg. Res. 55(2),
233–247 (1993) 47, 353
1244. Tranquillo, R.T., Murray, J.D.: Continuum model of fibroblast-driven wound contraction:
inflammation-mediation. J. Theor. Biol. 158(2), 135–172 (1992) 353, 565
1245. Treloar, L.R.G.: The elasticity of a network of long-chain molecules—II. Trans. Faraday
Soc. 39, 241–246 (1943) 293
1246. Treloar, L.R.G.: The elasticity of a network of long-chain molecules. I. Rubber Chem.
Technol. 16(4), 746–751 (1943) 293
1247. Treloar, L.R.G.: The Physics of Rubber Elasticity. Clarendon Press, Oxford (1975) 241
1248. Triantafyllidis, N., Abeyaratne, R.: Instabilities of a finitely deformed fiber-reinforced elastic
material. J. Appl. Mech. 50, 149–156 (1983) 321, 324
1249. Trinci, A.P.J., Saunders, P.T.: Tip growth of fungal hyphae. J. Gen. Microbiol. 103, 243–248
(1977) 239
1250. Truesdell, C.: Sulle basi della termomeccanica, rend. accad. lincei 22, 33–88, 158–166.
english transl. Rational Mechanics of Materials. Intl. Sci. Rev. Ser pp. 292–305 (1957) 393
1251. Truesdell, C., Noll, W.: The Non-Linear Field Theories of Mechanics. Springer (2004) 262,
281, 288, 291
1252. Truesdell, C., Toupin, R.: The Classical Field Theories. Springer (1960) 393
1253. Truesdell, C.A.: A program of physical research in classical mechanics. ZAMP 3(2), 79–95
(1952) 261, 291
1254. Tsapis, N., Dufresne, E.R., Sinha, S.S., Riera, C.S., Hutchinson, J.W., Mahadevan, L., Weitz,
D.A.: Onset of buckling in drying droplets of colloidal suspensions. Phys. Rev. Lett. 94(1),
018,302 (2005) 432
1255. Tsoularis, A., Wallace, J.: Analysis of logistic growth models. Math. Bisoc. 179(1), 21–55
(2002) 13
1256. Tsui, Y., Clyne, T.: An analytical model for predicting residual stresses in progressively
deposited coatings part 1: Planar geometry. Thin Solid Films 306(1), 23–33 (1997) 45
deposited coatings part 2: Cylindrical geometry. Thin solid films 306(1), 34–51 (1997) 45
deposited coatings part 3: Further development and applications. Thin Solid Films 306(1),
52–61 (1997) 45
1259. Turing, A.M.: The chemical basis of morphogenesis. Philos. Trans. Roy. Soc. London B
237(641), 37–72 (1952) 564
1260. Turner, L., Ryu, W.S., Berg, H.C.: Real-time imaging of fluorescent flagellar filaments. J.
Bacteriol. 182 (2000) 155
632 References
1261. Turner, M.S., Briehl, R.W., Ferrone, F.A., Josephs, R.: Twisted protein aggregates and dis-
ease: The stability of sickle hemoglobin fibers. Phys. Rev. Lett. 90, 128,103 (2003) 116
1262. Turner, R.T.: Invited review: what do we know about the effects of spaceflight on bone? J.
Appl. Physiol. 89(2), 840–847 (2000) 28
1263. Tvergaard, V.: Effect of void size difference on growth and cavitation instabilities. J. Mech.
Phys. Solids 44(8), 1237–1253 (1996) 451
1264. Twitty, V.C., Schwind, J.L.: The growth of eyes and limbs transplanted heteroplastically
between two species of Amblystoma. J. Exp. Zoology 59(1), 61–86 (1931) 557, 558
1265. Ubukata, T.: Stacking increments: a new model and morphospace for the analysis of bivalve
shell growth. Hist. Biol. 15(4), 303–321 (2001) 570
1266. Ueda, Y., Fukuda, K., Nakacho, K., Endo, S.: A new measuring method of residual stresses
with the aid of finite element method and reliability of estimated values. Transactions of
JWRI 4(2), 123–131 (1975) 400
1267. Urdy, S., Goudemand, N., Bucher, H., Chirat, R.: Allometries and the morphogenesis of
the molluscan shell: a quantitative and theoretical model. J. Exp. Zool. B 314(4), 280–302
(2010) 194
1268. Vaishnav, R.N., Vossoughi, J.: Distribution of stresses and of strain-energy density through
the wall thickness in a canine aortic segment. Circ. Res. 32, 577–583 (1973) 512
1269. Vaishnav, R.N., Vossoughi, J.: Estimation of residual strains in aortic segments. In: C.W.
Hall (ed.) Biomedical Engineering II: Recent Developments: Proceedings of the Second
Southern Biomedical Engineering Conference, pp. 330–333. Pergamon Press, NY (1983)
43
1270. Vaishnav, R.N., Vossoughi, J.: Residual stress and strain in aortic segments. J. Biomech.
20(3), 235–237 (1987) 43
1271. Valentín, A., Humphrey, J.D., Holzapfel, G.A.: A finite element-based constrained mix-
ture implementation for arterial growth, remodeling, and adaptation: Theory and numerical
verification. International J. Numer. Meth. Biomed. Engineering 29(8), 822–849 (2013) 396
1272. Van Veen, M.P., Van Pelt, J.: Neuritic growth rate described by modeling microtubule dy-
namics. Bull. Math. Biol. 56(2), 249–273 (1994) 85
1273. Vande Geest, J.P., Sacks, M.S., Vorp, D.A.: The effects of aneurysm on the biaxial mechanical
behavior of human abdominal aorta. J. Biomech. 39(7), 1324–1334 (2006) 293
1274. Vandiver, R., Goriely, A.: Tissue tension and axial growth of cylindrical structures in plants
and elastic tissues. Europhys. Lett. 84, 58,004 (2008) 40, 46, 475, 484
1275. Vandiver, R., Goriely, A.: Differential growth and residual stress in cylindrical elastic struc-
tures. Philos. Trans. Roy. Soc. Lond. A 367(1902), 3607–3630 (2009) 303
1276. Vandiver, R., Goriely, A.: Morpho-elastodynamics: The long-time dynamics of elastic
growth. J. Biol. Dyn. 3(2-3), 180–195 (2009) 78
1277. Vella, D., Bico, J., Boudaoud, A., Roman, B., Reis, P.M.: The macroscopic delamination
of thin films from elastic substrates. Proc. Natl. Acad. Sci. USA 106(27), 10,901–10,906
(2009) 573
1278. Vermeulen, J.J.: Notes on the non-marine molluscs of the island of Borneo 6. The genus
Opisthostoma (Gastropoda Prosobranchia; Diplommatinidae), part 2. Basteria 58 (1994)
156
1279. Vernerey, F.J.: A mixture approach to investigate interstitial growth in engineering scaffolds.
Biomech. Model Mechanobiol. pp. 1–20 (2015) 396
1280. Vianello, M.: Coaxiality of strain and stress in anisotropic linear elasticity. J. Elasticity 42(3),
283–289 (1996) 421
1281. Vignola, A.M., Mirabella, F., Costanzo, G., Giorgi, R.D., Gjomarkaj, M., Bellia, V., Bon-
signore, G.: Airway remodeling in asthma. Chest 123(3), 417S–422S (2003) 531
1282. Vilagio, P.: Mathematical models for elastic structures. Cambridge University Press (1997)
66, 97
1283. Vito, R.P., Dixon, S.A.: Blood vessels constitutive models–1995-2002. Annu. Rev. Biomed.
Eng. 5 (2003) 293
References 633
1284. Vologodskii, A.: Topology and Physics of Circular DNA. CRC Press, Boca Raton (1992)
136
1285. Von Bertalanffy, L.: Quantitative laws in metabolism and growth. Q. Rev. Biol. 32(3), 217–
231 (1957) 12, 14
1286. Von Kármán, T.: Enzyclopädie der mathematischen wissenschaften. Forschungsarabeiten
Berlin (1907) 256
1287. Wadee, M., Hunt, G.W., Whiting, A.: Asymptotic and Rayleigh–Ritz routes to localized
buckling solutions in an elastic instability problem. Proc. Roy. Soc. Lond. A 453(1965),
2085–2107 (1997) 145
1288. Waite, J.H.: Quinone-tanned scleroproteins. In: A.S.M. Saleuddin, K.M. Wilbur (eds.) The
Mollusca, Physiology, pp. 467–504. Academic Press, New York (1983) 188
1289. Wallman, J., Winawer, J.: Homeostasis of eye growth and the question of myopia. Neuron
43(4), 447–468 (2004) 557, 568
1290. Wang, A.S.D., Ertepinar, A.: Stability and vibrations of elastic thick-walled cylindrical and
spherical shells subjected to pressure. Int. J. Non-linear Mech. 7 (1972) 456, 457
1291. Wang, C.C., Truesdell, C.: Introduction to Rational Elasticity. Springer Science & Business
Media (1973) 291
1292. Wang, I., Politi, A.Z., Tania, N., Bai, Y., Sanderson, M.J., Sneyd, J.: A mathematical model
of airway and pulmonary arteriole smooth muscle. Biophys. J. 94(6), 2053–2064 (2008) 533
1293. Wang, J.H.C., Jia, F., Gilbert, T.W., Woo, S.L.Y.: Cell orientation determines the alignment
of cell-produced collagenous matrix. J. Biomech. 36(1), 97–102 (2003) 419
1294. Wang, J.H.C., Thampatty, B.P.: An introductory review of cell mechanobiology. Biomech.
Model Mechanobiol. 5(1), 1–16 (2006) 561, 562
1295. Wang, W., Zhao, X.: A three-dimensional phase diagram of growth-induced surface insta-
bilities. Scientific reports 5, 8887 (2015) 573
1296. Wang, Y., Son, S., Swartz, S.M., Goulbourne, N.C.: A mixed Von Mises distribution for
modeling soft biological tissues with two distributed fiber properties. Int. J. Solids Struct.
49(21), 2914–2923 (2012) 340
1297. Wang, Z., Rudraraju, S., Garikipati, K.: A three-dimensional field formulation, and isogeo-
metric solutions to point and line defects using Toupin’s theory of gradient elasticity at finite
strains. J. Mech. Phys. Solids 94, 336–361 (2016) 576
1298. Ward, F.: Outlines of Human Osteology. Henry Renshaw, London (1938) 35
1299. Wartlick, O., Mumcu, P., Jülicher, F., Gonzalez-Gaitan, M.: Understanding morphogenetic
growth control—lessons from flies. Nat. Rev. Mol. Cell Biol. 12(9), 594–604 (2011) 559
1300. Waterkeyn, L.: Cotton Fibres: their development and properties: a technical monograph from
the Belgian Cotton Research Group. International Institute for Cotton, Manchester, U. K.
(1985) 155
1301. Wayman, B.H., Taylor, W.R., Rachev, A., Vito, R.P.: Arteries respond to independent control
of circumferential and shear stress in organ culture. Ann. Biomed. Eng. 36(5), 673–684
(2008) 32
1302. van der Weele, C.M., Jiang, H.S., Palaniappan, K.K., Ivanov, V.B., Palaniappan, K., Baskin,
T.I.: A new algorithm for computational image analysis of deformable motion at high spatial
and temporal resolution applied to root growth. roughly uniform elongation in the meristem
and also, after an abrupt acceleration, in the elongation zone. Plant Physiol. 132(3), 1138
(2003) 25
1303. Weibel, E.R.: Physiology: the pitfalls of power laws. Nature 417(6885), 131–132 (2002) 21
1304. Weibel, J., Fields, W.S.: Tortuosity, kinking and coiling of the internal carotid artery. Neu-
rology 15, 7–11 (1965) 510
1305. Weickenmeier, J., Butler, C., Young, P.G., A., G., Kuhl, E.: The mechanics of decompressive
craniectomy: Personalized simulations. Comput. Methods Appl. Mech. Eng. (2016) 295, 568
1306. Weickenmeier, J., Kuhl, E., Goriely, A.: The mechanics of decompressive craniectomy:
Bulging in idealized geometries. J. Mech. Phys. Solids (2016) 568
1307. Weickenmeier, J., Saez, P., Kuhl, E., Goriely, A.: Bulging brains. Preprint (2016) 295
634 References
1308. Weinkove, D., Neufeld, T.P., Twardzik, T., Waterfield, M.D., Leevers, S.J.: Regulation of
imaginal disc cell size, cell number and organ size by Drosophila class I A phosphoinositide
3-kinase and its adaptor. Current biology 9(18), 1019–1029 (1999) 558
1309. Weissenberg, K.: A continuum theory of rheological phenomena. Nature 159(4035), 310–
311 (1947) 556
1310. Weliky, M., Oster, G.: The mechanical basis of cell rearrangement. I. Epithelial morphogen-
esis during Fundulus epiboly. Development 109(2), 373–386 (1990) 57
1311. Wente, H.C.: The symmetry of sessile and pendent drops. Pacific J. Math 88(2), 387–397
(1980) 217
1312. Werblin, F., Roska, B.: The movies in our eyes. Sci. Am. 296(4), 72–79 (2007) 568
1313. Wesolowski, Z.: Stability of an elastic, thick-walled spherical shell loaded by an external
pressure. Arch. Mech. Stosow. 19, 3–23 (1967) 453, 457
1314. Wessels, J.G.H.: Role of cell wall architecture in fungal tip growth generation. In: I.B. Heath
(ed.) Tip growth in plant and fungal cells. Academic Press, San Diego (1990) 235, 237
1315. West, G.B., Brown, J.H., Enquist, B.J.: A general model for the origin of allometric scaling
laws in biology. Science 276(5309), 122–126 (1997) 15
1316. West, G.B., Brown, J.H., Enquist, B.J.: A general model for ontogenetic growth. Nature
413(6856), 628–631 (2001) 15, 16
1317. Westgate, M., Boyer, J.: Osmotic adjustment and the inhibition of leaf, root, stem and silk
growth at low water potentials in maize. Planta 164(4), 540–549 (1985) 96
1318. Wettlaufer, J.S.: Accretion in protoplanetary disks by collisional fusion. The Astrophysical
Journal 719(1), 540 (2010) 570
1319. Wettlaufer, J.S., Jackson, M., Elbaum, M.: A geometric model for anisotropic crystal growth.
J. Phys. A: Math. Gen. 27(17), 5957–5967 (1994) 570
1320. White, C.R., Seymour, R.S.: Allometric scaling of mammalian metabolism. J. Exp. Biol.
208(9), 1611–1619 (2005) 22
1321. White, J.F., Gould, S.J.: Interpretation of the coefficient in the allometric equation. Am. Nat.
pp. 5–18 (1965) 20
1322. White, J.L.: Dynamics of viscoelastic fluids, melt fracture, and the rheology of fiber spinning.
J. Appl. Polym. Sci. 8(5), 2339–2357 (1964) 556
1323. Wiggs, B.R., Hrousis, C.A., Drazen, J.M., Kamm, R.D.: On the mechanism of mucosal
folding in normal and asthmatic airways. J. Appl. Physiol 83, 1814–1821 (1997) 532, 536,
535, 538
1324. Wilber, J.P., Walton, J.R.: The convexity properties of a class of constitutive models for
biological soft issues. Math. Mech. Solids 7(3), 217–235 (2002) 293
1325. Wilkes, E.W.: On the stability of a circular tube under end thrust. Q. J. Mech. Appl. Math.
8(1), 88–100 (1955) 489
1326. Wilkie, A.O.M.: Craniosynostosis: genes and mechanisms. Human molecular genetics 6(10),
1647–1656 (1997) 572
1327. Wineman, A.: Some results for generalized neo-Hookean elastic materials. Int. J. Non Linear
Mech. 40(2), 271–279 (2005) 294
1328. Winkler, E.: Die Lehre von der Elustizitat und Festigkeit. Dominicus Prague, Czechoslovakia
(1867) 145
1329. Winsor, C.P.: The Gompertz curve as a growth curve. Proc. Natl. Acad. Sci. USA 18, 1–8
(1932) 16
1330. Withers, P., Bhadeshia, H.: Residual stress. part 1: Measurement techniques. Mater. Sci.
Tech. 17(4), 355–365 (2001) 45
1331. Withers, P., Bhadeshia, H.: Residual stress. part 2: Nature and origins. Mater. Sci. Tech.
17(4), 366–375 (2001) 45
1332. Withers, P.J.: Residual stress and its role in failure. Rep. Prog. Phys. 70(12), 2211–2264
(2007) 45
1333. Wold, M.P., Gamow, R.I.: Fiber-composite model for helical growth in the Phycomyces cell
wall. J. Theor. Biol. 159(1), 39–51 (1992) 543
References 635
1334. Wolf, S.D., Silk, W.K., Plant, R.E.: Quantitative patterns of leaf expansion: comparison of
normal and malformed leaf growth in Vitis vinifera cv. Ruby Red. Am. J. Bot. pp. 832–846
(1986) 25
1335. Wolff, J.: Über die innere architektur der knochen und ihre bedeutung für die frage von
knochenwachstum. Arch. Pathol. Anat. Physiol. Klin. Med. 50(389–453) (1870) 35
1336. Wolfram, S.: Statistical mechanics of cellular automata. Rev. Mod. Phys 55(3), 601 (1983)
51
1337. Wolgemuth, C.W., Goldstein, R.E., Powers, T.R.: Dynamic supercoiling bifurcations of
growing elastic filaments. Physica D 190(3), 266–289 (2004) 98, 141, 171
1338. Wolpert, .: Positional information and the spatial pattern of cellular differentiation. J. Theor.
Biol. 25(1), 1–47 (1969) 558
1339. Woods, R.H.: A few applications of a physical theorem to membranes in the human body in
a state of tension. Trans. Roy. Acad. Med. Ireland 10(1), 417–427 (1892) 32
1340. Woolley, T.E., Gaffney, E.A., Oliver, J.M., Baker, R.E., Waters, S.L., Goriely, A.:
Cellular blebs: pressure-driven, axisymmetric, membrane protrusions. Biomech. Model.
Mechanobiol. 13(2), 463–476 (2014) 208
1341. Wosilait, W., Luecke, R., Young, J.: A mathematical analysis of human embryonic and fetal
growth data. Growth, development, and aging: GDA 56(4), 249–257 (1991) 17
1342. Wright, G.D., Arlt, J., Poon, W.C.K., Read, N.D.: Optical tweezer micromanipulation of
filamentous fungi. Fungal Genet. Biol. 44(1), 1–13 (2007) 238
1343. Wu, M., Kirchner, H.: Nonlinear elasticity modeling of biogels. J. Mech. Phys. Solids 58(3),
300–310 (2010) 403
1344. Wyczalkowski, M.A., Chen, Z., Filas, B.A., Varner, V.D., Taber, L.A.: Computational models
for mechanics of morphogenesis. Birth Defects Research Part C: Embryo Today: Reviews
96(2), 132–152 (2012) 410
1345. Xu, G., Bayly, P.V., Taber, L.A.: Residual stress in the adult mouse brain. Biomech. Model.
Mechanobiol. 8(4), 253–262 (2009) 44, 45
1346. Yager, D., Cloutier, T., Feldman, H., Bastacky, J., Drazen, J.M., Kamm, R.D.: Airway surface
liquid thickness as a function of lung volume in small airways of the guinea pig. J. Appl.
Physiol. 77(5), 2333–2340 (1994) 538
1347. Yamada, S., Tadano, S., Fujisaki, K.: Residual stress distribution in rabbit limb bones. J.
Biomech. 44(7), 1285–1290 (2011) 45
1348. Yamashita, I., Hasegawa, K., Suzuki, H., Vonderiszt, F., Mimori-Kiyosue, Y., Namba, K.:
Structure and switching of bacterial flagellar filaments studied by X-ray diffraction. Nature
Struct. Biol. 5, 125–132 (1998) 155
1349. Yang, W., Fung, T.C., Chian, K.S., Chong, C.K.: Instability of the two-layered thick-walled
esophageal model under the external pressure and circular outer boundary condition. J.
Biomech. 40(3), 481–490 (2007) 524, 525
1350. Yavari, A.: A geometric theory of growth mechanics. J. Nonlinear Sci. 20, 781–830 (2010)
48, 363
1351. Yavari, A.: Compatibility equations of nonlinear elasticity for non-simply-connected bodies.
Arch. Rat. Mech. Anal. pp. 1–17 (2013) 358
1352. Yavari, A., Goriely, A.: Riemann–Cartan geometry of nonlinear dislocation mechanics. Arch.
Ration. Mech. Anal. 205(1), 59–118 (2012) 365, 369, 579
1353. Yavari, A., Goriely, A.: Weyl geometry and the nonlinear mechanics of distributed point
defects. Proc. Roy. Soc. Lond. A 468(2148), 3902–3922 (2012) 367, 579
1354. Yavari, A., Goriely, A.: Nonlinear elastic inclusions in isotropic solids. Proc. Roy. Soc. Lond.
A 469(2160), 20130,415 (2013) 365, 450, 476
1355. Yavari, A., Goriely, A.: Riemann–Cartan geometry of nonlinear disclination mechanics.
Math. Mech. Solids 18(1), 91–102 (2013) 579
1356. Yavari, A., Goriely, A.: On the stress singularities generated by anisotropic eigenstrains and
the hydrostatic stress due to annular inhomogeneities. J. Mech. Phys. Solids 76, 325–337
(2015) 400, 450
636 References
1357. Yavari, A., Goriely, A.: The twist-fit problem: finite torsional and shear eigenstrains in
nonlinear elastic solids. Proc. Roy. Soc. Lond. A 471(2183), 20150,596 (2015) 400
1358. Yavari, A., Sozio, F.: Nonlinear mechanics of surface growth for cylindrical and spherical
elastic bodies. J. Mech. Phys. Solids (2016) 572
1359. Yin, F.: Ventricular wall stress. Circulation Res. 49(4), 829–842 (1981) 32
1360. Young, T.: An essay on the cohesion of fluids. Phil. Trans. R. Soc. Lond. pp. 65–87 (1805)
33
1361. Zachos, L.G.: A new computational growth model for sea urchin skeletons. J. Theor. Biol.
259(3), 646–657 (2009) 217, 219
1362. Zajac, E.E.: Stability of two planar loop elasticas. ASME J. Applied Mech. March, 136–142
(1962) 63
1363. Zecca, M., Basler, K., Struhl, G.: Sequential organizing activities of engrailed, hedgehog
and decapentaplegic in the Drosophila wing. Development 121(8), 2265–2278 (1995) 558
1364. Zecca, M., Basler, K., Struhl, G.: Direct and long-range action of a wingless morphogen
gradient. Cell 87(5), 833–844 (1996) 558
1365. Zhao, J., Chen, X., Yang, J., Liao, D., Gregersen, H.: Opening angle and residual strain in a
three-layered model of pig oesophagus. J. Biomech. 40(14), 3187–3192 (2007) 45
1366. Zheng, J., Lamoureux, P., Santiago, V., Dennerll, T., Buxbaum, R.E., Heidemann, S.R.:
Tensile regulation of axonal elongation and initiation. J. Neurosci. 11, 1117–1125 (1991)
30, 80, 81
1367. Zhu, G.L., Boyer, J.S.: Enlargement in Chara Studied with a Turgor Clamp. Growth Rate Is
Not Determined by Turgor. Plant Physiol. 100(4), 2071–2080 (1992) 89, 92
1368. Zou, Y., Zhang, Y.: An experimental and theoretical study on the anisotropy of elastin
network. Ann. Biomed. Eng. 37(8), 1572–1583 (2009) 320
1369. Zwietering, M., Jongenburger, I., Rombouts, F., Van’t Riet, K.: Modeling of the bacterial
growth curve. Appl. Environ. Microb. 56(6), 1875–1881 (1990) 17
Index
A sea urchin (Lytechinus variegatus), 216,

Acceleration, 277 432
Actinomycetes, 234, see also Bacteria Anisotropic material, 320
Active force, 574 Antimarginal ornamentation, 197, 200
Active strain, 574 Antman, Stuart S., 262
Active stress, 574 Aortic valve, 340
Adenosine triphosphate (ATP), 575 Apical swelling, 246
Adscititious inequalities, 291 Appressorium, 224
Aerenchyma, 480 Arabidopsis thaliana, 27
Aerial phase, 237 Arc length, 98
Airways, 525 Area growth, 434
Allometric law, 12 Arnett, Tim, 36
Allometry, 20 Artery, 32, 303, 340, 508
Allomyces, 6 Asthma, 530
Ampère, André-Marie (1775-1836), 151 Auxanometer, 30
Axon, 30, 80, 556
Anand, Lallit, 262
Anelasticity, 86
Angular integration method, 340
B
Angular momentum, 282
Bacteria, 65
Animalia Actinomycetes, 234
Chinese hamster (Cricetulus griseus), Kitasatospora setae, 6
432 Bacillus subtilis, 156, 168
beetle (Dynastes), 21 Escherichia coli, 116, 155, 376
bream (Abramis brama), 14 Salmonella, 155
chameleon (Chamaeleo calyptratus), Spirochaetes, 116
577 Streptomyces violascens, 65
dogfish (Anchovia brownii), 18 Streptomyces, 6, 65
fiddler crabs (Uca), 21 Streptomyces A3(2), 242
fruit fly (Drosophilia), 558 Balance law, 278
Mollusca, see Mollusca Balance of angular momentum, 106, 282,
mouse 380
(Mus musculus), 558, 566 Balance of energy, 284, 381
New Zealand white rabbit, 25 Balance of entropy, 382
rat (Rattus norvegicus), 17, 80 Balance of linear momentum, 104, 280, 380
salamander (Ambystoma), 558 Balance of mass, 279, 377
salamander (Desmognathus), 21 Ball, John MacLeod, 450
Interdisciplinary Applied Mathematics 45, DOI 10.1007/978-0-387-87710-5
638 Index
Bat wing, 340 Cellular automata, 51

Beading, 246 Cellulose, 89, 155
Bean plant experiments, 483 Center dynamics models, 54
Beatty, Millard F., 490 Centroid, 104
Ben Amar, Martine, 252 Chen, Yi-Chao, 454
Berengarius, Jacobus (1460-1530), 154 Chitin, 542
Bernoulli, Jacob (1655-1705), 115, 487 Chladni, Ernst Florens Friedrich (1756-
Bernoulli–Euler equations, 122 1827), 251
Bifurcation, 132, 133, 138, 310, 487 Christoffel symbols, 361, 367
Bilby, Bruce Alexander (1922-2013), 353, Circumferential buckling, 524
363 Circumnutation, 150
Bingham model, 87, 90 Clausius, Rudolf Julius Emanuel (1822-
Bingham, Eugene C. (1878-1945), 87, 88, 1888), 382
90, 555 Clausius–Duhem inequality, 406
Biot, Maurice Anthony (1905-1985), 318, Coelomic pressure, 217
394 Cohn, Ferdinand Julius (1828-1898), 47, 234
Blastulation, 432 Coiling rate, 194
Body, 263 Coleman, Bernard, 383, 405
Body couple, 107 Coleman–Noll procedure, 383, 405
Body force, 105, 281 Collagen, 4, 320
Body Mass Index, 12 Colloidal suspension, 432
Boltzmann constant, 294 Commarginal ornamentation, 196, 197
Boudaoud, Arezki, 228 Compatibility, 358
Bourgery, Jean-Baptiste Marc (1797-1849), Compatible, 353
34, 47 Compliant source, 380
Brücke, Ernst Wilhelm von (1819-1892), 41, Configuration, 100, 263
47 current, 66, 100, 127, 208, 263
Bracts, 565 Eulerian, 264
Brand, Alexandra, 238 initial, 67, 127, 263
Buffon, Georges-Louis Leclerc (1707- Lagrangian, 264
1788), 10 material, 264
reference, 78, 100, 209
spatial, 264
C stress-free, 100
Cactus, 565 unstressed, 66
Cannon, Walter Bradford (1871-1945), 48, virtual, 78, 127, 231
409 Conidium, 224
Capillary length, 218 Connection, 365
Carnoy, Jean-Baptiste (1836–1899), 542 Constitutive law, 67, 214, 285, 383
Cartilage, 340 Constitutive relationship, 285
Cauchy equation, 282 Constrained mixture theory, 396
Cauchy’s stress principle, 281 Contact force, 104, 281
Cauchy, Augustin-Louis (1789-1857), 276, Contact inhibition, 38
281, 282 Contorted aleotropy, 361
Cauchy–Green tensor Conway, John Horton, 51
left, 277 Coral, 570
right, 276 Cornea, 340
Cavitation Cosserat, Eugène Maurice Pierre (1866-
plants, 480 1931), 251
rubber, 449 Cosserat, François-Nicolas (1852-1914),
Cayley, Arthur (1821-1895), 290 251
Cayley–Hamilton theorem, 290 Covariant derivative, 367
Cell track, 176 Cowin, Stephen, 47, 49
Index 639
Cranial growth, 571 Differential growth, 152

Craniectomy, 568 Diffusion, 564
Craniosynostosis, 572 Dislocated body, 352
Cuboid, 385, 386, 415 Dispersion parameter, 342
Culmann, Carl (1821-1881), 34 Dissipation, 383
Curl, 371 Dissipation inequality, 383, 407
Curvature, 98, 103 Divergence, 267
Gaussian, 254 Divergence theorem, 267
geodesic, 253 DNA, 63, 119, 135, 234
mean, 254 Doubling time, 377
normal, 253 Dubois, Marie Eugène François Thomas
pea test, 42 (1858-1940), 20, 47
principal, 254 Duhamel du Monceau, Henri-Louis (1700-
tensor, 367 1782), 5, 47
Curves, 98 Duhem, Pierre Maurice Marie (1861-1916),
Curvilinear coordinates, 268 382
Cuteness factor, 17 Dumais, Jacques, 6, 258
Cuvier, Georges (1769-1832), 20 Dutrochet, Henri (1776-1847), 151
Cylinder
bifurcation, 487
cavitation, 449, 480 E
effective rigidity, 496 Eckart, Carl H. (1902-1973), 363
fiber-reinforced, 328 Eigenstrains, 400
growth, 476 Einstein tensor, 362
incompatibility, 373 Elastic cube, 292
inflation–extension, 299 Elastic deformation tensor, 233
Elastic stretch, 128
instability, 262
Elastica, 122
straightening, stretching, and shearing,
Emmetropization, 567
307
Energy, 381
Cytoskeleton, 237
Entropy, 382
Epstein, Marcelo, 49
Equation of rod dynamics, 113
D Equations of elasticity, 287
Darboux Equations of morphoelasticity, 384
matrix, 102 Ericksen, Jerald, 290, 291, 305
vector, 100 Ertepinar, Aybar, 454
Darcy, Henry Philibert Gaspard (1803- Euclidean norm, 264
1858), 393, 395 Euler buckling, 487
Darcy’s law, 395 Euler’s laws, 103
Darwin, Charles (1809-1882), 152, 162 Euler, Leonhard (1707-1783), 115, 487
da Vinci, Leonardo (1452-1519), 565 Eulerian, 25
Davis’ law, 36, 37 Eulerian growth rate, 24
Davis, Henry Gassett (1807-1896), 36, 47, Eulerian growth velocity, 23
409 Evolution law, 69, 233, 399
Deban, Stephen, 577 Excess twist, 103
Deborah number, 555 Extrinsic curvature tensor, 254
Deformation, 263
Deformation gradient, 272
de Montbeillard, Philibert Guéneau (1720- F
1785), 10, 47 Fairbairn crane, 35
Destrade, Michel, 490 Fairbairn, Willian (1789-1874), 34
Determinant method, 459 False-twist technique, 152
de Vries, Hugo (1848-1935), 153 Fiber growth, 434
640 Index
Fiber-tensile region, 326, 423 Geodesic curvature vector, 253

Fibonacci series, 567 Geometric stretches, 232
Fick, Adolf Eugen (1829-1901), 393 Germain, Marie-Sophie (1776-1831), 251
Flagella, 156 Gibberellin, 483
Flory, Paul John (1910-1985), 353, 403, 563 Gompertz’ law, 16
Fluid membrane, 208, 210 Gompertz, Benjamin (1779-1865), 16, 47
Föppl, August (1854-1924), 256 Gradient, 267
Fosdick, Roger L., 490 Gravitropic response, 65
Fourier heat flux, 381 Gray, Asa (1810-1888), 152
Fourier, Joseph (1768-1830), 381 Green, Albert E. (1912-1999), 318
Frame Growing beams, 414
Frenet, 98 Growing ring, 138, 355, 363, 369, 388, 389,
Frenet–Serret, 98 391
general, 100 Growing shell, 431
Frenet Growth, 4
equations, 99 accretive, 7, 175, 570
helix frame, 116 allometric, 17
Frenet, Jean Frédéric (1816-1900), 98 apical, 5, 7, 234
Fried, Eliot, 262 appositional, 7
Fu, Yibin, 454 area, 176, 402, 434
Functional adaptation, 35 axis, 181, 185
Fundamental form bulk, 9
first, 252 cambial, 6, 506
second, 254 classification, 4
Fundamental theorem of curves, 99 cone, 80
Fung model, see model connection, 366
Fung, Yuan-Cheng (Bert), 295, 409, 509 crystal, 570
Fungi, 5, 12, 63, 65 differential, 22, 39
Allomyces, 6 exponential, 69
Candida albicans, 238 fibre, 402
filamentous, 237 gradient, 22
Neurospora, 237 helical, 65, 542
Phycomyces blakesleeanus, 543 hoop, 356
rice blast fungus (Magnaporthe grisea), human, 10
224 interstitial, 9
isometric, 17
isotropic, 402
G kinematics, 345
Galileo, Galilei (1564-1642), 17, 47 law, 559
Game of Life, 51 metric, 359
Garikipati, Krishna, 49 normal, 234
Gasser, T. Christian, 344, 512 of the skull, 571
Gastrulation, 432 on a line, 63
Gauss orthotropic, 401
Theorema Egregium, 254 plant cell, 88, 90
Gauss, Carl Friedrich (1777-1855), 254 primary, 6
Gene, 27 radial, 356
decapentaplegic, 558 regulation, 557
wingless, 558 relative, 17
Generalized structure tensor, 341 scaling, 10
Gent model, see model secondary, 6, 64
Gent, Alan Neville (1927-2012), 295, 318, source, 22
445, 450 stems, 130
Index 641
stress, 42 Incompatibility, 345, 348

stretch, 69, 127, 232 Incompatibility tensor, 362
surface, 7 Incompatible state, 352
tensor, 9, 233 Incremental deformations, 311
time scale, 376 Inequalities
tip, 5, 7, 234 adscititious, 291
transversely isotropic, 402 Baker-Ericksen, 291, 294, 414
vector, 25 empirical, 292
velocity, 379 generalized empirical, 293
velocity gradient tensor, 379 ordered-forces, 291, 414
volumetric, 9 Inflation jump, 443
Growth-induced instability, 425 Inherent strains, 400
Growth tensor, 400 Initial configuration, 127
Guard cells, 89 Initial strains, 400
Gurtin, Morton, 262 Initiation, 30
Instantaneous elastic moduli, 428, 429
Invariants, 220
H Inversion line, 518
Half-plane, 298, 314 Isochoric, 274
Hamilton, William Rowan (1805-1865), 290 Isotropic function, 289
Haughton, David M., 454 Isotropic material, 289
Hay, Angela, 321
Healey, Timothy, 454
J
Heart, 32
Jacobi’s relations, 271
Heat setting, 152
Helical spring, 159
Helical spring formulas, 119
K
Heteroclinic orbit, 157
Kelvin solid, 87
Hill, Rodney M. (1921-2011), 454
Kepler, Johannes (1571-1630), 565
His, Wilhelm Sr. (1831-1904), 28, 47
Kinematics, 23, 127, 262
History of growth, 46
Kirchhoff kinetic analogy, 115
Hoger, Anne, 48, 353 Kirchhoff rod, 97, see also rod
Holzapfel, Gerhard, 344, 509, 512 Kirchhoff, Gustav Robert (1824-1887), 97,
Homeostasis, 32, 409 251, 284
Homeostatic stress, 78 Kirchhoff–Love shell, 208
Homeostatic stress assumption, 409 Klapper, Isaac, 171
Homogeneous deformations, 296, 410 Kleiber’s law, 15
Human growth, 10 Kleiber, Max (1893-1976), 15, 48
Humphrey, Jay, 48, 49, 396, 446, 514 Kondo, Kazuo (1911-2001), 363
Huxley, Julian (1887-1975), 20, 48 Kröner, Ekkehart, 353, 363
Hydrostatic pressure, 286 Kronecker symbol, 266
Hydrostatic skeleton, 336 Kuhl, Ellen, 49, 320, 420, 422, 508
Hyper-restoration, 410
Hyperelastic, 285
Hyperplasia, 9 L
Hypertrophy, 9 Lagrange, Joseph-Louis (1736–1813), 487
Lagrangian, 25
Lagrangian growth rate, 23
I Lagrangian growth velocity, 23
Ideal fluid, 86 Lake Tanganyika, 376
Ideal spring, 159 Langers’ lines, 320
Immersion, 361 Lapicque, Louis (1866-1952), 20, 47
Impotent stresses, 365 Laplace’s equation, 212
642 Index
Laplace’s law, 33, 223 M

Laplace, Pierre-Simon marquis de (1749- Magic angle, 327, 338
1827), 33, 216 Maini, Philip Kumar, 47
Law Maize (Zea mays L.), 92
allometric, 12 Mandel stress tensor, 407
balance, 278 Mass balance equation, 378
constitutive, 67, 214, 383 Mass conservation, 280
Darcy’s, 395 Material, see model
Davis’, 36 Material constraint, 286
Euler’s, 281, 282 Material-frame indifference, 288
evolution, 69, 233, 399 Material symmetry, 288
first, 381 Material time derivative, 277
for a growing continuum, 375 Maxwell–Bingham model, 87, 95
Gompertz’, 17 Maxwell fluid, 86
growth, 187, 405, 559 Maxwell, James Clerk (1831-1879), 86, 116,
Hookean, 68 150, 280
Kleiber’s, 15 Maxwell transport, 280
Laplace’s, 32, 223 McCulloch, Andrew D., 353
power, 14, 20 Mechanotransduction, 30, 561
scaling, 13 Medawar, Peter Brian (1915-1987), 16, 48
second, 382, 560 Membrane, 208
Thoma’s, 32 Membrane osmometer, 89
Wolff’s, 34 Mendelson, Neil, 169
Woods’, 32 Menzel, Andreas, 49
Young’s, 33 Metric
Le Gros Clark, Wilfrid Edward (1895-1971), compatibility, 367
28 Euclidean, 359
Léon, Isidore, 151 flatness, 359
Levi-Civita connection, 368 growth, 359
Levi-Civita permutation symbols, 265, 362 Metric tensor, 252
Levi-Civita, Tullio (1873–1941), 265 Michell’s instability, 133
Limit-point instability, 304, 442 Michell, John Henry (1863-1940), 133
Lineal density, 104 Microtubules, 321
Linear affine connection, 366 Minimal surface, 254
Linear momentum, 280 Mixed loading, 288
Linnaeus, Carl (Carl von Linné. 1707-1778), Mixture theory, 393
153 Model
Listing, Johann Benedict (1808-1882), 150 Demiray, 295
Loading Fung, 221, 295, 303, 318, 444, 501, 504,
dead, 288 518
hydrostatic, 288 generalized neo-Hookean, 294
mixed, 288 Gent, 295, 318, 445
rigid, 288 Mooney–Rivlin, 294, 317, 444
Local elastic deformation tensor, 351 neo-Hookean, 303, 423, 438, 443, 445,
Localization procedure, 106, 280 447, 450, 451, 454, 455, 460, 480, 490,
Lockhart’s equation, 91 491, 501, 504, 505, 518
Locking stretches, 320 Ogden, 294, 319, 446
Logarithmic spiral, 190 Modulus
Logistic model, 13 shear, 68, 294
Love, Augustus Edward Hough (1863- Young’s, 46, 67, 96, 109, 214, 221, 224,
1940), 251 255, 294, 449, 451, 484, 488, 496
Lysigenous, 481 Mollusca
Lysis, 246 Neptunea angulata, 189
Index 643
Architectonica, 189 O
Bolinus, 205 O(3), 266
Ceratostoma, 205 Objectivity, 288
Chicoreus, 198 Oesophagus, 525
Dictomosphinctes, 197 Off-lattice models, 54
giant clam shell (Tridacna gigas), 204 Ogden, Ray, 262, 294, 344, 454, 512
Hexaplex, 205 On-lattice models, 51
Murex, 198 One-director Cosserat surfaces, 208
Nipponite, 195 Oomycetes, 237
peanut snail (Cerion), 150 Opening angle, 43, 509
Peltoceras, 199, 201 Orientation density function, 340
Promicroceras, 7 Ornamentation, 196
Pterynotus, 205 Osmotic pressure, 89
Stephanoceras, 201 Osteoblasts, 35
Turritella terebra, 189 Osteoclasts, 35
Turritella, 195 Osteocytes, 35
Xipheroceras, 201
Mooney–Rivlin model, see model
Morphoelastic shells, 251 P
Parallel transport, 366, 367
Morphoelasticity, 45, 46, 76, 126, 232, 353
Pennybacker, Matt, 566
Morphogenesis, 4
Periostracum, 188
Morphogens, 558, 564
Permutation symbols, 265
Morphospace, 218
Perversion, 150, 151
Moseley, Henry (1801- 1872), 47, 188, 192
Phenomenological laws, 408
Motion, 263
Phycomyces, 321, 542
Moulton, Derek, 176
Phyllotaxis, 565
Mucosal folding, 524
Piola, Gabrio (1794- 1850), 284
Müller, Nicolaus Jacob Carl (1843-1901),
Pipkin, Allen Compere (1931-1994), 556
24, 26, 47
Pith, 40
Multiplicative decomposition, 353
Plant cells, 88
Multiscale approach, 561
Plant stems, 30
Murray, James Dickson, 47, 353
Plantae
Mycelium, 237 Arabidopsis thaliana, 27
Myopia, 568 barrel clover (Medicago truncatula), 239
blue bindweed (Solanum dulcamara),
163
N Cardamine hirsuta, 321
Nanson’s formula, 275 cat-tail (Sparganium emersum), 482
Necrotic core, 74 chives (Allium Schoenoprasum), 482
Neo-Hookean model, 68, 293 chocolate vine (Akebia quinata), 163
Neo-Hookean, see also model grape vine (Vitis vinifera), 41
Neoplasia, 9 green algae, 321
Neurites, 30, 80 Acetabularia, 256
Neurons, 80 Chara corallina, 92
Noll, Walter, 262, 383, 405 Nitella, 92
Non-compliant entropy, 382 Hollyhock (Althaea rosea), 482
Non-polar, 282 horsetail (Equisetum), 482
Normal curvature vector, 253 ivy morning glory (Pharbitis nil), 164
Normal growth hypothesis, 247 maize (Zea mays L.), 93, 482
Normal stress, 283 moso bamboo (Phyllostachys edulis),
Normal stretch, 210 376
Nutt, John Joseph, 37, 47 passion flowers (Passiflora), 131
Nye’s dislocation tensor, 372 pea (Pisum sativum), 42
644 Index
policeman’s helmet (Impatiens glan- Resultant force, 105

dulifera), 65 Richards model, 13
purple morning glory (Ipomoea pur- Riemann-Cartan manifolds, 367
purea), 163 Riemannian curvature tensor, 361
rhubarb (Rheum rhabarbarum), 39 Riemannian curvature tensor associated with
rockcress (Arabidopsis thaliana), 27 the metric M, 368
snapdragon (Antirrhinum), 559 Rigid loading, 288
sunflower (Helianthus annuus), 501, 565 Rivlin–Ericksen representation, 290
wapas (Eperua venosa), 39 Rivlin, Ronald (1915-2005), 290, 294, 321,
white bryony (Bryonia dioica), 152 563
Plasticity, 85, 88 Rod, 99
Plate, 208 buckling, 132
Podia, 217 constitutive laws, 108
Poisson’s ratio, 109 evolution law, 129
Polar decomposition theorem, 276, 401 extensible, 108
Pollen tube, 5, 12 Frenet, 103
Potts models, 52 helical, 116
Potts, Renfrey Burnard (1825-2005), 53 in one dimension, 66
Poynting effect, 293, 403 inextensible, 102, 108
Poynting, John Henry (1852-1914), 403 initially straight, 115
Pre-strains, 400 Kirchhoff elastic, 97
Primordia, 565 mechanics, 103
Principal bending stiffnesses, 109 morphoelastic, 125
Principal curvature matrix, 254 on a foundation, 142
Principle of objectivity, 288 planar, 122
Protoplanetary disks fusion, 570 scaling, 110
Pütter, August Friedrich Robert (1879- summary, 113
1929), 12, 48 unshearable, 102
Rodriguez, Edward K., 48, 353
Root hair, 5, 63, 238
Q
Roux, Wilhelm (1850-1924), 35, 47
Quetelet coefficient, 12
Roy, Charles Smart (1854-1897), 47, 508
Quetelet, Lambert Adolphe Jacques (1796-
1874), 11, 47
S
R Saccomandi, Giuseppe, 445
Régis Chirat, 197, 198 Sachs, Julius von (1832-1897), 31, 42, 47,
Radial stretch, 210 163, 481
Rajagopal, Kumbakonam, 48, 396 Saltatory spurts, 12
Rameaux, Jean-François (1805-1878), 14, Sarrus, Pierre-Frédéric (1798-1861), 14, 47
47 Scalar product, 264
Reaction stress, 286 Scale factors, 268
Reference density, 280 Schizogenous, 481
Register angle, 102 Seashell, 187, see also Mollusca
Reinhardt, M.O., 47, 234, 238 Self-similarity, 187, 191, 194, 242
Reinking, Robert, 543 tip growth, 242
Reisner, Markus (1886-1976), 555 Semi-inverse problem, 305
Relative elemental growth rate, 25 Serret, Joseph Alfred (1819-1885), 98
Remodeling, 4, 125, 419 Shear modulus, 68, 294
maximum stretch, 421 Shear stress, 283, 292
Residual strains, 39 Shell, 208
Residual stress, 39, 43, 348 Sherman-Morrison formula, 402
Resultant couple, 106 Shipman, Patrick, 567
Index 645
Signorini’s mean-stress theorem, 350 Strictly convex, 352

Singularity, 448 Stroh, Allan N. (1926-1962), 353
Size regulation, 557 Structure tensor, 322
Skalak, Richard (1923-1997), 48 Surface
Skull, 571 elliptic, 254
Slow-growth assumption of morphoelastic- hyperbolic, 254
ity, 377 parabolic, 254
Small-on-large deformations, 311 Swelling gel, 403, 525
Smith, Richard, 567
S O(3), 266
Soft-spot hypothesis, 241 T
Southwell, Richard V. (1888–1970), 490 Taber, Larry, 49, 514
Space curve, 98 Tabor, Michael, 236, 543
Specific growth rate, 16 Tait, Peter Guthrie (1831-1901), 133
Spencer, Anthony J.M. (1929-2008), 321 Tangential stretch, 210
Spheroid, 72, 432 Tendon, 63
Spin matrix, 102 Tendril, 63, 149
Spin vector, 100 Tension
Spine, 63, 565 uniaxial, 291, 320, 326, 343, 413
Spiral, 190 Tensor
Spitzenkörper, 238 coaxial, 407, 421
Sporangiophore, 542 curvature, 367
Standard fiber-reinforcing model, 324, 333, derivatives, 266
545 determinant, 265
Stiffness divergence in curvilinear coordinates,
bending, 111 269
torsional, 111 gradient in curvilinear coordinates, 269
Stokes’ theorem, 371 growth, 9, 233
Stokes, George Gabriel (1819-1903), 371 inverse, 266
Strain, 67, 78, 276 orthogonal, 265
biaxial, 293 product, 264
equibiaxial, 298 proper orthogonal, 265
plane, 299 second-order, 265
uniaxial, 293, 296, 299, 501 symmetric, 265
Strain-energy density function, 109, see trace, 265
model transpose, 265
Strain-energy function, see model 285 Theorema Egregium, 254
Mooney–Rivlin, 221 Thermodynamics, 405
Streptomyces, 6, 241 Thoma, Richard Andreas Thoma (1847-
Stress, 67 1923), 32, 47, 409
Stress function, 111 Thompson, D’Arcy Wentworth (1860-
Stress power, 285 1948), 20, 47, 150, 188, 216
Stress tensor Thomson, Willian (Kelvin) (1824-1907), 87,
Cauchy, 282 133
engineering, 284 Time scales, 376
first Piola-Kirchhoff, 284 Timoshenko, Stephen (1878-1972), 64
nominal, 284 Tip growth, 5, 69
Stretch, 67, 210, 275 Tissue tension, 39, 498
elastic, 67 Torsion, 99, 103
principal, 277 Torsion tensor, 367
Stretch locking, 320 Torsional stiffness, 109
Stretch tensor, 276 Total stretch, 128
Stretch vector, 100, 101 Transformation of area, 274
646 Index
Transformation of volume, 274 Von Mohl, Hugo (1805-1872), 151

Transformation strains, 400 Von Neumann, John (1903-1957), 51
Transport formulas, 280 Voronoi tessellation, 54
Transversely isotropic material, 322
True incompatibility tensor, 372
Truesdell, Clifford Ambrose (1919-2000), W
262, 393 Wall loosening, 90
Tsapis, Nicolas, 432 Wang, Chao-cheng, 454
Tumor, 432 Warping function, 111
Tumor growth, 37, 71, 376 Water potential, 89
Turgor pressure, 89 Weissenberg number, 556
Turing, Alan (1912-1954), 564 Weitzenböck manifolds, 368
Turitella, 183 Wharton’s jelly, 154
Twining vines, 162 Wolff’s law, 34, 419
Twistless spring, 159 Wolff, Julius (1836-1902), 34, 47, 409
Woods’ law, 34
Woods’ number, 33
U Woods, R.H., 32, 47, 409
Uniform, 108 Work conjugate, 285
Universal deformations, 305 Worm, 336
Unstrained, 276 Wound healing, 353
Urchin, 216 Wrench, 118
Urchin test, 217 Wrinkling instability, 573
V X
Vectors and tensors, 264 Xylem, 90
Vegetative phase, 236
Velocity, 277
Velocity gradient tensor, 278 Y
Vertex models, 56 Yavari, Arash, 358, 372
Vesicle supply center, 238 Young’s equation, 212
Vines, 149 Young’s law, 33
Virtual configuration, 127 Young’s modulus, 67, 294, 297, 298
Virtual state, 352, 384 effective, 326
Viscoelasticity, 88 Young, Thomas (1773-1829), 33, 67
Voigt, Woldemar (1850-1919), 87 Young–Laplace’s equation, 212
Volume density, 279
Von Bertalanffy, Karl Ludwig (1901-1972),
14, 48 Z
Von Kármán, Theodore (1881-1963), 256 Zerna, Wolfgang (1916-2005), 318

Goriely 2017

Uploaded by

Copyright:

Available Formats

Goriely 2017

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

Goriely 2017

Uploaded by

Copyright:

Available Formats

What is the topic of research covered in the book?

What is the topic of research covered in the book?

What mathematical techniques are used in the book?

What mathematical techniques are used in the book?

Interdisciplinary Applied Mathematics 45

More information about this series at http://www.springer.com/series/1390

ISSN 0939-6047 ISSN 2196-9973 (electronic)

© Springer Science+Business Media LLC 2017

Printed on acid-free paper

This Springer imprint is published by Springer Nature

describe phenomena appearing in different scientiﬁc disciplines. This monograph is

particularly dangerous as it is the ideal ecological zone for highly specialized

Oxford, UK Alain Goriely

I am deeply indebted to a number of friends and collaborators. First, it is with great

Part I Introduction: Where It All Starts

3.2 Off-Lattice Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

Part II Filament Growth: A One-Dimensional Theory

5.8 The Planar Elastica: Bernoulli–Euler Equations . . . . . . . . . . . . 122

Part III Surface Growth: A Two-Dimensional Theory

8.1.3 Constitutive Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

Part IV Volumetric Growth: A Three-Dimensional Theory

11.3 Constitutive Equations for Hyperelastic Materials . . . . . . . . . . . 285

13.9 Mixture Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392

16.2.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484

Part V Conclusion: Where It Does Not End

17.8 Dynamics and Post-bifurcation Behavior . . . . . . . . . . . . . . . . . 572

At first the infant,

Our study of growth will mostly be limited to non-diffusive processes relevant to

Figure 1.1 Growing root as

evolution is often described by modeling tissues as fluid or viscoelastic rather than

1.1.1 Tip Growth

In his book, “The Physics of Trees” published in 1758, Duhamel du Monceau

1.1.2 Accretive Growth

Accretive, surface, or appositional growth describe mechanisms, such as deposition

Figure 1.3 The fossil shell

1.1.3 Volumetric Growth

1.2 The Scaling of Growth

Naturelle” [1123]. A quarter of a millennium later, it still captures the essence of

Ṁ = M(a M − p − b), (1.2)

M(t) = M0 e(a−b)t , (1.3)

where M0 = M(t = 0) is the mass at birth.

Figure 1.10 Example of

Dimensionless mass ratio

Figure 1.13 Gompertz’ law. Weight (gm)

which leads to −e−kt

1.3 Relative Growth

Figure 1.15 Isometric

By expressing m as a function of M, the explicit dependence on time is removed

C. claw weight (mg)

body weight (mg)

1.4 The Kinematics of Growth

In particular, if growth is taking place independently of the position, we would have

t=15 days t=28 days t=49 days t=87 days

(a) (b) (c)

(d) (e) (f)

2.1 Growth is Influenced by Stress

From a biologist’s perspective, growth is mediated by gene activation and regulation.

© Springer Science+Business Media LLC 2017 27

Figure 2.3 Ear gauging

σ = E = E(α − 1). (4.2)