Physical Chemistry of Surfaces Sixth Edition ARTHUR W. ADAMSON Department of Chemistry, University of Southern California Los Angeles, California and ALICE P. GAST Department of Chemical Engineering, Stanford University Stanford, California ®) [A WILEY-INTERSCIENCE PUBLICATION John Wiley & Sons, Inc. NEW YORK / CHICHESTER / WEINHEIM / BRISBANE / SINGAPORE / TORONTOAbout the Front Cover Design ‘The fer-like shapes are two-dimensional dendritic crystals of the protein streptavidin bound to a lipid monolayer. These dendrites were produced by Andrew Ku with the help of Seth Darst, Roger Kornberg, and Channing Robertson and are described in Chapter XV; note Fig. XV-5. ‘The island pattern is from a marbled paper, for which we thank Mrs. Phoebe Easton. Marbling is an ancient process whereby colored oils with surfactant are spread on the surface of water stiffened by agar-agar and the pattern lifted onto the paper. The pattern could be that of domains ‘of 2 and of CO chemisorbed on PX(100); note Fig. XVILI-28. is book is printed on acid-free paper. © Copyright © 1997 by John Wiley & Sons, In. All rights reserved. Published simultaneously in Canada. [No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning or ‘otherwise, except as permitted under Sections 107 or 108 of the 1976 United States Copyright ‘Act, without either the prior written permission of the Publisher, or authorization through ‘payment of the appropriate per-copy fee to the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923, (508) 750-8400, fax (508) 750-4744. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., (605 Third Avenue, New York. NY 10158-0012, (212) 850-6011, fax (212) 850-6008. E-Mail: PERMREQ @ WILEY.COM. Library of Congress Cataloging in Publication Data: ‘Adamson, Arthur W. Physical chemistry of surfaces / Arthur W. Adamson and Alice P. Gast. —6th ed. P. om, “A Wiley-Interscience Publication.” Includes bibliographical references and index. ISBN 0-471-14873-3 (cloth : alk. paper) 1. Surface chemistry. 2. Chemistry, Physical and theoretical L. Gast, Alice P. (Alice Petry), 1958- Il, Tile. QD506.A3 1997 541.3°3—de21 Printed in the United States of America 10987654321XVI The Solid-Gas Interface—General Considerations 1. XVII. Adsorption of Gases and Vapors on Solids 1 . The Adsorption Time .. |. Experimental Procedures . ). The BET and Related Isotherm . The Potential Theory. . The Characteristic Isotherm and Related Concepts |. Chemical Physics of Submonolayer Adsorption . Phase Transformations in the Multilayer Region . . Thermodynamics of Adsorption ... CONTENTS Introduction . ‘The Surface Area of Solids . The Meaning of Surface Area . Surfaces as Having a Fractal Geometry . 5 Methods Requiring Knowledge of the Surface Free Energy or Total Energy . Rate of Dissolving The Mercury Porosimeter.. Other Methods of Surface Area Determination mmo ADP . The Structure and Chemical Nature of Solid Surfaces. The Nature of the Solid-Adsorbate Complex .. A. Effect of Adsorption on Adsorbate Properties B. Effect of the Adsorbate on the Adsorbent C. The Adsorbate-Adsorbent Bond... Problems .. General References Textual References Introduction . The Langmuir Adsorption Isotherm .. A. Kinetic Derivation. .. B. Statistical Thermodynamic Derivation. C. Adsorption Entropies D. Lateral Interaction. E, Experimental Applications of the Langmuir Equation ‘A. Derivation of the BET Equation B. Properties of the BET Equation... C. Modifications of the BET Equation . . Isotherms Based on the Equation of State of the Adsorbed Film A. Film Pressure—Area Diagrams from Adsorption Isotherms ... B. Adsorption Isotherms from Two-Dimensional Equations of State... A. The Polanyi Treatment .. B. Isotherms Based on an Assumed Variation of Potential with Distance .. . The Polarization Model. . Comparison of the Surface Areas from the Various Multilayer Models . xv sm si 372 372 374 516 577 377 580 381 582 582 589 591 592 593 594 627 629 630 631 634 639 641xvi CONTENTS ‘A. Theoretical Considerations. B. Experimental Heats and Energies of Adsorption. 13. Critical Comparison of the Various Models for Adsorption A. The Langmuir-BET Model... B. Two-Dimensional Equation of Statc Treatments . C. The Potential Model 14. Physical Adsorption on Heterogeneous Surfaces .. A. Distribution of Site Energy Patches . B. Thermodynamics of Adsorption on Heterogeneous Surfaces. . : C. Point versus Patch Site Energy Distributions D. Geometric Heterogeneity 15. Rate of Adsorption. ... Msi asia ce Paw Soe ta ‘A. Molecular Sieves....... B. Capillary Condensation. C. Micopore Analysis 17. Problems .... General References . Textual References. XVIII. Chemisorption and Catalysis... 1. Introduction 2. Chemisorption: A. LEED Structures....... B. Surface Microscopies..... = C. Spectroscopy of Chemisorbed Species . D. Work Function and Related Measurements. E. Programmed Desorption sues 3. Chemisorption Isotherms. A. Variable Heat of Adsorption. B. Effect of Site and Adsorbate Coordination Number. C. Adsorption Thermodynamics seve 4. Kinetics of Chemisorption A. Activation Energies. B. Rates of Adsorption C. Rates of Desorption .. 5. Surface Mobility 6. The Chemisorption ‘A. Some General Aspects. B. Metals....... C. Semiconductors . D. Acid-Base Systems 7. Mechanisms of Heterogeneous Catalysis. ‘A. Adsorption or Desorption as the Rate-Determining Step... B, Reaction within the Adsorbed Film as the Rate-Determining Step. secsveeeee ee 641 647 652 652 653 654 655 65s 703 705 707 709 n2 12 ns 7 m8 70 720 72CONTENTS 8. Influence of the Adsorption Isotherm on the Kinetics of Heterogeneous Catalysis ‘A. Unimolecular Surface Reactions B. Bimolecular Surface Reactions..... 9. Mechanisms of a Few Catalyzed Reactions ‘A. Ammonia Synthesis .... B. Fischer-Tropsch Reactions. C. Hydrogenation of Ethylene D, Catalytic Cracking of Hydrocarbons and Related Reactions. E. Oxidation of CO... F. Photochemical and Photoassisted Processes at Surfaces. 10, Problems ...... General References . Textual References... Index... xviiCHAPTER XVI The Solid-Gas Interface—General Considerations 1. Introduction ‘These concluding chapters deal with various aspects of a very important type of situation, namely, that in which some adsorbate species is distributed between a solid phase and a gaseous one. From the phenomenological point of view, cone observes, on mechanically separating the solid and gas phases, that there is a certain distribution of the adsorbate between them. This may be expressed, for example, as n,, the moles adsorbed per gram of solid versus the pressure P. The distribution, in general, is temperature dependent, so the complete empirical description would be in terms of an adsorption function n, = f(P. 7). While a thermodynamic treatment can be developed entirely in terms of J(P,T), to apply adsorption models, it is highly desirable to know n, on a per square centimeter basis rather than a per gram basis or, alternatively, to know 0, the fraction of surface covered. In both the physical chemistry and the applied chemistry of the solid~gas interface, the specific surface area L is thus of extreme importance. All gases below their critical temperature tend to adsorb as a result of general van der Waals interactions with the solid surface. In this case of Physical adsorption, as it is called, interest centers on the size and nature of adsorbent-adsorbate interactions and on those between adsorbate molecules. There is concem about the degree of heterogeneity of the surface and with the ‘extent to which adsorbed molecules possess translational and internal degrees of freedom. If the adsorption energy is large enough to be comparable to chemical bond energies, we now speak of chemisorption. The adsorbate tends to be localized at particular sites (although some surface diffusion or mobility may still be present), and the equilibrium gas pressure may be so low that the adsorbent- adsorbate system can be studied under high-vacuum conditions. This allows the many diffraction and spectroscopic techniques described in Chapter VIII to be used to determine what actual species are present on the surface and their packing and chemical state.) This is also true for physisorption systems if the surface is well defined and the temperature low enough that the equilibrium Pressure is very low, see Fig. XVII-17 for example. It has become increasingly appreciated in recent years that the surface struc ture of the adsorbent may be altered in the adsorption process. Qualitatively, 571572 XVI THE SOLID-GAS INTERFACE. TABLE XVI-1 ‘Types of Adsorption Systems ‘Type of Adsorbent ‘Type of Adsorbate- Adsorbent Interaction Molecular Refractory Van der Waals (physical Surface restructuring on 'No change in surface adsorption) adsorption structure on adsorption Chemical Chemical reaction Surface restructuring (chemisorption) modifying the adsorbent on adsorption such structural perturbation is apt to occur if the adsorption energy is compa- rable to the surface energy of the adsorbent (on a per molecular unit basis). As summarized in Table XVI-1, physical adsorption (sometimes called physisorp- tion) will likely alter the surface structure of a molecular solid adsorbent (such as ice, paraffin, and polymers) but not that of high-surface-energy, refractory solids (such as the usual metals and metal oxides and carbon black). Chemisorp- tion may alter the surface structure of refractory solids. ‘The function of this chapter is to summarize some of the general approaches to the determination of the physical and chemical state in both of the types of adsorption systems described. 2. The Surface Area of Solids The specific surface area of a solid is one of the first things that must be determined if any detailed physical chemical interpretation of its behavior as an adsorbent is to be possible. Such a determination can be made through adsorp- tion studies themselves, and this aspect is taken up in the next chapter; there are a number of other methods, however, that are summarized in the following material. Space does not permit a full discussion, and, in particular, the meth- ods that really amount to a particle or pore size determination, such as optical and electron microscopy, x-ray or neutron diffraction, and permeability studies are largely omitted. A. The Meaning of Surface Area It will be seen that each method for surface area determination involves the ‘measurement of some property that is observed qualitatively to depend on the extent of surface development and that can be related by means of theory to the actual surface area. It is important to realize that the results obtained by different methods differ, and that one should in general expect them to differ. ‘The problem is that the concept of surface area tums out to be a rather elusive ‘one as soon examined in detail. The difficulty in stating just what is meant by the “surface area” of a solid, inXVI-2 THE SURFACE AREA OF SOLIDS 373 (o) (b) (e) Fig. XVI-1. Successive enlargements of a hypothetical coastline. (From Ref. 1.) ‘one aspect, can be illustrated by considering the somewhat analogous question of what is meant by the “length” of a particular section of coastline. Super- ficially, it seems easy to say that the length is simply the distance along the shore between two points. Geography books, however, are prone to make such statements as “because of its many indentations, the coast of Maine is actually some 2000 miles long.” It seems reasonable that the coast of Maine should be considered as being greater than that of a straight line drawn from border to border and that the added length due to harbors, bays, points, and so on should be included. However, once one begins to consider irregularities, the problem is when to stop. Figure XVI-1 shows how successive magnifications of a hypo- thetical stretch of coastline might look. Quite obviously there are indentations on indentations, and with each successive magnification one increases the com- puted coastline by a more or less constant factor. There seems to be no logical stopping point to the sequence of successive enlargements if one is after some “absolute” value for the length of the shore- line. One thus arrives at the point of considering irregularities due to individual stones on the beach, due to the individual grains of sand, and due to the sur- face roughness of each grain and, finally, of considering the surface area of individual atoms and molecules. In our particular example, the tides and waves would prevent the exact carrying out of the sequence, but apart from this, the Heisenberg principle comes in eventually to inform us that there is considerable ‘uncertainty in the location of electrons and that the “surface area” of an atom is a somewhat philosophical concept. Quite obviously, the geographer solves this logical difficulty by arbitrarily deciding that he will not consider irregularities smaller than some specified size. A similar and equally arbitrary choice is frequently made in the estimation of surface areas. The minimum size of irregularities to be considered may be574 XVI THE SOLID-GAS INTERFACE defined in terms of the resolving power of a microscope or the size of adsorbate molecules.+ ‘Another approach is the iterative one. In the case of the Maine coastline, one might count the number of coastal villages or landing jetties and take this number as a measure of the length of a stretch of coast. In the case of a solid surface, one might apply a method that attempts to count the number of atoms or molecules in the surface. The result obtained is undoubtedly useful, but again, it does not have the quality of being a unique measure of absolute surface area. Perhaps the closest approach to such a measure of surface area is the geometric area of a plane liquid surface or of a perfect crystalline plane. Finally, in the case of solids, there is the difficulty that surface atoms and molecules differ in their properties from one location to another. The discussion in Section VII-4 made clear the variety of surface heterogeneities possible in the case of a solid. Those measurements that depend on the state of surface atoms or molecules will generally be influenced differently by such heterogeneities. Different methods of measuring surface area will thus often not only give dif- ferent absolute values, but may also give different relative values for a series of solids. ‘The preceding discussion serves to emphasize the point that, although the various topics to be discussed in this and succeeding chapters relate to the common property, surface area, the actual meaning of the answer arrived at is generally characteristic of the method and associated theory. Fortunately, one is very commonly more interested in relative surface areas than in accurate “abso- lute” values. Any given method will usually be valuable in determining ratios of areas between various materials (subject to the reservation made in the preced- ing paragraph). The comparisons, moreover, will be particularly valid when they are to be applied to the prediction or correlation of properties closely related to the one used in the measurement. Thus surface areas estimated by adsorption studies are more likely to lead to valid predictions of how additional, similar adsorbates will behave than to accurate predictions of, say, how the relative heats of immersion or the relative microscopic areas should vary. B. Surfaces as Having a Fractal Geometry Figure XVI-1 and the related discussion first appeared in 1960 [1], and since then a very useful mathematical approach to irregular surfaces has been applied to the matter of surface area measurement. Figure XVI-1 suggests that a coast- line might appear similar under successive magnifications, and one now pro- ceeds to assume that this similarity is exact. The result, as discussed in Section VII-4C and illustrated in Fig. VIL-6, is a self-s ilar line, or in the present case, a self-similar surface. Equation VII-21 now applies and may be written in the form ‘Reference 9 shows a needlepoint embroidery (by V. L. Adamson) illustrating the point that a large adsorbate molecule would "see" less of a rough surface than would a small oneXVI-2 THE SURFACE AREA OF SOLIDS 575 B= No) = Co“) (XVE-L) where N(a) is the number of adsorbate molecules of area ¢ needed to “cover” cone gram of the surface. For a perfect plane surface, D = 2 and E = C’, that is, does not depend on the area of the adsorbate molecules. In the case of a self- similar surface such that 2 < D < 3, E will increase with decreasing « and D may be obtained from a plot of In E versus In a. To the extent that Eq. XVI-1 is obeyed, C’ rather than © now characterizes the adsorbent. The specific surface area E will also vary with the particle size, and a more general equation is iD Cagumoa Rua (XVI-2) where R is the particle radius. If the particle surface is smooth, D = 2 and E varies inversely as R, as expected. In the limit of very porous particles, D—> 3, however, and L becomes independent of the particle size. A plot of In E versus In R thus provides an alternative method for determining D. The question of what @ to use is discussed in Ref. 2. The above methods for obtaining D, as well as other ones, are reviewed in Refs. 3-12, and Refs. 7-9 give tables of D values for various adsorbents. For example, D is close to 3 for the highly porous silica gels and close to 2 for nonporous fumed silica and for graphitized carbon black; coconut charcoal and alumina were found to have D values of 2.67 and 2.79, respectively [7]. ‘The chemical reactivity of a self-similar surface should vary with its frac- tional dimension. Consider a reactive molecule that is approaching a surface to make a “hit.” Taking Fig. VII-6d as an illustration, it is evident that such a molecule can “see” only a fraction of the surface. The rate of dissolving of quartz in HF, for example, is proportional to R-3), where D,, the reactive dimension, is about the same as D determined from adsorption studies (see Refs. 7-12). Values of D have also been obtained from bimolecular energy transfer reactions where acceptor and donor molecules must make an encounter by dif- fusing along the surface and also from the variation of x-ray scattering with scattering angle (see Refs. 4-15). The common factor in all of these approaches is that an analysis of the change in some measurable quantity as the yardstick size (adsorbent size, adsorbate size, diffusion distance, scattering distance) is varied. It might be thought that self-similarity is too demanding a mathematical property to be likely to occur in nature, and some reservations are, indeed, appropriate. If the range is small, data may only appear to fit the linear log-log plot used to obtain a D value. Other explanations, including surface chemical heterogeneity, may be important. Yet it does appear that approximate self-sim- ilarity over a fair range of yardstick size is not uncommon. Nor is it actually unexpected. Computer simulations of particle growth by nonequilibrium accre- tion or aggregation processes do lead to self-similar surfaces, for example (see Refs. 11-18).576 XVL_ THE SOLID-GAS INTERFACE, C. Methods Requiring Knowledge of the Surface Free Energy or Total Energy ‘A number of methods have been described in earlier sections whereby the surface free energy or total energy could be estimated. Generally, it was nec- essary to assume that the surface area was known by some other means; con- versely, if some estimate of the specific thermodynamic quantity is available, the application may be reversed to give a surface area determination. This is true if the heat of solution of a powder (Section VII-5B), its heat of immersion (Section X-3A), or its solubility increase (Section X-2) are known, ‘Two approaches of this type, purporting to give absolute surface arcas, might be mentioned. Bartell and Flu [19] proposed that the heat of immersion of a powder in a given liquid Gimm = L(Esy — Es.) (XVI-3) can be combined with data on the temperature dependence of the contact angle for that liquid to obtain the area ¥ of the sample. Thus from Eq. X-39, (XVE-4) dos @ imm = 2 (& cos @- Ty Fr ) where ginm is on a per gram basis so that E is area per gram, The actual use of Eq. XVI-4 to obtain 5 values is far from easy (see Ref. 20). ‘The general type of approach, that is, the comparison of an experimental heat of immersion with the expected value per square centimeter, has been discussed and implemented by numerous authors [21,22]. It is possible, for example, to estimate Esy - Es. from adsorption data or from the so-called isosteric heat of adsorption (see Section XVII-12B). In many cases where approximate rel- ative areas only are desired, as with coals or other natural products, the heat of immersion method has much to recommend it. In the case of microporous adsorbents surface areas from heats of immersion can be larger than those from adsorption studies [23], but the former are the more correct [24]. A second method that has been suggested as giving absolute surface areas involves the following procedure. If the solid is first equilibrated with saturated vapor, the molar free energy of the adsorbed material must be equal to that of the pure liquid, and it is a reasonable (but nonthermodynamic) argument that therefore the interfacial energy £, for the adsorbed film-vapor interface will be the same as for the liquid-vapor interface. Note that the adsorbed film is thought of as being duplex in nature in that the film-vapor interface is regarded as distinct from the film-solid interface. If the solid is now immersed in pure liquid adsorbate, the film-vapor interface is destroyed, and the heat liberated should correspond to 4 E,, thus giving the area A. This assumption was invoked by Harkins and Jura [21] in applying the method, with some success, to a nonporous powder. A concem, however, isXVI-2_ THE SURFACE AREA OF SOLIDS 577 that if the adsorbed film is thick, interparticle and capillary condensation will occur to give too low an apparent area. Roquerol and co-workers, however, have shown that in the case of a ground quartz—water system, the heat of immersion became constant after only about 1.5 statistical monolayers of adsorbed water [25]; see also Ref. 24. D. Rate of Dissolving A rather different method from the preceding is that based on the rate of dissolving cof a soluble material. At any given temperature, one expects the initial dissolving rate to be proportional to the surface area, and an experimental verification of this expectation hhas been made in the case of rock salt (see Refs. 26, 27). Here, both forward and reverse rates are important, and the rate expressions are ANAC) _ a = ha (XVES) R ah = keane (KVI-6) where aysci denotes the activity of the dissolved salt. By employing a single cubic piece of rock salt, it was possible to confirm the above rate expressions. In such cases, the rate constants ki and k2 are found to depend on the degree of stirring, although their ratio at equilibrium, of course, does not. The rate-controlling step is therefore probably one of diffusion through a boundary liquid film, since the liquid layers next to the solid surface will possess very litle velocity normal to the surface. In dissolving, a concentration Co, equal to that of the saturated solution, is built up in the layer immediately adjacent to the solid, whereas at some distance 7 away the concentration is equal to that of the bulk solution C. The dissolving rate is determined by the rate of diffusion of solute across this concentration gradient. Although the rate of dissolving measurements do thus give a quantity identified as the total surface area, this area must include that of a film whose thickness is on the order of a few micrometers but basically is rather indeterminate. Areas determined by this procedure thus will not include microscopic roughness (or fractal nature). The slow step in dissolving, alternatively, may be that of the chemical process of dissolving itself. This apparently was the case in a study (28] of the rate of dissolving of silica by hydrofluoric acid. The method was calibrated by determining the rate of dissolving of silica rods of geometrically determined area and was then applied to sam- ples of powdered material. The rate was not affected by stirring rate and, moreover, the specific surface area of the silica powder so obtained, about 0.5 m?/g, was found to agree well with that estimated from adsorption isotherms. As noted in Section XVI-2B, however, chemical dissolving rate constants may vary with particle size. E. The Mercury Porosimeter The method to be described determines the pore size distribution in a porous material or compacted powder; surface areas may be inferred from the results,578 XVI THE SOLID-GAS INTERFACE however. The method is widely used and, moreover, does make use of one of the fundamental equations of surface chemistry, the Laplace equation (Eq. 11-7). It will be recalled that Bartell and co-workers [29,30] determined an average pore radius from the entry pressure of a liquid into a porous plug (Section X-6A), ‘A procedure that is more suitable for obtaining the actual distribution of pore sizes involves the use of a nonwetting liquid such as mercury—the contact angle on glass being about 140° (Table X-2) (but note Ref. 31). If all pores are equally accessible, only those will be filled for which 2y\ cos 6} rete (XVE7) 80 that each increment of applied pressure causes the next smaller group of pores to be filled, with a concomitant increase in the total volume of mercury penetrated into the solid. See Refs. 32-35 for experimental details; commercial equipment is available (see Refs. 36, 37). The analysis of the direct data, namely, volume penetrated versus pressure, is as follows. Let dV be the volume of potes of radii between r and r - dr; dV will be related to r by some distribution function D(r): aV=-D(r) dr (XVI-8) For constant 9 and + (the contact angle was found not to be very dependent on pressure), one obtains from the Laplace equation, Pdr+rdP=0 (XVI-9) which, in combination with Eq. XVI-8, gives aV- Di) 5 P= DO) nee? ap (XVE-10) or _ @/nav . pn: Sa (XVI) Thus D(r) is given by the slope of the V versus P plot. The same distribution function can be calculated from an analysis of vapor adsorption data showing hysteresis due to capillary condensation (see Section XVII-16). Joyner and co- workers [38] found that the two methods gave very similar results in the case of charcoal, as illustrated in Fig. XVI-2. See Refs. 36 and 39 for more recent such comparisons. There can be some question as to what the local contact angle is [31,40]; an error here would shift the distribution curve.XVI-2 THE SURFACE AREA OF SOLIDS 579 rasan TS) nen emt arog water 58 | 0.002 = Porosineter #130") 645 Mo) 22 | Masaures (BET) a8 ese ° 700 200 300 Pore radius, + (A) Fig. XVI-2. Comparison of the pore volume distribution curves obtained from porosimeter data assuming contact angles of 140° and 130° with the distribution curve obtained by the isotherm method for a charcoal. (From Ref. 38.) ‘The foregoing analysis regards the porous solid as equivalent to a bundle of capillaries of various size, but apparently not very much error is introduced by the fact that such solids in reality consist of interconnected channels, provided that all pores are equally accessible (see Refs. 41, 42); by this it is meant that access to a given-sized pore must always be possible through pores that are as large or larger. An extreme example of the reverse situation occurs in an “ink bottle” pore, as illustrated in Fig. XVI-3. These are pores that are wider in the interior than at the exit, so that mercury cannot enter until the pressure has risen to the value corresponding to the radius of the entrance capillary. Once this pressure is realized, however, the entire space fills, thus giving an erroneously high apparent pore volume for capillaries of that size. Such a situation should also lead to a hysteresis effect, that is, on reducing the pressure, mercury would leave the entrance capillary at the appropriate pressure, but the mercury in the ink-bottle part would be trapped. Indeed, there is often some hysteresis in that Fig. XVI-3. An ink bottle pore.580 XVI THE SOLID-GAS INTERFACE. the plot of V versus P is not retraced on depressurization (see Ref. 43). In some cases, hysteresis can be explained as being due to a difference between the advancing and receding contact angle [44]. Interestingly, a general thermodynamic relationship allows the surface area of a porous system (without ink bottles) to be calculated from porosimetry data, note Section XVII-I6B. The equation is [45] 1 Vn j Pav (XVI-12) 0 yoos 6 where V» is the total volume penetrated. There is good agreement between areas obtained this way and from nitrogen adsorption [37]. E. Other Methods of Surface Area Estimation There is a number of other ways of obtaining an estimate of surface area, including such obvious ones as direct microscopic or electron-microscopic examination. The rate of charging of a polarized electrode surface can give relative areas. Bowden and Rideal [46] found, by this method, that the area of a platinized platinum electrode was some 1800 times the geometric or apparent area. Joncich and Hackerman [47} obtained areas for platinized platinum very close to those given by the BET gas adsorption method (see Section XVI-5). The diffuseness of x-ray diffraction patterns can be used to esti- ‘mate the degree of crystallinity and hence particle size [48,49]. One important general approach, useful for porous media, is that of permeability determination; although some- what beyond the scope of this book, it deserves at least a brief mention. A simple law, known as Darcy's law (1956), states that the volume flow rate per unit area is proportional to the pressure gradient if applied to the case of viscous flow through a porous medium treated as a bundle of capillaries, 2, _ A AP _, AAP : fase XVEI3) et W G y where Q is the volume flow rate, r is the radius of the capillaries, A is the total cross- sectional area of the porous medium, / is its length, AP is the pressure drop, 7 is the viscosity of the fluid, and K is called the permeability of the medium. Tn applying Eq. XVI-13 to an actual porous bed, r is taken to be proportional to the volume of void space Ale, where ¢ is the porosity, divided by the amount of surface; alternatively, then, ___te Ao(l =) where Ao is the specific surface area of the particles that make up the porous bed, that is, the area divided by the volume of the particles A/(1-«). With the further assumption that the effective or open area is eA, K in Eq. XVI-13 becomes a (XVE-14)XVI-3_ NATURE OF SOLID SURFACES, 581 1(#)?_¢ XVI-IS) i (a) d-# ove) ‘The detailed consideration of these equations is due largely to Kozeny [50]; the reader is also referred to Collins [51]. However, itis apparent that, subject to assumptions con- cerning the topology of the porous system, the determination of K provides an estimate of do. It should be remembered that Ao will be the external area of the particles and will not include internal area due to pores (note Ref. 52). Somewhat similar equations apply in the case of gas flow; the reader is referred to Barrer [53] and Kraus and co-workers {54]. 3. The Structural and Chemical Nature of Solid Surfaces We have considered briefly the important macroscopic description of a solid adsorbent, namely, its specific surface area, its possible fractal nature, and if porous, its pore size distribution. In addition, it is important to know as much as possible about the microscopic structure of the surface, and contemporary surface spectroscopic and diffraction techniques, discussed in Chapter VIII, pro- vide a good deal of such information (see also Refs. 55 and 56 for short general reviews, and the monograph by Somorjai [57]). Scanning tunneling microscopy (STM) and atomic force microscopy (AFT) are now widely used to obtain the structure of surfaces and of adsorbed layers on a molecular scale (see Chapter VIII, Section XVIII-2B, and Ref. 58). On a less informative and more statistical basis are site energy distributions (Section XVII-14); there is also the somewhat large-scale type of structure due to surface imperfections and dislocations (Sec- tion VII-4D and Fig. XVIII-14). The composition and chemical state of the surface atoms or molecules are very important, especially in the field of heterogeneous catalysis, where mixed- surface compositions are common. This aspect is discussed in more detail in Chapter XVIII (but again see Refs. 55, 56). Since transition metals are widely used in catalysis, the determination of the valence state of surface atoms is important, such as by ESCA, EXAFS, or XPS (see Chapter VIII and note Refs. 59, 60). Many solids have foreign atoms or molecular groupings on their surfaces that are so tightly held that they do not really enter into adsorption-desorp- tion equilibrium and so can be regarded as part of the surface structure, The partial surface oxidation of carbon blacks has been mentioned as having an important influence on their adsorptive behavior (Section X-3A); depending on conditions, the oxidized surface may be acidic or basic (see Ref. 61), and the surface pattern of the carbon rings may be affected (62]. As one other exam- ple, the chemical nature of the acidic sites of silica-alumina catalysts has been a subject of much discussion. The main question has been whether the sites Tepresented Bronsted (proton donor) or Lewis (electron-acceptor) acids. Hall382 XVI THE SOLID-GAS INTERFACE et al. [63] have addressed this type of question in the case of zeolites; see also Refs. 64 and 65. 4. The Nature of the Solid~Adsorbate Complex Before entering the detailed discussion of physical and chemical adsorption in the next two chapters, it is worthwhile to consider briefly and in relatively general terms what type of information can be obtained about the chemical and structural state of the solid-adsorbate complex. The term complex is used to avoid the common practice of discussing adsorption as though it occurred on an inert surface. Three types of effects are actually involved: (1) the effect of the adsorbent on the molecular structure of the adsorbate, (2) the effect of the adsorbate on the structure of the adsorbent, and (3) the character of the direct bond or local interaction between an adsorption site and the adsorbate. A. Effect of Adsorption on Adsorbatet Properties Statistical Thermodynamics of Adsorbates. First, from a thermodynamic or statistical mechanical point of view, the internal energy and entropy of a molecule should be different in the adsorbed state from that in the gaseous state. This is quite apart from the energy of the adsorption bond itself or the entropy associated with confining a molecule to the interfacial region. It is clear, for example, that the adsorbed molecule may lose part or all of its freedom to rotate. It is of interest in the present context (and is useful later) to outline the sta- tistical mechanical basis for calculating the energy and entropy that are associ- ated with rotation [66]. According to the Boltzmann principle, the time average energy of a molecule is given by = ENE geet? aS Se (XVE16) where the summations are over the energy states €;, and g; is the statistical weight of the jth state. It is the assumption of statistical thermodynamics that this time-average energy for a molecule is the same as the instantaneous aver- age energy for a system of such molecules. The average molar energy for the system, B, is taken to be Nog. Equation XVI-16 may be written in the form ang =RT? E-RT! (XVI-17) An altemative term for adsorbate is adsorptive; it is more commonly used in European literature than in the U.S. literature.XVI-4 THE SOLID-ADSORBATE COMPLEX. 583 where Q is the partition function and is defined as the term in the denominator of Eq. XVI-16. The heat capacity is then a\ _R FinQ G-= (#) (FMF (XVE-18) from which the entropy is found to be 7 E s=f Cydin T= F+RInQ (XVE-19) 0 if the entropy at 0 K is taken to be zerv. The Helmholtz free energy A is just E-TS, so A=-RTINQ (XVI-20) The rotational energy of a rigid molecule is given by J(J + 1)h?/8x7IKT, where J is the quantum number and fis the moment of inertia, but if the energy level spacing is small compared to kT, integration can replace summation in the evaluation of Q,.., which becomes mp : ( Sel (XVI-21) ees aa Equation XVI-21 provides for the general case of a molecule having n inde- pendent ways of rotation and a moment of inertia J that, for an asymmetric molecule, is the (geometric) mean of the principal moments. The quantity o is the symmetry number, or the number of indistinguishable positions into which the molecule can be turned by rotations. The rotational energy and entropy are [66,67] nRT Era = "5 (KVI-21a) 1 (8eUkT\"?| on su-n {in| ( Re ) ay (XVI-22) Molecular moments of inertia are about 10~°? g/cm’; thus / values for ben- zene, Nz, and NH; are 18, 1.4, and 0.28, respectively, in those units. For the case of benzene gas, o = 6 and n = 3, and S,ox is about 21 cal K"! mol"! at 25°C. On adsorption, all of this entropy would be lost if the benzene were unable to rotate, and part of it if, say, rotation about only one axis were possible (as might be the situation if the benzene was subject only to the constraint of lying flat584 XVI THE SOLID-GAS INTERFACE ‘on the surface). Similarly, all or part of the rotational energy of 0.88 kcal/mol would be liberated on adsorption or perhaps converted to vibrational energy. Vibrational energy states are too well separated to contribute much to the entropy or the energy of small molecules at ordinary temperatures, but for higher temperatures this may not be so, and both internal entropy and energy changes may occur due to changes in vibrational levels on adsorption. From a somewhat different point of view, it is clear that even in physical adsorption, adsorbate molecules should be polarized on the surface (see Section VI-8), and in chemisorption more drastic perturbations should occur. Thus internal bond energies of adsorbed molecules may be affected. Vibrational and Related Spectroscopies. The determination of the charac- teristic vibrational frequencies of adsorbed molecules is useful in at least two general ways. The spectrum can serve to identify what molecular species are present, which is important when a chemical reaction may have occurred; changes in molecular symmetry and bond strengths may manifest themselves as well as the new adsorbent-adsorbate bond that has formed. A common experi- ‘mental approach is simply that of infrared absorption spectroscopy—see Refs. 68 and 69 and, more recently, Refs. 57, 70, and 71. As an example, the pro- portion of Brdnsted and Lewis acid sites on a promoted ZrO, catalyst was determined using pyridine as an indicator {71a}. While transmission infrared spectroscopy can often be used for high-surface-area oxide-type adsorbents, in other cases (such as with metal adsorbents) RAIRS (reflection—absorption infrared spectroscopy) or EELS (electron-energy-loss spectroscopy) or VEELS (vibrational EELS) is necessary. Figure XVI-4 shows the RAIRS spectrum for propene adsorbed on a Pt(111) surface along with transmission IR spectra for propene and propane adsorbed on SiO2-supported Pt. Of interest was the spec- tral evidence for the presence of surface propylidyne. Other examples are found in Section XVIII-2C, and see Ref. 72 for a review on EELS. ‘An interesting point is that infrared absorptions that are symmetry-forbidden and hence that do not appear in the spectrum of the gaseous molecule may appear when that molecule is adsorbed. Thus Sheppard and Yates [74] found that normally forbidden bands could be detected in the case of methane and hydrogen adsorbed on glass; this meant that there was a decrease in molecular symmetry. In the case of the methane, it appeared from the band shapes that some reduction in rotational degrees of freedom had occurred, Figure XVII-16 shows the IR spectrum for a physisorbed Hy system, and Refs. 69 and 75 give the IR spectra for adsorbed N2 (on Ni) and OQ; (in a zeolite), respectively. Vibrational spectra of adsorbed species may also be obtained by Raman spec- troscopy (see Refs. 75, 76, and 78). Here, irradiation in the wavelength region of an adsorbate electronic absorption band can show features representing con- comitant changes in vibrational states, or a resonance Raman effect. Figure XVI-S shows spectra obtained with p-nitrosodimethylaniline, NDMA, as the absorbing or probe molecule, which was adsorbed on ZnO [79]. The intensity changes of the 1600, 1445, and 1398 cm”! peaks indicated that on adsorption of NH, the NDMA was displaced from the acidic sites that it originally occupied.XVI-4 THE SOLID-ADSORBATE COMPLEX 585 fone je (b) v/em" Fig. XVI-4. (a) RAIRS for propene on PU(II1) at 340 K. (6) Transmission IR of propane on Pr/SiO at ~300 K; (c) same as (6) but with propene adsorbed. IT denotes pro- Pylidyne assignments. (From Ref. 73.) (Reproduced by permission of Royal Society of Chemistry.) See Refs. 80 and 81 for other examples. Surface-enhanced Raman spectroscopy is discussed in Section XVI-4C. Information about molecular perturbations may also be found from UV- visible spectroscopy of adsorbed species. Kiselev and co-workers [82] observed shifts of 200-300 cm™! (about 700 cal/mol) to longer wavelengths in the ultra- violet absorption spectra of benzene and other aromatic species on adsorption on Aerosil (a silica). Leermakers and co-workers [83] made extensive studies of the spectra and photochemistry of adsorbed organic species. The absorp- tion spectrum of Ru(bipyridine)3* is shifted slightly on adsorption on silica gel {84}. Absorption peak positions give only the difference in energy between the ground and excited states, so the shifts, while relatively small, could reflect larger but parallel changes in the absolute energies of the two states. In other586 XVI THE SOLID-GAS INTERFACE RAMAN INTENSITY —> 1600 1800r400 180012001100 0 (emt) Fig. XVI-5. Resonance Raman spectra of NDMA adsorbed on ZnO; (a) in the presence of 100 torr of NH3; (b) after evacuation of the NH3 from the cell. [Reprinted with permission from J. F. Brazdil and E. B. Yeager, J. Phys. Chem., 85, 1005 (1981) (Ref. 79). Copyright 1981, American Chemical Society.] words, the shifts suggest that appreciable changes in the ground-state energies ‘occurred on adsorption, As with the lost rotational energy, these would appear as contributions to the heat of adsorption. Electron Spin Resonance Spectroscopy. Several ESR studies have been reported for adsorption systems 85-90]. ESR signals are strong enough to allow the detection of quite small amounts of unpaired electrons, and the shape of the signal can, in the case of adsorbed transition metal ions, give an indication of the geometry of the adsorption site. Ref. 91 provides a contemporary example of the use of ESR and of electron spin echo modulation (ESEM) to locate the environment of Cu(II) relative to 3'P in a microporous aluminophosphate molecular sieve. Nuclear Magnetic Resonance. NMR studies are limited to atoms having a nonzero nuclear magnetic moment, common ones being 'H, 7H, '8C, !°N. '70, !9F, ?°Si, *'P, and more esoteric ones such as °°Mo and '2°Xe. There are two types of problems. One is that the intensity of the NMR signal may not be strong enough unless high surfaceXVI-4 THE SOLID-ADSORBATE COMPLEX 3587 area adsorbents are used, such as a silica gel or a zeolite, and the other is that line widths tend to be such as to obscure chemical shifts (but see further below). The early work concentrated on proton NMR (e.g., Ref. 92). In the case of n-heptane adsorbed con Spheron or diamond, the line width for the first adsorbed layer was very large, but with three or more adsorbed layers, a line structure developed [93]. Again, a narrow line is associated with sufficient mobility that local magnetic fields are averaged, and the interpretation was that adsorbed layers beyond the second were mobile. The lon- gitudinal and transverse relaxation times T and T> can be interpreted in terms of a correlation time 1, which is about the time required for a molecule to tum through a radian or to move a distance comparable with its dimensions. Zimmerman and Lasater 194], studying water adsorbed on a silica gel, found evidence for two kinds of water for surface coverages greater than 0.5 and, as illustrated in Fig. XVI-6, a variation of catculated correlation time with coverage, which suggested that surface mobility was fully developed at about the expected monolayer point. A tentative phase identification can be made on the basis that ¢ for water is about 10~' sec at 25°C, while for ice (or “rigid” water) it is about 10-5 sec. That is, inspection of Fig. XVI-6 would suggest that below 0 = 0.5, the adsorbed water was solidlike, while above 6 = 1 it was liquidlike. As an example of the sophistication of more recent work, Dybowski et al. {95] studied the tf] § 10° 10? = ‘ wr eh 10 107? 107" 10° x eho. M gS, Fig. XVI-6. Nuclear correlation times for water adsorbed on silica gel. (From Ref. 94.)588 XVI THE SOLID-GAS INTERFACE. heteronuclear coupling of 'H and '95Pt for benzene adsorbed on a Pt/AlzOs catalyst to draw conclusions as to the nature of the adsorption site and the Pt-benzene distance. ‘Some useful general references are those by Hunter, Dybowski, and co-workers [96-98]. There has been much interest in the mobility of species adsorbed on or in layer and cage minerals such as clays and zeolites (see Fig. XVII-26 for an illus- trative structure of the latter). As one example, in the case of benzene adsorbed on a zeolite, '"C NMR showed a line broadening (and hence correlation time) which increased with n, the number of molecules per cage, up to n = 4, Beyond this point there was a rapid line narrowing, indicating that a new, highly mobile phase was present [99]. In a study of chemisorbed methanol on MgO, "°C NMR indicated molecules to be rigidly bound below a half monolayer coverage but that higher coverages produced isotropically rotating molecules [100]. As another illustration, zeolite rho is a relatively fiexible zeolite with useful cat- alytic properties. Xenon-129 NMR spectroscopy indicates this flexibility and also that there can be rapid exchange between two kinds of adsorption sites [101-103]. Finally, it has been possible to determine the long-range diffusion coefficient of Nz in commercial zeolite crystals, using pulsed field gradient NMR. Values of D ranging from 10°? to 10-4 cm?/sec were found [103a]. ‘An important development has been that of “magic angle” spinning (MAS) [97]. A major source of line broadening is that of dipole-dipole relaxation, due to fluctuating local magnetic fields imposed on the externally applied field; the effect may be due to protons in the same molecule, for example. The rapid tumbling of molecules in fluid solution tends to average out the local field, giving sharp NMR lines. Molecules adsorbed on a solid, however, may not be able to tumble (or move around) on the NMR time scale—hence the typical line broadening that is observed. The local field Big. at a nucleus A generated by a neighboring nucleus B is given by Bice = tpar,y(3.00s? 8 ~ 1) (XVI-23) where jp is the magnetic moment of B, rag the internuclear distance, and 0 the angle between the internuclear vector and the applied field. The (3 cos? 6 ~ 1) term is zero for 6 = 54°44’. In solution, drifts rapidly over all values, averaging out the effect, but essentially the same result can be obtained with a solid sample by spinning it rapidly at the magic angle (about 3 Hz is needed). In many cases, relatively sharp lines can now be obtained. One may now see chemical shifts due to sites of differing acidity, for example [104]. Other examples are found in Refs. 96 and 105-109. Dielectric Behavior of Adsorbed Water. Determination of the dielectric absorption of adsorbed water can yield conclusions similar to those from proton NMR studies and there is a considerable, although older literature on the subject. Figure XVI-7 illustrates how the dielectric constant for adsorbed water varies with the frequency used as well as with the degree of surface coverage. A characteristic relaxation time 7 can be estimatedXVI-4 THE SOLID-ADSORBATE COMPLEX 389 rr a a a) Rettve pressure, PIP? Fig. XVI-7. Dielectric isotherms of water vapor at 15°C adsorbed on a-Fe203 (solid points indicate desorption). A complete monolayer was present at P/P® = 0.1, and by P/P® = 08 several layers of adsorbed water were present. (From Ref. 110.) from the frequency dependence of the dielectric constant, especially in combination with that of the dielectric loss. This relaxation time is essentially that for the molecule to move or to reorient so as to follow a changing electric field. For water on a-Fe>Ox, 1 varied from about 1 sec at monolayer coverage to 10° sec when several adsorbed layers were present [110]. Since the characteristic time was far larger than for liquid bulk water, 10~'? sec, the conclusion was that the adsorbed water was present in an icelike, hydrogen-bonded structure. A more recent study indicated adsorption steps for water on FeOOH-type oxides [111]. Similar, very detailed studies were made by Ebert [112] on water adsorbed on alu- ‘mina with similar conclusions. Water adsorbed on zeolites showed a dielectric constant of only 14~21, indicating greatly reduced mobility of the water dipoles {113]. Similar results were found for ammonia adsorbed in Vycor glass [114]. Klier and Zettlemoyer {114a] have reviewed a number of aspects of the molecular structure and dynamics of water at the surface of an inorganic material. The state of an adsorbate is often described as “mobile” or “localized,” usually in connection with adsorption models and analyses of adsorption entropies (see Section XVII-3C). A more direct criterion is, in analogy to that of the fluidity of a bulk phase, the degree of mobility as reflected by the surface diffusion coefficient. This may be estimated from the dielectric relaxation time; Resing [115] gives values of the diffusion coefficient for adsorbed water ranging from near bulk liquids values (10°Scm? sec) to as low as 10-%cm?/sec. B. Effect of the Adsorbate on the Adsorbent It is evident from the preceding material that a great deal of interest has centered on the chemical and physical state of the adsorbate. There is no reason not to expect the adsorbate to affect properties of the adsorbent. For example,590 XVI THE SOLID-GAS INTERFACE as discussed in Section XVII-12Aii, adsorbent thermodynamic properties must change as adsorption occurs. Of more interest is the matter of adsorbent surface structural changes accompanying adsorption. First, it is entirely possible that surface heterogeneities and imperfections undergo some reversible redistribution during adsorption. As noted in Section ‘VII-4B, Dunning [116] has considered that since the presence of an adsorbed molecule should alter the energy of special sites (such as illustrated in Fig. ‘VII-5), above some critical temperature for surface mobility, their distribution should depend on the extent of adsorption. Also, the first few layers of the crys- talline surface of a solid are distorted (see Section VII-3B), and this distortion should certainly be altered if an adsorbed layer is present; here no more than motions perpendicular to the surface may be involved. There is a scattering of carly literature indication that changes in surface structure, specifically, surface reconstruction, can occur on adsorption. In the 1960’s Lander and Morrison [117] concluded from a LEED study that considerable surface rearrangement of germanium surfaces took place on adsorption of iodine. Similarly, field emis- sion work with tungston led Ehrlich and Hudda {118] to conclude that recon- struction could accompany the adsorption or desorption of nitrogen. Hydrogen atom adsorption on metals can lead to complex surface reconstructions, as for H on Ni(110) [56] and on W(100) [119]. Somorjai [120] gives a somewhat more detailed account of such effects. ‘Some fascinating effects occur in the case of CO on Pt(100). As illustrated in Fig. XVI-8, the clean surface is reconstructed naturally into a quasi-hexag- onal pattern, but on adsorption of CO, this reconstruction is lifted to give the bulk termination structure of (110) planes [56]. As discussed in Section XVIII- 9E very complicated changes in surface structure occur on the oxidation of CO P1100) co hex == 1x1 Fig. XVI-8. (a) The quasi-hexagonal surface structure of clean Pt(100) surface. (b) Adsorption of CO lifts this reconstruction to give the structure corresponding to the termination of (100) planes (from LEED studies). (Reprinted with permission from G. Ertl, Langmuir, 3, 4 (1987) (Ref. 56). Copyright 1987, American Chemical Society:]XVI-4 THE SOLID-ADSORBATE COMPLEX 591 on various metal surfaces. The picture that emerges is that the adsorbent is not just a chemically active platform for catalyzed reaction but may execute a complicated pas 4 deux with the adsorbate. As discussed in connection with Table XVI-1, one can expect important surface structural changes to accompany adsorption whenever the adsorbent-adsorbate bond is comparable in energy to the heat of sublimation of the adsorbent, that is, to the adsorbent-adsorbent bond energy. In the case of refractory solids, such as metals and ionic crystals and oxides, sublimation energies are in the range of 20-100 kcal/mol, which is also that for chemisorption. One thus expects surface reconstruction to be likely under chemisorption conditions. The energy of physical adsorption, which is due to van der Waals forces, is typically around 5-10 kcal/mol, about the same as the cohesion energy of molecular solids and of chain-chain interactions in polymers. ‘One concludes that physical adsorption should not perturb the structure of a refractory adsorbent but should be able to do so in the case of a molecular solid. As an exam ple, surface restructuring is indicated when n-hexane is adsorbed on ice above -35°C {121,122}. The surface structure of polymers has been estimated from LEIS (see p. 296) ‘measurements; here, one can expect the surface packing of polymer chains to be affected when physical adsorption of a vapor occurs. C. The Adsorbate-Adsorbent Bond The immediate site of the adsorhent-adsorbate interaction is presumably that between adjacent atoms of the respective species. This is certainly true in chemisorption, where actual chemical bond formation is the rule, and is largely true in the case of physical adsorption, with the possible exception of multilayer formation, which can be viewed as a consequence of weak, long-range force fields. Another possible exception would be the case of molecules where some electron delocalization is present, as with aromatic ring systems. It has been customary to assign all of the observed heat of adsorption to the adsorption bond, although the preceding discussion has made it clear that other, less localized contributions can be important. Furthermore, the matter is complicated by the surface reconstruction that can occur. The various spec- troscopies discussed in Section XVIII-2 give some information about chemisorption bond energies and bond lengths, and the chemisorption bond itself is covered in Section XVIII-6. There are no great surprises but there are complications if the adsorbate has more than one possible bonding atom and if bridging bonds must be considered, as in Fig. XVIII-5. The physisorption bond, being relatively weak, is even more difficult to char- acterize than the chemisorption one. Some aspects of this are covered in Section XVII-10. SERS. A phenomenon that certainly involves the adsorbent-adsorbate inter- action is that of surface-enhanced resonance Raman spectroscopy, or SERS. ‘The basic observation is that for pyridine adsorbed on surface-roughened sil- ver, there is an amazing enhancement of the resonance Raman intensity (see Refs. 124-128). More recent work has involved other adsorbates and colloidal592 XVI THE SOLID-GAS INTERFACE silver particles [129-132]. The intensity enhancement, around a factor of 10°, has been attributed to a large electric fields generated by oscillations of the elec- trons in small metal structures [129,133,133]. There is also the possibility that much of the effect is due to the actual surface area being larger than thought (see Ref. 134). 5. Problems 1. The following data have been obtained for powdered periclase (see Ref. 2) (mean particle diameter, um; surface area, m?/g): (3.39; 2.280), (4.77; 1.541), (8.03; 0.854), (11.8; 0.543), (18.0; 0.421), (22.6; 0.295). Make the appropriate plot and determine D. 2. The following data have been obtained for a sample of coal mine dust (see Ref. 2) (mean particle diameter, um; surface area, m?/g): (2.40; 8.61), (2.82; 6.93), (5.95; 5.28), (8.00; 4.51). Make the appropriate plot and obtain D. Comment on the result. 3. The following data have been obtained for a succession of adsorbates on a silica gel (see Ref. 4) (adsorbate; n, moles adsorbed at monolayer point; ¢, molecular area of adsorbate, A): (Nz; 5.29; 16.2), (CH3OH; 4.09; 17.8), (CH3CH2OH; 2.82; 22.9), [(CHs)3COH; 1.80; 31.6], [(CHs)(CH3CH2)COH; 1.50; 37.5)), ((CH3CH2)3COH; 1.26; 42.7]. Make the appropriate plot and obtain D, Comment on the result, 4. Derive Eq. XVI-12. 5. Ina mercury porosimetry experiment Hig is taken to be 480 dyn/cm, @ is 130°. If pressure is measured in psi and volume intruded in cm?/g, show that Eq. XVI-12 becomes E = 0.0223 j PV, where Eis in m?/g. The following data were obtained for a Sterling FT carbon (volume intruded, cm?/g; applied pressure, psi): (0; 2), (0.2; 2.05), (0.35; 2.1), (0.36; 5), (0.37, 10), (0.39; 35), (0.52; 100), (0.85; 500), (1.10; 1000). Estimate the specific surface area of this sample. 6. The contact angle for water on single-crystal naphthalene is 87.7° at 35°C, and d0/dT is -0.13 deg/K. Using data from Table III-1 as necessary, calculate the heat of immersion of naphthalene in water in cal/g if a sample of powdered naphthalene of 10 m?/g is used for the immersion study. (Note Ref. 135.) 7. Bartell and Flu [19] were able to determine the adhesion tension, that is, ysy — -YSL, for the water-silica interface to be 82.8 ergs/cm* at 20°C and its temperature change to be -0.173 erg cm? K~!, The heat of immersion of the silica sample in water was 15.9 cal/g. Calculate the surface area of the sample in square centimeters per gram. 8. Harkins and Jura [21] found that a sample of TiO having a thick adsorbed layer of water on it gave a heat of immersion in water of 0.600 cal/g. Calculate the specific surface area of the TiO» in square centimeters per gram. 9. The rate of dissolving of a solid is determined by the rate of diffusion through a boundary layer of solution. Derive the equation for the net rate of dissolving. Take Co to be the saturation concentration and d to be the effective thickness of the diffusion layer; denote diffusion coefficient by D. 10. Using the curve given by the square points in Fig. XVI-2, make the qualitative reconstruction of the original data plot of volume of mercury penetrated per gram versus applied pressure.XVI GENERAL REFERENCES 593 I. Optical microscopic examination of a finely divided silver powder shows what appears to be spheres 2 um in diameter. The density of the material is 4 g/cm. The sur- face is actually rather rough on a submicroscopic scale, and suppose that a cross section through the surface actually has the appearance of the coastline shown in Fig. XVI-1, where the left arrow represents a distance of 0.1 4m and the successive enlargements are each by 10-fold. Make a numerical estimate, with an explanation of the assumptions involved, of the specific surface area that would be found by (a) a rate of dissolving study, (b) Harkins and Jura, who find that at P° the adsorption of water vapor is 6.5 cm’ STP/g (and then proceed with a heat of immersion measurement), and (c) a measurement of the permeability to liquid flow through a compacted plug of the powder. 12, Calculate the rotational contribution to the entropy of adsorption of benzene ‘on carbon at 35°C, assuming that the adsorbed benzene has one degree of rotational freedom. 13, Calculate the rotational contribution to the entropy of adsorption of ammonia on silica at -30°C, assuming (a) that the adsorbed ammonia retains one degree of rotational freedom and (b) that it retains none. In case (a) assume that the nitrogen is bonded to the surface. 14, Explain why Bioe in Eq. XVI-23 should average to zero for a freely tumbling molecule. 15. According to Eq. XVI-2 the specific surface area of an adsorbate should vary with particle size R as R~3, using the same adsorbate for each sample. Justify this relationship by means of a qualitative argument (an exact derivation is not required). General References D. Avnir, ed., The Fractal Approach to Heterogeneous Chemistry, Wiley, New York, 1989. ‘Advances in Catalysis, Vol. 10, Academic, New York, 1958; Vol. 12, 1960; Vol. 16, 1966. R.K. Chang and T. E. Furtek, eds., Surface Enhanced Raman Scattering, Plenum, New York, 1982. R. J. Clark and R. E. Hester, eds., Advances in Infrared and Raman Spectroscopy, Vol. 9, Heyden, London, 1982. R. E. Collins, Flow of Fluids through Porous Materials, Reinhold, New York, 1961 C. Dyhowski and R. L. Lichter, eds., NMR Spectroscopy Techniques, Marcel Dekker, New York, 1987, A. E, Flood, ed., The Solid-Gas Interface, Vol. 1, Marcel Dekker, New York, 1967. S. J. Gregg and K. 8. W. Sing, Adsorption, Surface Area, and Porosity, Academic, New York, 1982. M. L. Hair, Infrared Spectroscopy in Surface Chemistry, Marcel Dekker, New York, 1967. T. L. Hill, Introduction to Statistical Thermodynamics, Addison-Wesley, Reading, MA, 1962. R.R. Irani and C. F. Callis, Particle Size, Wiley, New York, 1963.594 XVI THE SOLID-GAS INTERFACE S. Ross and J. P. Olivier, On Physical Adsorption, Interscience, New York, 1964, J. K. M, Sanders and B. K. Hunter, Modern NMR Spectroscopy, 2nd ed., Oxford Uni- versity Press, New York, 1993. G. Somorjai, Introduction to Surface Chemistry and Catalysis, Wiley-Interscience, New York, 1994. J.M. Thomas and R. M. Lambert, eds., Characterization of Catalysts, Wiley, New York, 1980, A. Zangwill, Physics at Surfaces, Cambridge University Press, New York, 1988. ‘Textual References 1. A. W. Adamson, The Physical Chemistry of Surfaces, Imterscience, New York, 1960, p. 427. 2. A. ¥. Meyer, D. Farin, and D. Avnir, J. Am. Chem. Soc., 108, 7987 (1986). 3. D. Avnir and D. Farin, J. Colloid Interface Sci., 103, 112 (1985). 4. P. Pfeifer, in Preparative Chemistry Using Supported Reagents, P. Laszlo, ed., ‘Academic, New York, 1987. 5. H. Van Damme, P. Levitz, F. Bergaya, J. F. Alcover, L. Gatineau, and J. J. Fripiat, J. Chem. Phys., 85, 616 (1986). 6. P. Meakin, Multiple Scattering of Waves in Random Media and Random Rough Surfaces, The Pennsylvania State University Press, State College, PA, 1985. 7. D. Avnir, D. Farin, and P. Pfeifer, Nature, 308, 261 (1984) 8. D. Avnir, in Better Ceramics Through Chemistry, C. J. Brinker, ed., Material Research Society, Pittsburgh, PA, 1986. 9. D. Avnir, D, Farin, and P. Pfeifer, New J. Chem., 16, 439 (1992). 10. D. Farin and D. Avnir, J. Phys. Chem. 91, 5517 (1987). 11, D. Avnis, ed., The Fractal Approach to Heterogeneous Chemistry, Wiley, New York, 1989. 12. D. Farin and D. Avnir, J. Chromatog., 406, 317 (1987). 13. D. Rojanski, D. Huppert, H. D. Bale, X. Dacai, P. W. Schmidt, D. Farin, A. Seri- Levy, and D. Avnir, Phys, Rev. Lett, 56, 2505 (1986). 14. D. W. Schaefer, B. C, Bunker, and J. P. Wilcoxon, Phys. Rev. Lett., 58, 284 (1987). 15. P. Meakin, CRC Critical Reviews in Solid State and Materials Science, Vol. 13, Grand Rapids, MI, 1986, p. 143. 16. P. Meakin, H. E. Stanley. A. Coniglio, and T. A. Witten, Phys. Rev A, 32, 2364 (1985) 17. R. Kopelman, Science, 241, 1620 (1988). 18. P. Meakin and J. M. Deutsch, J. Chem. Phys., 85, 2320 (1986). 19, FE, Bartell and Y. Fu, in The Specific Surface Area of Activated Carbon and Silica, in Colloid Symposium Annual, H. B. Weiser, ed., Vol. 7, Wiley, New York, 1930. 20. J. W. Whalen and W. H. Wade, J. Colloid Interface Sci., 24, 372 (1967), 21. See G. Jura, in The Determination of the Area of the Surfaces of Solids in Physical29. 30. ai. 32. 33. 34, 35. 37, £588 42. 43. 45. 47. 48. 50. Si. 52. . RJ. Matyi, L. H. Schwartz, and J. B, Butt, Catal. Rev. Sci. Er XVI TEXTUAL REFERENCES 3595 Methods in Chemical Analysis, W. G. Berl, ed., Vol. 2, Academic, New York, 1951. . 8. Ergun, Anal. Chem, 24, 388 (1952). . A.J. Tyler, J. A. G. Taylor, B. A. Pethica, and J. A. Hockey, Trans. Faraday Soc., 67, 483 (1971). |. R. Denoyel, J. Fernandez-Colinas, Y. Grillet, and J. Rouquerol, Langmuir, 9, 515 (1993). . S. Partyea, F, Rouguerol, and J. Rouquerol, J. Colloid Interface Sci., 68, 21 (1979). A.B. Zdanovskii, Zh. Fiz. Khim., 2, 170 (1951); Chem. Abstr, 48, 4291c (1964). . A. R, Copper, Jr, Trans. Faraday Soc., $8, 2468 (1962). . W. G, Palmer and R. E, D. Clark, Proc. Roy. Soc. (London), A149, 360 (1935); see also E. Darmois, Chim. Ind. (Paris), 59, 466 (1948) and I. Bergman, J. Cartwright, and R. A. Bentley, Nature, 196, 248 (1962). FE, Bartell and C, W. Walton, Jr., J. Phys. Chem., 38, 503 (1934). See F. E. Bartell and L. S. Bartell, J. Am. Chem. Soc., 56, 2202 (1934). J. Kloubek, Powder Tech., 29, 89 (1981). H.L. Ritter and L. C. Drake, Ind, Eng. Chem., Anal. Ed., 17, 782 (1945). L. C. Drake, Ind. Eng. Chem., 41, 780 (1949). N. M. Winslow and J. J. Shapiro, ASTM Bull., No. 236, p. 39 (1959). L. A. de Wit and J. J. F. Scholten, J. Catal., 36, 36 (1975). . J. Kloubek and J. Medek, Carbon, 24, 501 (1986). D. R. Milburn, B. D. Adkins, and B. H. Davis, in Fundamentals of Adsorption, A. I. Liapis, ed., Engineering Foundation, New York, 1987. . L. G. Joyner, F. P. Barrett, and R. Skold, J. Am. Chem. Soc., 73, 3155 (1951), . B. D. Adkins and B. H. Davis, Adsorp. Sci. Technol., 5, 76, 168 (1988). . R. J. Good and R. Sh. Mikhail, Powder Tech., 29, 53 (1981). . D. H. Everett, in The Solid-Gas Interface, E. A. Flood, ed., Marcel Dekker, New York, 1967. D.N. Winslow, J. Colloid Interface Sci., 67, 42 (1978). H. M. Rootare and C. F. Prenzlow, J. Phys. Chem., 71, 2733 (1967); M. P. Astier and K. S. M. Sing, in Adsorption at the Gas-Solid and Liquid-Solid Interface, J. Rouquerol and K. S. W. Sing, eds., Elsevier, Amsterdam, 1982. . S. Lowell and J. E. Shields, J. Colloid Interface Sci., 80, 192 (1981). L. C. Drake and H. L. Ritter, Ind. Eng. Chem., Anal, Ed., 17, 787 (1945). .. F. P, Bowden and E. K. Rideal, Proc. Roy. Soc. (London), A120, 59 (1928). M. J. Joncich and N. Hackerman, J. Electrochem. Soc., 111, 1286 (1964). D. J.C. Yates, Can, J. Chem., 46, 1695 (1968). , 29, 41 (1987). J. Kozeny, Sitzber. Akad. Wiss. Wien, Wasserwirtsch. Math, Nat., Kl, Wa, 136,271 (1927); Wasserwirtschaft, 22, 67, 86 (1927). R. E. Collins, Flow of Fluids Through Porous Materials, Reinhold, New York, 1961. B. V. Derjaguin, D. V. Fedoseev, and S. P. Vnukov, Powder Tech., 14, 169 (1976).596 XVI THE SOLID-GAS INTERFACE 53. R. M. Barrer, Discuss. Faraday Soc., 3, 61 (1948); see also R. M. Barrer and D. M, Grove, Trans. Faraday Soc., 47, 826 (1951), 54, G. Kraus, J. W. Ross, and L. A. Girifalco, J. Phys. Chem., 57, 330 (1953); see also G. Kraus and J. W, Ross, J. Phys. Chem., $7, 334 (1953). 59. G. A. Somorjai and S. M. Davis, Chemtech, 13, 502 (1983); S. R. Bare and G. A. Somorjai, Encyclopedia of Physical Science and Technology, Vol. 13, Academic, New York, 1987. 56, G, Ertl, Langmuir, 3, 4 (1987). 57. G, A. Somorjai, Introduction t0 Surface Chemistry and Catalysis, Wiley-Inter- science, New York, 1994. 58. S.N. Magonov and M.-H. Whangbo, Surface Analysis with STM and AFM, VCH, New York, 1996. 59. J. N. Fiedor, A. Proctor, M. Houalla, P.M. A. Sherwood, F. M. Mulcahy, and D. M. Hercules, J. Phys. Chem., 96, 10967 (1992). 60. A. Yu. Stakheev and W. M. H. Sachtler, J. Chem, Soc., Faraday Trans., 87, 3703 (1991). 61. H. P. Boehm, Adv. Catal., 16, 179 (1966). 62. G. R. Henning, Z Elektrochem., 66, 629 (1962). 63. B. Umansky, J. Engelhardt, and W. K. Hall, J. Catal., 127, 128 (1991). 64, F. Boccuzzi, S, Coluccia, G. Ghiotti, C. Morterra, and A. Zecchina, J. Phys. Chem, 82, 1298 (1978). 65. B. A. Morrow, I. A. Cody, and L. S.M. Lee, J. Phys. Chem., 80, 2761 (1976). 66. A. W. Adamson, Textbook of Physical Chemistry, 3rd ed., Academic Press, New York, 1986, p. 137. 67. C. Kemball, Proc. Roy. Soc. (London), A187, 73 (1946), 68. L. H. Little, Infrared Spectra of Adsorbed Molecules, Academic, New York, 1966. 68a, M. L. Hair, Infrared Spectroscopy in Surface Chemistry, Marcel Dekker, New York, 1967. 69. R. P. Eischens, Acc. Chem, Res., 5, 74 (1972). 70. M. K. Weldon and C. M. Friend, Chem. Rev., 96, 1391 (1996). 71. N, Sheppard, Annu. Rev. Phys. Chem., 39, 589 (1988). Tha. K. Ebitani, H. Hattori, and K, Tanabe, Langmuir, 6, 1743 (1990). 72. M. A. Chesters and N. Sheppard, in Spectroscopy of Surfaces, R. J. H. Clark and R. E. Hester, eds., Wiley, New York, 1988, p. 377. 73, M. A. Chesters, C. De La Cruz, P. Gardner, E. M. McCash, P. Pudney, G. Shahid, and N. Sheppard, J. Chem. Soc. Faraday Trans., 86, 2757 (1990). 74, N. Sheppard and D. J.C. Yates, Proc. Roy. Soc. (London), A238, 69 (1956). 75. F, Jousse and E, C. De Lara, J. Phys. Chem., 100, 233 (1996), 76. P. J. Hendra, I. D. M. Tumer, E, J. Loader, and M. Stacey, J. Phys. Chem., 78, 300 (1974). 77. 5. F, Rabott, R. Santo, and J. D. Swalen, Appl. Spectrosc., 13, 549 (1979). 78. R. B. Quincy, M. Houalla, and D. M. Hercules, Fresenius J. Anal. Chem., 346, 676 (1993).1. 80. 81. 82. 83. 4. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101 102. 103. XVI TEXTUAL REFERENCES 597 J. F. Brazdil and E, B. Yeager, J. Phys. Chem., 85, 1005 (1981). T. Takenaka and K. Yamasaki, J. Colloid Interface Sci, 78, 37 (1980). J. F. Brazdil and E. B. Yeager, J. Phys. Chem., 85, 995, 1005 (1981). VN. Abramov, A. V. Kiselev, and V. I. Lygin, Russ. J. Phys. Chem., 37, 1507 (1963). See L. D. Weis, T. R. Evans, and P. A. Leermakers, J. Am. Chem. Soc., 90, 6109 (1968) and references therein. J. Namnath and A. W. Adamson, unpublished work. G. P. Lozos and B. M. Hoffman, J. Phys. Chem., 78, 200 (1974). M. B. McBride, T. J. Pinnavia, and M. M. Mortland, J. Phys. Chem., 79, 2430 «as75). S. Abdo, R. B. Clarkson, and W. K. Hall, J. Phys, Chem., 80, 2431 (1976). M. F, Ontaviani and G. Martini, J. Phys. Chem., 84, 2310 (1980), L. Kevan and S. Schlick, J. Phys, Chem., 90, 1998 (1986). 7 Che and A. J. Tench, Advances in Catalysis, Academic, New York, 1983, Vol. 2. T. Wasowica, S. J. Kim, S. B. Hong, and L. Kevan, J. Phys. Chem., 100, 15954 (1996). W. S. Brey, Jr. and K. D. Lawson, J. Phys. Chem., 68, 1474 (1964). D. Graham and W. D. Phillips, Proc. 2nd Int. Congr. Surf. Act., London, Vol. 1, p.22, J.R. Zimmerman and J. A. Lasater, J. Phys. Chem., 62, 1157 (1958). C.F. Tirendi, G. A. Mills, and C. Dyhowski, J. Phys. Chem., 96, 5045 (1992). J. K. M, Sanders and B. K. Hunter, Modern NMR Spectroscopy, Oxford University Press, Oxford, 1987. C. Dybowski and R. L. Lichter, eds., NMR Spectroscopy Techniques, Marcel Dekker, New York, 1987. J. KM, Sanders and B. K. Hunter, Modern NMR Spectroscopy, Oxford University Press, New York, 1993. V. Ya. Borovkov, W. K. Hall, and V. B. Kazanski, J. Catal., 51, 437 (1978), 1. D. Gay, J. Phys. Chem., 84, 3230 (1980). €. Tsiao. J. S. Kauffman, D. R. Corbin, L. Abrams, E. F. Carroll, Jr, and C. Dybowski, J. Phys. Chem., 95, 5586 (1991). M. L. Smith, D. R. Corgin, and C. Dybowski, J. Phys. Chem., 97, 9045 (1993). M, L, Smith, D. R, Corbin, L. Abrams, and C. Dybowski, J. Phys. Chem., 97, 7193 (1993). 103a. P. L. McDaniel, C. G. Coe, J. Kiirger, and J. D. Moyer, J. Phys. Chem., 100, 104. 105. 106, 16263 (1996). W. L. Earl, P. O, Fritz, A. A. V. Gibson, and J. H. Lunsford, J. Phys. Chem., 91, 2091 (1987). W. P, Rothwell, W. Shen, and J. H, Lunsford, J. Am. Chem. Soc., 106, 2452 (1984). D, Slotfeldt-Ellingsen and H. A. Resing, J. Phys. Chem., 84, 2204 (1980); J. B. Nagy, E. G. Derouane, H. A. Resing, and G. R. Miller, ibid., 87, 833 (1983).598 107. 108. 109. 110. Mm. 112. 13. 44, XVI THE SOLID-GAS INTERFACE E. C. Kelusky and C. A. Fyfe, J. Am. Chem, Soc., 108, 1746 (1986). D. Slotfeldt-Ellingsen, H. A. Resing, K. Unger, and J. Frye, Langmuir, 5, 1324 (1989). J.C, Edwards and P, D. Ellis, Langmuir, 7, 2117 (1991). E, McCafferty and A. C. Zettlemoyer, Disc. Faraday Soc., 239 (1971). K. Kaneko, M. Serizawa, T. Ishikawa, and K. Inouye, Bull. Chem. Soc. Jpn., 48, 1764 (1975). G. Ebert, Kolloid-Z., 174, 5 (1961). K.R. Foster and H. A. Resing, J. Phys. Chem., 80, 1390 (1976). I. Lubezky, U. Feldman, and M. Folman, Trans. Faraday Soc., 61, 1 (1965), 114a. K. Klier and A. C. Zettlemoyer, J. Colloid Interface Sci., $8, 216 (1977). us. 116. 47. 18. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130, 131. 132. 133. H. A. Resing, Adv. Mol. Relaxation Processes, 1, 109 (1967-68); ibid., 3, 199 1972). W. J. Dunning, J. Phys. Chem., 67, 2023 (1963). J.J. Lander and J. Morrison, J. Appl. Phys., 34, 1411 (1963) and preceding papers. G. Ehrlich and F, G. Hudda, J. Chem. Phys., 35, 1421 (1961). J.J. Amecis, Y. J. Chabal, and S. B. Christman, Phys. Rev. B, 33, 7906 (1986). G. A. Somorjai, Introduction to Surface Chemistry and Catalysis, Wiley-Inter- science, New York, 1994, pp. SSff, 412 ff. ‘A. W. Adamson and M. W. Orem, in Progress in Surface and Membrane Science, Vol. 8. D. A. Cadenhead, J. F. Danielli, and M. D. Rosenberg, eds., Academic, New York, 1974. A. W, Adamson, L. M. Dormant, and M. W. Orem, J. Colloid Interface Sci., 28, 206 (1966); M. W. Orem and A. W. Adamson, ibid., 31, 278 (1969). K. J. Hook, J. A. Gardella, Jr. and L. Salvati, Jr., Macromolecules, 20, 2112 (1987). J. A. Creighton, C. G. Blatchford, and M. G. Albrecht, J. Chem. Soc., Faraday Trans. Il. 75, 790 (1979), €.S. Allen and R. P. van Duyne, Chem. Phys. Lett., 63, 455 (1979). A. Otto, Surf. Sci., 75, 1392 (1978). I. Pockrand, in Springer Tracts in Modern Physics, G. Hihler. ed., Vol. 104, Springer-Verlag, 1984. M. Moskovits, Rev. Mod. Phys., 57, 783 (1985). ©. Siiman, R. Smith, C. Blatchford, and M. Kerker, Langmuir, 1, 90 (1985). M. Kerker, Acc. Chem. Res., 17, 271 (1984). ‘A. Lepp and O. Siiman, J. Phys. Chem., 89, 3493 (1985). N. Neto, M. Muniz-Miranda, and G. Sbrana, J. Phys. Chem., 100, 9911 (1996). D. A. Weitz and S. Garoff, J. Chem. Phys., 78, 5324 (1983). 133a. T. Xiao, Qi Ye, and Li Sun, J. Phys. Chem. B, 101, 632 (1997). 134, 135. M. R. Mahoney, M. W. Howard, and R. P. Cooney, Chem. Phys. Lett., 71, 59 (1980); M. W. Howard, R. P. Cooney, and A. J. McQuillan, J. Raman Spectrosc., 9, 273 (1980). T. C. Wong and T. T. Ang, J. Phys. Chem., 89, 4047 (1985).CHAPTER XVII Adsorption of Gases and Vapors on Solids 1. Introduction The subject of gas adsorption is, indeed, a very broad one, and no attempt is made to give complete coverage to the voluminous literature on it. Instead, as in past chapters, the principal models or theories are taken up partly for their own sake and partly as a means of introducing characteristic data. {As stated in the introduction to the previous chapter, adsorption is described phenomenologically in terms of an empirical adsorption function n = f(P,T) where n is the amount adsorbed, As a matter of experimental convenience, one usually determines the adsorption isotherm n = fr(P); in a detailed study, this is done for several temperatures. Figure XVII-I displays some of the exten- sive data of Drain and Morrison [1]. Itis fairly common in physical adsorption systems for the low-pressure data to suggest that a limiting adsorption is being reached, as in Fig. XViI-la, but for continued further adsorption to occur at pressures approaching the saturation or condensation pressure P (which would be close to 1 atm for Nz at 75 K), as in Fig. XVII-1b. Alternatively, data may be plotted as n versus T at constant pressure or as P versus T at constant n, One thus has adsorption isobars and isosteres (note Problem XVII-2). As also noted in the preceding chapter, it is customary to divide adsorp- tion into two broad classes, namely, physical adsorption and chemisorption. Physical adsorption equilibrium is very rapid in attainment (except when lim- ited by mass transport rates in the gas phase or within a porous adsorbent) and is reversible, the adsorbate being removable without change by lowering the pressure (there may be hysteresis in the case of a porous solid). It is supposed that this type of adsorption occurs as a result of the same type of relatively nonspecific intermolecular forces that are responsible for the condensation of a vapor to a liquid, and in physical adsorption the heat of adsorption should be in the range of heats of condensation. Physical adsorption is usually important only for gases below their critical temperature, that is, for vapors. Chemisorption may be rapid or slow and may occur above or below the critical temperature of the adsorbate. It is distinguishable, qualitatively, from physical adsorption in that chemical specificity is higher and that the energy of adsorption is large enough to suggest that full chemical bonding has occurred. Gas that is chemisorbed may be difficult to remove, and desorption may be 599600 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS Pm He () Fig. XVII-1. Adsorption of Nz on rutile; temperatures indicated are in degrees Kelvin. (a) Low-pressure region; (b) high-pressure region. (From Ref. |.). accompanied by chemical changes. For example, oxygen adsorbed on carbon is held very strongly; on heating it comes off as a mixture of CO and COp [2]. Because of its nature, chemisorption is expected to be limited to a mono- layer; as suggested in connection with Fig. XVII-1, physical adsorption is not so limited and, in fact, may occur on top of a chemisorbed layer (note SectionXVII-2_ THE ADSORPTION TIME 601 Fig. XVI-2. Transition between phys- ical and chemical adsorption. XI-IC) as well as alongside it. The infrared spectrum of CO, adsorbed on -y-alu- mina suggests the presence of both physically and chemically adsorbed mole- cules (3). Chemisorption may be slow and the rate behavior indicative of the presence of an activation energy; it may, in fact, be possible for a gas to be physically adsorbed at first, and then, more slowly, to enter into some chemical reaction with the surface of the solid. At low temperatures, chemisorption may be so slow that for practical purposes only physical adsorption is observed, whereas, at high temperatures, physical adsorption is small (because of the low adsorp- tion energy), and only chemisorption occurs. An example is that of hydrogen on nickel, for which a schematic isobar is shown in Fig. XVII-2. Curve 1 shows the normal decrease in physical adsorption with temperature, and curve 2, that for chemisorption. In the transition region, curve 3, the rate of chemisorption is slow, but not negligible, so the location of the points depends on the equilibra- tion time allowed; curve 3 is therefore not an equilibrium one and is not retraced on cooling, but rather some curve between 3 and 4 is followed, depending on the rate. As is made evident in the next section, there is no sharp dividing line between these two types of adsorption, although the extremes are easily distinguishable. It is true that most of the experimental work has tended to cluster at these extremes, but this is more a reflection of practical interests and of human nature than of anything else. At any rate, although this chapter is ostensibly devoted to physical adsorption, much of the material can be applied to chemisorption as well. For the moment, we do assume that the adsorption process is reversible in the sense that equilibrium is reached and that on desorption the adsorbate is recovered unchanged. 2. The Adsorption Time A useful approach to the phenomenon of adsorption is from the point of view of the adsorption time, as discussed by de Boer [4]. Consider a molecule in the602 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS gas phase that is approaching the surface of the solid. If there were no attractive forces at all between the molecule and the solid, then the time of stay of the molecule in the vicinity of the surface would be of the order of a molecular vibration time, or about 10-'? sec, and its accommodation coefficient would be zero. This means the molecule retains its original energy. A “hot” molecule striking a cold surface should then rebound with its original energy, and its reflection from the surface would be specular. The definition of accommodation coefficient, , as given by Knudsen [5] (sce Ref. 6), is _B-t ~ TN (XVII) where T1, T2, and T; are the temperatures of the gas molecules before they strike the surface, of the surface, and of the molecules that leave the surface, respectively. Yasumoto [6] cites experimental « values of 0.48 and 0.18 for He on carbon and Pt, respectively, as compared to 0.90 and 0.71 for Ar, again respectively. In the case of polyatomic molecules, one may consider separately the accommoda- tion coefficients for translational and for vibrational energy. Values of the latter, ay, are discussed by Nilsson and Rabinovitch [7]. If attractive forces are present, then according to an equation by Frenkel (see Ref, 2), the average time of stay 7 of the molecule on the surface will be r= 198k (XVII-2) where 79 is 10-' to 10! sec and Q is the interaction energy, that is, the energy of adsorption. If r is as large as several vibration periods, it becomes reasonable to consider that adsorption has occurred; temperature equilibration between the molecule and the surface is approached and, on desorption the molecule leaves the surface in a direction that is independent of that of its arrival; the accom- modation coefficient is now said to be unity. In addition to Q and z, a quantity of imerest is the surface concentration I’, where T=Zr (XVII-3) Here, if Z is expressed in moles of collisions per square centimeter per second, T is in moles per square centimeter. We assume the condensation coefficient to be unity, that is, that all molecules that hit the surface stick to it. At very low @ values, I as given by Eq, XVII-3 is of the order expected just on the basis that the gas phase continues uniformly up to the surface so that the net surface concentration (e.g., I in Eq. XI-24) is essentially zero. This is the situationXVU-3 THE LANGMUIR ADSORPTION ISOTHERM 603 TABLE XVIL-1 ‘The Adsorption Spectrum Q (kcal/mol) 25°C (sec) Comments 0.1 0 Adsorption nil; specular reflection; accommodation coefficient zero 15 10-2 Region of physical adsorption; 35 4x 107!" accommodation coefficient 9.0 4x 10-7 unity 20.0 100 Region of chemisorption 40.0 10!7 prevailing in the first two rows of Table XVII-1. The table, which summarizes the spectrum of adsorption behavior, shows that with intermediate Q values of the order of a few kilocalories, I’ rises to a level comparable to that for a complete monolayer. This intermediate region corresponds to one of physical adsorption. The third region is one for which the Q values are of the order of chemi- cal bond energies; the 7 values become quite large, indicating that desorption may be slow, and I’ as computed by Eq. XVI-3 becomes preposterously large. Such values are evidently meaningless, and the difficulty lies in the assump- tion embodied in Eq. XVU-3 that the collision frequency gives the number of molecules hitting and sticking to the surface. AS monolayer coverage is approached, it is to be expected that more and more impinging molecules will hit occupied areas and rebound without experiencing the full Q value. One way of correcting for this effect is taken up in the next section, which deals with the Langmuir adsorption equation. 3. The Langmuir Adsorption Isotherm The following several sections deal with various theories or models for adsorption. It turns out that not only is the adsorption isotherm the most con- venient form in which to obtain and plot experimental data, but it is also the form in which theoretical treatments are most easily developed. One of the first demands of a theory for adsorption then, is that it give an experimentally cor- rect adsorption isotherm. Later, it is shown that this test is insufficient and that ‘4 mote sensitive test of the various models requires a consideration of how the energy and entropy of adsorption vary with the amount adsorbed, Nowadays, a further expectation is that the model not violate the molecular picture revealed by surface diffraction, microscopy, and spectroscopy data, see Chapter VIII and Section XVIII-2; Steele [8] discusses this picture with particular reference to physical adsorption.604 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS A. Kinetic Derivation The derivation that follows is essentially that given by Langmuir [9] in 1918, in which one writes separately the rates of evaporation and of condensation. The surface is assumed to consist of a certain number of sites S of which S, are occupied and So = S$ — S; are free. The rate of evaporation is taken to be proportional to S,, or equal to ki, and the rate of condensation proportional to the bare surface So and to the gas pressure, or equal to kzPSo. At equilibrium, Si = kyPSo = k2P(S — Si) (XVII-4) Since $,/S equals 8, the fraction of surface covered, Eq. XVII-4 can be written in the form bP oO TP Se) where & (XVI-6) ky Alternatively, 8 can be replaced by 7/tm, where rin, denotes the moles per gram adsorbed at the monolayer point. Thus __timbP - -14+bP (XVII-7) It is of interest to examine the algebraic behavior of Eq. XVII-7. At low pressure, the amount adsorbed becomes proportional to the pressure N= MmbP (XVII-8) whereas at high pressure, n approaches the limiting value 1m. Some typical shapes are illustrated in Fig. XVII-3. For convenience in testing data, Eq. XVII-7 may be put in the linear form Pi? (XvII-9) 2 Bim Mm A plot of P/n versus P should give a straight line, and the two constants ry, and b may be evaluated from the slope and intercept. In turn, ny, may be related to the area of the solid:XVII-3_ THE LANGMUIR ADSORPTION ISOTHERM. 605 e Fig. XVII-3. Langmuir isotherms. _= © Noo? Mm (XVI-10) where E denotes the specific surface area of the solid, and o° is the area of a site. If 6° can be estimated, E can be calculated from the experimental n, value. If there are several competing adsorbates,t a derivation analogous to the foregoing ives MmibiPi © T+ Dy BP; The rate constants k, and k may be related to the concepts of the preceding section as follows. First, k, is simply the reciprocal of the adsorption time, that is, nm (xvi) k= (3) err (XVII-12) To In evaluating ka, if a site can be regarded as a two-dimensional potential box, then the rate of adsorption will be given by the rate of molecules impinging on the site area oo. From gas kinetic theory, +The term adsorbate is used here to denote the species adsorbed: an altemative term, more popular in Europe than in the United States, is adsorptive. The distinction may he made that adsorbate refers to the species actually adsorbed, while adsorptive refers to a gas-phase species that is capable of being adsorbed. (See Gregg and Sing (1983) in General References.)606 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS Noo? > Samet Matee) and the Langmuir constant b becomes Noo rpe2/®T CeMRTE (XVII-14) It will be convenient to write b as Noo b= bye? — by MEE (XVII-I5) where by is of the nature of a frequency factor. Thus for nitrogen at its normal boiling point of 77 K, bo is 9.2 x 10°*, with pressure in atmospheres, 0° taken to be 16.2 A? per site (actually, this is the estimated molecular area of nitrogen), and 79, as 10"'? sec. B. Statistical Thermodynamic Derivation ‘The preceding derivation, being based on a definite mechanical picture, is easy to follow intuitively; kinetic derivations of an equilibrium relationship suf- fer from a common disadvantage, namely, that they usually assume more than is necessary. It is quite possible to obtain the Langmuir equation (as well as other adsorption isotherm equations) from examination of the statistical ther- modynamics of the two states involved. The following derivation is modified from that of Fowler and Guggenheim [10,11]. The adsorbed molecules are considered to differ from gaseous ones in that their potential energy and local partition function (see Section XVI-4A) have been modified and that, instead of possessing normal translational motion, they are confined to localized sites without any interactions between adjacent molecules but with an adsorption energy Q. Since translational and internal energy (of rotation and vibration) are inde- pendent, the partition function for the gas can be written Q = Qans Qn (XVII-16) ‘We write for the adsorbed or surface state QO = Vie Qne7 (XVI-I7) where the significance of the site partition function Qf, is explained later andXVU-3 THE LANGMUIR ADSORPTION ISOTHERM 607 the inclusion of the term ¢2/®7 means that Q’ is referred to the gaseous state. Furthermore, Q’ is a function of temperature only and not of the degree of occupancy of the sites. The complete partition function is obtained by multi- plying Q* by the number of distinguishable ways of placing N molecules on S sites. This number is obtained as follows: there are S ways of placing the first, molecule, (S— 1) for the second, and so on; and for N molecules, the number of ways is S(S—1)---[S-(N-1)} or SI/(S—N)! Of these, N! are indistinguishable since the molecules are not labeled, and the complete partition function for N molecules becomes -—5__ @y ? a= Gop OY (XVH-18) The Helmholtz free energy of the adsorbed layer is given by -kT'In Qi, (Eq. XVI-20), and with the use of Stirling’s approximation for factorials x! = (x/e" ‘one obtains At = kT{-Sln S+Nin N+(S-N)ln(S-N)-NInQ’] — (XVII-19) The chemical potential u‘ is given by (@A‘/ON)r, so that N wi kTIn -kTIN@ (XVH-20) For the gas phase, pi = -kTinQ@ (XVII-21) and on equating the two chemical potentials (remembering that @ = N/S) one obtains =~ (XVI-22) It is now necessary to examine the partition function in more detail. The energy states for translation are assumed to be given by the quantum-mechanical picture of a particle in a box. For a one-dimensional box of length a, _ ie ~ 8am (XVII-23) €n608 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS so that the one-dimensional translational partition function is a 242 Ors = exp ( a) -f aed (- ar) us ( 2nmkT ) te - a (XVII-24) re where ordinarily the states are so close together that the sum over the quantum numbers n may be replaced by the integral (perhaps contrary to intuition, this is true even if a is of molecular magnitude). For a two-dimensional box, Quans amir a (XVI-25) and in three dimensions, QnmkT \*? kT Qn (202)" HF where a’ is now replaced by the volume and, in turn, by k7/P. Substitution of Eqs. XVII-26 and XVII-17 into Eq, XVII-22 gives (XVII-26) 2 3/2 = @ Peek) QreQin orp (XVII-27) oP (XVII-28) which is the same as the Langmuir equation, Eq. XVI-5, but with bo of Eq. XVII-I5 given by _ BPamkTY? Su ~ kT Ca The two expressions for by may be brought into formal identity as follows. On adsorption, the three degrees of translational freedom can be supposed to appear as two degrees of translational motion within the confines of a two- % (XVH-29)XVU-3 THE LANGMUIR ADSORPTION ISOTHERM 609 dimensional box of area a” = 0°, plus one degree of vibration in the adsorption bond, normal to the surface. The partition function for the first is given by Eq. XVII-25. For one degree of vibrational freedom the energy states are en = (n+ 5 )hw? (XVII-30) in the case of a harmonic oscillator. The partition function is 2 ear Quin = Yer ET * To (XVII-31) If the adsorption bond is weak so that hv®/kT « 1, expansion of Eq. XVII-31 gives Qy = kT/hv”. If we now make the following identification: Que = igns X Qin (XVII-32) box ads bond then (XVIL-33) Since v° corresponds to 1/79, we have »_ No’ (Qu \_, Qa , > Gann (qe) -® 9 own) Thus the kinetic and statistical mechanical derivations may be brought into identity by means of a specific series of assumptions, including the assumption that the internal partition functions are the same for the two states (see Ref. 12). As discussed in Section XVI-4A, this last is almost certainly not the case because as a minimum effect some loss of rotational degrees of freedom should occur on adsorption. C. Adsorption Entropies i. Configurational Entropies. The factorial expression in Eq. XVII-18 may be called the configurational partition function; it is that part of the partition function that has to do with the ways of arranging a given state. Thus s! Qenie = GoM (XVII-35)610 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS Since Qos, has no temperature dependence, we have from Eq. XVII-20 that Sin = Ft Orgy: On applying Stings approximation and avidng trough by N, to obtain Sijq5, on a per molecule basis, the result is Shantig In(1 - 6) + In 4| (XVII-36) This is an integral entropy; the differential entropy is obtained by the operation 5 = a(WS)/AN = S+N(@S/AN), which yields Sootg = —* In (XVII-37) (The same result can be obtained from Eq. XVII-20 since 5 = -dn/2T, consid- ering only the configurational term in that equation.) Thus the thermodynamic description of the Langmuir model is that the energy of adsorption Q is constant and that the entropy of adsorption varies with 6 according to Eq. XVI-37. To further emphasize the special nature of the configurational entropy assumption embodied in the Langmuir model, let us repeat the derivation assuming instead that the surface molecules are mobile. In terms of the kinetic derivation, this amounts to setting the rate of condensation equal to k:PS rather than to k2PSo, with the result that @ = bP. The effect on the statistical thermo- dynamic derivation is first that the factorial grouping in Eq. XVH-18 becomes just 1/N! multiplied by S (sinc there are now S ways of placing each of the N molecules), with the result that Eq. XVII-22 becomes g 6=— XVII-38) e ; 2 and second that in obtaining Qi,,., a? in Eq. XVII-25 is replaced by the total area.A = So° rather than just by the site area 0°. On making these substitutions, the result is O=bP b= bye@/®T (XVII-39) The configurational entropies are now Srontig = kind +k (XVII-40) rate ‘ois (XVIL-41) All of these entropies may be put on a per mole basis by replacing k by R.XVU-3| THE LANGMUIR ADSORPTION ISOTHERM oul The case of a vapor adsorbing on its own liquid surface should certainly correspond to mobile adsorption. Here, 6 is unity and P = P®, the vapor pres- sure. The energy of adsorption is now that of condensation Q,, and it will be convenient to define the Langmuir constant for this case as b°; thus, from Eq. XVI-39, 1 = b°P® = boP exp o. (XVI-42) R If, furthermore, we write c = b/b® and x = P/P®, the Langmuir equation can be put in the form ox T+ex (XVII-43) ii, Entropies of Adsorption. ‘The Langmuir model is not usually considered to imply any particular value for the total entropy change on adsorption, but the statistical thermodynamic approach makes it easy to postulate various possible values. In considering, for example, the differential entropy change that should occur when | mole of gas at | atm pressure adsorbs at surface coverage @, the matter becomes one of assembling the various possible contributions. First, the total translational partition function for 1 mol of gas is fee (zeit? : = aap > (XVIL-44) On applying Stirling's approximation, replacing V/N by kT/P, and using Eq. XVII-20 and setting P = | atm, the final result, known as the Sackur-Tetrode ‘equation, is SOE, = Rin(T5?M/) — 2.30 (XVIT-45) For nitrogen at 77 K, 5/8, is 29.2 BU(cal K! mol), For localized adsorption, we need the contribution from the adsorption bond. Per degree of vibrational freedom, we obtain on applying Eq. XVII-20 to Eq. XVIL31, In(l =e | (XVIL-46) which approximates to R[1-In(hv°/kT)]. If T is 77 K and v” is taken to be 10!612 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS sec“! (see Ref. 13), then Si, is 2.9 EU per degree of freedom. Second, if the adsorption site is regarded as a two-dimensional potential box, as implied by the kinetic derivation given above, the corresponding translational entropy must be evaluated. The total partition function for N molecules is just [(2xmkT/h?)o°]", and the usual operation yields Strans = RIn(MTo°) +63.8 (XVII-47) oP box For nitrogen at 77 K, with 0° = 16.2 A?, this entropy becomes 11.4 BU. If the adsorbed gas is mobile, the total partition function is (S¥/N!) [(2xmkT/h)A]", which gives (see Ref. 11) Sings = -RIn 0 + Rin To°) + 65.8 (XVII-48) remembering that A/N = 0°/8. Alternatively, Sans = Rin 6 + Rin(MTo°) + 63.8 (XVIL-49) fam Notice that the nonconfigurational part of Eq. XVII-49 is just the entropy given by Eq. XVII-47. We can now proceed with various estimates of the entropy of adsorption AS®,,. Two extreme positions are sometimes taken (see Ref. 14). First, one assumes that for localized adsorption the only contribution is the configurational entropy. Thus ee 6 5a = ain( 1-8 =0-29.2 = -29.2 cal K' mol”! (XVII-50) where the numbers are for nitrogen at 77 K and 6 = 0.5. This contrasts with the value for a mobile film: AS%, = RIN 0+ Sian, — Seibs Mise ui = 144 114-292 = 164 cal K"! mol! (XVI-S1)XVI-3 THE LANGMUIR ADSORPTION ISOTHERM 613 This difference looks large enough to be diagnostic of the state of the adsorbed film. However, to be consistent with the kinetic derivation of the Lang- muir equation, it was necessary to suppose that the site acted as a potential box and, furthermore, that a weak adsorption bond of v° corresponding to 1/ro was present. With these provisions we obtain = oe AS®,, =-Rh (x25) +5 +S - BO) ee Bs = 04114429 -29.2 14.9 cal K~! mot! (XVII-52) Thus the entropy of “localized” adsorption can range widely, depending on whether the site is viewed as equivalent to a strong adsorption bond of neg- ligible entropy or as a potential box plus a weak bond (sce Ref. 12). In addi- tion, estimates of AS®,, should include possible surface vibrational contribu- tions in the case of mobile adsorption, and all calculations are faced with pos- sible contributions from a loss in rotational entropy on adsorption as well as from change in the adsorbent structure following adsorption (see Section XVI- 4B). These uncertainties make it virtually impossible to affirm what the state of an adsorbed film is from entropy measurements alone; for this, additional independent information about surface mobility and vibrational surface states is needed. (However, see Ref. 15 for a somewhat more optimistic conclusion.) D. Lateral Interaction It is assumed in the Langmuir model that while the adsorbed molecules ‘occupy sites of energy Q they do not interact with each other. An approach due to Fowler and Guggenheim [10] allows provision for such interaction. The probability of a given site being occupied is N/S, and if each site has z neigh- bors, the probability of a neighbor site being occupied is zV/S, so the fraction of adsorbed molecules involved is 26/2, the factor one-half correcting for double counting. If the lateral interaction energy is w, the added energy of adsorption is zw0/2, and the added differential energy of adsorption is just 2w8. The modified Langmuir equation becomes BY = boexp ( oun? ) = bexp a (XVIE-53) It is convenient to illustrate the lateral interaction effect by plotting @ versus614 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS ‘O01 Or T 0 op Fig. XVII-4. Langmuir plus lateral interaction isotherms. the Langmuir variable bP for various values of 6 = zw/RT, as shown in Fig. XVII-4. For 8 < 4, the isotherms are merely steepened versions of the Langmuir equation, but if 6 > 4, a maximum and a minimum in bP appear so that a two- phase equilibrium is implied, with phases of @ values lying at the ends of the dotted line. The quantity zw will depend very much on whether adsorption sites are close enough for neighboring adsorbate molecules to develop their normal van der Waals attraction; if, for example, zw is taken to be about one-fourth of the energy of vaporization [16], 8 would be 2.5 for a liquid obeying Trouton’s rule and at its normal boiling point, The critical pressure P,, that is, the pressure corresponding to @ = 0.5 with 6 = 4, will depend on both Q and T. A way of expressing this follows, with the use of the definitions of Eqs. XVII-42 and XVIE-43 [17]: (XVII-54) This is useful since ¢ can be estimated by means of the BET equation (see Section XVII-5). A number of more or less elaborate variants of the preced- ing treatment of lateral interaction have been proposed. Thus, Kiselev and co-workers, in their very extensive studies of physical adsorption, have pro- posed an equation of the form Kix= ~ KiK,6, (XVII-55) 6 where K,, reflects the degree of adsorbate-adsorbate interaction and c = K) + KK, so that Eq. XVII-55 reduces to Eq. XVII-43 if K,, is zero [18]. Misra [19] has summarized other (and semiempirical) Langmuir-like adsorption isotherms.XVII-4 EXPERIMENTAL PROCEDURES 615 ‘A fundamental approach by Steele [8] treats monolayer adsorption in terms of interatomic potential functions, and includes pair and higher order interactions. Young and Crowell [11] and Honig [20] give additional details on the general subject; a recent treatment is by Rybolt [21]. E. Experimental Applications of the Langmuir Equation A variety of experimental data has been found to fit the Langmuir equation reasonably well. Data are generally plotted according to the linear form, Eq. XVIL9, to obtain the constants b and nq from the best fitting straight line. The specific surface area, E, can then be obtained from Eq. XVII-10. A widely used practice is to take @° to be the molecular area of the adsorbate, estimated from liquid or solid adsorbate densities. On the other hand, the Langmuir model is cast around the concept of adsorption sites, whose spacing one would suppose to be characteristic of the adsorbent. See Section XVII-SB for an additional discussion of the problem. A true fit to the Langmuir equation implies that nq, and Q are independent of temperature, and systems obeying the form of Eq. XVII-9 often fail this more severe test. In some cases, multilayer formation, discussed in Section XVII-5, is a source of trouble, but in general the Langmuir model is too simple for really detailed agreement to be expected with experimental systems. However, more sophisticated models also introduce more semiempirical parameters, so the net gain, apart from data fitting, has not been great. The simple Lang- muir mode! has therefore retained great general utility as well as providing the point of departure from many of the proposed refinements (as, for example, Eq. XVILS3). 4. Experimental Procedures The remainder of the chapter is concemed with increasingly specialized developments in the study of gas adsorption, and before proceeding to this mate rial, it seems desirable to consider briefly some of the experimental techniques that are important in obtaining gas adsorption data. See Ref. 22 for a review of traditional methods, and Ref. 23 for IUPAC (International Union of Pure and Applied Chemistry) recommendations for symbols and definitions. ‘Adsorption isotherms conventionally have been determined by means of a vacuum line system whereby pressure-volume measurements are made before and after admitting the adsorbate gas to the sample, For some recent experi- ‘mental papers, see Refs. 24 and 25. If the total surface area is small (say, a few hundred square centimeters), the amount adsorbed becomes so little that measurements are difficult by normal procedures. Thus the change in pressure-volume product on admitting gas to the adsorbent becomes so small that precision is impaired. One way of avoiding this problem is to set T so that the vapor pressure of the liquid adsorbate is very low (e.g., krypton at -195°C (26, 27]. Monolayer616 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS Signal a in liquid Np Fig. XVIL-S. Schematic detector response in a determination of nitrogen adsorption and desorption. A flow of He and Nz is passed through the sample until the detector reading. is constant; the sample is then cooled in a liquid nitrogen bath. For desorption, the bath is removed. (From Ref. 28. Reprinted with permission from John Wiley & Sons, copyright 1995.) formation usually occurs when a few tenths of saturation pressure is reached, regardless of its absolute value; and if P? is very small, even small amounts of adsorption will cause relatively large changes in the pressure-volume product of the gas in the vacuum line, Alternatively, gas chromatography may be used; Fig. XVII-5 shows a schematic readout of the thermal conductivity detector, the areas under the peaks giving the amount adsorbed or desorbed. A second general type of procedure, due to McBain [29], is to determine n by a direct weighing of the amount of adsorption. McBain used a delicte quartz spiral spring, but modem equipment generally makes use of a microbalance or a transducer. An illustra- tive schematic is shown in Fig. XVII-6. Ultrahigh vacuum techniques have become common, especially in connec tion with surface spectroscopic and diffraction studies, but also in adsorption on very clean surfaces. The techniques have become rather specialized and the reader is referred to Ref. 8 and citations therein. The heat of adsorption is an important experimental quantity. The heat evo- lution with each of successive admissions of adsorbate vapor may be measured directly by means of a calorimeter described by Becbe and co-workers [31]. Alternatively, the heat of immersion in liquid adsorbate of adsorbent having various amounts preadsorbed on it may be determined. The difference between any (wo values is related (o the integral heat of adsorption (see Section X-3A) between the two degrees of coverage. See Refs. 32 and 33 for experimental papers in this area.XVII-S THE BET AND RELATED ISOTHERMS 617 Cahn 1000 eloktrobalance () adsorptive supply | ij mnarmostat | Fig. XVII-6, Schematic of gravimetric apparatus for adsorption measurements. (From Ref. 30. Reprinted with permission ftom American Chemical Society, copyright 1995.) 5. The BET and Related Isotherms Adsorption isotherms are by no means all of the Langmuir type as to shape, and Brunauer [34] considered that there are five principal forms, as illustrated in Fig. XVI-7. Type I is the Langmuir type, roughly characterized by a mono- tonic approach to a limiting adsorption that presumably corresponds to a com- plete monolayer. Type Il is very common in the case of physical adsorption 1 { | | ! | I! i i | i Pr Type V ? Pe F - Fig, XVII-7. Brunauer's five types of adsorption isotherms. (From Ref. 34.)618 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS | og ° Pir io Pipe r Tyee vt Type Vit ‘a o Fig. XVII-8. Two additional types of adsorption isotherms expected for nonwetting adsorbate-adsorbent systems. (From Ref. 37.) and undoubtedly corresponds to multilayer formation. For many years it was the practice to take point B, at the knee of the curve, as the point of completion of a monolayer, and surface areas obtained by this method are faily consistent with those found using adsorbates that give type I isotherms. Type III is relatively rare—an example is that of the adsorption of nitrogen on ice [35]—and seems to be characterized by a heat of adsorption equal to or less than the heat of liquefaction of the adsorbate. Types IV and V are considered to reflect capillary condensation phenomena in that they level off before the saturation pressure is reached and may show hysteresis effects. This description is traditional, and some further comment is in order. The flat rezion of the type I isotherm has never been observed up to pressures approach- ing P ; this type typically is observed in chemisorption, at pressures far below ‘Types Il and III approach the P? line asymptotically; experimentally, such. behavior is observed for adsorption on powdered samples, and the approach toward infinite film thickness is actually due to interparticle condensation [36] (see Section X-6B), although such behavior is expected even for adsorption on a flat surface if bulk liquid adsorbate wets the adsorbent. ‘Types IV and V specifically refer to porous solids. There is a need to recognize at least the two additional isotherm types shown in Fig. XVII-8. These are two simple types possible for adsorption on a flat surface for the case where bulk liquid adsor- bate rests on the adsorbent with a finite contact angle [37, 38). A. Derivation of the BET Equation Because of their prevalence in physical adsorption studies on high-energy, powdered solids, type II isotherms are of considerable practical importance. Brunauer, Emmett, and Teller (BET) [39] showed how to extent Langmuir's approach to multilayer adsorption, and their equation has come to be known as the BET equation. The derivation that follows is the traditional one, based on a detailed balancing of forward and reverse rates.XVI-5 THE BET AND RELATED ISOTHERMS 619 fs |os | s 1 s % Fig. XVIL-9. The BET model. The basic assumption is that the Langmuir equation applies to each layer, with the added postulate that for the first layer the heat of adsorption Q may have some special value, whereas for all succeeding layers, it is equal to Q,, the heat of condensation of the liquid adsorbate. A further assumption is that ‘evaporation and condensation can occur only from or on exposed surfaces. AS illustrated in Fig. XVU-9, the picture is one of portions of uncovered surface So, of surface covered by a single layer Sy, by a double-layer S2, and so on.t The condition for equilibrium is taken to be that the amount of each type of surface reaches a steady-state value with respect to the next-deeper one. Thus for So G1PSo = b Sie" OV/RT (XVII-56) and for all succeeding surfaces, GPS,—1 = biSje 2/*T (XVI-57) It then follows that Si=yS — S2= x8) and S) =x!" 'S) = yal 'Sp = ex!So where oe peat a PeQe/RT (XVII-58) and +On the right of Fig. XVII-9 is shown an isolated stack of two molecules. While physically unrealistic, such stacks are implicit in the BET model and, in fact, have been reproduced in a ‘Monte Carlo simulation [40].620 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS A101 0) -Qu/RT = g(01-O0v/RT ECrise bia, Then 20S, ey it (XVIL-60) = 0 Te Soe De a) Insertion of the algebraic equivalents to the sums yields n ex/(1 — xP oP eHf= a XVI hn 1+cx/(1-x) Coe) which rearranges to von cx P Umm =e ba ore where v denotes cm? STP (standard temperature and pressure) of gas adsorbed. ‘The essential next step is to take the ratio of frequency terms a;/b; to be the same as for liquid adsorbate-vapor equilibrium so that combination of Eqs. XVII-58 and XVII-42 (where bp = a;/b;) identifies x as equal to P/P°. Usu- ally, also, it is assumed that a1/b, = a;/b; in that the approximate form of Eq. XVII-59 is used in interpreting the constant c. Although the preceding derivation is the easier to follow, the BET equation also may be derived from statistical mechanics by a procedure similar to that described in the case of the Langmuir equation [41,42]. B. Properties of the BET Equation The BET equation filled an annoying gap in the interpretation of adsorption isotherms, and at the time of its appearance in 1938 it was also hailed as a general method for obtaining surface areas from adsorption data. The equation can be put in the form (¢~ Ix Pe er (XVII-63) so that ny and c can be obtained from the slope and intercept of the straight Tine best fitting the plot of x/n(I — x) versus x. The specific surface area can then be obtained through Eq. XVII-10 if o° is known. In the case of multilayer adsorption it seems reasonable to take o° as an adsorbate (as opposed to anXVI-5 THE BET AND RELATED ISOTHERMS 621 adsorbent site) area, based on either the solid or the liquid density, depending on the temperature, Values that give reasonably self-consistent areas are (in square angstroms per molecule): Nz, 16.2; Op, 14.1; Ar, 13.8; Kr, 19.5; n-CgHio, 18.1. These and values for other adsorbates are reviewed critically by McClellan and Harnsberger [43]. The preceding values are close to the calculated ones from the liquid densities at the boiling points and are, in this respect, reasonable for a multilayer adsorption situation. There may be special cases. Pierce and Ewing [44] suggest that for graphite surfaces it is the lattice spacing of the adsorbent that controls and that the effective molecular area for N2 becomes 20 A? rather than the usual 16.2 A?, Amati and Kovéts [45] found the effective a9 for Np to vary from 16.2 to 21 A? as y, was decreased by surface modification of a silica. Rouquerol et al. [46] conclude that, depending on the adsorbent, the effective 0° for Nz may vary from 11.2 to 16.2 A? since the No may in some case be adsorbed end-on. From the experimental point of view, the BET equation is easy to apply, and the surface areas so obtained are reasonably consistent (see Section XVII- 8 for further discussion). The equation in fact has become the standard one for practical surface area determinations, usually with nitrogen at 77 K as the adsorbate, but in general with any system giving type II isotherms. On the other hand, the region of fit usually is not very great—the linear region of a plot according to Eq. XVII-63 typically lies between a P/P® of 0.05 and 0.3, as illustrated in Fig. XVU-14. The typical deviation is such that the best-fitting BET equation predicts too little adsorption at low pressures and too much at high pressures. ‘The BET equation also seems to cover three of the five isotherm types described in Fig. XVII-7. Thus for c large, that is, Q1 >> Q,, it reduces to the Langmuir equation, Eq. XVII-43, and for small c values, type III isotherms result, as illustrated in Fig. XVII-10. However, the adsorption of relatively inert gases such as nitrogen and argon on polar surfaces generally gives c values around 100, corresponding to type II isotherms. For such systems the approx- imate form of Eq. XVI-62, (XVII-64) works quite well in the usual region of fit of the BET equation, and a “one point” method of surface area estimation thus follows (see Ref. 47). The method has been incorporated in commercial rapid surface area determination equipment. C. Modifications of the BET Equation The very considerable success of the BET equation stimulated various investigators to consider modifications of it that would correct certain approximations and give a beter ft to type I isotherms. Thus if itis assumed that multilayer formation is limited to n layers, perhaps because of the opposing walls of a capillary being involved, one622 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS Fig. XVII-10. BET isotherms. obtains (1) [umex/(1 ~ a] = (n+ Dx" enxtt E 1+ exit (XVII-65) By choosing the appropriate value, the amount of adsorption predicted at large P/P? values is reduced, and a better fit to data can usually be obtained. Kiselev and co- workers have proposed an equation paralleling Eq. XVII-5S, but that reduces to the BET equation if K,, the lateral interaction parameter, is zero [18]. There are other modifications that also give better fits to data but also introduce one or more additional parameters. Young and Crowell [47a] and Gregg and Sing [48] may be referred to for a more detailed summary of such modifications; see Sircar [49] for a recent one. One interesting example is the following. If the adsorbed film is not quite liquidlike so that ai/b; 4 Bp, its effective vapor pressure will be PO” rather than P° (note Section X-6C), and the effect is to multiply x by a factor k (50). Redhead (51) discusses a com- posite model that approaches the BET equation at one limit and the Frenkel-HalseyHill ‘one at another (see Section XVII-7B). 6. Isotherms Based on the Equation of State of the Adsorbed Film It was shown in Section X-3B that a formal calculation can be made of the film pressure, or reduction in the surface free energy of the solid, by evaluatingXVII-6 ISOTHERMS BASED ON EQUATIONS OF STATE 623 the integral of nd In P. Conversely, if it is assumed that the adsorbed material can be treated as a two-dimensional film of one of the types found with mono- layers on liquid substrates, one has a large new source of isotherm equations. A. Film Pressure-Area Diagrams from Adsorption Isotherms ‘According to Eq. X-12, straightforward application of the Gibbs equation gives PL ran = BP eee (XVIL-66) If the isotherm is linear, Eq. XVII-66 reduces to RTny r =IRT or r0=RT (xvIL-67) Gregg [52] (see also Ref. 48) has surveyed the types of force-area equations obtainable from adsorption isotherms. B. Adsorption Isotherms from Two-Dimensional Equations of State Numerous types of equations of state apply to solids, liquids, and gases; a few of these are considered here. Ideal-Gas Law. Here Eq. XVII-67 applies, and on reversing the procedure that led to it, one finds kP=6 (XVII-68) The ideal gas law equation of state thus leads to a linear or Henry's law isotherm. A natural modification adds a co-area term: xo ~0°)=RT (XVII-69) from which one obtains (remembering that § = 09/0) kP= wa A-o (XVII-70) ‘Equation XVII-70 bears a strong resemblance to the Langmuir equation (see Ref. 4}—to the point that it is doubtful whether the two could always be distinguished experimen- tally. An equivalent form obtained by Volmer [53] worked well for data on the adsorp- tion of various organic vapors on mercury [54] (see Problem XVII-40). Van der Waals Equations of State. A logical step to take next is to consider equations of state that contain both a covolume term and an attractive force term, such as the van der Waals equation. De Boer [4] and Ross and Olivier [55] have given this type of equation much emphasis.624 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS TABLE XVII-2 ‘Two-Dimensional Equations of State and Corresponding Isotherms Equation of State Corresponding Isotherm Tdcal-gas type x0 =RT InkP=In0 x(o - 0°) =RT In kP = 8/(1 — 6) + In[0/(1 - 8)] Van der Waals type (x+a/0?)(o - 0°) = RT (+ a/o3)(0 - 0°) = RT (x+a/0?\(o - 0°/o)= RT Viral type 0 = RT+ on - Bx? Condensed film x=b-ao In KP = 8/(1 ~ 8) + Inf0/(1 - 8)] - 8 In kP = 6/(1 — 8) + In{0(1 - 8)] - c6? In P= 1/(1— 8) + } Inf/(1 ~ 69) - 0B (¢ = 2a/oRT) In kP= (6/20) + (4a) + D [6-1 +20]! = In{@~ 1) +1 1? + 20}'7) @= 1/8, = 2BRT/?) (see Ref. 56) In(P/P9) = = A/n?) It must be remembered that, in general, the constants a and b of the van der ‘Waals equation depend on volume and on temperature. Thus a number of vari- ants are possible, and some of these and the corresponding adsorption isotherms are given in Table XVII-2. All of them lead to rather complex adsorption equa- tions, but the general appearance of the family of isotherms from any one of them is as illustrated in Fig. XVUI-L1. The dotted line in the figure represents the presumed actual course of that particular isotherm and corresponds to a two-dimensional condensation from gas to liquid. Notice the general similarity to the plots of the Langmuir plus the lateral interaction equation shown in Fig. XVII-4. cor or ro Fig. XVIL-11. The van der Waals equation of state isotherm. r 0XVII-7 THE POTENTIAL THEORY 625 Equations such as in Table XVI-2 may well apply both experimentally and in con- cept to selected cases of adsorption in the submonolayer region (see Ref. 57 for exam- ple). The difficulty is that, as with the various modified BET equations, there are enough possibilities for algebraic variations and enough parameters that it is difficult to know how much significance should be attached to ability to fit data. The matter of possible phase changes is discussed in Section XVII-11 and that of the complicating effect of surface heterogeneity and of surfaces having a fractal dimension in Section XVII-15. 7. The Potential Theory A. The Polanyi Treatment A still different approach to multilayer adsorption considers that there is a potential field at the surface of a solid into which adsorbate molecules “fall.” ‘The adsorbed layer thus resembles the atmosphere of a planet—it is most com- pressed at the surface of the solid and decreases in density outward. The gen- eral idea is quite old, but was first formalized by Polanyi in about 1914—see Brunauer [34]. As illustrated in Fig. XVII-12, one can draw surfaces of equipo- tential that appear as lines in a cross-sectional view of the surface region. The space between each set of equipotential surfaces corresponds to a definite vol- ume, and there will thus be a relationship between potential U and volume ¢. If we consider the case of a gas in adsorption equilibrium with a surface, there must be no net free energy change on transporting a small amount from ‘one region to the other. Therefore, since the potential U, represents the work donc by the adsorption forces when adsorbate is brought up to a distance x from the surface, there must be a compensating compressional increase in the free energy of the adsorbate. Thus Fig. XVII-12. Isopotential contours, (From Ref. 34.)626 XVIl ADSORPTION OF GASES AND VAPORS ON SOLIDS T<% 2, however, there will be regions accessible to small moleculesXVU-15 RATE OF ADSORPTION 661 TABLE XVII-5 Fractal Dimensions for Selected Adsorbents” Adsorbent D_ iRef. Adsorbent D Ref. ‘Active caron i 30-173 (coconut shell) 27 169 30 167 Coal 24 170 Vitreous silica 20 174 Graphitized a Fe203 24-27 167 carbon black 20-21 167 Kaolinite 21,24 175 Ab-illared Montmorillonite 2.1,2.3 175 ‘montmorillonite 19 171 ‘Synthetic faujasite 20° 172 “Taken from Ref. 167; D values obtained by varying adsorbate size, Eq. XVI-1. but not to larger ones; the point is illustrated in Ref. 167,t which also tabu- lates D values for a number of adsorbents, determined by various means. Table XYVIL-S gives a few of the cases where D was determined by varying the adsor- bate size. In the case of mixtures of gases of different molecular size, an adsorbent of D > 2 will effect some segregation by size. This segregation will also affect the probability of bimolecular reactions between molecules of different sizes (168). 15. Rate of Adsorption The rate of physical adsorption may be determined by the gas kinetic sur- face collision frequency as modified by the variation of sticking probability with surface coverage—as in the kinetic derivation of the Langmuir equation (Section XVII-3A)—and should then be very large unless the gas pressure is small, Alternatively, the rate may be governed by boundary layer diffusion, a slower process in general. Such aspects are mentioned in Ref. 146. Rate effects may not be chemical kinetic ones. Benson and co-worker [84], in a study of the rate of adsorption of water on lyophilized proteins, comment that the empirical rates of adsorption were very markedly complicated by the fact that the samples were appreciably heated by the heat evolved on adsorption. In fact, it appeared that the actual adsorption rates were very fast and that the time dependence of the adsorbate pressure above the adsorbent was simply due to the time variation of the temperature of the sample as it cooled after the initial heating when adsorbate was first introduced. Deitz and Carpenter [176] found that argon and nitrogen adsorbed only ‘One of us is pleased that this paper shows an illustrative needlepoint embroidery created by his wife.662 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS slowly on diamond under conditions such that the final @ was only a few per- cent. They were able to rule out heat transfer at rate controlling and concluded that while the adsorption process per se was very rapid, the nature of the dia- mond surface changed slowly to a more active one on the cooling of the sample prior to an adsorption run, The change appeared to be reversible and may con- stitute a type of illustration of the Dunning effect discussed in Sections VII-4B and XVI-4B. A related explanation was invoked by Good and co-workers [177]. In studies of the heat of immersion of Al,O3 and SiO» in water, they noticed a slow residual heat evolution, which they attributed to slow surface hydration. In conclusion, any observation of slowness in attainment of physical adsorp- tion equilibrium should be analyzed with caution and in detail. When this has been done, the phenomenon has either been found to be due to trivial causes or else some unsuspected and interesting other effects were operative. 16. Adsorption on Porous Solids—Hysteresis ‘As a general rule, adsorbates above their critical temperatures do not give multilayer type isotherms. In such a situation, a porous absorbent behaves like any other, unless the pores are of molecular size, and at this point the distinc tion between adsorption and absorption dims. Below the critical temperature, multilayer formation is possible and capillary condensation can occur. These two aspects of the behavior of porous solids are discussed briefly in this sec- tion. Some IUPAC (International Union of Pure and Applied Chemistry) rec- ommendations for the characterization of porous solids are given in Ref. 178. A. Molecular Sieves Some classes of adsorbents have internal surface accessible by pores small enough to act as molecular sieves, so that different apparent surface areas are obtained according to the size of the adsorbate molecule. This is a screening effect and not one of surface roughness as might be described in terms of fractal geometry (Section XVI-2B). Zeolites have been of much interest in this con- nection because the open way in which the (Al, Si)Q, tetrahedra join gives rise to large cavities and large windows into the cavities. This is illustrated in Fig. XVII-26 [179]. As a specific example, chabasite (CaAlpSi,Q}2) has cages about 10 A in diameter, with six openings into each or windows of about 4-A diam- eter, Monatomic and diatomic gases, water, and n-alkanes can enter into such cavities, but larger molecules do not. Thus isobutane can be separated from n-alkanes and, on the basis of rates, even propane from ethane [180]. The replacement of the calcium by other ions (zeolites have ion exchange properties—note Section XI-6C) considerably affects the relative adsorption behavior. Various synthetic zeolites having various window diameters in the range of 4 to 10 A are available under the name of Linde Molecular Sieves (see Ref, 179). Listings of types of zeolites and their geometric properties may be found in Refs. 181 and 182 and the related properties of expanded clayXVII-16 ADSORPTION ON POROUS SOLIDS—HYSTERESIS 663 Fig. XVII-26. The arrangement of (Al, SiO, tetrahedra that gives the cubooctahedral cavity found in some felspathoids and zeolites. (From Ref. 179.) minerals in Ref. 183. A more recent addition to commercial porous materials is called M41S (MCM-41). These are silicious but formed on surfactant liquid crystal templates (184,185). The pores are parallel cylinderlike, a cross-section looking much like a honey comb. It can be synthesized with pore diameters in the mesoporous range, varying from 15 to 100 A, each type of preparation having a narrow size distribution. FTIR and other studies of water adsorption indicated a relatively hydrophobic character to the pores [185a] The adsorption isotherms are often Langmuirian in type (under conditions such that multilayer formation is not likely), and in the case of zeolites, both ry, and b vary with the cation present. At higher pressures, capillary condensation typically occurs, as discussed in the next section. Some N2 isotherms for M41S materials are shown in Fig. XVII-27; they are Langmuirian below P/P® of about 0.2. In the case of a microporous carbon (prepared by carbonizing olive pits), the isotherms for He at 4.2 K and for N; at 77 K were similar and Langmuirlike up to P/P® near unity, but were fit to a modified Dubninin-Radushkevich (DR) equation (sce Eq. XVII-75) to estimate micropore sizes around 40 A [186]. A special case of adsorption in cavities is that of clatherate compounds. Here, cages are present, but without access windows, so for “adsorption” to occur the solid usually must be crystallized in the presence of the “adsorbate.” Thus quinol crystallizes in such a manner that holes several angstroms in diame- ter occur and, if crystallization takes place in the presence of solvent or gas664 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS 0.0 00 402 04 08 08 10 p/p? Fig. XVII-27. Nitrogen adsorption at 77 K for a series of M41S materials. Average pore diameters: squares, 25 A; triangles, 40 A; circles, 45 A. Adsorption: solid symbols; desorption: open symbols. The isotherms are normalized to the volume adsorbed at P/P? = 0.9. (From Ref. 187. Reprinted with kind permission from Elsevier Science-NL, Sara Burgerhartstraat 25, 1055 KV Amsterdam, The Netherlands.) molecules of small enough size, one or more such molecules may be incor- porated into each hole. The union may thus be stoichiometric, but topological rather than specific chemical factors are involved; in the case of quinol, such diverse species as sulfur dioxide, methanol, formic acid, and nitrogen may form clathrate compounds [188]. A particularly interesting clathrating substance is ice; the cavities consist of six cages of 5.9 A diameter and two of 5.2 A per unit cell [189], anda variety of molecules are clathrated, ranging from rare gases to halogens and hydrocarbons. The ice system provides an illustration that it is not always necessary to carry ‘out an in situ crystallization to obtain a clathrate. Barrer and Ruzicka (190] report a spontaneous clathrate formation between ice powder and xenon and krypton at ~78°C, and in an attempted adsorption study of ethane on ice at -96°C it was found that, again, spontaneous ethane hydrate formation occurred [191); similar behavior was found for the adsorption of CO> by ice [192]. These cases could be regarded as extreme examples of an adsorbate-induced surface rearrangement! B. Capillary Condensation Below the critical temperature of the adsorbate, adsorption is generally mul- tilayer in type, and the presence of pores may have the effect not only of lim- iting the possible number of layers of adsorbate (see Eq. XVII-65) but also of introducing capillary condensation phenomena. A wide range of porous adsor- ents is now involved and usually having a broad distribution of pore sizes and shapes, unlike the zeolites. The most general characteristic of such adsorption systems is that of hysteresis; as illustrated in Fig. XVII-27 and, more gener-XVII-16 ADSORPTION ON POROUS SOLIDS—HYSTERESIS 665 I] oy ° 1 0 1 ‘Pipe Pipe Pipe te o © Fig. XVI-28. Hysteresis loops in adsorption. ally, in Fig. XVII-28, the desorption branch lies to the left of the adsorption ‘one. The isotherms may tend to flatten as P/P° approaches unity (type IV of Fig. XVII-7). We are concemed in this section mainly with loops of the type shown in Fig. XVII-28b and c. The open loop of Fig. XVI-28a has at least two types of explanation. One is that “ink-bottle” pores (see Fig. XVI-3) may be present, that can trap adsorbate. A perhaps more plausible explanation in the case of vapors (as contrasted to liquid mercury) is that irreversible change may occur in the pore structure on adsorption, so that the desorption situation is truly different from the adsorption one [193]. The adsorption branch of isotherms for porous solids has been variously modeled. Again, the DR equation (Eq. XVII-75) and related forms have been used [186,194]. With respect to desorption, the variety of shapes of loops of the closed variety that may be observed in practice is illustrated in Fig. XVII-29 (see also Refs. 195 and 197). Explanations of hysteresis in capillary condensation probably begin in 1911 with Zsigmondy [198], who attributed the effect to contact angle hysteresis due to impurities; this might account for behavior of the type shown in Fig. XVII-28a, but not in general for the many systems having retraceable closed hysteresis loops. Most of the early analyses and many current ones are in terms of a model representing the adsorbent as a bundle of various-sized cap- illaries. Cohan [199] suggested that the adsorption branch—curve abc in Fig. XVII-28b—represented increasingly thick film formation whose radius of cur- vature would be that of the capillary r, so that at each stage the radius of cap- illaries just filling would be given by the corresponding form of the Kelvin equation (Eq. IlI-19): xq = eC TVIRT (XVII-134) Ate all such capillaries would be filled and on desorption would empty by retreat of a meniscus of curvature 2/r, so that at each stage of the desorption branch dea the radius of the capillaries emptying would be666 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS 16 26 56 A a a Fig. XVI-29. Nitrogen isotherms; the volume adsorbed is plotted on an arbitrary scale. ‘The upper scale shows pore radii corresponding to various relative pressures. Samples: A, Oulton catalyst; B, bone char number 452; C, activated charcoal; F, Alumina catalyst F12; G, porous glass; S, silica aerogel. (From Ref. 196). xg = eR V/RT (XVII-135) The section ed can be regarded as due to relatively large cone-shaped pores that would fill and empty without hysteresis. At the end of section cd, then, all pores should be filled, and the adsorbent should hold the same volume of any adsorbate. See Ref. 200 for a discussion of this conclusion, sometimes known as the Gurvitsch rule, The effect of the bundle of capillaries picture is to stress the use of the de- sorption branch to obtain a pore-size distribution. The basic procedure stems from those of Barrett et al. [201] and Pierce [196]; see also Ref. 58 wherein the effective meniscus curvature is regarded as given by the capillary radius minus the thickness of ordinary multilayer adsorption expected at that P/P?. This last can be estimated from adsorption data on similar but nonporous material, and for this de Boer's t-curve (see Table XVI-4) is widely used. The general cal- culation is as follows. After cach stepwise decrease in Pg (pressure on desorp- tion), an effective capillary radius is calculated from Eq. XVII-135, and the true radius is obtained by adding the estimated multilayer thickness. The exposed pore volume and pore area can then be calculated from the volume desorbed in that step. For all steps after the first, the desorbed volume is first corrected for that from multilayer thinning on the sum of the areas of previously exposed pores. In this way a tabulation of cumulative pore volume of pores of radius greater than a given r is obtained and, from the slopes of the corresponding plot, a pore-size distribution results. Such a distribution was compared in Fig. XVI-2XVII-16 ADSORPTION ON POROUS SOLIDS—HYSTERESIS 667 to the one from mercury porosimetry, with excellent agreement. See Refs. 202 and 203 for additional examples. Also, Dubinin [204] has elaborated on the de Boer type of approach. It might be noted that nonporous adsorbents may show hysteresis due to interparticle capillary condensation (see Ref. 205). A gen- eral complication in calculating pore volumes is that the density of capillary condensed liquid may be significantly diminished because of being under a high negative pressure [206]. Sing (see Ref. 207 and earlier papers) developed a modification of the de Boer t-plot idea. The latter rests on the observation of a characteristic isotherm (Section XVII-9), that is, on the conclusion that the adsorption isotherm is inde- pendent of the adsorbent in the multilayer region. Sing recognized that there were differences for different adsorbents, and used an appropriate “standard isotherm” for each system, the standard isotherm being for a nonporous adsor- bent of composition’ similar to that of the porous one being studied. He then defined a quantity a, = (n/n,), where n, is the amount adsorbed by the non- porous reference material at the selected P/P®. The a, values are used to correct pore radii for multilayer adsorption in much the same manner as with de Boer. Lecloux and Pirard [208] have discussed further the use of standard isotherms. Everett [209] has pointed out that the bundle of capillaries model can be outrageously wrong for real systems so that the results of the preceding type of analysis, while internally consistent, may not give more than the roughest kind of information about the real pore structure. One problem is that of “ink bottle” pores (Fig. XVI-3), which empty at the capillary vapor pressure of the access channel but then discharge the contents of a larger cavity. Barrer et al. [210] have also discussed a variety of geometric situations in which filling and emptying paths are different. In a rather different perspective, Seri-Levy and Avnir [210a] have shown that hysteresis can occur with adsorption on a fractal surface due to “pores” that are highly irregular ink bottle in shape. Brunauer and co-workers [211, 212] proposed a “modelless” method for obtaining pore size distributions; no specific capillary shape is assumed. Use is made of the general thermodynamic relationship due to Kiselev [213] da = dp dn (XVII-136) where dA is the surface that disappears when a pore is filled by capillary con- densation, Ay: is the change in chemical potential, equal to RT In P/P®, and dn is the number of moles of liquid taken up by the pore. It can be shown that the Kelvin equation is a special case of Eq. XVII-136 (Problem XVII-33). The integral form of Eq. XVII-136 is aT j ” in (P/P)\dn (XVII-137) Y dn One now defines a hydraulic pore radius ry as668 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS v naz (XVIL-138) where V is the volume of a set of pores and A, their surface area. The remain- ing procedure is similar to that described. For each of successive steps along the desorption branch, the volume of that group of pores that empty is given by the change in 7, and their area by Eq. XVII-138. Their hydraulic radius follows from Eq. XVII-137. The total pore surface area can be estimated from appli- cation of the BET equation to the section of the isotherm before the hysteresis loop, and a check on the pore distribution analysis is that the total pore area calculated from the distribution agrees with the BET area. In one test of the method, the pore size distribution for a hardened Portland cement paste agreed well with that obtained by the traditional method [211]. More detailed information about the pore system can be obtained from “scanning curves,” illustrated in Fig. XVII-28c. Thus if adsorption is carried only up to point a and then desorption is started, the lower curve ab will be traced; if at b absorption is resumed, the upper curve ab is followed, and so on. Any complete model should account in detail for such scanning curves and, conversely, through their complete mapping ‘much more information can be obtained about the nature of the pores. Rao [214] and Emmett (215) have summarized a great deal of such behavior. A potentially powerful approach is that of Everett [216], who treats the pore sys tem as a set of domains, independently acting in a first approximation. Each domain consists of those elements of the adsorbent that fill at a particular xqqj) and empty at a particular other relative pressure x, the associated volume being V;. Each domain is thus characterized by these three variables, and a plot of the function V(x,,x4) would produce a surface in three dimensions something like a relief map. On increasing the ‘Vapor pressure from xq t0 xq +dxq, all domains of filling pressure in this interval should fill, but these domains can and in general would have a range of emptying pressure xg ranging from xq = 0 to xq = xg. Since the xq of any domain cannot exceed its xq (it cannot empty at a higher pressure than it fills!), the base of the topological map must, be a 45° triangle. The detailed map contains in principle full information about the adsorption and desorption branches, as well as about all possible scanning loops. The problem of deducing such a map from data is a massive one, however, and seems not yet to have been done. Other approaches to the problem treat a porous adsorbent as a network of various size capillaries [217, 218]. Finally, capillary condensation where the vapor consists of two or more components—a rather complex subject—is discussed in several papers in Ref. 219. Acconcluding comment might be made on the temperature dependence of adsorption in such systems. One can show by setting up a piston and cylinder experiment that mechanical work must be lost (i.e., converted to heat) on carrying a hysteresis system through a cycle. An irreversible process is thus involved, and the entropy change in a ‘small step will not in general be equal to q/T. As was pointed out by LaMer [220], this means that second-law equations such as Eq. XVII-107 no longer have a simple ‘meaning. In hysteresis systems, of course, two sets of qs, values can be obtained, from the adsorption and from the desorption branches. These usually are not equal and neitherXVII-16 ADSORPTION ON POROUS SOLIDS—HYSTERESIS 669 of them in general can be expected to equal the calorimetric heat. Another way of stating. the problem is that the system is not locally reversible. The adsorption following an increase of x by 5x is not retraced on decreasing the pressure by 5x. This means that extreme caution should be exercised in treating qg, values as though they represented physical heat quantities, although it is certainly possible that in individual cases or in terms of particular models the discrepancy between qx, and a calorimetric heat may not, be serious (see Ref. 221). C. Micropore Analysis ‘Adsorbents such as some silica gels and types of carbons and zeolites have pores of the order of molecular dimensions, that is, from several up to 10-15 A in diameter. Adsorption in such pores is not readily treated as a capillary condensation phenomenon—in fact, there is typically no hysteresis loop. What happens physically is that as multilayer adsorption develops, the pore becomes filled by a meeting of the adsorbed films from opposing walls. Pores showing this type of adsorption behavior have come to be called micropores—a conven- tional definition is that micropore diameters are of width not exceeding 20 A (larger pores are called mesopores), see Ref. 221a. Adsorption isotherms in the micropore region may start off looking like one of the high BET c-value curves of Fig. XVII-10, but will then level off much like a Langmuir isotherm (Fig. XVII-3) as the pores fill and the surface area available for further adsorption greatly diminishes. The BET-type equation for adsorption limited to n layers (Eq. XVII-65) will sometimes fit this type of behavior. Currently, however, more use is made of the Dubinin-Raduschkevich or DR equation. This is Eq. XVII-75, but now put in the form eoela(s) ee (B]] orem (see Refs. 222-224). Here, 8 is a “similarity” coefficient characteristic of the adsorbate and B is a constant characteristic of the adsorbent; V and Vo are the amounts adsorbed expressed as a liquid volume, at a given P/P®, and when micropore filling is complete. For a single set of slitlike pores, B appears to be proportional to the pore width [224, 225]. Some recent work [226] on micro- porous carbon having slitlike pores about three adsorbate molecules thick sug- gests that pore filling may be autocatalytic. As opposing walls acquire one adsorbed layer each, filling of the middle region then becomes easy. This was the explanation of S-shaped isotherms, that is, sections where addition of more Np adsorbate led to a decrease of the final pressure (note Problem XVII-30). ‘Most microporous adsorbents have a range of micropore size, as evidenced, for example, by a variation in qq or in calorimetric heats of adsorption with amount adsorbed [227]. As may be expected, a considerable amount of effort has been spent in seeing how to extract a size distribution from adsorption data.670 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS Fig. XVII-30. Adsorption of No on a silica gel at 77.3 K, expressed as a u-+ plot, illustrating a method for micropore analysis. (From Ref. 230.) One approach has been to build on the DR equation by adding a Gaussian [228] or gamma function-type (224] distribution of B values. Another approach makes use of a modified Fowler-Guggenheim equation (see Eq. XVII-53) (229]. A rather different method for obtaining a micropore size distribution was pro- posed by Mikhail, Brunauer, and Bodor [230], often known as the MP method. ‘The method is an extension of the t-curve procedure for obtaining surface areas (Section XVII-9); a plot of cubic centimeters STP adsorbed per gram v versus the value of t for the corresponding P/P° (as given, for example, by Table XVIL-4) should, according to Eq. XVII-84, give a straight line of slope propor- tional to the specific surface area E. As illustrated in Fig. XVII-30, such plots may bend over. This is now interpreted not as a deviation from the character- istic isotherm principle but rather as an indication that progressive reduction in surface area is occurring as micropores fill. The proposal of Mikhail et al. was that the slope at each point gave a correct surface area for the P/P° and v value. The drop in surface area between successive points then gives the vol- ume of micropores that filled at the average P/P? of the wo points, and the average f value, the size of the pores that filled. In this way a pore size distri- bution can be obtained. Figure XVII-31 shows adsorption isotherms obtained for an adsorbent consisting of a-FeQOH dispersed on carbon fibers, and Fig. XVII-31b, the corresponding distribution of micropore diameters [231]. Finally, as might be expected, a fractal geometry approach has been made.XVII-16 ADSORPTION ON POROUS SOLIDS—-HYSTERESIS 671 0.75; Nz Adsorbed (mi/a) & OT thy @ os! oth) 0.10) CEL 0.08: AV (Volume adsorbed)/Ar (1 > meter 2r (A) Micropore oy Fig. XVII-31. (a) Nitrogen adsorption isotherms expressed as t-plots for various sam- ples of a-FeOOH dispersed on carbon fibers. (6) Micropore size distributions as ‘obtained by the MP method. [Reprinted with permission from K. Kaneko, Langmuir, 3, 357 (1987) (Ref. 231.) Copyright 1987, American Chemical Society.]672 XVII ADSORPTION OF GASES AND VAPORS ON SOLIDS This may be based on Eq. XVI-2 [232] or on related equations with film thick- ‘ness given by some version of the Frenkel—Halsey—Hill equation (Eq. XVII-79) (233,234). There has been fierce debate (see Refs. 232, 235-237) over the usefulness of the preceding methods and the matter is far from resolved. On the one hand, the use of algebraic models such as modified DR equations imposes artificial constraints, while on the other hand, the assumption of the validity of the ?-plot in the MP method is least tenable just in the relatively low P/P° region where micropore filling should occur. 17. Problems 1. The accommodation coefficient for Kr on a carbon filament is determined exper- imentally as follows. The electrically heated filament at temperature T2 is stretched down the center of a cylindrical cell containing Kr gas at T;. Gas molecules hittin the filament coo! it, and to maintain its temperature a resistance heating of Q cal sec~' cm” is needed. Derive from simple gas kinetic theory the expression 1/2 - a= AQ(MT;) laa I (XVII-140) Evaluate the constant A if M is molecular weight and P is in dyn/cm’, 2. Read off points from Fig. XVII-I and plot a set of corresponding isosteres and isobar. 3. Derive the general form of the Langmuir equation (Eq. XVII-11). 4, The separate adsorption isotherms for gases A and B on a certain solid obey the Langmuir equation, and it may be assumed that the mixed or competitive adsorption obeys the corresponding form of the equation. Gas A, by itself, adsorbs to a 6 of 0.02 at P = 200 mm Hg, and gas B, by itself, adsorbs to @ = 0.02 at P = 20 mm Hg; T is 77 K in both cases. (a) Calculate the differ ence between Qa and Qp, the two heats of adsorption. Explain briefly any assumptions or approximations made. (b) Calculate the value for @ when the solid, at 77 K, is equi- librated with a mixture of A and B such that the final pressures are 200 mm Hg each (©) Explain whether the answer in b would be raised, lowered, or affected in an unpre- dictable way if all of the preceding data were the same but the surface was known to be heterogeneous. The local isotherm function can still be assumed to be the Langmuir equation. 5. Calculate the value of the first three energy levels according to the wave mechanical picture of a particle in a one-dimensional box. Take the case of nitrogen ina 3.5-A box. Calculate Qrans ‘it 78 K using the integration approximation and by directly evaluating the sum Ey exp(—¢n/KT). 6. Calculate A352 at 6 = 0.1 for argon at 77 K that forms a weak adsorption bond with the adsorbent, having three vibrational degrees of freedom. 7. Discuss the physical implications of the conditions under which bo as given byXVII-17 PROBLEMS 673 the kinetic derivation of the Langmuir equation is the same as 6, the constant as given by the statistical thermodynamic derivation. 8 The standard entropy of adsorption A32 of benzene on a certain surface was found to be ~25.2 EU at 323.1 K; the standard states being the vapor at 1 atm and the film at an area of 22.5 x T A? per molecule. Discuss, with appropriate calculations, what the state of the adsorbed film might be, particularly as to whether it is mobile or localized. Take the molecular area of benzene to be 22 A? 9. Derive Bq. XVII-S4. 10. Show that the critical value of B in Eq, XVII-S3 is 4, that is, the value of 6 above which a maximum and a minimum in bP appear. What is the critical value of or 11. Drain and Morrison (1) report the following data for the adsorption of Nz on rutile at 75 K, where P is in millimeters of mercury and v in cubic centimeters STP per gram. P v P v P ev P v 117 600.06 275.0 1441.14 455.2 2418.34 498.6 3499.13 14.00 719.54 310.2 1547.37 464.0 2561.64 501.8 3628.63, 45.82 821.77 341.2 1654.15 471.2 2694.67 87.53 934.68 368.2 1766.89 477.1 2825.39 127.7 1045.75 393.3, 1890.11 482.6 2962.94 1644 1146.39 413.0 2018.18 487.3. 3107.06 204.7 © 1254.14 429.4 2144.98 491.1 3241.28 239.0 1343.74 443.5 2279.15 495.0 3370.38 Plot the data according to the BET equation and calculate v,, and c, and the specific surface area in square meters per gram. ‘The saturation vapor pressure P? of No is given by 339.8 og +7.71057 — 0.0056286T (Ref, 48), in mm Hg. 12. When plotted according to the linear form of the BET equation, data for the adsorption of Nz on Graphon at 77 K give an intercept of 0.004 and a slope of 1.7 (both in cubic centimeters STP per gram). Calculate E assuming a molecular area of 16 A? for Np. Calculate also the heat of adsorption for the first layer (the heat of condensation of Nz is 1.3 kcal/mol). Would your answer for vm be much different if the intercept were taken to be zero (and the slope the same)? Comment briefly on the practical significance ‘of your conclusion. 13. Consider the case of the BET equation with c = 1. Calculate for this case the heat of adsorption for the process: A(liquid adsorbate at 7) = A(adsorbed, in equilibrium with pressure P, at T) for @ values of 0.1 and 1.5, Calculate also the entropies of adsorption for the same