Untitled

Copyright 1999 by Marcel Dekker. All Rights Reserved.
URFACTANT S IENCE SERIES
FOUNDING EDITOR
MARTIN J. SCHICK
Consultant
New York, New York
SENIOR ADVISOR
ARTHUR T. HUBBARD
Department of Chemistry
University of Cincinnati
Cincinnati, Ohio
ADVISORY BOARD
DANIEL BLANKSCHTEIN ERIC W. KALER

Department of Chemical Engineering Department of Chemical Engineering
Massachusetts Institute of Technology University of Delaware
Cambridge, Massachusetts Newark, Delaware
S. KARABORNI CLARENCE MILLER

Shell International Petroleum Department of Chemical Engineering
Company Limited Rice University
London, England Houston, Texas
LISA B. QUENCER DON RUBINGH

The Dow Chemical Company The Proctor & Gamble Company
Midland, Michigan Cincinnati, Ohio
JOHN F. SCAMEHORN BEREND SMIT

Institute for Applied Surfactant Shell International Oil Products B. V
Research Amsterdam, The Netherlands
University of Oklahoma
Norman, Oklahoma
P. SOMASUNDARAN
Henry School of Mines
Columbia University
New York, New York

Nonionic Surfactants, edited by Martin Schick (see also Volumes 19, 23,
and 60)
2. Solvent Properties of Surfactant Solutions, by Kozo Shinoda (see
Volume 55)
3. Surfactant Biodegradation, Swisher (see Volume 18)
4. Surfactants, edited by Eric (see also Volumes 34, 37,
and 53)
5. Detergency: Theory and Test Methods (in three parts), edited by G.
Cutler and Davis (see also Volume 20)
6. Emulsions and Emulsion Technology (in three parts), edited by Kenneth
7 . Anionic Surfactants (in two parts), edited by Wamer M. Linfield (see Volume
8. Anionic Surfactants: Chemical Analysis, edited by John Cross (out of print)

Stabilization of Colloidal Dispersions by Polymer Adsorption, Sato
and Richard (out of print)
10. Anionic Surfactants: Biochemistry, Toxicology, Dermatology, edited by
Christian Gloxhuber (see Volume 43)
Anionic Surfactants: Physical Chemistry of Surfactant Action, edited by H.
(out of print)
12. Amphoteric Surfactants, edited by and Clifford L. (see
Volume 59)
13. Demulsification: Industrial Applications, Kenneth Lissant (out of print)
14. Surfactants in Textile Processing, Datyner
15. Electrical Phenomena at Interfaces: Fundamentals, Measurements, and
Applications, edited by Ayao Kitahara and Akira Watanabe
16. Surfactants in Cosmetics, edited by Martin Rieger (seeVolume 68)
17. Interfacial Phenomena: Equilibrium and Dynamic Effects, Clarence A. Miller
and P. Neogi
18. Surfactant Biodegradation: Second Edition, Revised and Expanded,
Swisher
19. Nonionic Surfactants: Chemical Analysis, edited by John Cross
20. Detergency: Theory and Technology, edited by Gale Cutler and Erik
21. Interfacial Phenomena in Apolar Media, edited by Hans-Friedrich Eicke and

Geoffrey
22. Surfactant Solutions: New Methods of Investigation, edited by Zana
23. Nonionic Surfactants: Physical Chemistry, edited by Martin J. Schick
24. Microemulsion Systems, edited by Henri L Rosano and Marc
25. Biosurfactants and Biotechnology, by W. Cairns, and
Neil C. Gray
26. Surfactants in Emerging Technologies, edited by Milton Rosen
27. Reagents in Mineral Technology, edited by P. Somasundaran and
Moudgil
28. Surfactants in Engineering, edited by
Martin and Dinesh 0. Shah
29. Thin Liquid Films, edited by 1. Ivanov

30. Microemulsions and Related Systems: Formulation, Solvency, and Physical
Properties, edited by Maurice and Robert Schechter
31. Crystallization and Polymorphism of Fats and Fatty Acids, edited by
and Kiyotaka Sato
32. Interfacial Phenomena in Coal Technology, edited by Gregory Botsaris
and Glazman
33. Surfactant-Based Separation Processes, edited by John F. Scamehom and
Jeffrey Harwell
34. Surfactants: Organic Chemistry, edited by James M. Richmond
35. Alkylene Oxides and Their Polymers, Bailey, and Joseph V.
Koleske
36. Interfacial Phenomena in Petroleum Recovery, edited by Norman Morrow
37. Surfactants: Physical Chemistry, edited by Donn and
Paul Holland
38. Kinetics and Catalysis in Microheterogeneous Systems, edited by
and K. Kalyanasundaram
39. Interfacial Phenomena in Biological Systems, edited by Max Bender
40. Analysis of Surfactants, Thomas M.
Light Scattering by Liquid Surfaces and Complementary Techniques, edited
by Langevin
42. Polymeric Surfactants,
43. Anionic Surfactants: Biochemistry, Toxicology, Dermatology. Second Edition,
Revised and Expanded, edited by Christian Gloxhuber and Klaus Kunstler
44. Organized Solutions: Surfactants in Science and Technology, edited by Stig
Friberg and Bjom Lindman
45. Defoaming: Theory and Industrial Applications, edited by P. Garrett
46. Mixed Surfactant Systems, edited by Keizo Ogino and Masahiko Abe
47. Coagulation and Flocculation: Theory and Applications, edited by
48. Biosurfactants: Production Properties Applications, edited by Naim
49. Wettability, edited by John Berg

50. Fluorinated Surfactants: Synthesis Properties Applications, Erik Kissa
51. Surface and Colloid Chemistry in Advanced Ceramics Processing, edited by
Robert J. Pugh and Lennart Bergsfrom
52. Technological Applications of Dispersions, edited by Robert
53. Surfactants: Analytical and Biological Evaluation, edited John
Cross and Edward Singer
54, Surfactants in Agrochemicals, Tharwat Tadros
55. Solubilization in Surfactant Aggregates, edited by Sherril D. Christian and
John Scamehom
56. Anionic Surfactants: Organic Chemistry, edited by Helmut
57. Foams: Theory, Measurements, and Applications, edited by Robert K.
homme and Saad A. Khan
58. The Preparation of Dispersions in Liquids, H. Stein
59. Amphoteric Surfactants: Second Edition, edited by Eric
60. Nonionic Surfactants: Polyoxyalkylene Block Copolymers, edited by Vaughn
Nace

Emulsions and Emulsion Stability, edited by Johan Sjoblom
62. Vesicles, edited by Morton Rosoff
63. Applied Surface Thermodynamics, edited by A. Neumann and Jan K.
Spelt
64. Surfactants in Solution, edited by Arun K. Chaifopadhyay and K.
65. Detergents in the Environment, edited by Milan Johann Schwuger
66. Industrial Applications of Microemulsions, edited by Conxita Solans and
Hironobu Kunieda
67. Liquid Detergents, edited by Kuo-Yann Lai
68. Surfactants in Cosmetics: Second Edition, Revised and Expanded, edited by
Martin Rieger and Linda Rhein
69. Enzymes in Detergency, edited by Jan H. van Ee, and Erik
Baas
70. Structure-Performance Relationships in Surfactants, edited by Kunio
and Ueno
Powdered Detergents, edited by Michael
72. Nonionic Surfactants: Organic Chemistry, edited by M. van
73. Anionic Surfactants: Analytical Chemistry, Second Edition, Revised and
Expanded, edited by John Cross
74. Novel Surfactants: Preparation, Applications, and Biodegradability, edited by
Krister
75. Biopolymers at Interfaces, edited by Martin Malmsten
76. Electrical Phenomena at Interfaces: Fundamentals, Measurements, and Ap-
plications, Second Edition, Revised and Expanded, edited by Hiroyuki
and Kunio
77. Polymer-Surfactant Systems, edited by Jan T. Kwak
78. Surfaces of Nanoparticles and Porous Materials, edited by James A.
and I.
79. Surface Chemistry and Electrochemistry of Membranes, edited by Torben
Smith
ADDITIONAL VOLUMES IN PREPARATION
Interfacial Phenomena in Chromatography, edited by Pefferkom
Solid-Liquid Dispersions, Xueping and Wolfgang von

Rybinski
Modern Characterization Methods of Surfactant Systems, edited by P.
Interfacial Forces and Fields, edited by Jyh-Ping

Silicone Surfactants, edited by Randal M. Hill
Surface Characterization Methods: Principles, Techniques, and Applications,

edited by Andrew Milling

edited by

ISBN: 0-8247-1933-6
This book is printed on acid-free paper.
Headquarters
Marcel Dekker
270 Madison Avenue, New York, NY 10016
tel: 212-696-9000; fax: 212-685-4540
Eastern Hemisphere Distribution
Marcel Dekker AG
4, Postfach 812, Switzerland
tel: 44-61-261-8482; fax: 44-61-261-8896
orld Wide Web
http://www.dekker.com
The publisher offers discounts on this book when ordered in bulk quantities. For
more information, write to Special Marketing at the headquar-
ters address above.
Copyright 1999 by Marcel Dekker All Rights Reserved.
Neither this book nor any part may be reproduced o r transmitted in any form or by
any means, electronic or mechanical, including photocopying, microfilming, and
recording, or by any information storage and retrieval system, without permission
in writing from the publisher.
Current printing (last digit):
1 0 9 8 7 6 5 4 3
PRINTED IN THE UNITED STATES OF AMERICA

They knew each other, they liked each other, and they
departed from their families and us almost at the same
time, in the last 1997.
They were our mentors. W e dedicate this volume to
Edward J . and I. with the
assurance that they knew the roles they have
played in our lives.

as it goes, a small may give analogy o f great things, and show the tracks
of knowledge.
from De (On The Nature of Things) BC)
In the world of nanoparticles and microporous materials, the interaction forces

between nanosized particles and molecules from the surrounding medium, or the
forces between particles themselves, may exceed the mechanical forces between
bodies of the macroscopic world. This is caused by the high surface-to-volume
ratio of nanoparticles and microporous materials. When familiar materials become
mainly surface, they acquire new optical, magnetic, electrical, chemical, and trans-
port properties. Thus, dispersions tend to agglomerate, fine particles show increased
mechanical strength, and microporous solids develop tremendous sorption and
molecular sieving properties.
Materials with a high surface-to-volume ratio have played important roles in
evolution and in our own lives. Extensive surface area provides optimal conditions
for chemical transformations to proceed with high reaction rates and high product
selectivity. The organization and stability of nanosized structures are controlled by
interactions at the molecular electrical or magnetic-rather than
by the mechanical forces that shape the macroscopic world. Our ability to control
and use for our benefit all kinds of special properties developed at the extended
surface that characterizes the objects of the nanoworld depends on our understand-
ing of phenomena at and across the interfaces.
Nanoparticles and porous materials exist essentially as an extended surface. The
line of structural similitude, however, ends at this point. Nanoparticles derive their
properties from their small solid size in their condensed phase, while porous mate-
rials derive their properties from the absence of solid material in their condensed
state. In other words, the shape of the is either convex or concave,

depending on whether the object is a collection of microscopic particles or a collec-
tion of microscopic voids.
In either case, the surface can strongly influence the chemistry that occurs within
the surrounding media. The gas-solid and liquid-solid interfaces of porous materi-
als have been the focus of many studies, over decades. In recent years,
materials have become the focus of many researchers. Such materials have
interesting properties, but they d o not definitively promise to change the state of
science or technology. What might change is our use of design strategies for con-
verting theoretical materials into practical materials for use.
The purpose of this volume, therefore, is to collect state-of-the-art procedures for
construction and design of nanoparticles and porous materials, where their applica-
tions might be most appropriate. To that end, synthesis and characterization pro-
cedures are presented. The ultimate test is their practical utilization in world"
environments that exist at the gas and liquid interfaces of these materials. Case
studies are presented and, in some instances, conclusions and projections for opti-
mal design procedures of nanoparticles and porous materials are offered. The scope
of this volume is inherently multidisciplinary from the viewpoint of usage of mate-
rials. The common factor, however, is that their surface chemical behavior
mates, and thus, unification of purpose and scope becomes a reality.
The volume is organized into three sections, each of which addresses fundamen-
tal and practical realization of the production of nanostructured materials. The first
section deals with the preparation, characterization, and transport properties of this
unique class of materials. Structural and chemical heterogeneity are the result of
preparation protocols, and various spectroscopies can be used to characterize these
properties. Transport of adsorbates is affected by both intraparticle and interpar-
ticle resistance, which can greatly influence applications in practical processes. Each
of these topics is represented as a case study that is general enough in scope that
cautious application of the reported results can be extended to other systems of
technological importance.
The remaining two sections deal with the fundamental and practical utilization
of nanostructural materials in gaseous and liquid environments. The second section
deals with the former case. A balanced blend of theoretical and practical applica-
tions is presented. Both theory and application emphasize the importance of struc-
tural and energetic heterogeneity and its influence on the performance of three-
dimensional materials that are essentially only two-dimensional due to their high
surface-to-volume ratio.
The final section presents case studies of the adsorption properties of surfaces
under the influence of a solvent that can alter the surface and chemical heteroge-
neity by mechanisms that have been studied by colloidal scientists for many years.
Again, in this section, there is a complimentary blend of theory and application.
Applications involving both inorganic and organic adsorbates are considered.
The subject areas in all three sections are quite diverse, which is a reflection of the
versatility of nanoparticles and porous materials in technologies that demand super-
ior performance, yet still require latitude for optimization strategies. Each of the
contributions has been peer-reviewed and we feel confident that the information
contained is complete and can be used by seasoned researchers and newcomers

to the field to condition their own research objectives in alignment with their own
expertise.
We wish to first acknowledge Professor Arthur for the suggestion that
a volume of this type was appropriate at this time. Each of the authors has
responded under undue pressure from both of us to submit their contributions in
a timely fashion. This has resulted in a final document that many should find of
interest. We appreciate the assistance of Dr. Adriana Contescu in maintaining up-
to-date folders of our correspondence with the contributors. Finally, the attention
to detail and the electronic communication to get everything in place for publication
came from Ms. Dawn Long. She has been responsible for helping us both remain
honest and on time, so that the final submission to the publisher occurred on
schedule and with minimum confusion.
James A .
I.

Preface
Contributors xiii
Part I Preparation, Characterization, and Transport Properties of Nanoparticles

and Porous Solids
Synthesis of a Polysilazane Coating on a Silica Gel via Chemical Surface
Coating Comparing Liquid- and Gas-Phase Chlorosilylations 1
Nathalie R. E. N . Ztnpens and Etienne F. Vansant
Preparation of Molecular Sieves by Pillaring of Synthetic Clays 15
Soon- Yong Jeong
Engineering of Nanosize Superparamagnetic Particles for Use in Magnetic
Carrier Technology 31
Zhenghe Xu, Qingxia Liu,and James A. Finch
Acid-Base Behavior of Surfaces of Porous Materials 51
Cristian I. Contescu and James A . Schirarz
Electro-Optical Spectroscopy of Colloidal Systems 103
Maria Stoimenova and Tsuneo Okuho
N M R Studies of Colloidal Oxides 125
Edisson Morgado, Jr., Sonia Maria Cabral de Menezes, and Carlos Roberto
Nogueira Pacheco
Polymer Surface Dynamics Using Surface-Modified Glasses via Dynamic
Contact Angle Measurements 169
Joung-Mafz Park

Contents
8. Microporous Structure of Collagen Fibers 185

Keito Boki
9. Adsorption onto Oxides: The Role of Diffusion 199
Lisa Axe and Paul R. Anderson
10. Electrokinetic Phenomena in Porous Media and Around Aggregates 21 1
Pierre M . Adler, David Coelho, Jean-Francois Thoven, and Michael Shapu'o
1 1 . Transport Processes in Microemulsions 259
Satya P. Moulik and Bidyut K. Paul
12. Structural Effects on Diffusivity Within Aggregates of Colloidal Zirconia 281
David H. Reeder, Alon V . McCormick, and Peter W . Carr
Part I I Adsorption from the Vapor/Gas Phase onto Nanoparticles and Porous
Solids
13. Characterization of Energetically Heterogeneous Surfaces from Experimental
Adsorption Isotherms 295
James P. Olivier
14. Computer Simulations of the Structural and Thermodynamic Properties of
Adsorbed Phases 319
William Stede
15. Surface Heterogeneity Effects o n Adsorption Equilibria and Kinetics:
Rationalizations of the Elovich Equation 355
WladyslawRudzinski and Tornasz Panczyk
16. Single- and Multicomponent Adsorption Equilibria of Hydrocarbons on
Activated Carbon: The Role of Micropore Size Distribution 391
K. and D. Do
17. Surface and Structural Properties of Modified Porous Silicas 443
Michal Kruk and Mietek Jaroniec
18. Nanodimensional Magnetic Assembly of Confined O2 473
Katsumi Kaneko
19. Heat of Adsorption of Pure Gas and Multicomponent Gas Mixtures on
Microporous Adsorbents 501
Shivaji Sircar and Madhukar B. Rao
art 111 Adsorption from the Liquid Phase onto Nanoparticles and Porous Solids
20. Surface Chemistry of Activated Carbon Materials: State of the Art and
Implications for Adsorption 529
Ljubisa R. Radovic
21. Charge Regulation at the Surface of a Porous Solid 567
Boris V. Zhrnud and Lennart Bergstrom

Contents
22. Surface Ionization and Complexation 593

Zhenghe Xu, Qingsong Zhang, and James A . Finch
23. The Surface Charge of Alkali Halides in Their Saturated Solutions 613
Jan D. Miller and Srinivas Veeraimsuneni
24. Ionic Adsorbates on Hydrophobic Surfaces 645

Richard L. Zollars
25. Adsorption of Metal Ions onto Humic Acid 661

Hideshi Seki and Akira Suzuki
26. Hydrous Metal Oxides as Adsorbents for Aqueous Heavy Metals 675
Russell Crawford, David E. Maimvaring, and Ian H . Hording
27. Adsorption of Ions onto Alumina 71 1

Edwin Baumgarten
28. Protein Adsorption onto Latex Particles 743

Jeffrey Leaver
29. Adsorption of Pharmaceutical Organic Compounds onto Porous
Materials 763
Keiji Yamamoto and Siriporn Okonogi

Pierre M. Adler Equipe Milieux Poreux, Institut de Physique du Globe de Paris,
Paris, France
Paul R. Anderson Department of Chemical and Environmental Engineering,

Illinois Institute of Technology, Chicago, Illinois
Lisa Axe Department of Civil and Environmental Engineering, New Jersey

Institute of Technology, Newark, New Jersey
aumgarten Institut fur Physikalische Chemie und Elektrochemie,

Heinrich-Heine-Universitat, Dusseldorf, Germany
ergstrom Institute for Surface Chemistry, Stockholm, Sweden
oki Department of Pharmaceutical Sciences, Kinki University, Osaka,

Japan
. Carr Department of Chemistry, University of Minnesota, Minneapolis,

Minnesota
oelho Equipe Milieux Poreux, Institut de Physique du Globe de Paris,

Paris, France
Cristian I. Contescu Department of Chemical Engineering and Materials

Science, Syracuse University, Syracuse, New York
xiii

xiv Contributors
ussell Crawford Centre for Applied Colloid and BioColloid Science,

Swinburne University of Technology, Melbourne, Australia
D. D. Do Department of Chemical Engineering, University of Queensland, St.

Lucia, Australia
dames A. Finch Department of Mining and Metallurgical Engineering, McGill

University, Montreal, Quebec. Canada
lan H. Harding Centre for Applied Colloid and BioColloid Science, Swinburne
University of Technology, Melbourne, Australia
. Impens
Chemistry Laboratory of Adsorption and Catalysis,
University of Antwerp, Antwerp, Belgium
Mietek daroniec Department of Chemistry, Kent State University, Kent, Ohio
Soon-Yong deong Chemical Process and Engineering Center, Korea Research

Institute of Chemical Technology, Taejeon, Korea
Katsumi Kaneko Department of Chemistry, Chiba University, Chiba, Japan
Michal Kruk Department of Chemistry, Kent State University, Kent, Ohio
Jeffrey Leaver Molecular Recognition Group, Hannah Research Institute, Ayr,

Scotland
xia Liu Cominco Research. Cominco Ltd., Trail, British Columbia, Canada
Department of Applied Chemistry, Royal Melbourne

Institute of Technology, Melbourne, Australia
Ion V. McCormick Department of Chemical Engineering and Materials

Science, University of Minnesota, Minneapolis, Minnesota
onia Maria Cabral e Menezes Chemistry Division, Petrobras Research and

Development Center, Rio de Janeiro, Brazil
. Miller Department of Metallurgical Engineering, University of Utah, Salt

Lake City, Utah
disson Morgado, Jr. Catalysts Division, Petrobras Research and Development

Center, Rio de Janeiro, Brazil
atya P. Moulik Center for Surface Science, Department of Chemistry, Jadavpur

University, Calcutta, India

ontributors xv
konogi Department of Pharmaceutical Technology, Chiang Mai

University, Chiang Mai, Thailand
Tsuneo Okubo Department of Applied Chemistry, Gifu University, Gifu, Japan
livier Micromeritics Instrument Corporation, Inc., Norcross,

Georgia
Carlos Roberta Nogueira Pacheco Chemistry Division, Petrobras Research

and Development Center, Rio de Janeiro, Brazil
anczyk Department of Theoretical Chemistry, Marie-Curie-

Sklodowska University, Lublin, Poland
Joung-Man Park Department of Polymer Science and Engineering, Gyeongsang

National University, Chinju, Korea
idyut K. Paul Geological Studies Unit, Indian Statistical Institute, Calcutta,

India
Ljubisa R. Radovic Department of Materials Science and Engineering, The

Pennsylvania State University, University Park, Pennsylvania
ao Corporate Science and Technology Center, Air Products and

Chemicals, Inc., Allentown, Pennsylvania
er Department of Chemical Engineering and Materials Science,

University of Minnesota, Minneapolis, Minnesota
Wiadyslaw Rudzinski Department of Theoretical Chemistry. Marie-Curie-

Sklodowska University, Lublin, Poland
chwarz Department of Chemical Engineering and Materials Science,

Syracuse University, Syracuse, New York
ideshi Seki Department of Marine Bioresources Chemistry, Hokkaido

University. Hakodate, Japan
hapiro Department of Mechanical Engineering, Technion-Israel

Institute of Technology, Haifa, Israel
Shivaji Sircar Corporate Science and Technology Center, Air Products and
Chemicals, Inc., Allentown, Pennsylvania
teele Department of Chemistry, The Pennsylvania State University,

University Park, Pennsylvania

xvi
Maria Stoimenova Institute of Physical Chemistry. Bulgarian Academy of

Sciences, Sofia, Bulgaria
kira Suzuki Department of Marine Bioresources Chemistry, Hokkaido

University, Hakodate, Japan
Thovert Laboratoire de Combustion et de Detonique, Centre

National de la Recherche Scientifique, Futuroscope, France
Etienne F. Vansant Chemistry Laboratory of Adsorption and Catalysis,

University of Antwerp, Antwerp, Belgium
rinivas Veeramasuneni Department of Metallurgical Engineering, University

of Utah, Salt Lake City, Utah
. Wang Department of Chemical Engineering, University of Queensland, St.

Lucia, Australia
Zhenghe Xu Department of Chemical and Materials Engineering, University of

Alberta, Edmonton, Alberta, Canada
Keiji Yamamoto Department of Pharmaceutical Technology, Chiba University,

Chiba, Japan
Zhang Hewlett Packard, Palo Alto, California
Institute for Surface Chemistry, Stockholm, Sweden
ollars Department of Chemical Engineering, Washington State

University, Pullman, Washington

NATHALIE and ETIENNE
Chemistry Laboratory of Adsorption and Catalysis, University of
Belgium
I. Introduction
II. Experimental
III. Results and Discussion
A. Liquid-phase modification of silica gel with tetrachlorosilane
and the first ammoniation
B. Second and higher reaction cycles
C. Porosity study
IV. Conclusions
References
I. INTRODUCTION
Due to the widespread use of ceramic coatings, several synthesis techniques have
been developed in recent decades. The majority of these techniques, such as che-
mical vapor deposition (CVD) and physical vapor deposition (PVD) and their
variants, are focused on the synthesis of flat coatings. Recently, the preceramic
polymeric synthesis route has offered the possibility to impregnate preceramic
materials into porous matrices prior to pyrolysis in order to create coated or
composite materials This technique, however, leads to pore filling and altera-
tions of the original substrate texture. The impregnation does not guarantee a
chemical bond between the substrate and the preceramic material. This can lead
to an unstable composite material after pyrolysis. T o avoid this. another technique,
called chemical surface coating (CSC) can be used. Based on the successive
chemisorption of gaseous reagents in a cyclic way at relatively low temperatures,
the technique offers clarity in the reaction mechanisms [4,5] and the evolution of
porosity [6] during the preceramic synthesis. It has been shown that the preceramic

and
coating is chemically bound to the substrate surface and that is has grown in a very
homogeneous way in spite of the very irregular substrate texture. The so-called
coating precursor is afterwards pyrolyzed, and a chemically bound ceramic coating,
e.g., or BN; is formed on silica gel. Characterization of the thermally con-
verted Si- N or B-N polymer shows that a very thin homogeneous silicon
oxynitride or boron nitride coating, respectively, is formed, whereas the original
substrate texture is practically unchanged [7,8].
In the procedure described in this chapter, a preceramic polysilazane coating is
synthesized on silica gel via CSC using S i c &and NH3, but the chlorosilylations are
performed in the liquid phase. In this way, a very small amount of the hazardous
reagent is used in comparison with gas-phase reactions, and the total precursor
synthesis can be performed at room temperature.
The solid-gas reaction of with thermally pretreated silica gel has been
optimized, and the reaction mechanisms were studied earlier [4]; a reaction tem-
perature of 663 K is needed to modify all the surface hydroxyl groups. T o reduce
the reaction temperature, and Hair [9] studied the effect of the presence of
during reaction with At room temperature, a
very reaction yield was obtained. The reaction mechanism was deduced from
the of silica proposed by Blitz et [see reaction
The nitrogen-containing base promotes the reaction by rendering the

group more nucleophilic. When the base strength is high enough (as in
mine), proton transfer is enabled. However, is formed at the
substrate surface and must be removed by sublimation or dissolution in a
methanol which results in the hydrolysis of groups. This has to be
avoided in the case of the CSC polysilazane synthesis, as these groups are especially
created to enable a reaction with that creates amine groups.
The effects of solvent and amine base on the solid-liquid reactions of
chlorosilanes with the silica surface hydroxyls were throroughly investigated by
and Unger They concluded that the best yield of was
obtained by using strong bases together with solvents with high Lewis donating
and accepting properties. T o our knowledge, none of the authors reported a pure
with the formation and conservation of groups without the
formation of ammonium salt on the silica surface.

In this study, the reaction parameters were chosen in such a way that the salt
formed was dissolved in the solvent and the yield of groups was comparable
to that obtained with the gas-solid reactions at high temperature. For the reaction
of with the substrate in the following reaction cycles, a solvent was chosen
that would not affect the precursor. A comparison of the chemical and morpholo-
gical composition of the coating precursor using liquid- and gas-phase
synthesized in solid-gas and solid-liquid conditions was conducted using
elemental analysis, Fourier transform infrared spectroscopy with photoacoustic
detection (FTIR-PAS), and nitrogen
The silica (KG60-Merckj, thermally pretreated at 973 was allowed to cool in an

glove box to avoid any possible rehydration of the sample. This treat-
ment results in 0.7 isolated hydroxyl groups on the silica surface. This is
determined by chemical probe reaction with as explained
elsewhere The were performed using products,
which were stored in an atmosphere. (99.8%) was used as
received, whereas the solvents toluene cyclohexane and
methane (99.8%) were zeolite-dried prior to use 0.5 g of silica, and excess of 0.8
and varying amounts of base were added to 25mL of
solvent. After 3 h of stirring, the silica was filtered, then washed four times with
of pure solvent. Afterwards, the silica was dried using a cryogenic trap and a
vacuum pump.
Ammoniation was carried out in a dynamic volumetric adsorption apparatus, as
described elsehwere The ammonia uptake was measured
whereas the surfce chlorine concentration was determined
Extreme care was taken to prevent the samples from hydrolysis, by handling
them in a glove box or in The total chlorine concentration was deter-
mined after direct hydrolysis of the modified surface following
reaction.
FTIR-PA spectra were recorded on a 5DXB FTIR spectrophotometer
with photoacoustic detection. The PA detector is a prototype of the MTEC-100 cell
constructed by F. The photoacoustic cell was flushed with
dried helium in an glove box. The mirror velocity was 0.08 Comparison of
the amount of hydroxyls was performed by integrating the band at
3 7 4 7 c m 1 after Kubelka-Munk correction of the spectrum and normalizing the
OH band by using the integrated reference band at assigned to a
combination mode.
Nitrogen adsorption-desorption isotherms at 7 7 K were measured with a
Quantachrome 1 instrument. The sample surface area was calculated
using the Brunaner-Emmett-Teller (BET) model whereas the pore size dis-
tribution was obtained using the Barrett-Joyner-Halenda (BJH) model The
plot method was used to check the microporosity

and
Modification of
Ammoniation
The creation of chlorosilyl groups on the substrate surface is very important in the
synthesis of a chemically bound preceramic polysilazane coating, because the
amounts and types of chlorosilyl species present influence the reaction behavior
of the reactions. In Table 1 , a comparison of the elemental analysis data is shown
for the gas- and liquid-phase synthesis routes as a function of the number of
reaction cycles. Infrared spectroscopy of the modified silica gel was used to study
the reaction yield. In Figure the infrared spectra of pure silica (973 spectrum
a) and silica modified with in both the gas and liquid phases (spectra are
shown. When toluene (spectrum c) or cyclohexane were used as solvents, all the
huge bands between 3000 and 2000 c m ' and the sharp bands superimposed on the
Si 0 vibrations between 1500 and 800 exactly correspond to
chloride, which was formed on the silica surface and could not be removed by
severe washing. When dichloromethane was used (lines d and e) no salt formation
was detected, and very fast removal of the solvent was possible due to its volatility.
This solvent's high Lewis accepting property is shown to promote the reaction of
silica gel with octadecyldimethylchlorosilane a t room temperature The nitro-
gen-containing base used in this study has to be a tertiary amine, avoiding any
polymerization reactions with Triethylamine was chosen, as its conjugated
base has a relatively high value (10.8). When the base was added in a catalytic
amount (5 1/20 relative to the OH groups present), the reaction amount
was very low (Fig. 1, spectrum e). The reason is that after protonation the base has
no further influence on the reaction, which implies that it does not act as a catalyst,
as was erroneously reported in the literature An excess of base (800 resulted
in almost complete modification of the surface hydroxyl groups (Fig. 1, spectrum
d). As a blank, the reaction was also carried out under similar conditions but
Nitrogen and Chlorine Uptake (in

1 Unmodified Silica Gel) and
Ratio as a Function of the Number of Reaction Cycles for the Liquid- and Gas-Phase
Synthesis Routes
Reaction Cycle3
Liquid phase route Cl 1.53 (0.05) 2.7 (0.1) 1.6 (0.1) 1.6 (0.3)
N
Gas phase route

N
1.8 (0.05) 1.5 (0.2) 1.65 (0.06) 1.8 (0.2)
Values in parentheses represent the absolute experimental errors.

. FTIR-PA spectra of KG 60 (a) pretreated at 973 unmodified, and
modified with (b) in gas phase, at optimal reaction conditions; (c) in liquid phase,
using toluene and in excess; (d) in liquid phase, using and in excess;
(e) in liquid phase. using and in catalytic amounts; (f) in liquid phase,
using without Relevant peak identification of silica gel: 3747 free
H stretch; combination; ,
overtone; asymmetric stretch: 800 c m ' , SiO-symmetric
stretch.
without triethylamine, resulting in no modification at all, as shown in Fig. 1, line

In the liquid phase the highest yield is obtained with a high amine concentration in
combination with a (spectrum d). The vibration is positioned at the
1
same frequency ( 6 2 0 c m ) for both a gas-phase reaction at 633 K (spectrum a) and
a liquid-phase reaction at 298 K (spectrum d), indicating that monodentate binding
occurs, resulting in groups The small H band in the case
of the liquid-phase modification (spectrum d) represents 25% of the total amount

of hydroxyl groups at the surface (0.53 x of O H groups per gram of
substrate are modified). As only groups are expected from data, the
total chlorine content (Table 1) could be used to calculate the amount of
groups created, being 0.51 x of groups per gram of substrate.
These two independently calculated values are in good agreement. Longer reaction
times or higher amounts of reactants or amine base did not improve the reaction,
probably due to steric hindrance, as both the and the must be in the
neighborhood in order to have any reaction.
Similar to the gas-phase chlorosilylated silica gel, the ammoniation was per-
formed for 20 at room temperature, as these were shown to be the optimal
reaction conditions, resulting in the reaction [5]
After spectroscopy showed no difference between the gas and

liquid synthesis routes. However, during the liquid-phase route, the nitrogen uptake
is twice the chlorine uptake within the limit of experimental error. whereas only
90% of the chlorines are ammoniated after a gas-phase chlorosilylation
1.8). From a steric point of view, this is a logical result, as in the liquid synthesis
route both and must be near the attacked group. Therefore, a
smaller amount of chlorosilyl groups is formed in the liquid-phase synthesis route,
situated at the most easily reached hydroxyls. After the liquid-phase reaction,
leaves the reaction site, leaving an empty space that can be reached by during
the In the gas-phase synthesis route, no is needed for the
and higher proton mobility is expected at the high temperature, resulting
in silylation of O H groups at sterically hindered sites. Anyway. a significant frac-
tion of these chlorosilyl groups synthesized in the liquid phase could not be mod-
ified by ainmoniation due to steric reasons.
In Fig. 2, the spectra of the samples after the first (spectrum a) and
the second liquid chlorosilylation are shown for different reaction conditions.
the second chlorosilylation on, dichloromethane could not be used and was
replaced by cyclohexane, because dichloromethane reacts with the amino groups.
The addition of to the reaction mixture (spectrum b) leads to the
deposition of triethylammonium chloride a t the sample surface, which could not be
removed with the solvent used or by sublimation, because thermal treatment also
leads to condensation reactions of the amino groups Therefore, the
are performed without the addition of triethylamine. Spectrum c of Fig. 2 is
that of the preceramic precursor after the second chlorosilylation. In the Si-
region, maxima are observed at 624 and This means that both
and groups are formed. The total disappearance of the asymmetric N-H
bending vibration at assigned to groups indicates a total
modification of these groups. Therefore, an estimation of the ratio of mono
and bidentate reactions could be calculated Elemental analysis and iteration
showed that of the groups are modified by bidentate binding and
by monodentate binding.

FTIR-PA spectra of KG 60. pretreated at 973 (a) Reacted with in
and excess and afterwards reacted with in the gas phase. (b. c)
as (a) but modified with (b) in the presence of using cyclohexane, and (c)
without using cyclohexane.
The second does not lead to completion, as an ratio of only

1.33 is reached. A similar phenomenon was observed in the gas-phase synthesis
route, where an ratio of 1.5 was found. In the gas-phase synthesis route, a
higher uptake of Cl is observed during the third and fourth chlorosilylations than in
the liquid phase. This is certainly due to the presence of the solvent in the liquid
case, which leads to a sterically unfavorable situation. Therefore, chlorosilylation is
expected only at the most easily reached sites. From the third reaction cycle on, the
ammoniation leads to a complete modification of the groups in the
phase synthesis route, whereas the gas-phase synthesis route never leads to an
ratio of 2. After the solvent evacuation in the liquid case, there is enough space left
for to reach and react with all the groups. The general trend of

and
nitrogen, chlorine, and weight uptake during the four reaction cycles for both the
liquid- and gas-phase synthesis routes is shown in Fig. 3.. In general, all the curves
are linear except for the chlorine uptake in the liquid-phase route. where a deflec-
tion is observed at two cycles. However, this trend is not followed by the
uptake, as the ammoniation is not disturbed by solvent hindrances. From this
figure, it is clear that in the liquid case a smaller amount of polymer is formed
on the silica surface, whereas all the gas-phase slopes are higher than in the case of
liquid-phase modification. The total increase in weight gain after four reaction
cycles if 0.55 and 0.8 g per gram of unmodified silica in the liquid and gas synthesis
routes, respectively.
From the rough analytical data of Table 1, the amounts of
and can be calculated (Table 2). These data can give an idea of the
cross-linking in the polymeric preceramic material. The amount of reflects
roughly the ratio between the increasing weight of the two synthesis routes (weight:
= 0.69; whereas the amount of unreacted
groups is relatively lower in the liquid-phase synthesis route, even when the
increased weight ratio is taken into account = 0.47). This result was
expected from the trend difference in the and uptakes. Another important
factor is the percentage of primary amines that are not modified into In
the gas phase, only 26% of the total population has not reacted,
whereas 36% is unreacted in the liquid-phase synthesis route. This means that a
lower cross-linking efficiency is reached in the liquid-phase route. This value, in
combination with the lower uptake of nitrogen in the liquid-phase synthesis route,
shows that a gas-phase synthesis is more suitable for the synthesis of a dense cross-
linked precursor.
With each modification cycle during the synthesis of the structure, the
amount of product left at the silica pore walls increases. To study the effect on the
substrate porosity of both types of precursors, adsorption-desorption isotherms
are recorded at 77 K as a function of the number of reaction cycles. The isotherms
presented in Fig. 4 are from the liquid-phase synthesis route and are all of type IV
according to the (BDDT) classification
Therefore, both the substrate and the coated material are mesoporous materials.
The total pore volume decreases with increasing number of reaction cycles. The
type of hysteresis changes from type to according to the IUPAC classifica-
tion Type corresponds to the pore shape of pure silica gel, whereas the
type indicates ink-bottle or irregularly shaped pores. The evolution in the hysteresis
shows the occurrence of pore shape modification during the formation of the
polymer. This phenomenon earlier observed was for the gas-phase synthesis
route The behavior of the constants for the liquid-phase reaction shows
the same trend as in the gas phase; from the first cycle on, an important
decrease is observed, indicating total monolayer coverage.
Using the calculation methods pointed out by et al. a difference
between the theoretically expected and experimentally observed pore volume
decreases can be used to calculate the amounts of pore blocking and regular

0
0 1 2 3
reaction cycle
reaction cycle
FIG. 3 Total (a) and weight uptakes as a function of the number of

reaction cycles for gas and (+) liquid synthesis routes, with corresponding linear
regressions. A deflection point is introduced in the regression of the (+) curve of (a). The
experimental errors for the first three reaction cycles are smaller than the symbols used.

1 and
Composition of the Coating After Four Reaction Cycles for

the Liquid- and Gas-Phase Synthesis Routes
in coating unmodified
silica
Species Liquid-phase route Gas-phase route
"'Values in parantheses represent the absolute experimental errors
pore narrowing. This is independent of the pore shape In Figs. 5a and

the theoretical and experimental pore volumes are shown as a function of the
amount of chemisorption for both synthesis routes. As a n example, the
amount of chemisorbed is highlighted in the figures when a pore volume of
0.2 is left. In the liquid-phase synthesis route (Fig. only 13
is chemisorbed. In the gas-phase synthesis route (Fig. the chemisorbed is
already 15 This indicates a higher amount of pore blocking in the
phase synthesis route.
.4 Nitrogen adsorption-desorption isotherms at 77 K for the polysilazane-coated

silica gel after zero, one, three and four reaction cycles of the liquid-phase synthesis
route. The experimental error on the amount of adsorbed is 0.1

uptake
FIG. 5 Pore volume decreases as a function of the of reacted. (

Theoretical and (+) experimental values, (a) Liquid-phase synthesis route; (b)
phase synthesis route. The experimental errors are smaller than the symbols used.
The linear regressions serve only as a guide for the eyes.
A similar trend was observed when the decreases in pore length were calculated
using integration of the pore volume distributions obtained by the BJH method
as explained earlier However, this method is valid exclusively for cylind-
rical pores. The results are shown in Fig. 6. A systematically pore length
value is obtained for the liquid-phase route. These data show that more pore
blocking occurs in this synthesis route.
The calculated amounts of pore-blocking and pore narrowing and their contri-
bution to the total pore volume decrease are listed in Table 3. The results are based
o n the data of Fig. 5 . Systematically higher absolute and relative amounts of pore
blocking are seen for the liquid-phase synthesis route in spite of the lower uptake of
N and per reaction cycle, resulting in a lower coating homogeneity. Moreover,
from the higher absolute and relative narrowing values in the gas-phase synthesis
route, it can be concluded that a higher coating efficiency is obtained using the
phase synthesis route.
Successive reactions of in the liquid phase and in the gas phase were
performed on silica gel. The reaction parameters for the
solvent choice and base promotion) were optimized to get as much uptake
as possible. The chemical composition of the polysilazane layer as a function of the
number of reaction cycles is obtained by using a combination of elemental and
infrared analysis.
A comparison of these results with the composition of a polysilazane coating
fully synthesized in the gas phase showed less uptake of N and and a greater
number of groups due to the presence of the solvent. This implies that the

0 2 4 6 8 10 12 14 16
uptake
FIG. 6 Pore length decreases for (+) gas and liquid synthesis routes, with
respective linear regressions. The experimental errors are smaller than the symbols
used. The linear regressions serve only as a guide for the eyes.
Contribution of Pore Narrowing and Pore Blocking to the Decrease in Pore
Cycles Blocking Narrowing
(gas phase)
0.70 0.00 0.00
0.65 0.12 0.05
0.58 0.17 0.12
0.49 0.20 0.21
0.39 0.26 0.31
(liquid phase)
0.70 0.00 0.00
0.66 0.13 0.04
0.61
0.57 0.26 0.13
0.53 0.33 0.17
PV = pore volume.
"The experimental errors on the pore volumes range from 0.0005 to The errors in the
percentages are smaller than 1

gas-phase synthesis route is more suitable for the creation of a dense, cross-linked
polysilazane polymer o n the silica surface.
The porosity was studied by using adsorption- desorption isotherms recorded
a t 77 A higher amount of pore blocking in the liquid-phase synthesis route was
observed, although a smaller amount of product was A combination
of the pore blocking and pore narrowing results shows that the gas-phase synthe-
sized polysilazane coating is more homogeneously spread on the silica surface.
1. Seyferth, Adv. Chem. 224: 565.

2. D. Seyferth and H. Weiseman, Polymer Prepr. (1984).
3. Vansant, P. Van Der Voort and Vrancken. Characterization and
Chemical o f the Silica Surface, Elsevier. Amsterdam. 1995.
4. Possemiers, and Vansant, Chem. Faraday
Trans (1996).
5. Possemiers, N. and Vansant, Chem. Faraday
Trans. (1997).
6. E. Possemiers, and E. F .
Vansant, Porous Mater. 4121 (1997).
7. Vansant, P. Grobet, Riga and 0.
Brouxhon, Mater. Chem. (1997).
8. Possemiers, thesis, 1996.
9. P. Tripp, and Hair, Chem. (1993).
10. J. Blitz, S. Murthy, and Leyden, Colloid Interface
(1988).
11. and Unger, J. Chhromatogr. (1984).
S. P. Emmett, and Teller, Am. (1938).
13. P. Barrett, Joyner, and P. P. Am. (1951).
14. and De Boer, Catal. (1965).
and K . Unger, Chromatogr. (1984).
16. Brunauer, S. and Teller, Am. Chem.
62:1723 (1940).
17. and Sing, Adsorption, Surface Area and Porosity, Academic,
New York, 1976, p. 287.

Chemical Process and Engineering Center, Korea
Research Institute of Chemical Technology, Taejeon, Korea
I. Introduction
I I . Experimental Methods
A. Synthesis of aluminum-free clays
B. Synthesis of H-magadiite and H-kenyaite
C. Silica-intercalated layered silicate
D. Analytical methods
III. Results and discussion
A. Synthesis of aluminum-free clays
B. Octylamine/octylammonium layered silicate gel
C. Reactions with TEOS
D. Gelation effects
IV. Conclusions
References
In recent years, the synthesis of aluminum-free clays such as magadiite

kenyaite and kanemite
has been of increasing interest due to their catalytic, absorptive, and
exchange properties Synthetic aluminum-free clays are also called layered sili-
cates because they have a layered structure. They have the characteristic of being free
of aluminum, unlike zeolites, and exhibit acid-resistance and thermal stability.
Magadiite and kenyaite were first found by [3] in the lake beds of Lake
Magadi in Kenya. Since then they have been reported to occur in various regions
Most of these deposits were found in sodium carbonate-rich alkaline
lake waters. Also they have been successfully synthesized under hydrothermal

conditions. Their basic structures are composed of duplicated tetrahedral
sheets and are similar to those of clay minerals with no aluminum Lagaly
and Beneke discussed the preparation of Na-magadiite in a hydrothermal system
at employing 4 weeks of reaction. More recently, Muraishi [9] synthesized K-
magadiite at with 2-5 days of reaction in KOH solution containing
various amounts of Kosuge et sythesized at
with h of reaction in the presence of organic solvent. However, the preparation
of kenyaite has rarely been reported. and Lagaly indicated that the
formation of at required several months. At higher temperature, the
reaction time for the formation of kenyaite was much decreased, but quartz was
formed simultaneously. More recently, Fletcher and preliminarily inves-
tigated the effects of anions the formation of kenyaite.
Silica pillaring of magadiite was reported by a few researchers. et
found that the pillaring of magadiite could be facilitated by using a preswelling step
in which the interlayers are exposed to ion or amine. The cal-
cined sample obtained from tetraethoxysilane ITEOS; pillaring exhib-
ited a high surface area, Scholzen et described the change in the
ratio of to groups caused by coupling of silicate
species with groups on the interlayered surface of magadiite. Sprung et
reported that pillared derivatives of magadiite can be obtained the calci-
nation of hydrolyzed Daily and
synthesized derivatives on the basis of H-magadiite by of
TEOS with EtOH suspension. After calcining to remove organic compounds, pil-
lared magadiite with surface areas of about depending on the
amount of gelled TEOS, was formed. However, the pillaring of kenyaite has rarely
been reported. Recently, et al. prepared pillared derivatives on the basis
of H-kenyaite by of TEOS with EtOH suspension. The calcined sample
exhibited a high surface area,
In the sol-gel process, a solvent such as EtOH is added to prevent liquid-liquid
separation during the initial stage of the hydrolysis reaction and to control the
concentration of silicate species and water that influence the kinetics
However, the reaction of TEOS by alcohol suspension occurs very
slowly and because TEOS is alcohol-soluble it can be released outside the
layered phases during gelation. Aelion et al. observed that the rate of hydrolysis
of TEOS was influenced by the strength and concentration of the acid and base
catalysts. Fast using a catalyst such as acid or base can minimize the
release of TEOS from the layered phases during gelation. Generally, TEOS
by acid- or base-catalyzed hydrolysis could diversify the interfacial properties of
products and results in such products as bulk gel, film, fiber, powder, and catalyst
support. In particular, by acid and base catalyst into interlayers could
bring out the pillaring effect as well as the diversity of interfacial properties of
pillared layered phases. According to and Scherrer the use of catalyst
tends to develop degree of of the pillar in interlayers, so that
gallery height can be increased.
We synthesized pure and well-crystallized Na-magadiite and Na-kenyaite under
the carbonate system in a short reaction time. I report here the effects related to the
compositions of reactants, reaction and temperature on the preparation of

these clays in the sodium carbonate system. Also, I report the effects of acid and
base catalysts on the hydrolysis and condensation polymerization of intercalated
TEOS in layered silicates.
Na-magadiite and Na-kenyaite were prepared by the reaction of the

system under hydrothermal conditions. The materials used were silica gel
(Wakogel Q-63) and NaOH and of analytical reagent grade. Experiments
were carried out in a stainless steel autoclave without stirring for the various reac-
tion times and ratios of reactants at under autogenous pressure.
The molar ratio of NaOH to was fixed at 1/2 in all cases. The products
were filtered, washed with deionized water to remove excess NaOH or and
then dried at
of and
H-magadiite was prepared by titrating with 0.1 using the
method of Beneke and Lagaly A suspension of 40g of per
of deonized water was stirred for 1 h. The suspension was then
with 0.1 N to a final pH of 2.0 and then maintained at the same pH value for 1
week in a refrigerator. H-magadiite was recovered by filtering, washed with deio-
nized until chlorine free, and then dried in air at The preparation of H-
kenyaite was the same as that of
magadiite gel was reacted for 24 h at room tempera-

ture by adding of excess to of air-dried An
organic pillar precursor, 40 g of TEOS, was added to
magadiite gel and then stirred for 24 h at room temperature. TEOS was then
absorbed into the organophilic interlayer region. TEOS-intercalated magadiite
was separated by centrifugation the mother liquid.
kenyaite gel was formed by allowing air-dried H-
kenyaite to react at room temperature with excess octylamine for 24 h.
During octylamine addition, absorbs the liquid immediately
forming a gelatinous mixture that will not flow. Silica-intercalated derivatives of
kenyaite were prepared by the reaction of excess TEOS with a gel
of octylammonium kenyaite by excess octylamine for 24 h at room tem-
perature. The TEOS-intercalated product was separated by the
mother liquid.
of the intercalated TEOS without catalyst was carried out by drying
suspension of TEOS-intercalated magadiite of kenyaite at in
air. EtOH was mixed with 3 and 0.1 N in order to examine the
effects of base and acid catalysts during the gelation. The compositions of acid and
base catalysts are shown in Table 1. was conducted with stirring for 60

1 Compositions of Catalysts
No catalyst 95.0 5.0

Base catalyst 16.5 79.2 4.3
Acid catalyst 64.8 35.2 0.01
Source: Ref 29
min after addition of of each catalyst to the TEOS-intercalated magadiite or

kenyaite at room temperature. The and methodology of of
TEOS are well known, and the physical characterization of gelled silicate has been
studied by several researchers The gelled samples were filtered from the
mother liquid, dried in air, and then calcined at for 4 h in air to remove water,
intercalated ion, and organic by products from TEOS hydrolysis.
. Analytical Methods
Identification of samples was carried out by X-ray poweder diffraction (Rigaku
diffractometers equipped with radiation) and scanning electron
microscopes (Jeol, The chemical compositions of synthetic magadiite
and kenyaite were determined with an energy-dispersive X-ray spectrometer (Link
and analysis (TGA). EDX analysis was
carried out using an electron beam with the sample of carbon-coated pellets. For the
quantitative analysis of silica and sodium, quartz and sodium chloride were used as
references. analyses were performed in air on a Dupont 9900
analyzer. Basal spacings of samples were determined from the
001 X-ray powder diffraction using a equipped with Cu
radiation. Samples of layered silicate gel and TEOS-
intercalated layered silicate gels were prepared by smearing a thin film across a
microscope slide and then recording diffraction patterns of wet samples. The pro-
ducts gelled by EtOH suspension. base-catalyzed reaction. and acid-catalyzed reac-
tion were prepared by depositing a gelled suspension on a glass plate and allowing
the suspension to air dry at The calcined silica-pillared magadiite and kenyaite
for X-ray diffraction analysis were prepared by calcining the gelled products on glass
plates at in air for 4 Nitrogen adsorption-desorption isotherms were
determined by ASAP 2000 at All samples were outgassed at
under a vacuum for 4 Surface area was determined using the
Emmett-Teller (BET) equation. The pore size distributions of silica-pillared pro-
ducts were determined by using the (BJH) equation
Figure 1 shows the development of crystalline magadiite with increasing reaction

time at in the reactant compositions of + 5 and

FIG. 1 X-ray diffraction patterns of synthetic magadiites at according to
reaction Reactant molar ratio: 5;
= 100. (a) 15 (b) 36 (c) 72 Magadiite.
+ = 100. The X-ray diffraction pattern of synthetic

magadiite exhibited 001 X-ray reflections of the film sample corresponding to a
basal spacing of The peak positions for synthetic magadiite agree well with
values reported previously for synthetic and natural magadiite. Pure
magadiite is obtained after 36 h of reaction, and then the intensity of the peak
increases as the reaction time increases from 36 to 72 h. Magadiite is synthesized
directly from an amorphous form of silica, and the pure magadiite can be synthe-
sized easily by adjusting reaction time and temperature. Figure 2 shows XRD
patterns of synthetic samples at at fixed reactant compositions of
+ = 5 and + = 100 for the different
reaction times. As reaction time increases from 20 to 72 magadiite is transformed
into kenyaite, and then kenyaite is transformed into quartz. Beneke and
suggested that at amorphous silica is transformed into magadiite and then
into kenyaite. Above the rate of kenyaite formation is much increased, and
the rate of quartz formation is increased simultaneously. After 55 pure crystal-
lized kenyaite is formed. The XRD pattern of synthetic kenyaite exhibits 001 X-ray
reflections of the film sample corresponding to a basal spacing of 2.03 nm. The peak
positions for synthetic kenyaite agree with values reported previously for
synthetic kenyaite. After 72 however, the quartz phase appears. This result
indicates that it is very difficult to obtain pure kenyaite by controlling reaction

eon
2 X-ray diffraction patterns of synthetic samples at according to

reaction time. Reactant molar ratio: + = 5;
= 100. (a) 20 h; (b) 45 (c) 50 (d) 55 (e) 72 h.
kenyaite; quartz. (From Ref. 30.)
time because pure kenyaite can be synthesized within only a short range of reaction
time.
Figure 3 shows XRD of synthetic samples obtained at in the
higher molar ratio of + = 20 and
= with reaction time. can be synthesized directly from
amorphous silica, and magadiite was not observed as a n intermediate.
The higher molar ratio accelerates the formation of kenyaite. Silica is first
dissolved in alkali solution a t elelvated temperature, and then the nucleation of
kenyaite occurs gradually. Therefore, the ratio in the solution is a very
important factor in the formation of kenyaite. Also, the higher content of silica
in the composition of starting materials may cause a large ratio in the
solution. These conditions could accelerate the nucleation of kenyaite that has a

of Synthetic
IG. 3 X-ray diffraction patterns of synthetic at according to reaction

Reactant molar ratio: + = 20;
200. (a) 96 (b) 65 (c) 45 h. (From Ref.
large ratio. However, if the ratio in the soltuion is large enough,

kenyaite is capable of forming directly from amorphous silica without the forma-
tion of magadiite as an intermediate. The higher molar ratio and the direct
synthesis without an intermediate could be taken advantage of for obtaining pure
kenyaite industrially.
Figure 4 shows scanning electron micrographs of the synthetic magadiite and
kenyaite. Na-magadiite shows a particle morphology composed of silicate layers
intergrown to form spherical nodules resembling rosettes. The shape of kenyaite is
very similar to that of magadiite.
The chemical compositions of Na-magadiite and Na-kenyaite were obtained by
T G A and EDS analyses for silica and sodium. T G A and EDS data obtained from
and are illustrated in Fig. 5 and Table 2. The weight
losses from the two clays are coincident with an associated endotherm on a DTA
scan from dehydration. The weight losses below are attributed to dehydra-
tion, while those above are presumably from water losses via condensation
of the groups to form bonds The water loss and sodium
and silica contents give an approximate unit cell composition of
which compares well with that given by Graces et
(The degree of hydration can differ due to differences in sample treatment.)

FIG. 4 Scanning electron micrographs of synthetic samples. (A) Magadiite; (B)
kenyaite.

of Synthetic Clays
Temperature
5 Thermogravimetric analyses of (a) and (b) Na-kenyaite.
Ref, 30.)
The unit cell of Na-kenyaite deduced from TGA and the analyses of
sodium and silica is which compares favorably with the
approximate composition supported by the early work of
Lagaly and Beneke
Lagaly and Beneke reported that intercalated with

oxide (DMSO) reacts with to form ordered of
ions and molecules between silicate layers. Daily and
found that air-dried H-magadiite reacts directly with octylamine without the
intercalation with DMSO. In our experiment, air-dried H-magadiite and
ite reacted directly with excess octylamine. The X-ray diffraction pattern of the
magadiite gel, as shown in Table 3, is indicative of
3.40 nm basal spacing. Since the layer thickness of is almost the same as
the basal spacing of air-dried (1.12 nm). the corresponding gallery
height can be calculated to be 2.28 nm. In the case of
2 Compositions of Synthetic Magadiite and Kenyaite

Weight percent Atomic ratio
Synthetic
sample
E D X analysis.
Source: Ref. 30.

3 Basal Spacings of Samples
Basal spacings Gallery height
Sample
Na-magadiite
H-magadiite (air-dried)
H-kenyaite (air-dried)
magadiite gel
kenyaite gel
TEOS-intercalated magadiite
TEOS-intercalated kenyaite
kenyaite gel, the gallery height is 2.15 indicating that this value is similar t o the
gallery height of H-magadiite. This result indicates that the expan-
sion of the gallery height of the layered phase is caused by the formation of
of ion and molecules between the silicate
layers.
Reactions with
The reactions of magadiite gel with T E O S afford
TEOS-intercalated magadiite with well-ordered basal spacings. This product exhi-
bits reflections correspoding to a basal spacing of 3.80 nm (gallery height, 2.68 nm).
The gallery height for the siloxane intercalated derivative is about larger
than the gallery height of observed in the
magadiite precursor gel. the case of TEOS- intercalated kenyaite, the basal
spacing of this product is (gallery height, The gallery height
for the siloxane intercalated derivative is 0.39 n m larger than the gallery height of
2.15 n m observed in magadiite precursor gel. Kwon
et reported the effects of organic solvents o n the intercalation of octylamine
into According to this result, the expansion of gallery height could be
effected not only by the solvation of intercalated octylamine but also by the steric
effects of solvated solvent molecules and solvent itself. In the
layered silicate gel reacting with excess TEOS, T E O S could be exchanged
with solvated free octylamine molecules in the interlayers a n d hydrogen-bonded
ion on the inner surface. The interlayer and structure of
TEOS could cause a n increase in gallery height.
The of intercalated TEOS by catalyst produces siloxane-intercalated deri-

vatives with well-ordered basal spacings as well as the expansion of gallery height.
The basal spacings of the products gelled by suspension, base-catalyzed
reaction, and acid-catalyzed reaction are shown in Table 4. These products exhibit

4 Basal Spacings of Gelled Products
Basal spacing Gallery height
Sample
TEOS-intercalated by 2.50
EtOH suspension
TEOS-intercalated magadiite by 4.02
base-catalyzed reaction
TEOS-intercalated by 3.42
acid-catalyzed reaction
TEOS-intercalated kenyaite by 3.09
EtOH suspension
base-catalyzed reaction
acid-catalyzed reaction
reflections corresponding to basal spacings of 2.50, 4.02 and 3.42 for magadiite
and 3.09. 4.23, and for kenyaite. The silica-pillared products gelled by
base- and acid-catalyzed reactions exhibit a large increase in the gallery height,
compared with the gallery height of the product gelled by E t O H sus-
pension. A distinctive increase in the gallery height is related to the size and struc-
ture of pillared silica, which could be associated with the amount of intercalated
TEOS, the condition (catalyst type, solvent composition, p H , etc.), and the
rate of gelation. The calcination of the intercalated derivatives a t
yields silica-intercalated products, showing a 1-3 A reduction of basal spacings
(Table 5).
by base and acid catalysts could minimize the release of TEOS outside
the layered phase during the of the intercalated TEOS, because
time is markedly reduced. The hydrolytic polycondensation of intercalated TEOS
by acid o r base catalyst could form silica clusters of a highly branched and stiff
network structure. These conditions could bring about the effective of
intercalated TEOS in the interlayer and contribute to develop the size and structure
of pillared silica clusters that bring about the the large expansion of the gallery
height of pillared kenyaite.
Adsorption- desorption isotherms of nitrogen were obtained a t Several
pore characteristics calculated from them are listed in Table 5. The specific surface
areas were calculated by using the BET equation from the adsorption isotherm
below = 0.1. The calcined silica-pillared products have high surface areas of
533-845 depending on the catalyst type. As the gallery height of the calcined
sample decreases, the specific surface area is increased owing to the development of
micropores. The total pore volume of the base-catalyzed sample is the largest

ABLE 5 Physical Properties of Silica-Pillared and Kenyaite
Specific External
Basal spacing Gallery height surface area surface area 'Total pore
volume
Silica-pillared
suspension 2.30
Base catalyst 3.92
Acid catalyst 3.33
Silica-pillared kenyaite
suspension 2.95
Base catalyst 3.99
Acid catalyst 3.95
Source: Adapted from Refs. 29 and 31

of Synthetic 27
among the three pillared samples, indicating that the average pore size of that
is the largest.
Figure 6 shows the pore size distributions of silica-pillared The sam-
ple treated with the base-catalyzed has the greatest microporosity (dia-
meter 2 and shows a sudden increase in volume near 3.6 nm with a
narrower pore size distribution. These results indicate that abrupt by base
catalyst is closely related to the formation of a more uniform pore. O n the other
hand, the sample gelled by the suspension of E t O H has a broader pore size dis-
tribution. In the case of by a n acid catalyst, decreases and
mesoporosity increases between 4.0 and nm compared with with no
catalyst. It is interesting that acid- and base-catalyzed products are very similar in
physical properties to surfactant-templated mesoporous silica Figure 7
shows the pore size distributions of silica-pillared The base-catalyzed
sample shows a sudden increase in volume near 2.2 with a
.6 Pore size distributions of the calcined silica-pillared inagadiite gelled by

suspension, base-catalyzed reaction. and acid-catalyzed reaction. (Adapted from
Ref. 3 1 .)

. 7 Pore size distributions of the calcined silica-pillared gelled by
suspension, base-catalyzed reaction. and acid-catalyzed reaction. (Adapted from Ref.
29.)
tively narrow pore size distribution. The acid-catalyzed sanple shows a narrow pore
size distribution with a sharp peak near 1.8 These values are similar to the
gallery heights of the samples. These results indicate that the pore size of base- or
acid catalyzed samples is closely related to the gallery height. The abrupt of
acid and base catalysts is closely related to the formation of uniform pores.
On the other hand, the sample gelled by the suspension of EtOH has a broader pore
size distribution and more (diameter 2.0 nm). Therefore, the
or base-catalyzed reaction not only allows for high surface area through optimiza-
tion of silica pillaring but also permits the synthesis of materials with narrow pore
size distributions.
In the suspension of and the effects of the compositions of

reactants, reaction time, and temperature on the formation of and

Na-kenyaite were investigated. Magadiite was easily synthesized after a short
reaction time of 36 h at various compositions of reactants and reaction tempera-
tures. In the higher molar ratio, kenyaite can be directly synthesized from
amorphous silica without the formation of magadiite as an intermediate. A higher
silica content in the composition of starting materials may cause a larger
ratio in the solution. These conditions could accelerate the nucleation of kenyaite.
which has a large ratio. These results indicate that the composition of reac-
tant is a n important factor in the formation of layered silicate.
The reaction of H-type layered silicate with results in a gel-like
intercalation with a basal spacing swollen by the intercalation of
neutral octylamine. The solvated gel-like derivative has a bilayered structure. The
reaction of silicate gel with T E O S yields
TEOS-intercalated layered silicate. The increase of basal spacing in the TEOS-
intercalated layered silicate may be caused by interlayer solvation and the
factor of TEOS molecules The of intercalated T E O S by acid o r base
catalyst produces siloxane-intercalated derivatives with well-ordered basal spacing
as well as a n increase in the gallery. T h e calcined samples have high specific surface
areas between 533 and depending on the catalyst type. Base- and
catalyzed samples exhibit narrow pore size distributions near 1.8 and
respectively. The pore size distributions of the calcined samples are closely related
to the gallery heights.
Finally, the acid- or base-catalyzed reaction of hydrolysis and condensation
polymerization of T E O S into a layered silicate gallery could affect the physical
properties of silica-pillared magadiite and kenyaite. The samples that were
pillared by acid- and base-catalyzed reactions show a large increase in basal spa-
cing. Also, they exhibit relatively narrow pore size distributions and show high
surface areas, depending o n the type of catalyst and layered silicate. These results
indicate that variations in the conditions of contribute to the improvement
in the physical properties of silica-pillared molecular sieves.
Bergk, Schweiger, and Porsch, Chem. Tech. (1987).

H. Werner, Beneke, and (1980).
Eugster, Science (1967).
House. and Eugster, Mineral. (1968).
P. Rooney, Jones, and Neal, Am. Mineral. (1969).
Brondley, Am. Mineral. (1969).
A. Brandt, and Bergk, Rev. Chem. Miner. (1987).
Lagaly and Beneke, Am. Mineral. (1975).
Am. Mineral. (1989).
A. and Otsuka, Chem. (1991).
Beneke and Lagaly, Mineral. (1983).
A. Fletcher and Clays Clay Minerals (1987).
A. Aufdembrink, P. I. Johnson, Kirker. and
Am. Chem. (1991).

30
Scholzen, Beneke, and Lagaly, Anorg. Chem. (1991).

Sprung, Davis, and C. Dybowski. Chem. Res.
(1990).
J. Daily and Chem. Mater. (1992).
Brinker and W. Scherer, Sol-Gel Science The Physics and Chemistry
Processing, Academic, London, 1990, pp.
A. Loebel, and Am. (1950).
M. and Solids (1980).
J. Brinker, D. Keefer, W. and S. Ashley, Solids
(1982).
P. Barrett. and P. P. J. Am. Chem. (1951).
J. and Inorg. Chem. (1988).
Graces, Rocke, Crowder, and L. Clays Clay Minerals
(1988).
Lagaly and Beneke. Am. Mineral. (1975).
0. Kwon, Jeong, H. Jin, and Lee, Colloid Interface
(1996).
D. I. Margolese, Ciesla, P. E. Gier, Sieges. Leon, P. M.
Petroff, F. and Stucky, Nature (1994).
and Pinnavaia, Nature (1994).
Chen. X. Li. and Davis, Mater. (1993).
Jeong, J. H. J. Lee, and 0. Kwon, Colloid Interface Sci.
996).
0. Kwon, Jeong, and Lee, Bull. Korean Chem.
(1995).
Jeong, 0. Kwon, Suh, Jin, and M. Lee, J. Colloid Interface
(1995).

XU Department of Chemical and Materials Engineering, University
of Alberta, Edmonton, Alberta. Canada
Cominco Research, Cominco Ltd.. Trail, British Columbia,
CH Department of Mining and Metallurgical Engineering,

University, Montreal, Quebec, Canada
I. Introduction
I I . Molecular Self-Assembly
III. Direct Silanization
IV. Silica Coatings
V. Applications
A. Self-assembled monolayers
B. Silanized monolayers
VI. Summary
References
Magnetic particles have shown great promise as a base material for engineering
carriers in separation science and technology. A potential area of application of
magnetic particles is as supports for catalysts. Magnetic particles used in these areas
are called magnetic carriers, although a distinction is sometimes made between
magnetic carriers and magnetic tags. The former applies to magnetic particles
significantly larger than the targets so that the targets coat on the particles (carriers)
and the latter refers to magnetic particles smaller than the targets so they (the tags)
deposit on the targets In this chapter, both are referred to collectively as
magnetic carriers.
The objective of magnetic carrier technology is to confer the magnetic property
to a naturally nonmagnetic target so that the target can be separated from the
stream using magnetic separators. This is particularly important for separations

2 et
in a complex multiphase system. Figure 1 summarizes schematically the processes

involved in a typical magnetic carrier technology application. To make a target
magnetic, the target has to be recognized by and attached to a magnetic carrier.
This requires the magnetic particles to be usually by a
surface or putting a specific on the particle (step I) or by in situ formation
of a magnetic core in a structured template such as vesicles, porous silica, or
polymer (step The particles prepared in this way, termed magnetic
carriers (b, Fig, are added to a biological system containing unwanted targets
such as infected cells (white circles in The selective attachment of infected cells to
the magnetic carriers is accomplished through molecular recognition, which confers
the magnetic property to the infected cells only (step By exposing the biological
system to an external magnetic field. the labeled cells are directed to a desired
location (the top of the system in this case), where they can be readily
without disturbing the system (step IV). It should be noted that the application is
not limited to biological cell separations. The targets can be toxic species of metal
ions in industrial effluents. The use of magnetic separation in combination with
magnetic carrier technology, as shown schematically in Fig. 1, isolation of
targets in a complex system feasible and attractive.
For most applications, such as biological cell separation, effluent detoxification,
ferrofluids, and the preparation of magnetic inks and magnetic recording media,
the use of fine size carriers is either essential or beneficial. In biological cell separa-
tion, for example, it is important to avoid mechanical instability sedimentation
under gravity) of magnetic particles in a biological system in addition to the "invi-
sibility" of the carriers to most biological cells. This requires the particles to be
extremely small (in the to nanometer range) to allow thermal energy
Magnet
. 1 Schematics of magnetic carrier technology in separation science and

technology.

and surface forces to prevail. Another importnt requirement for these various
applications is to avoid magnetic a phenomenon in which magnetic
particles agglomerate due to magnetic forces. With particles being in the
rometer to nanosize range, the residual magnetic field can be considered negligible,
so permanent magnetization can be avoided and magnetic flocculation minimized.
The major challenge to the application of carrier technology in various
areas, however, lies in the engineering of a robust, tailored surface functionality. A
few methods have been developed to attach functional groups to magnetic particles,
including surface precipitation of ferrihydrite in situ polymerization the
formation of magnetic cores in porous media molecular self-assembly using a
direct silanization and silica coating followed by
nization on nanosize magnetic particles. In this chapter, two surface engineering
technologies-molecular self-assembly and silanization developed in our research
laboratory are described. The particles prepared are characterized by Fourier
transform infrared spectroscopy (FTIR), X-ray photoelectron spectroscopy
electrokinetics, and vibrating sample magnetometry (VSM) in combination
with acid leaching.
Molecular (SA) is a process of spontaneous formation of an orga-

nized organic molecular (surfactant) layer on an inorganic substrate from an
organic solvent In SA, in contrast to (LB) film deposi-
tion, molecules are anchored on the surface through chemical binding between
functional groups of molecules and surface active sites, and assembled by hydro-
carbon interchain associations. As a result, the organic film deposited on the
substrate by is more stable than those produced by LB film deposition.
The method has been used widely in preparing organic films for
developing optical sensors, electronic devices, biosensors, and corrosion- and
wear-resistant coatings For example, SA has been used to deposit
surfactant on hydroxylated surfaces such as silica and aluminum oxide;
nethiolates on gold, silver, and copper; alcohol and amines on platinum; and
carboxylic and hydroxamic acids on aluminum and silver oxides In
these examples, the terminal group has been limited to either methyl
or trifloromethyl on flat substrates. Recently, SA has been used to
-
fabricate other types of terminal groups 22-26], including
and halides (X) on flat gold, silica, and aluminum oxide sur-
faces. A bolaamphiphile of having a reactive functional group at the
in position was used in these cases. The halide terminal group allows for further
derivatization to a desired functionality. Most of the fabrications of SA
layers are limited to flat substrates despite broad potential applications of SA
monolayers on nanosize and submicrometer magnetic particles.
In our laboratory, 16-mercaptohexadecanoic acid MHA]
was assembled onto nanosize superparamagnetic particles (40 nm from
either ethanol o r chloroform solutions The XPS narrow-scan spectra of the
particles with and without M H A coatings are shown in Fig. 2. For comparison, the

-732 -724 -716 -708 -535 -532 -529 -526
Binding Energy Binding Energy
, ,
-287 -283 -279 -170 -166 -162 -158
.2 Narrow-scan XPS spectra for the elements of interest on with
assembled (a) MHA and (b) acid, in comparison with (c) untreated
particles.

Superpara 5
spectra of with stearic acid are also shown in this figure.

Compared with uncoated particles, a significant increase in band intensity
suggests the deposition of (spectrum a) and stearic acid (spectrum on
In addition, a sulfur band was detected on particles treated with
MHA, further confirming the deposition of A detailed analysis revealed
two different carbon states, one at 284.6 as in a hydrocarbon chain (C- -C) and
the other at 288.3 corresponding to a carbon in a carboxylic acid environment
(COO). The area ratio of the band at high binding energy to that at low binding
energy was found to be and 15, respectively, for stearic and M H A coated on
the particles, in excellent agreement with that derived from the molecular structure.
It is interesting to note a similar band area ratio, 3.7 and 3.6 for
coated with stearic acid and MHA, respectively, suggesting that the densities of
stearic acid and M H A on are comparable.
T o determine the orientation of molecules in the a
surface-sensitive, diffuse-reflectance infrared Fourier transform spectroscopy
(DRIFTS) was used, and the spectra are shown in Fig. 3. Over the high
region, the stretching vibration bands at 2924 and 2851 are observed,
suggesting the presence of a hydrocarbon chain as anticipated. The bands
appeared at the same band positions for bulk M H A and for M H A coated on
However, the bands are sharper for the M H A on than for bulk
MHA, indicating a more ordered polymethylene chain and confirming the assembly
of a densely packed surfactant monolayer. Over the low wavenumber region, on the
other hand, the M H A on resulted in a different spectral feature than bulk
MHA. The absence of the carbonyl band at and the presence of a
carboxylate band at for the M H A on suggest not only the
anchoring of carbonyl groups on transferring a to the
late functionality, but also the absence of the second M H A layer, only a
monolayer coverage. Should the second layer be present, a mixed functionality
of carboxylate from the first layer (band at and carbonyl from the
second layer (band at would be observed. From film flotation a
critical surface tension of MHA-coated surfaces was found to be 38
with a difference between the surface tensions for complete wetting and
compared to with a difference of 5 for stearic
coated particles. These results further confirm that the terminal group of the MHA
coated on is thiol or bisulfide. rather than a carbonyl. sulfate, or
functionality, which would result in a much higher critical surface tension
Acid leaching tests in pH 2 solutions for 24 showed that no
iron was detected by atomic absorption in the leachate. It is important to note
that a coating of on only reduced the saturation mag-
netization of magnetic particles marginally (2 as determined by VSM.
The particles remained after coating, as anticipated. The char-
acteristics of coatings on are summarized in Table 1. It is clear that the
strong binding of the carboxyl groups of M H A with iron sites on com-
pounded with the interchain association, allows M H A to be
producing particles with the desired character-
istics for the biological and environmental applications.

acteristics of Nanosize Superparamagnetic Particles Coated with
Stearic Acid and 16-Mercaptohexadecanoic Acid Using Self-Assembly from
Chloroform
0.2 60.3 72.8 52.7

MHA-coated 3.6 nil 38 50.6
Stearic acid-coated 3.7 nil 30 50.6
The SA method using bolaamphiphiles for preparing magnetic carriers appears

attractive due to the stability of the monolayer and the high density of surface
functional groups. control of the reactivity between the two functional
groups and the substrate is crucial in determining the quality of the coatings. In the
ideal case, one functional group anchors chemically to the surface while the other is
relatively unreactive to the surface, which ensures a unique surface functionality.
This requirement limits the types of functional groups that can be directly intro-
duced onto the magnetic particles. A more general approach is to use a silane
coupling agent of a general structure Y Si where X represents
an alkoxy or halide group and Y an organic functional group. including amine,
carboxylic, phosphate, vinyl, cyanide, or methacrylate. The strong reactivity
between Si and surface metal hydroxyl groups causes silane coupling agents to
condense readily on metal hydroxide surfaces, forming Si- 0- Mbonds upon
dehydration, whereas the free groups polymerize laterally, a two-
dimensional network. It is therefore clear that silanization is just a variation of SA,
in that the interchain association that occurs in SA is replaced by lateral
ization between condensed silane coupling agents through siloxane bonds.
was initially introduced to modify silica surfaces for use in synthetic
transformation preconcentration of trace metals immobilization of arti-
ficial membranes molecular recognition chromatographic separations
and preparation of It has since been extended to appli-
cations such as the protection of photocatalysts and the promotion of adhesion
between different materials such as polymers, metals, glass fibers. and polymeric
resins A large volume of literature is available on silanization using silane
coupling agents to tailor the surface properties of oxides Whitehead
et al. used direct silanization in functionalizing paramagnetic particles. The
functional groups they attempted include amino,
aliphatic, and mixed functionalities. We examined the use of silanization in
engineering superparamagnetic carriers For the purpose of illustra-
tion, the silanization of using 3-aminopropyl triethoxysilane (APTES) is
presented below.

It is recognized that the coating of silane coupling agents by silanization is a
multistep process The controlled hydrolysis of both surface and coupling agent
is essential for successful coatings. The presence of nitrogen at 399.4 and
silicon at XPS bands seen in Fig. 4 suggests that APTES can be
deposited on from both water and toluene solutions. Fig. 4, two
different oxygen and nitrogen environments are seen. The bands centered at 399.4
and 401.3 are attributed to nitrogen in protonated and nonprotonated amines,
respectively; those a t 531.8 and are assigned to oxygen from APTES film
and particle matrix. respectively. Band fit analysis revealed that the pro-
portion of oxygen in the Si - 0 environment is higher for the film silanized in water
(24%) than in toluene suggesting a higher degree of APTES deposition on
from water. Semiquantitative analysis of the XPS band area shows that a
larger fraction of amines are protonated in the films silanized water (26%)
than in those silanized from toluene (17%). The higher APTES deposition is
Narrow-scan XPS spectra for the elements of interest on with silanized

APTES (a) from water and from toluene. in comparison with untreated
particles.

attributed to a higher degree of hydrolysis of APTES, while the higher fraction of
amine protonation is related to the stronger interactions of amine groups with
in water than in Also interesting to note is that the APTES forms
a monolayer on from both solvents
Although the XPS analysis confirmed the deposition of APTES on the
technique is unable to determine the degree of hydrolyzation of ethoxy groups and
subsequent cross-linking, which have a significant impact on the density and sta-
bility of the film. DRIFTS spectra were therefore collected. In contrast to
in water, DRIFTS spectra of the APTES film deposited from toluene solutions
showed vibrational bands of methyl groups at 2974 and indicating the
presence of small fraction of unhydrolyzed ethoxy groups in the film. The
drolyzed ethoxy groups are anticipated to inhibit the lateral polymerization of
APTES, resulting in a poorer packing density and lower surface coverage than
those of the films silanized from water in which the hydrolysis is more effective.
This observation is consistent with the findings from XPS analysis, which showed a
lower APTES coverage on toluene.
T o further confirm the orientation of APTES on the electrokinetics
(zeta potentials) of the coated particles were measured. As shown in Fig. 5, an
point the point where zeta potential is zero, of the particles
silanized with APTES from water and toluene was found at about pH 8.5 and
A
in water
in toluene
.5
Zeta potential of without silanization and with silanization by
APTES in water and in toluene.

9.2, respectively, in contrast to pH for uncoated The measured
electrokinetics of the silanized particles resembles that of air bubbles in
(a solutions, in which case the amine groups were
exposed to water. This similarity confirms the condensation of APTES on
from both water and toluene with amine groups facing the environment
and remaining reactive. The difference in the measured zeta potentials using the
particles silanized in water and toluene reflects mainly the variations of APTES
orientation on Since APTES was deposited on from water
than from toluene, a slightly lower iep value observed for the particles silanized in
water suggests that more amine groups were hidden inside the film compared to the
silanization in toluene. As a result, the contribution from amine groups to the
number of positive surface sites decreases while that from groups to the
number of negative sites increases. These two effects result in a lower iep as
experimentally observed, consistent with XPS analysis, which showed a higher
degree of protonation of amine groups considered to bind with the surface
The stability of the silanized APTES films on was investigated by
leaching the particles in p H 2 solutions for 20 It was found that the
of iron leached out reduced from for unsilanized to for
silanized Compared to the particles coated with using the SA
method, the amount of iron leached out is significant, suggesting that the original
particles were not fully covered, compounded with some degree of detachment of
APTES from the surface in acid solutions. As there was only a slight shift in iep
p H 9.2 to 8.7 upon leaching of the particles silanized from toluene, it is
concluded that the surface was not fully covered by APTES as expected
for insufficient lateral polymerization due to unhydrolyzed ethoxy groups. For
particles silanized water, a significant shift of iep from 8.6 to 6.2 was
observed, although the of APTES deposited was greater than with
in toluene. It appears that the protonated APTES, presumably attached
to the surface through hydrogen bonding was not stable, detaching readily in an
acidic environment. After exposure to an alkaline environment, on the other
hand, the electrokinetics of silanized approached that of unsilanized
suggesting a much more significant detachment of APTES from
Whether the detachment of APTES is due to the breakage of S i - 0 of F e - 0
bonds remains to be further investigated. The appears to occur at the
F e - 0linkage to account for the observed electrokinetics. (If the breakage were
at the S i - 0linkage, the attacked surface should have resembled a silica surface
with an iep around pH 2.) The comparative information for the direct silanization
is summarized in Table 2.
Clearly, the direct silanization of APTES from water and toluene solutions on
described above is not suitable for engineering magnetic particles in biolo-
gical and environmental applications. It is well documented that the Si- 0bond is
more stable than bonds in acidic solutions. A well-known example is that
silica is insoluble but iron oxides dissolve in strong acids. One of the alter-
natives, therefore, is to coat a surface with a thin layer of silica, followed
by the well-established silanization on silica, to improve the coating density and
film stability.

2 Coating Characteristics and Stability (in Acid) of APTES Films Silanized on from Water and Toluene Solutions
XPS
DRIFTS Fe in
Before leaching leaching
60.3 4.5 --
in water 0.32 2.85 no 34 8.6 6.2

in toluene 0.16 4.88 yes 34.1 9.2 8.7

Silica coating been used on of carbon, steel, alumina, and polymer
resins to proinote adhesion, minimize photodegradation, or prevent material oxi-
dation and corrosion. In view of the strong hydration of silica surfaces in aqueous
solutions and hence high colloidal stability, silica coating also been used to
prepare stable ferrofluids. Silica-coated particles have been for the synthesis of
conductive polymer Dense liquid sol-gel processes are the
methods used to coat silica films on finely dispersed particles. Silica
coating using a dense liquid process is a complicated process
achieved by controlling the supersaturation level of silica in aqueous solutions.
At least three competitive processes, homogeneous nucleation followed by surface
growing and heterogeneous and coating, are present simultaneously.
In general. homogeneous coating requires the lowest level, fol-
lowed by heterogeneous coating and finally
requires excess energy as predicted by the Kelvin equation to account for the
extremely curvature of nuclei. Although homogeneous nucleation can be
avoided by carefully the level just above the critical
concentration of heterogeneous coating, as desired, the surface coat-
ing often presents a challenge to uniform surface coatings. It is evident that soon
as substrate is coated with silica, the growth of coated area, a process similar
to coating, prevails as it requires a level,
resulting in a (islands) coating as shown in
Fig.
Silica coating by the sol-gel process, on hand, is an
organic solvent. The process is based on hydrolysis of the precursors such as
tetraethoxysilane (TEQS) in the presence of water and by con-
densation surface metal hydroxyls. With controlled hydrolysis of TEQS, an
linkage is established between surface metal (M) and
TEQS, followed by lateral polyinerization and finally the formation of a three-
dimensional network via the of bonds
increasing TEQS concentration and hydrolysis, A drawback to the
sol-gel process for engineering magnetic carriers is the porous nature of the silica
chematics of silica films coated on by (a) sol-gel, (b) dense liquid,

and (c) two-step a followed by b) processes.

film as shown schematically in Fig. although this could be advantageous for
certain applications.
A compromise between two methods is a novel two-step coating process
developed in our laboratory This approach is based on the idea that the sol-gel
process coat a uniformly, the is often porous, as in
Fig. In the second step, using dense liquid process, the residual ethoxy
groups in silica films prepared using sol-gel process are
and the pores are anticipated to be closed and filled silica under
saturation conditions. It is clear that the two-step silica coating process integrates
the advantages of uniform coatings by sol-gel process and the
levels required for homogeneous coating by the dense process. As a
result, a uniform thin silica layer as shown in Fig. can be coated onto
to protect the particles with a reduction saturation magnetization
and to provide a surface for further by
For the silica coated on at the 1 (by weight) silica
level using these three Narrow-scan XPS spectra of the coated particles
are shown in Fig. 7. The presence of a silicon band and additional
oxygen band (01) at 532.8 on the spectra of coated particles confirms the coat-
ing of silica on Semiquantitative analysis showed an area ratio of silicon
to iron bands of 0.7, 4.1, and 4.1 for particles coated using the dense liquid,
sol-gel, and two-step processes, respectively, A higher ratio indicates a higher
two
sol-gel
DL
-532 -530 -526 -730 -720 -710 -700
Narrow-scan spectra for the elements of interest on silica-coated

particles.

degree of silica coating. It is evident that the silica coating efficiency is comparable
for the sol-gel and two-step processes, higher than that of the dense liquid process.
The DRIFTS spectra showed formation of bonds in all three processes.
In contrast to the dense liquid or two-step coating process, the
ethoxy group was detected for the particles coated the sol-gel process. The
presence of terminal groups is partially responsible for the of the
films coated with the sol-gel process.
Leaching tests conducted to further the state of silica film on
The coated particles dispersed in an acid solution of 2.5 for 20
and tlie amount of iron in the was analyzed. The results are summarized in
Table 3. Also shown in table is the saturation magnetization of coated particles.
It is evident that the amount of iron leached out became undetectable for the
particles coated the two-step process in contrast to tlie single-step, sol-gel,
dense liquid process, in which case, 1.1 or 2.8 mg of Fe per gram of particles
detected. However, the saturation is comparable for particles coated
with silica using sol-gel and two-step processes, suggesting that the amount of
silica on the surface is virtually the and confirming the important role of the
film structure in protecting the matrix component of magnetic carriers biolog-
ical and applications. It is evident that the two-step silica coating
process is successful making base materials that be further by
the process to produce carriers of desired
functionality.
The M H A coated prepared by the method contain reactive

groups that are known to have strong affinity to
metal such as gold, silver, and copper. The engineered carriers of large surface
area can therefore be used to capture silver and copper industrial
effluents. The metal-loaded particles can be readily separated from the effluent
3 Stability as Measured by of Iron Leached in Acid (pH 2.5)

Saturation Magnetization of Silica-Coated by Sol-Gel, and
Two-step Coating Processes at 11
Coating method
60.3
Dense liquid 2.8
Sol-gel 1.1
Two-step nil
nil

by using magnetic separation methods. application is illustrated in our loading
tests of Cu on MHA-coated particles from or
solutions, The narrow-scan spectra of the loaded particles separated a
hand magnet are in Fig. The load of on the
particles is evident from the presence of (934 and and (368 and
374 on spectra c and respectively. The area ratio of the copper
satellite band to its band is than expected, some of the
copper ions are in the cuprous state. It appears that of ions
reduced to cuprous ions, accompanied by the oxidation of thiol to
which accounts for the presence of an additional sulfur band of higher binding
energy but intensity, The metal-to-sulfur atomic ratio found to be
0.6 and 0.8 for copper and silver, respectively. These results suggest a
sulfur) binding for copper and binding for monovalent silver. It is clear
that the loading efficiency is sufficient for MHA-coated magnetic carriers
to be used in the or recovery of Ag and Cu from industrial effluents. For
(a)
-955 -950 -945 -940 -935 -930 -925 -375 -370 -365
Energy Binding Energy
Narrow-scan XPS spectra for
8 elements of interest on with various
treatments: (a) Non; (b) MHA coating; MHA coating by loading; and
(d) MHA coating by Ag

other metal ions, the thio or disulfide group can be further oxidized to obtain the
sulfonate or functionality
Magnetic particles with reactive groups show special affi-
nity to monoclonal antibodies containing in the basic unit
Therefore, the magnetic carriers can be readily sensitized with an of specific
such as recognizes glycophorin, one kind of
protein (M and blood group antigens) in red blood cell membranes. When
carriers are introduced into a biological system, the cells containing glycophorin are
captured and separated along the carriers the system by magnetic separa-
tion methods as shown in Fig. A principal advantage of using magnetic carriers
in cell separation applications is that there is no need to destory the target in
in biological systems. As a result, the side effects of conventional chemo treat-
ments can be minimized.
It has been found that reactive amine groups are capable of attracting and adsorb-
ing such as arsenate, dichromate, tungstate, and
at the microgram per liter concentration level These species can then
be effectively from an effluent by using magnetic separation magnetic
carriers to detoxify the effluent. magnetic carriers with reactive amine groups
are found to confer specific adsorption for heavy metal ions, including
mercury, lead, copper, zinc, manganese, and monovalent silver in contaminated
aqueous and nonaqueous solutions Copper loading the silanized magnetic
particles of reactive groups is shown Fig. 9. For comparison, copper
loading on silica-coated magnetic particles is also shown in this figure. It is evident
that at a pH copper an be removed effectively from concentra-
tion solutions 100% removal from a With increasing initial
copper ion concentrations, the copper removal efficiency decreases as copper load-
ing approaches the capacity limit of about of per gram of particles. In
contrast, the copper loading on the silica-coated magnetic particles is significantly
lower than on the silanized particles, suggesting the important role of reactive
amine groups in application. It should be noted that the metal ion loading is
pH-dependent. a recent communication we demonstrated the selective
loading of various metal ions such as copper zinc by varying the solution
at which loading is performed.
It is evident that the of magnetic carriers in separation science and technol-
ogy, whether in the area of biological cell or industrial effluent detox-
ification, offers great flexibility. Other areas of potential applications of magnetic
carriers with reactive amine groups include immobilization of enzymes, drug deliv-
ery, and biological sensors, among others, A detailed review of these applications
can be found elsewhere The application can be extended further to other areas
as long as a key-lock relationship (see Fig. 10) can be developed. The lock varies
from antigen or as in biological applications to metals or toxic species as
in environmental applications, the key could be an antibody or specific
The important consideration for these applications is to further
improve the stability of coatings, which significantly reduce the operating

100
90
0.15
magnetic carriers
80
70
0.10
60
a
50
0.05
silica coated
30
0.00
0.0 0.1 0.2 0.3 0.4
Equilibrium concentration of
FIG. 9 Copper loading on left-hand axis) and removal by right-hand axis)
APTES on silica-coated particles in comparison to silica-coated
particles left-hand axis) (loading at pH 5.3).
antibody biological cell

biotin antigen
ligand enzyme
particles
metal ions
toxic species
Key-lock relation in potential applications of magnetic carriers.

cost by recycling magnetic carriers. The use of a silane coupling agent of longer
hydrocarbon chain 14) is anticipated to increase the film stability due to a
stronger intechain association, a close-packed array through molecular
self-assembly o n substrate particles, magnetic carriers are expected to rapidly
expand into other areas of application in the near future.
Three basic requirements for magnetic carriers in separation technology are fine
particle size, and chemical stability in addition to tailored
functionality. Two different approaches for engineering nanosize
carriers, molecular self-assembly and silica coating followed by
have been described. Using molecular self-assembly, a bolaamphiphile of two dif-
ferent functional groups (16-mercaptohexadecanoic acid) was found to form a
packed and stable o n nanosize producing a surface of reactive
thiol and/or functionality. Following a novel two-step silica coating pro-
cess, reactive amino groups were successfully on particles.
Examples have been given for magnetic carriers of tailored functionalities in a
variety of applications, including industrial effluent detoxification, recovery,
cell separation, drug delivery, and the preparation of biological sensors.
Moffat, A. Williams, Webb. and Stirling, Miner, (1994).

de Cuyper and (1991).
Mann, J. P. and J. P. Williams, Nature (1986); Mann
and P. Hannington. Colloid Interface (1988).
J. Fendler, Opinion Colloid Interface (1996).
M. John, and Herman.
Mater. (1996).
Anderson. and Res, WPCF (1991).
(1982).
Ellingsen, A. and U.S. Patent 4,654,267
1987).
Liu and Xu. (1996).
A. Whitehead, Chagnon, E. V. and Lee, Patent
4,695,393 (1987).
Xu, Liu, and A. Finch, Surf. (1997).
A. An Introduction to Organic Films to
Self-Assembly, Academic, San 199 1.
A. Netzer, and Sagiv. Thin Solid Films
(1985); Netzer, and Thin Solid Films (1983).
L. Allara and R. Nuzzo, Langmuir (1985).
Schlotter, Porter. B. Bright, and D. L. Lett,
(1986).

P. Laibinis, J. Hickinan, Wrighton, and Whitesides, Science.
(1989).
Troughton, Whitesides. and
J. Am. Chem. (1989).
M. C. Stole, A. and M. D . Porter, Am.
(1991).
A. Parikh, L. I. Azouz, and F. Rondelez.
(1994).
and A. Guzonas, Colloids Surf. (1994).
P. L. P. Laibinis, Whitesides, and R .
Nuzzo, Langmuir 13 (1995).
A, Nuzzo, and D.
(1983).
A. Ihs and Liedberg. Colloid Interface (1991); K. Uvdal,
Bodo, and Liedberg, Colloid Interface Sci. (1992).
A. Goss, Charych. and Majda, Anal. (1991).
smith, A. Andergg. D. Porter, and Siperko.
Langmuir (1992).
Y. Okahata, Ito. and Ebara, (1996).
Martinetti, and P. Tetrahedron Lett.
(1988).
Leyden and Anal. (1975).
X. Nichols, Invergo, and
Anal. Chem. 63:1985 (1
M. R. Stone, and Thompson,
Anal. (1990).
and Lodkowski, Liquid (1991).
M. Dicks. E. and I. Anal. Chem.
(1987).
and Pure (1991).
Matteucci and Am. (1981).
A. V. and Adhes. (1988).
Ondrus, Boerio, and K. Grannen. Adhes. (1989).
Poly. Compos. (1984).
Lund and P. Murphy, Adhes. (1992).
Scott, Ishida, and J. Maurer, Reinf. (1991).
P. Plueddemann, Coupling Agents, Plenum, New York, 1982.
Leyden Surfaces and Interfaces, Gordon and Breach. New
York. 1985.
Mittal and Other Utrecht, The Netherlands,
1992,
Dwight, A. Cole, and Huang. Solids
(1990).
D. S. A. Bell, P. Armes. and A. Colloid
Interface (1996).
Liu, An innovative approach in magnetic carrier technology, Thesis,
University, Montreal, 1996.

46. and C. (1995).
47. Thomas, Abraham. A. Otter. and
Lansdorp, J. Methods. (1992).
48. D. E Leyden, W. and Anal.
(1976).
49. Xu, and A. Finch., in Composites at Lake Louise Nicholson,
Trans. Tech Publications. Toronto, 1998.
50. A. Colloid and Surface Engineering: Applications in the Process
Butterworth Heinemann, Oxford, UK, 1992.

Department of Chemical
Engineering and Materials Science, Syracuse University, Syracuse,
New York
I. Introduction 52
I I . Acid-Base Properties of Solid Surfaces 53
A. General definitions 53
B. Extension to solid surfaces 56
C. The acid-base behavior of species confined in porous solids 58
111. Origin of Acid-Base Properties of Solid Surfaces 59
A. Ideally polarizable surfaces: metals 60
B. Open surfaces: oxides 61
C. Surfaces with variable composition: composite oxides 71
D. Layered materials: clays 75
E. Porous materials: zeolites and zeotypes 76
F. Microporous materials: activated carbons 78
G. Pseudo-liquid materials: heteropolyacids and salts 80
IV. Measuring Acid-Base Properties of Solid Surfaces 81
A. Nature, number, and strength of acid-base surface sites 81
B. Experimental methods 81
C . Theoretical concepts 88
V. Manifestation of Acid-Base Properties of Microporous Solids 91
A. Solid/gas interface: the role of acid-base sites in heterogeneous
catalysis 91
B. Solid/liquid interface: acid-base reactions in interlayers of
layered minerals 93
VI. Summary and Conclusions 94
References 95

The concepts of acidity and basicity are used by scientists from all branches of
theoretical, experimental, and engineering chemistry. They are fundamental in nat-
ure and are of unquestionable utility as a powerful classification tool. However, it is
interesting to observe that the meaning of these two simple words may vary depend-
ing on the compounds or materials discussed, experimental conditions or methods
of study. particular applications, state of matter or size of particles, and, in some
instances, even the context in which the terms are used. Indeed, any discussion about
acid-base properties is an assertion in comparative terms: a debate about relativity,
By definition, acidity and basicity are opposite properties (or principles) that man-
ifest themselves only in relationship to each other. Acids are revealed and measured
by bases; the presence of bases is revealed and quantified by acids.
A simple, rarely mentioned rule requires that acids and bases to be compared be
confined to a single, homogeneous phase, either liquid or gaseous. When this rule is
adopted, there are numerous methods in chemistry and physics adequate for the
identification and evaluation of the quality and quantity of either an acid or a base.
As measurable properties, acidity and basicity are intrinsically connected. Indeed,
all methods of quantitative characterization of acidity or basicity give only relative
values; the strengths of acids measured in aqueous solutions differ from those mea-
sured in nonaqueous solvents, and the latter vary from one solvent to another. Two
scales of acidity were selected for practical purposes, the aqueous scale and a scale
based on gas-phase measurements, but they d o not coincide. Both scales provide
valuable theoretical insights into the molecular origin of acid-base properties. In
fact, due to the inherent relativity of the measurements, any discussion of acid-base
chemistry depends on the standards used or is specific to experimental conditions.
Fortunately this fact is very well understood. As a result, the whole edifice of pure
and applied chemistry was built upon the simple rule mentioned above: Acids and
bases must be confined to the same homogeneous phase if they are to be compared.
As long as this convention is observed, the reliability of analytical methods in acid-
base chen~istryor the meaning of their results is unquestionable.
This review focuses on the acid-base properties of surfaces of porous solids. In
the context of the above discussion, it is inevitable that established practices will
require some modifications. It is obvious that solids possess acidity and basicity.
The challenge in characterizing their acid-base behavior results from the presence
of two phases and from location of the acid-base sites at the interface between the
solid and either a gaseous or liquid phase. Moreover, when acid-base chemistry
occurs in spaces confined to the micropores or interlayers of nanostructured mate-
rials, the rules are broken a second time because all references to acid-base proper-
ties of macroscopically honlogeneous phases based on the classical approach
become inconsistent.
The surface chemistry of acid-base interactions in the nanospaces of solids
comparable in size to the n~oleculardimensions of the reactants themselves is a
new and fascinating subject. It has been approached from various angles; attempts
were made to extend the concepts of classical chemistry, but new rules had to be
developed. The consequence is that there is still no consensus regarding the sig-
nificance of experimental results and their relationship to the theoretical notions.

An understanding of the acid-base properties of solid surfaces is extremely
important in almost all fields of science and engineering. For example, the solid/
gas interface intervenes when solid acids (or bases) are used as heterogeneous
catalysts in the oil refining and chemical industries. The solid/liquid interface
between rocks, minerals, and soils and natural waters is continuously transformed
by weathering processes that result in the geohydrochemical circuit of elements.
These are only two examples in which the primary reactivity factor is determined by
the acid-base chemistry at solid interfaces, usually confined to molecule-sized
spaces.
In the discussion that follows we attempt to organize the vast amount of infor-
mation available on the acid-base behavior of porous solids. In Sec. I definitions
from homogeneous systems are recalled and adapted to solid surfaces. Since the
focus of this review is the molecular scale interactions and structure-determined
acid-base properties at solid surfaces, the origin of these properties is discussed in
Sec. I I in relation to various types of solid interfaces. The discussion starts with the
flat surfaces of clean metals, but most references are made to environ-
ments in porous and microporous solids in contact with a gaseous or aqueous
phase. Section III deals with the experimental methods currently in use for char-
acterization of surface acidity. The reader may discover contradictions and con-
flicting arguments in the subsection dedicated to the theoretical problem of finding
a common scale for the acidity of solids. This unavoidable dilemma reflects the
"confined" state of current understanding in this "heterogeneous" field and the
inherently relative nature of most statements about acid-base properties.
Bronsted-Lowry
When formulated the classical views on acids and bases in 1887, his
theory dealt exclusively with reactions in aqueous media He defined as acids
all substances that upon reaction with water increase the concentration of
ion in solution; those that increase the hydroxyl ion ( O H ) concen-
tration were defined as bases.
In 1923 Bronsted [2] and [3] independently generalized these definitions
and emphasized the crucial role of protons. Acids were defined as substances able
to donate protons and bases as substances able to accept protons. Water was
longer required as a solvent, and bases no longer needed to donate
The following equilibrium defines the pairs of conjugate acids and bases and is
fundamental to the Bronsted-Lowry theory:
Acid A and its conjugate base B are two different but interrelated states in which a
chemical system may exist. The acid is the proton-rich (and proton donor) state,
while the base is the proton-deficient (and proton acceptor) state.

Because free protons do not exist in solution, above scheme is only a sim-
plification. A complete acid-base reaction actually takes place when two systems,
let them be Systems 1 and 2, of conjugate acid-base pairs interact with each other.
If most of the species of system 1 are in the proton-rich (acidic) state, their acidity
can be manifested only if most of the species in system 2 are in the proton-deficient
(basic) state. During the course of their interaction, a proton is transferred from the
acidic species of system 1, to the basic species of system 2,
The reaction in the forward direction leads to the formation of a base in system 1
the conjugate base of and an acid in system 2 the conjugate acid of
An important consequence is that the position of equilibrium depends on the
relative acidity basicity of conjugate acid-base pairs in the two systems,
and 2.
2. Lewis
In a further generalization, Lewis advanced a definition that was no longer
restricted to proton-exchange reactions. An acid was defined as any substance
that can accept a pair of electrons from a donor substance, and a base as any
substance able to donate an electron pair and form a "dative" covalent bond.
these more general the acid-base interaction includes the formation of cova-
lent bonds and applies to any chemical reaction in which an addition compound
is formed through a coordinative bond:
According to the above definition, both hydrogen bonding and bonding are
base interactions.
Between Lewis acid-base are possible displacement reactions that for-
mally resemble the reaction from the Bronsted-Lowry theory:
Thus the acids of the Bronsted-Lowry theory may be viewed as Lewis acid-base
in which the acidic component is (for example, can be viewed as
an The Bronsted-Lowry bases are also Lewis bases. One can
therefore conclude that the Lewis definition subsumes the definition
of acids and bases.
3. Strength of Acids and Bases

In the Bronsted-Lowry theory all substances are considered, to some extent,
ainphoteric. Theoretically, the strength of individual acids can be expressed
through the equilibrium constant of process (I), called the "acidity constant";
its reciprocal is termed the "basicity constant," Since neither acidity nor basi-
city of a (system l ) manifest themselves without a proton acceptor or a
proton donor (system 2), the individual constants and cannot be measured
directly. One can, however, measure the equilibrium constant of the
reaction

It follows that the acidity of an acid can be measured only with reference to a
base. If a common base is chosen, a relative acidity scale for several acids can be
established. Water is a very common solvent. A scale of values
= log very popular based on the convention that the common
base is water. The scale is largely used as an acceptable and reproducible
means for ranking the strength of acids (or of their conjugate bases) in dilute
aqueous solutions. In water there cannot exist acids stronger that
= -1.7) or bases stronger than O H = 15.7). In nonaqueous
solvents the acidity scales are different. Basic solvents enhance the
strength of acids, and acidic solvents (protogenic) raise the strength of bases; thus
the strength of acids (or bases) is leveled in highly basic or highly acidic solvents.
Alkanes are inert solvents that d o no accept or furnish protons; water is an
solvent that either accepts or donates protons.
The strength of Lewis acid-base interactions cannot be expressed in terms simi-
lar to the acidity and basicity constants. and of the theory.
Consequently an equilibrium constant resembling the protolytic constant of
Bronsted acid-base couples, cannot be specified. Because of the broad
variety of Lewis acid-base interactions there would be as many acid strength scales
as there are interacting bases.
However, the strength of Lewis acid-base interaction can be expressed in energy
terms, such as the exothermic molar heat, for the equilibrium of
formation. The enthalpy term is preferred because entropy effects accom-
panying the formation of coordinative bonds are difficult to determine. Various
models have been proposed for the theoretical estimation of the enthalpy term
based on molecular properties of reactants and are reviewed in Ref 5 .
The most significant developments have been the hard and soft acid-base principle
of the equation of Drago and the donor and
acceptor numbers of and the perturbation theory of Hudson and
The heats of formation with many Lewis acids have been measured
experimentally in the gas phase using the technique of ion cyclotron resonance
The absolute proton affinity (PA) of a gaseous base molecule has a precise
thermodynamic definition in relation to the negative of the enthalpy variation for a
hypothetical reaction of attachment of an isolated proton to a molecule M in the
gas phase,
in the absence of any interactions with the The corresponding free energy
variation is referred to as the gas-phase basicity. Many compilations of various
experimental results have been made I], and several useful correlations for ratio-
nalizing the acid-base chemistry have been revealed. Quantum-mechanical calcula-
tions for the enthalpy of in the gas phase generally give
overestimated values

5 and
The influence o f a solvent complicates any understanding o f acid-base interac-

tions in solutions. The calculation o f aqueous acidity constants would require the
calculation o f the free enthalpy o f solvation o f the acidic site, conjugate base, and
proton, in addition to the gas-phase term for the change o f free enthalpy during
deprotonation and, o f course, all entropy terms. While the entropy o f
is nearly independent o f the acid-base couple, the free enthalpy o f solvation
can be obtained only with difficulty from molecular dynamics calculations
Although the importance o f solvation effects for liquid-phase acid-base chem-
istry is evident, the experimental data are interpreted with difficulty.Thus. in aqu-
eous solutions where the reference acid is ammonia is a stronger base than
triethylphosphine, while the opposite is true when the reference Lewis acid is
Examples o f reversal in acid-base strength were also found for
phase formation, where interactions should be free o f any solvent effect
Several factors are responsible for these inversions, such as the ease o f forma-
tion o f hydrogen bonds in aqueous solutions or the occurrence o f favor-
ably oriented orbitals that the mutual overlap for gas-phase
intermolecular interactions.
to Solid
Bronsted and Lewis acids and bases are frequently encountered in the chemistry o f
solid surfaces, but in some instances these terms are used in a slightly different
context from their original proposals. From the point o f view o f the Bronsted
definition, a solid is acidic i f it is able to donate (or at least transfer partially) a
proton to a basic probe molecule to which the proton bonds (or with which it, at
least, becomes partially associated). According to the Lewis definition, a solid is
acidic i f it is able to accept an electron pair a basic molecule and also a
coordinative bond with it. The transfer o f an electron pair in the meaning o f the
Lewis definition should be clearly distinguished from surface processes in
which electron transfer ( o f one or two electrons) occurs without coordination.
Extension o f the above definitions to acid-base processes on solid surfaces is
further complicated by the "in between two phases" location o f the acid-base sites.
For example, little is known about acid-base interactions at inter-
faces. It has long been recognized that molecular interactions across the interface
between condensed phases may be split into physical and chemical terms. Physical
interactions (van der forces)contribute to the non-ideality o f fluids and have
been traditionally considered for interfaces. However, modern theories explain
interfacial phenomena on solid surfaces, such as adhesion or wetting, in terms o f
chemical interactions. Moreover, the Lewis definition o f acidity is so comprehen-
sive that it can easily be accepted that most chemical interactions at solid surfaces
may be effectivelydescribed as acid-base interactions A continuously growing
literature reinterprets interactions at interfaces in o f acid-base
properties. For example, their role was shown in relation to adhesion
in film-substrate or fiber-matrix systems as well as in wood and paper processing,

When solids interact with liquids and gases, the range of chemical interactions
that can possibly occur between a solid surface and molecules of the fluid phase is
very broad. For the interface and excluding from discussion all inter-
actions that may be destructive for the solid, one can envisage local interactions
that range from hydrogen bonding (such as between hydrogen atoms of chloroform
and carbonyl groups of a solid polyester), to coordinative bonding (such as between
nitrogen atoms of pyridine and electron acceptor sites that act as Lewis acids on the
solid) to chemical interactions with exchange of or O H on hydrated sur-
faces of oxides, mineral clays, or humic acids. The surface of activated carbon
contains a variety of oxygen-containing with either acidic or basic
(Bronsted) character, and even the basal planes of graphene layers in activated
carbons show properties of a Lewis base due to delocalized electrons
Moreover, for solids virtually free of functional groups, such as polyethylene,
adsorbed impurities are almost always present on the surface and may readily
react through specific chemical interactions. It is therefore warranted to consider
that nearly all chemical interactions between solid surfaces and liquids, many of
them resposible for wetting the solid, are acid-base reactions
The interface was traditionally studied with respect to adsorption and
catalysis. Here the assertion that the Bronsted definition of acidity is a particular
case of the Lewis definition is neither obvious nor even helpful. It suffices to say
that many reactions in heterogeneous catalysis require specifically the presence of
either Bronsted or Lewis acidic (or basic) sites, and the reaction mechanisms
depend on the nature of the surface site. A long-term goal of surface studies for
the characterization of solid catalysts was to distinguish and quantify the
of Bronsted or Lewis sites with potential catalytic activity for gas-phase reactants.
For that reason, when discussing the acid-base behavior of solid surfaces, it is no
longer possible, nor desirable, to adopt the viewpoint that subsumes Bronsted acid-
base properties in the more general Lewis definition.
Similar to homogeneous one-phase systems, the acidity of solid surfaces man-
ifests itself only in relation to a base that must be present; the strength of the acid-
base interaction therefore depends on the particular base involved. This is true for
both types of acid sites, Bronsted or Lewis. The reverse can be said, of course, for
surface basic sites.
The special nature of the solid interface raises an additional practical problem:
how to detect, in practice, the presence of acidity on solid surfaces. In general, any
solid able to change the colour of a basic indicator, or on which a basic probe is
chemically bound, should be regarded as a solid acid. Unfortunately, this very
simple definition introduces, from the very beginning, a degree of uncertainty.
How should those solid acids be classified that change the color of one indicator,
I, but are indifferent to another indicator, I'? How should those solid acids be
considered that strongly a base, B, but interact weakly with another
base, B'? Clearly, an important property of acids, their strength, is not uniquely
recognized and defined. The lack of an acceptable scale of solid acidity comparable
to the scale for aqueous solutions or to proton affinities for gas-phase acid-
base reactions is a primary problem that obscures the quantification of acidity on
solid surfaces.

Another factor that makes difficult the assessment of acid-base properties on
solid surfaces results the intrinsic property of interfaces to serve as accumula-
tion zones for adsorbed species. Returning to the above examples, one cannot take
for granted that the change of color observed for indicator I on the surface of a
solid acid is solely related to its acidity (in terms of classical definitions); the colour
change could be attributed to side effects such as electrostatic polarization of the
indicator induced by the local electric field on that particular surface. Also, if
adsorption of base B' is weaker than that of base this may be the result of
other surface-specific factors such as the presence of contaminants or steric hin-
drance of B' access to specific binding sites on the solid adsorbent. The latter
effect is most evident for solids, where the number and strength
ranking of acid sites detected experimentally may vary considerably with the nature
of the basic probe molecules.
The void space within porous materials constitutes a region with special properties,
many of them not completely understood. At the level of molecular interactions in
nanoscale-sized spaces it is difficult to distinguish between "pure" acid-base pro-
cesses and perturbations due to the confined environment. Both the solid matrix
and the molecules confined to the internal open spaces may show altered acid-base
behavior in comprison with their "free-space" counterparts.
T o illustrate this we select two examples that correspond to two classes of
interfacial phenomena, both important from the practical point of view.
A interface is present in the microporous environment inside the chan-
nels and cages characteristic of zeolite structures. All catalytic chemistry based on
the acid properties of crystalline (zeolites) is affected by two addi-
tional factors known as the "solvent" and sieve" effects. The chemical
contribution, the solvent effect, describes the interaction at the molecular level
between micropore-confined molecules and the coordination field created by all
atoms that constitute the pore walls. The molecular sieve or geometical effect
accounts for the perturbation of the dynamics of confined molecules and is parti-
cularly important in relation to the selectivity of catalytic reactions occurring over
this class of microporous solid acids. Both these effects may act concurrently to
distort the intrinsic acid-base properties of molecules confined in the micropore
spaces. In the case of alkane cracking, numerous observations demonstrate that
various zeolites show very different specific rates for the same catalytic reaction.
Two opposing views were expressed to account for this: (1) The -(OH)- Si
positions in zeolite lattices span a very broad range of intrinsic acid strengths
that determine various reactivities, or (2) all acidic sites are virtually identical, and
differences in reactivity result from geometric constraints in the micropores
The perturbation due to geometrical factors may be so important that the existing
experimental techniques cannot separate the effects. so any connection to the
base index of the solid zeolite surface becomes obscure It is known, for
example, that the heat of adsorption of gaseous hydrocarbons on zeolites depends
on the chain length or the diameter and shape of the micropores Catalysts with
comparable numbers and strengths of acid sites but with different pore geometry

show completely different reactivities. In the one-dimensional pore system of
denite, the rates of monomolecular reactions are singe-file diffusion-controlled, and
processes are virtually Because the interaction with
micropore walls is mainly due to dispersive van der Waals forces, which are typical
for without covalent bonds, these effects should not be considered in rela-
tion to zeolites intrinsic acid-base properties
As a second example, consider the interface in a nanosized environ-
ment representative of the interlayer space of clay minerals. The hydrophilic and
swelling properties of these minerals is the result of structurally induced perturba-
tions of acid-base chemistry in nanosized spaces.
The interlayer space in clays is bordered by oxygen from two planes of siloxane
(Si 0 Si) groups. The oxygen atoms are electron pair donors (basic sites); their
negative charge is compensated by exchangeable cations that are electron pair
acceptors (acidic sites). When clays are hydrated, water in the interlayer spaces is
structured. At the boundary of flat oxygen planes (zone and in the
hydration atmosphere of exchangeable ions (zone water forms ordered struc-
tures; a disordered zone occurs between the two ordered zones
Depending on their position in the interlayer space, water molecules may act either
as proton donors or as proton acceptors. Water acts as a hydrogen bond donor in
zone and may dissociate under the polarizing effect of metal cations in zone
Because of steric hindrance in the interlayer space, the hydration atmosphere of

ions in interlayer zone cannot extend as much as the hydration sphere of free
ions in solution. Consequently, water in zone is a stronger proton donor and
interlayer protons are more mobile than in a liquid; the proton-donating tendency
increases as the water content of the interlayer space decreases
The wettability of the oxygen planes and the resulting structure of the interlayer
water depend on the net balance between the kinetic energy of water molecules in
the interlayer environment, the electrostatic forces between molecules and
exchangeable cations, and the attraction and dispersion forces between the layers
The wettability increases with the tetrahedral substitution. In hectorite, where
charge deficit occurs in the octahedral sheet, the negative charge of the oxygen
plane is delocalized and no hydrogen bonds are formed with interlayer water. In
vermiculite the charge deficit occurs mainly in tetrahedral sheets, and thus there is a
greater localization of the negative charge that leads to the formation of strong
hydrogen bonds in zone Consequently, the interlayer of hectorite is the
most expandable, while that of vermiculite is the most difficult to expand.
Montmorillonite, saponite, and are intermediate between these extremes
In the last section of this review we discuss additional examples of these effects.
The development of a scientific basis to describe the chemistry of solid surfaces has
been impeded by their ill-defined nature; many factors can be responsible for
observed effects. An understanding of surface chemistry at the same level

in areas such as solution or gas-phase chemistry is desirable in that classical
concepts proven useful for rationalizing chemical bonding and reactivity in tradi-
tional chemistry could be extended to the chemistry of solid surfaces,
The acid-base concept, either in the limited interpretation of the
theory or in the more general sense of the Lewis theory, is one of the
most useful classification in chemistry and a tool for systematizing the
relationship between structure and reactivity. In this section we examine how this
general concept can be applied to solid surfaces From the most general (Lewis)
point of view, a surface is classified as acidic (electron acceptor) or basic (electron
donor) according to the direction of net electron transfer that results in the forma-
tion of new chemical bonds with an adsorbed
A. Ideally Polarizable Surfaces: Metals

For clean metal surfaces, the direction of net electron transfer can be determined
from measurements of the work function variation, A positive indicates
net electron transfer from the surface to adsorbed molecule (basic surface) while
a negative indicates electron transfer in the opposite direction (acidic surface).
As with all Lewis probes, a clean metal surface may act either as a base or as an
acid, depending on the nature of the interacting molecules. For example, a clean
surface is a Lewis base 0) toward carbon monoxide and a Lewis
acid 0) toward acetylene and benzene However, the criterion of net
electron transfer is not always sufficient to define the acid or base character of clean
metal surfaces. The reason is the substantial surface charge reorganization that may
accompany chemisorption on metals. Because electronic states of metal surfaces are
closely spaced in energy, acid-base interactions that require electron transfer and
the formation of new chemical bonds are often delocalized.
Doping clean metal surfaces with either electronegative or electropositive ele-
ments produces changes in the acidic or basic properties that can be explained by
inductive effects analogous to those in molecules Adsorbed foreign may
form either overlayers or incorporation structures. The latter may produce recon-
struction of the top metal layer and strong decoupling of the atoms in the first
layer from the bulk. This occurs, for example, during oxidation of metals and the
growth of surface layers of metal oxides. Incorporation of oxygen in the top metal
layer is accompanied by a significant electron transfer from metal atoms to the
(more electronegative) oxygen This creates a large gap between the HOMO
and LUMO energies of metal oxides and thus stabilizes the new charge distribution
and inhibits charge reorganization during chemisorption. The result is the occur-
rence of localized Lewis acid-base sites (from a geometrical point of view) at the
surface of oxidized metals.
The electron-rich oxygen anions at the top of oxide overlayers on metals show
basic (electron donor) properties; the metal cations exposed in coordinately unsa-
sites (cus) show acidic (electron acceptor) character. This is depicted by
structure I in Scheme 1. It is expected that the electron donors and acceptor orbitals
are more localized on metal oxide surfaces than on clean metal surfaces. This forms
the basis for the interpretation of the surface chemistry occurring on oxide surfaces

Fully dehydroxylated
dissociative surface
adsorption
Partially dehydroxylated
surface
Lewis Lewis
immersion
Fully
surface
low pH neutral pH high pH
IV. V. VI.
Transformation of surface structures and their acid-base characteristics
on oxide surfaces with various degrees of dehydroxylation.
in terms of acid-base sites associated with individual surface ions located at various
positions related to all possible terminations of the bulk crystal lattice.
1. Surface Morphology
The morphology of oxide crystallites depends largely on their preparation method
as well as on the prevailing crystalline phases. In general, the higher the tempera-
ture of calcination, the coarser and more crystalline is the final powdered oxide. For
most transition metal oxides the surface area cannot exceed certain limits that
reflect the microcrystalline organization of the solid For these materials of
low porosity the average pore size is much larger than molecular diameters. Their
surface is "open" for molecules impinging from the gas or liquid phase, and on the
molecular scale the interaction zone can be viewed as a "flat" surface.
Materials with "open" surfaces are ideal for characterization of intrinsic
base properties of individual surface sites. The interaction of impinging molecules is
essential bimolecular, and generally the surface does not show any peculiar

collective effect. Indeed, the electronic structure of most metal oxides is localized;
oxygen atoms (both in bulk and at surface sites) can be considered as closed-shell
ions that serve as ligands for metal cations This picture is strictly valid for
highly ionic oxides is an example); it may require corrections only in the very
few cases of semiconducting oxides (such as where bonding is largely covalent.
The metal and oxygen ions at the surfaces of oxides may or may not occupy the
positions defined by simple extension of the bulk lattice. Surface reconstruction by
migration and relaxation of surface plane atoms from their position in the
bulk stabilize the surface by minimizing the energy. Evidence exists that thermal
stability of oxide surfaces may be correlated with the polarity of the surface and the
degree of coordinative of surface cations; the extent of surface relaxa-
tion depends on the presence of lone-pair electrons on the surface However,
due to the paucity of existing experimental results gathered by techniques sensitive
to surface geometry (such as LEED), it is in general difficult to say, with any
certainty, where the really are on the surface. Yet surface models that neglect
surface reconstruction and relaxation can be regarded as valid.
2. Fully Dehydroxylated Surfaces

The surface of a fully dehydroxylated ionic oxide can be considered as an extended
array of coordinately unsaturated oxygen ions and metal cations. From the
base point of view, they act as Lewis base and Lewis acid centers, respectively
(structure I in Scheme 1). The net charge on each surface center depends on the
stoichiometry, ionicity, and local surface structure, which in turn vary from one
crystal plane to another. A few positions that represent geometrical defects on the
surface (corners, edges, dislocations) may be characterized by enhanced acid-base
character, as their energy is further increased due to additional chemical
The molecules impinging on the surface can have either a Lewis base or Lewis
acid character. For instance, CO with a filled orbital at the carbon end displays
Lewis basicity, while shows Lewis acidity The primary interaction is
strictly localized on oxides, and surface are formed with the acidic or
basic centers of the solid. An example is the nondissociative adsorption of
on exposed centers on oxide surfaces:
The adsorption of CO on oxides involves mainly a donation and little

bonding A similar character, though with a stronger interaction, is shown by
the adsorption of NO, which has an electronic structure similar to that of C O
except for one additional unpaired electron. Another example is the nondissociative
adsorption of water on coordinately unsaturated cations:

The newly formed adsorption complex (structure in Scheme 1) is a Lewis
and at the time a Bronsted acid site.
A11 important peculiarity of solid surfaces is the occurrence of adjacent Lewis
acid (unsaturated cation) and Lewis base (unsaturated anion) sites in repetitive
structures. If they interact in a concerted manner with the adsorbed species, the
result is dissociative adsorption. For example, adsorption of hydrogen on is
molecular at but at room temperature adsorption is dissociative. The
mechanism is most likely heterolytic, as suggested by new and 0- H
vibrations observed in the IR spectrum [reaction
(VIII)
Another example of dissociative adsorption by concerted action of a Lewis
base couple is bonding of alkenes on metal oxides, as in reaction (IX). However,
dissociative adsorption of water on ideally dehydroxylated oxide surfaces is by far
the most important example to cite, in view of the formation of surface
groups and their tremendous impact on the reactivity of most oxides under
or close to ambient conditions [reaction (X)].
In Scheme dissociative adsorption of water transforms the Lewis acid-base

pair from structure I into a pair of Bronsted acid-base sites (structure that
can be regarded, at the same time, as Lewis Also note the conversion of
the Lewis acid into a Bronsted base and that of the Lewis base into a Bronsted
acid.
3. Partially Dehydroxylated Surfaces

Fully dehydroxylated oxide surfaces, such as those considered above, can ideally be
prepared either by treatment at elevated temperature under vacuum or by oxide
cleavage in an inert atmosphere. These structures do not exist under normal con-
ditions because Bronsted acidity is generated upon reaction with water after pre-
paration. Under ambient conditions oxide surfaces are partially hydroxylated.
They correspond to structures and of Scheme 1 .
The validity of local models for acidic properties on oxide surfaces is supported
by quantum-mechanical calculations. acid sites are present only when

hydroxyl groups are present. The strength of Bronsted acid sites is controlled by
the electrostatics of system and is governed mainly by three structural para-
meters: the coordination number of the hydroxyl group, the charge on the
and the coordination of cations neighboring the hydroxyl
group
Surface hydroxyl groups may differ with respect to their coordination to under-
lying metal cations and the magnitude and sign of the uncompensated electron
charge. In turn, this results in variations of the strength of bonds at various
surface oxygen sites. The acid-base properties of surface oxygen ions are therefore
determined by surface structural factors. For example, the two Lewis of
structure differ with respect to their Bronsted properties: The monodentate
hydroxyl is basic whereas the bidentate hydroxyl is acidic. This fact can be recog-
nized in the fine structure of stretching vibrations in infrared spectra of
partially dehydoxylated oxide surfaces.
The literature on infrared spectroscopy abounds with information that reveals
the presence of several types surface hydroxyls with various characteristic
stretching frequencies on partially dehydroxylated oxides. Structural considerations
show that the number of all possible configurations for hydroxyl (and oxygen)
groups on low index planes of oxides is limited For example, in silica the
oxygen atoms are connected to two Si atoms, and therefore only one type of
hydroxyl is likely to exist on the surface (silanol). In every oxygen ion is
surrounded by three ions of and therefore on the surface there may exist only
terminal and bridging hydroxyls. Although the spinel structure of transitional alu-
m i n a ~ and is more complex, structural considerations suggest that since the
oxygen coordination number is 4, one can expect three types of surface hydroxyls.
Cerium oxide also. with tetrahedral coordination of oxygen atoms, is expected to
present three types of hydroxyl groups. Figure 1 illustrates the D R I F T spectra of
several simple oxides in the region of deuterium-exchanged hydroxyl vibrations
The spectra were collected using a Impact 400 spectrometer equipped
with a catalytic chamber. The rather than vibrations were investi-
gated in order to isolate the contribution of oxide surface groups from other
perturbing influences during the measurements.
Knozinger and assigned the bands on alumina to surface
groups in various configurations by assuming that only and
coordinated surface oxygen positions may exist in ratios and geometrical
that differ from one surface plane to another. With some modifications
their scheme was generalized by Busca for surface sites on several oxides
with spinel and corundum structures. In general, the bands depend on the
nature, coordination state, and number of directly neighboring cations. For OH
groups bonded to the same element, the frequencies decrease in the order:
terminal over a tetrahedral cation terminal over an octahedral cation bridging
triple bridging; the breadth of the OH stretching band increases in the order
terminal bridging << triple bridging
The spectra shown in Fig. 1 are in fairly good agreement with the pre-
dictions based on structural arguments; small variations arise due to the particular
structure of some oxides. Thus, a total of five bands or even more (instead of

ALUMINA
FIG. 1 Diffuse reflectance FTIR (DRIFT) spectra of the deuteroxyl stretching

vibrations for several oxides after partial surface dehydration at
three) were regularly found on transitional aluminas; the larger number is the result
of a distribution of ions between octahedral and tetrahedral positions, which
discriminates between O ‘ vibrations. Also, ceria is likely to present only densely
Ã
packed planes on its surface because of its type of lattice, and thus only two
(instead of three) types of hydroxyl groups are revealed in I R spectra
A dynamic, rather than static, picture of partially hydrated oxides is more
appropriate, as confirmed by proton NMR and dielectric spectroscopy studies
The results of revealed two types of protons o n partially
droxylated oxide surfaces: those associated with physisorbed water and those from
chemisorbed water. About 85% of the total protons are present in the first two
molecular layers. Subsequent NMR studies showed that a permanent
proton exchange occurs between chemisorbed and physisorbed water.
In conclusion, partially dehydroxylated oxide surfaces exhibit a large inventory
of surface O H groups and water molecules together with Lewis acidic and Lewis
basic sites with coordinative (structures and of Scheme 1). The
hydroxyl population is the source of protons that cause enhanced surface electrical
conductivity and catalytic activity. It is significant that the increase in the conduc-
tivity value is paralleled by increases in either the amount of weakly bound protons
or their mobility Almost all metal oxides are active in catalytic
of which is one of the least demanding reactions in terms of the require-
ments for the acid strength of active sites Studies on several oxide systems
show that the activity is lost after extensive dehydration and is partially restored by

rehydration In all these the catalytically active sites were proton
donors acids) of the oxide surface.
4. Fully Hydrated Surfaces
Complete hydration of oxides represented by structures and of Scheme 1
results in new surface structures (IV, V. and VI) at the interface. The
interfacial phenomena at the boundary include electrostatic adsorp-
tion, ion exchange, and ligand binding of species from solution. However, in most
instances the species that count as charge-determining (in aqueous media) are
always and O H ions. As the concentration of these ions is pH-dependent in
solution, the predominance of one or another of the surface structures IV, V, and
VI is also
the liquid side of the interface, the above phenomena are
best approached by the electrostatic or ligand chemistry description or by a com-
bination of both. However, from the solid side of the interface, the charge devel-
opment depends on proton binding or release processes that occur at the outermost
oxygen layer of the oxide lattice. On the solid side of the interface all proton
transfer processes are localized. They can be regarded as acid-base interactions
between the oxygen ions at the oxide bounday and one or two interface-structured
of hydrogen-bonded water molecules in the liquid phase.
Which site-binding model is best suited to describe the adsorption of H and
O H ions to specific sites at the interface is still a matter of contro-
versy. In the classical approach the surface is assumed to be monofunctional
and all surface oxygen ions are considered equivalent. Since most inorganic oxides
show amphoteric acid-base properties, it was postulated that the charge buildup
mechanism may be represented by the following two-step protonation process
model"):
The two proton-binding constants are interrelated through an experimentally mea-

surable quantity, the pH where the positive and negative charges compensate or the
pH of zero charge (PZC):
An alternative representation is the model" proposed by Bolt and van

In this model only one protonation step and only one value are
needed to describe the amphoteric character of surface oxygen groups:
+ '2
.
It can be shown that the model reduces to the model if
= log log 2 and the unique value obtained corresponds to
an experimental quantity, the point of zero charge: = log
The above approaches represent, of course, two idealized models. It has been
recognized for a long time that the surfaces of real metal oxides and hydroxides

are nonuniform, several types of oxo and hydroxo groups are present. It has
been assumed, as in the case of partially dehydroxylated surfaces, that the surface
chemistry of the interface can be derived from the structure of
solids using concept of bond valence. Recently, et al.
and others suggested an integrated approach based on the following
protonation scheme:
This representation takes the of a multisite complexation model (MUSIC).

It is assumed that surface oxygens on different surface planes may bind to one or
two protons, in separate equilibria, according to their and values (for
and hydroxo groups, respectively). In the above equations the bond valence v
depends on the coordination number of surface oxygens with metal cations, the
formal charge per site may be any fractional number, positive or negative. The
variability of surface groups, that may serve as sites for proton bonding is twofold:
First, on different surface planes exposed to the solution, surface oxygens are in
different coordination environments with respect to metal cations. Second, for each
type of surface oxygen (terminal or type I, bridging or type triple bridging or
type there are in principle two possible proton-binding equilibria, as shown by
reactions (XIV), above. The MUSIC model incorporates, therefore, the
and models as particular cases. In practice, however, it is
only one of the two reactions is in the p H range of interest (about pH
11) for most applications. Thus, in practical terms, a "one multisite" approach,
as introduced by Contescu et seems more appropriate for a description of
proton-binding equilibria and of corresponding thermodynamic properties at
the oxide solution interface. In Scheme 1, the equilibrium between structures IV
and V (deprotonation of monodentate water-bound complexes) occurs at a lower
pH than the equilibrium between structures V and VI (deprotonation of bridging
hydroxyl groups); proton dissociation of terminal of structures VI is
possible at even a higher pH.
Originally, the MUSIC model assumed an even distribution of charge between
metal cations and oxygens. A method was suggested for an a priori calculation of
the intrinsic proton-binding constants of surface oxygens based on an estimate
of the local electrostatic contributions to the standard energy of proton
adsorption. The method was calibrated with known values of equilibrium constants
for protonation of and hydroxo complexes in solution. In a later development
the calculation of the undersaturation of oxygen valence at the surface was
improved. Taking into account the structural details around surface oxygen ions
resulted in asymmetrical charge distributions in the coordination environment.
Following a suggestion by Bleam the redefined MUSIC model emphasizes
the role of hydrogen bonds (both donating and accepting) with neighboring
water molecules at the interface. The model is able to predict the proton affinity
constants for individual surface groups and can estimate correctly the PZC values
of several metal oxides and hydroxides.

An important point of the original MUSIC model and its subsequent refine-
ments is that proton-binding constants at the solid interface are
no longer regarded as free fitting parameters but as interface constants subjected to
the constraints of physical and chemical reality, such as the distribution of oxygen
ions on various crystallographic planes and the polarizing power of metal cations
Even though the model may not be totally quantitative, the tabulated values of
proton affinity constants have their own merit as reference values.
When experimental methods to measure proton affinity distribution spectra
from potentiometric titration data became available it was found
by Contescu et that the apparent values of proton binding constants iden-
tified from were in semiquantitative agreement with log K values for
binding predicted by the MUSIC model (Table 1). The characteristic for
several oxides are shown in Fig.
1 Acidity Constants for Hydroxyl Groups with Different Coordination

on Fully Hydrated Surfaces of Various Oxides
calculated from from PAD Other

Oxide MUSIC model measurements estimates
5.4 (CC)
7.7 (CC)
9.5 (CC)
7.2 (S)
6.7-7.8 (DL)
3.1-3.3 (DL)
6.0-6.3 (DL)
1.2 (DL)
Data predicted theoretically by the MUSIC model and measured experimentally by P A D and other
techniques.
DL = double-layer model; CC = constant capacitance model; spectroscopic methods.

Proton affinity distributions measured for several oxides in 0.1 N
5. Factors that Determine Surface Acidity

What are the factors that determine the acid-base properties of solid surfaces such
as metal oxides? On the basis of the discussion thus far it seems appropriate to
relate the appearance of Lewis acidity and disappearance of Bronsted acidity to the
increase in the degree of dehydroxylation. Indeed, the interconversion of Lewis and
Bronsted acid sites has been demonstrated for some oxides, such as or sup-
ported or by IR studies of pyridine or ammonia adsorption
But which factors determine the strength of acid sites?
The chemical factor, the chemical nature of metal cations, must be consid-
ered first. The nature of metal cations. their electronegativity and oxidation state,
determines the and geometry of the crystalline lattice.
Fully dehydroxylated covalent surfaces have a homopolar character, with no
sensible charge separation. Adsorption on covalent solids is mainly physical; cova-
lent surfaces tend to be hydrophobic, and extensive requires high
temperatures. Hydroxyl groups present on partially dehydroxylated surfaces show
very weak acidic character. They act as hydrogen bond donors with most bases and
can only very strong bases. A typical example is silica
The acid-base properties of solid oxide surfaces with ionic character are con-
trolled mostly by electrostatic factors. The stronger the ionic character, the stronger
is the electric field pointing outward the surface. The electric field of the
surface may enhance (or diminish) the local electron donor (or acceptor) properties

of surface ions and thus modify their Lewis acid-base properties; inductive effects
akin to those known for small molecules were proposed Various factors, such
as electronegativity effective charge on oxygen and the polarizing power
of metal ions have been correlated with the acid-base properties of ionic
oxides.
In stable oxide structures, the principle of requires that the
charge of the cation be neutralized by the charge of surrounding oxygens. The
distribution of charge over the surrounding can be expressed per bond,
leading to the concept of bond valence. Introduced by Pauling, the bond valence
is defined as the charge of the cation divided by its coordination number:
Pauling's bond valence rule has been used in various attempts to correlate acidic
properties of surface hydroxyls with their formal charge, as expressed by the above
relationship For example, the assignment of stretching vibrations of
surface hydroxyls already mentioned is based on calculations of the formal
charge on surface hydroxyls using Pauling's rule. Based on the possible relationship
between acidity and the stretching frequencies of hydroxyls in similar structures
the above reference to Pauling's rule illustrates, in fact, a relationship between
structure and properties (acidity) for surface hydroxyls on partially dehydrated
oxides. Pauling's valence bond rule served as a basis for the MUSIC model of
proton binding to surface hydroxyls on fully hydrated oxides. The results of
these semiempirical approaches are in remarkable agreement with exact quan-
tum-mechanical calculations based on local Hartree-Fock models and with
many experimental trends.
A comparison of Figs. 1 and 2 shows that FTIR spectra of isolated hydroxyls on
oxides have their counterpart in of the fully hydroxylated surfaces. The
parallelism is seen with respect to both the number and strength of hydroxyl groups
present on the surface. A plot of stretching frequencies groups (from Fig.
1) versus the values revealed by of several oxides (Fig. 2) is shown in Fig.
3. Despite some scatter of the data, a linear trend is evident. At this stage one can
only speculate that the slope of the straight describes the intensity of the
electrostatic at the oxide surface. It apparently depends on the sample's
chemical nature (possibly, through the cation's electronegativity or
radius ratio) and/or its surface structure (anion packing, preferential development
of crystal planes, surface imperfections). In any case, the correlation shown in Fig.
3 is similar to a relationship found for zeolites in which the increase in Bronsted
acidity correlates with decreasing bond strength and can be monitored
through the corresponding decrease in the stretching frequency in FTIR spectra
Besides chemical factors, the role of physical factors in determining the

base properties of solid oxides should also be considered. Thus, the symmetry type
of the crystalline lattice, the local geometry of surface ions, and the energetic
stability (and relative occurrence) of densely packed planes depend on the ratio
of the ionic radii of the cation and the ion. If polymorphic structures are likely,
they may have different acidities depending on the packing density of surface

FIG. 3 The relationship between stretching frequencies of isolated deuteroxyls on
partially dehydrated surfaces and the acidity constants determined by PAD at the
interface for several oxides. Alumina; silica;
anatase;
ions, the degree of exposure of surface cations, and their preference for defect sites.
For example, in the oxygen lattice is more loosely packed, the frequency of
stacking faults is greater, and the fraction of cations in tetrahedral positions is
slightly smaller than in although the local ionic configuration of
cations is the same in both modifications. Due to these structural differences, the
acidity, surface reactivity, and catalytic activity of are significantly higher
than those of
Surface imperfections of crystallographic structure may occur in small
oxide crystallites. Metal or oxygen vacancies are typical imperfections that cause
a local charge imbalance. Therefore, the acid properties may be strongly affected by
the crystalline size, in parallel with (and perhaps as a consquence of) variation of
electronic properties. For anatase particles with sizes in the range of 75-100 A , the
formation of new acid sites was observed with higher strength on smaller particles
1. Mixed Oxides
The acidity of materials can be controlled by mixing oxides with different proper-
ties and in various ratios. If phase separation does not occur, substitution of cations
of the minor component oxide in the host oxide lattice results in bulk phases of true
mixed oxides. Another method used to the acid-base properties of oxide

surfaces is surface deposition o f "impurity" ions with limited substitutional incor-
poration. In the practice o f catalyst preparation it is frequently found that the
surface acidity (or basicity) o f catalysts can be tuned by surface modifiers selected
from the group o f alkali or alkaline earth metals (to neutralize acidity) or from
anions o f oxyacids or electronegative elements (to promote acidity).
A number o f models have been proposed for prediction o f the generation o f new
acid sites in dilute oxide solid solutions. From the earliest models it was under-
stood that acid centers in mixed oxide structures owe their existence to localized
electric charges created through isoinorphous substitution. The model proposed by
Tanabe and coworkers to explain acid site formation is among the
frequently cited. This model attributes the generation o f surface acidity to the
charge imbalance that develops along the M 0 heterolinkages (where
and are differentcations) due to the differencesin cation valence and coordina-
tion number at the site o f lattice substitution o f the minor component cation. The
model is based on two postulates: (1) The coordination number o f cations from the
substituting metal oxide and the matrix oxide are maintained even when mixed; and
( 2 ) all the oxygens in the binary oxide maintain the coordination number o f oxygen
in the matrix oxide. Based on Pauling's valence rule, the formal charge o f each ion
is assumed to be evenly distributed over the coordinating bonds. I f the excess
charge at the cation site is negative or positive, a Bronsted acid site or a Lewis
acid site may appear. The former case is illustrated by, Ti substituted in a
matrix o f the latter by substituted in a matrix o f I f there is no excess
charge at the site o f the substituting ion then no acid sites are created, such as
in the system.
A differentapproach was taken by Kung This author proposed a model
that accounts for variations in the electrostatic potential experienced by the guest
cation from all ions in the oxide matrix. Although based on different assumptions,
in most cases Kung's model gives the same prediction as Tanabe's model. However,
an important differenceexists between the two models: Tanabe's model is a loca-
lized model, whereas the model by is delocalized. In other words, in Tanabe's
model the site o f acidity is fixed at the substituted cation, whereas in Kung's model
the newly created acid sites may be located at any position with respect to the
substitution site. Arguments supporting the localized model came recently from
the works o f Miller and on mixed oxide aerogels. Using preparation tech-
niques o f sol-gel chemistry that afford better control o f scale mixing,
they demonstrated that well-mixed two-component oxides
silica) display higher total site densities, larger Bronsted populations, and enhanced
activity as acidic catalysts than less homogeneously mixed samples o f the same
composition. These observations were explained in terms o f the much higher num-
ber o f linkages accessible in each well-mixed oxide pair.
2. Chemically Modified Oxides
A more efficient method for adjusting the surface acidity o f solids is chemical
modification o f the outermost atomic layers. and Dumesic
that acid centers (both Lewis and Bronsted) could be generated by doping small
amounts (less than 1%) o f a second metal cation on the surface o f several oxides
with various acid-base character, ranging from the very basic to acidic

or inert and In these systems the "impurity" ions were not
uniformly substituted in the bulk matrix oxide lattice but incorporated in defect
cation positions at the surface of the host oxide. The model used by and
Dumesic to explain generation of acidity on chemically surfaces differs
from or models for homogeneous solid oxide solutions. The
important factors in their case are the electronegativity and coordination environ-
ment of dopant metal cations on the host oxide surface. The
model assumes that the formal charge on the deposited cation is balanced by
coordinating oxygen ions from the surface lattice and that a deposited cation is
coordinately if its coordination number differs from that of cations in
the host oxide. Thus, the origin of both Lewis sites and sites was
explained: The former are localized at coordinately unsaturated dopant cations:
the latter are generated by substitution of a cation in the tetrahedral environ-
ment of an matrix.
Scokart and Rouxhet studied the influence of electropositive and electro-
negative dopants on various oxide adsorbents. They used Pauling's valence rule to
evaluate the net charge of nonprotonic acid sites and found agreement with the
strength of Lewis sites as measured by pyridine adsorption. For example, electro-
negative elements (such as fluorine enhance the acidity (both Bronsted and
Lewis) of alumina; electropositive elements (such as sodium or magnesium
partially neutralize alumina acidity. A scale of the effects of various modifying ions
on acid-base properties of alumina was summarized as follows by Delmon and
coworkers
increase of acidity
increase of basicity
The discovery of strong nonzeolitic catalysts obtained by chemical modification

was a breakthrough for the development of new solid acid catalysts. It was found
that metal oxides, especially and mixed
oxides modified with sulfate ions show unusually strong acidity. Sulfate may be
introduced either from a dilute solution of sulfuric acid or by impregnation with
ammonium sulfate. The strong "catalytic" acidity of these materials is evident in
their ability to activate paraffins at lower reaction temperatures than conventional
acid catalysts. Several authors denoted this property as "superacidity" and
defined superacids as materials with an acid strength stronger than that of 100%
sulfuric acid.
The mechanism for the generation of catalytically active centers in "superacidic"
catalysts is not fully understood. In many cases the properties of the catalysts are
strongly influenced by the preparation conditions. N o agreement has been reached
about the structure, strength, or nature of the acidic sites. Bronsted acidity is
commonly observed on zirconia, although strong Lewis acidity has also
been implicated in the high reactivity of these materials While nonsulfated
zirconia exhibits two types of isolated O H groups, modification with sulfuric acid
eliminates the terminal (most basic) O H group, enhances the acidity of the bridging

O H group, and introduces new, more acidic, Bronsted sites in which the O H
group is probably linked to sulfur and the proton is bridged to a neighboring oxide
ion on the support. It was also found that sulfur-containing species are bound
through two or more S- 0- Zrbridges and carry or S - 0bonds.
It is important to emphasize that spectroscopic evidence shows that water trans-
forms the Lewis acid sites of sulfated zirconia into Bronsted acid sites At the
same time, water promotes isomerization reactions over sulfated zirconia for a
moderate extent of catalyst dehydration. Similarities were reported between the
effect of rehydration on the isomerization activity of sulfated zirconia and
on that of other oxide catalysts that are consistent with the role of surface
donor sites in hydrocarbon isomerization reactions. However, when spectro-
scopic methods using basic probes were used to compare sulfated zirconia and
zeolites in terms of the strength of their acid sites, the results were inconsistent
with all catalytic data. These findings illustrate the danger of comparing the acidity
of catalyst systems that differ in structure and composition, such as zeolites and
sulfated zirconia; in these systems the "catalytic" and the "physicochemical" scales
for the strength of acid-base interaction may contain significantly different para-
meters.
3. Bidimensional Oxide Overlayers

Studies of bidimensional oxide overlayers has attracted increased interest in recent
years as these are intermediates in the preparation of of the
widely used industrial catalysts: hydrotreating catalysts
Bidimensional overlayers may be by deposition of one oxide
on a second, high surface area oxide
A peculiarity of the oxides that may form bidimensional overlayers is
their tendency to aggregate as polyoxoanions in acidic media and to disperse as
monomeric in basic media. Due to this fact, the factor that decisively
determines the stability of bidimensional structures for various systems investigated
so far is the match of acid-base properties between the dispersed and host oxide.
Using spectroscopy, and coworkers showed that the
structures stable under ambient conditions correspond to the chemical spe-
cies stable at the PZC of the composite oxide: Supports with basic character favor
molecular species that are stable in basic solutions and vice versa.
Because supported oxide species follow the same chemistry as they d o in aqu-
eous solution as a function of pH. the structure of the overlayer is controlled by the
of the composite oxide in contact with the ambient; under these conditions the
apparent and the structural conformation of the overlayer oxide depend on
the surface coverage For example, a strong increase in the proton acidity of
alumina was found on addition of small amounts (maximum 5%) of
and Moreover, when exposed to aqueous solutions of variable pH,
the structure of the overlayer changed with the solution's pH. Based on this prop-
erty, Adachi et al. showed that the number of protons released during
of correlated with the catalytic activity of
catalysts in a typical hydrodesulfurization reaction.

Clay minerals are hydrous layered The layer is
composed of two types of sheets: octahedral ( 0 ) and tetrahedral (T). Each sheet
is composed of parallel planes of atoms, either oxygen or aluminum in an
alternating sequence. The large variability in the properties of clay minerals is
related to the stacking sequence of the two basic types of sheets, T and 0 , and
to the uncompensated charge that develops by isomorphous substitution of
unevenly charged cations.
The hydration properties of clay minerals can be understood from the perspec-
tive of various types of acid-base interactions that may occur between several
structurally defined sites in the clay matrix and surrounding water molecules. To
facilitate discussion, a short description of clay structure is needed.
The tetrahedral sheet consists of a hexagonal network of corner-sharing
tetrahedra with all their bases in the same plane. In this plane the oxygen atoms
form an open hexagonal network, which is referred to as the hexagonal o r "perfo-
rated" plane. Each oxygen of this plane is covalently bound to two silicon atoms.
Substitution of for Si contributes a negative charge to the tetrahedral sheet, and
the newly formed 0 Si groups are more basic than the original Si 0
(siloxane) groups were. However, the electron-donating strength of the siloxane
oxygen depends on the existence of nonbonding electron pairs. Because of their
asymmetric charge distribution, only nonbonding hybridized orbitals may serve as
electron-pair donors. Consequently, structures with hybridization on oxygen
are expected to show greater hydrophilicity (because of the larger electron density
in nonbonding hybridized orbitals) than structures with oxygen in hybridiza-
tion, whereas hybridization leads to hydrophobicity. Measuring the Si- 0- Si
angle in minerals provides a good estimate of the oxygen hybridization state. In
good agreement with the above structural considerations, it was found that hydro-
philicity of siloxane groups increases with decreasing - 0 - angle. In most
silicates the 0 angle has values of 139-140" indicative of an intermediate
character between and hybridization. for which the characteristic angles are
and
The octahedral sheet consists of a densely packed hexagonal plane of ions
(which in brucite is substituted by sandwiched between two densely packed
hexagonal planes of hydroxyls. The latter are called the hydroxyl planes and make
an important contributions to the surface properties of minerals. The oxygen atoms
in the hydroxyl layer have all their orbitals in the valence shell occupied and
consequently cannot act as hydrogen bond acceptors with water molecules.
Therefore, the hydrophilicity of the hydroxyl plane is solely determined by the
interaction between protons of the hydroxyl layer and lone-pair electrons of
adsorbed water. Thus the hydroxyl group must be acidic. Although the acid
strength of M groups increases from Mg- OH to OH, it is believed
that the proton-donating ability of the hydroxyl layer is not great enough to allow
for the formation of hydrogen bonds with water molecules. In most minerals the
hydroxyl layer is therefore a hydrophobic plane, although substitution may
change the hydration properties. Thus, substitution of for increases the acid
strength of M- OH groups and the hydrophilicity of the hydroxyl plane.

Natural aluminosilicates, such as clays, were the first solid acid catalysis commer-
cially used on a large scale. They were the first cracking catalysts in the Houdry
process and were prepared by acid leaching natural clays. The negative impact of
impurities in natural clays stimulated research for new solid acid catalysts, and thus
amorphous-like synthetic were first prepared The natural
evolution led to the synthesis of crystalline aluminosilicates (zeolites),
and silica-alumina phosphates Zeolites are crystalline
aluminosilicates built from and tetrahedral units. There is a virtually
unlimited number of ways of linking tetrahedra, which give rise to zeolite crystals
with pores, cavities, and channels with a wide variety of shapes and sizes. Today,
crystalline zeolites and zeolite-type materials (zeotypes) are the most common acid
catalysts used in the chemical industry. In addition to the strong acidic character of
their amorphous predecessors, these materials have the benefits of high surface area
and controlled pore sizes in the range of molecular dimensions.
The acidity of is easily explained by model.
Substitution of a atom by tetrahedral atom in a silica-rich host lattice gen-
erates units with a localized negative charge that may be balanced by a cation
or a proton. There exists an extensive body of literature on quantum-mechanical
calculations of acidity from the proton affinity of small clusters that confirmed
this picture it was concluded that the 0 - Hbond weakens and the proton is
more easily ionized if the is located in bridging positions between and
atoms. A general discussion on the acidity concept in zeolites can be found in
Refs, 21 and 90.
In structure (a) of Scheme 2 the coordination of the 0 atom in the bridge
position to tetrahedral corresponds to a Lewis acid-base interaction; the stron-
ger the interaction, the stronger the Bronsted acid strength of the proton associated
with the bridging oxygen. The model for Bronsted acidity in zeolites shown as
structure (a) of Scheme 2 has since been improved by assuming that the bridging
oxygen atom is chemically bound to both and Si, making a real bridge between
them [structure (b) in Coordination of the bridging oxygen to both
and Si causes an increase in the length of the 0 - H bond and therefore
enhances the acidity of structure (b) with respect to the terminal depicted
Scheme 2, structure (c). groups (terminal hydroxyls) are only very weakly
acidic or nonacidic.
On the other hand, the increase in electronegativity in the neighborhood of the
O H induces a transfer of electron density from to 0 with a corresponding
increase in acidity. Thus, in silica-alumina phosphate materials (SAPO) the sub-
stitution of by creates a center of enhanced acidity as shown by structure
(d) of Scheme 2, where X = P
The chemical composition of the framework has a definite impact on the acid
strength of zeolites and zeotypes. The total number of Bronsted acid sites present in
a zeolite depends on the framework ratio or in general on the
cations. The variation of stretching frequency of bridging hydroxyls and results
from H N M R spectra suggest that the strength of Bronsted sites in zeolites
increases with decreasing ratio. This has been rationalized with the concept of

SCHEME 2 Structures proposed for the Bronsted acid sites in zeolites and substituted
zeolites.
electronegativity: Any change in the composition of a zeolite would modify the

electronegativity of the framework, and the higher the electronegativity, the greater
the acid strength. A correlation was found between Sanderson's average
negativity and the acid strength of zeolites and zeotypes
Geometrical effects, in addition to framework electronegativity, have a consid-
erable influence on zeolite acidity. It was argued that the long-range order imposed
by the crystalline structure is the reason the acid strength of crystalline zeolites is
higher than that of their amorphous silica-alumina counterparts. The geometry of
the crystalline lattice and the associated long-range interactions have a clear effect
on the acidity of bridging groups in O H Si linkages. Molecular
orbital calculations show that the deprotonation energy decreases (the acidity
increases) with increase of the T - 0 - Tangle. Acidic zeolites have a range of
angles larger than other, less acidic zeolites (cf. for ZSM-5,
for but only for By similar arguments,
structure (a) of Scheme 2 could be more representative for amorphous
mina, while structure (b) would be more appropriate for the acid site in a crystalline
zeolite with long-range symmetry stabilization
Besides long-range effects, the local structure of cage o r channel walls may cause
preferential location of some particular oxygens. For example, in the
structure of zeolite Y there are four different positions for covalent proton bonding
on oxygen next to aluminum atoms; they correspond to two different stretching
frequencies in infrared spectra, which indicates the occurrence of two groups with
different acid strengths
The acid-base properties of zeolites are further complicated by the confinement
of host molecules in narrow micropores. Several components of the interaction
energy between (basic) host molecules and the walls or acidic sites of zeolite
cages may be strongly enhanced in micropores with molecular dimensions.

Among them the most important are the attractive and repulsive contributions
from van der dipole-dipole, and interactions.
Derouanne suggested that molecules confined in zeolite micropores
tend to organize themselves by optimizing their van der Waals interaction with
the surroundings. In accordance with this rationale the guest molecules confined in
micropores of sizes close to the molecular dimension of the guest should be
regarded as "solvated" by the zeolite framework. The strong electric fields at the
internal surface of molecule-sized spaces must strongly affect the properties of
''solute" molecules, including the electronic energy levels of the guest molecule.
The theory predicts that confined guest molecules would show increased
basicities with respect to their intrinsic basicity in the gas phase. However, this
argument may equally be used to explain the unusually high acidity of zeolites
Lewis acidity may also be present on zeolites, but its precise nature remains an
open question. Initially it was thought that Lewis acidity is generated by extensive
and corresponds to threefold coordinated ions on the surface.
However, it is now established that dehroxylation is almost always accom-
panied by formation of extraframework aluminum species, that are
acidic in the sense and can be further dehydroxylated to form Lewis acid
sites similar to those present in active The number of Lewis acid
sites is a fraction (between 0.4 and 0.7, depending on dehydration temperature) of
the total content of nonframework aluminum In cation-exchanged zeolites,
another source of Lewis acidity is furnished by the exchanged cation; cations are
electron acceptors and are able to act as Lewis acids in the broader sense. It was
observed that the strength of Lewis acid sites in Na exchanged zeolites was depen-
dent on crystallinity, whereas the strength of Lewis basic sites was less so; the
collapse of the zeolite crystal caused a decrease in the numbers of both Lewis
acids and Lewis bases
Activated carbons consist of small graphene-like layers that are spatially distribu-
ted without regular organization and are separated by most of which
are slit-shaped. The most reactive sites for the attack of molecular oxygen are the
edges of the graphene layers and defect sites. Upon oxygen attack, and depending
on reaction conditions, two types of carbon surface oxides are Because of
their practical importance, surface oxides of carbons have been the subject of many
studies
Basic carbon oxides are formed upon the exposure to oxygen of a clean carbon
surface obtained by outgassing to followed by cooling to room tem-
perature. Aqueous suspensions of these carbons have a strongly basic pH, reducing
properties toward noble metals, and anion-exchange properties. Acidic carbon
oxides are when the carbon is oxidized a t lower temperatures
the presence of water vapor during the oxidation process increases the
amount of acids formed. When dispersed in water, these carbons have acidic pH
and cation-exchange properties. Acidic surface oxides can also be formed by oxida-
tion at room temperature with aqueous solutions of nitric acid,
hypochlorite, or hydrogen peroxide.

The reactivity of carbon surfaces is strongly influenced by their acid-base char-
acteristics. The broad spectrum of applications for activated carbons arises from
their wide range of properties, which can be systematically controlled.
Adsorbed oxygen makes the surface of carbon hydrophilic and aids dispersion in
water Adhesion between carbon fibers and the polymer in carbon-fiber com-
posites is vastly improved by the presence of carbon surface oxides Surface
acidic groups may act as anchoring sites for metal particles of carbon-supported
catalysts, limiting their growth, but they may also produce strong metal-support
interactions and modifications of the electronic nature of suppported metal parti-
cles
The chemical nature of acidic and basic oxides on carbons has been widely
investigated. The structure of basic oxides was first ascribed to cromene-like struc-
tures and later to pyrone-type groups Other workers assumed
that carbon basicity of the Lewis type can be associated with the high concentration
of electrons on basal graphene layers where bonding of water molecules may
occur in oxygen-free media:
The acidic surface oxides obey a more complex chemistry A large

variety of surface structures may occur. The most acidic are carboxyl groups, which
may either be free or be condensed with neighboring hydroxyl groups to form
or with neighboring groups to form lactols. At the other extreme
are phenol-type hydroxyls, which are weakly acidic groups. All these groups have
been identified by chemical reactions
In contrast with acidic groups of most other solids, the acidic groups of carbons
as described above belong to well-established classes of chemical compounds
described in any organic chemistry textbook. This fact permits their qualitative
identification as well as their quantitative determination. Counting the number of
acid sites on carbons poses fewer problems than with other solid acids, and a
classical analytical method developed by Boehm is still in use today. The
Boehm method uses the fact that carboxyl groups are strong enough to neutralize a
very weak base like = 6.37). the less acidic and
tones are neutralized by dilute = 10.25) but not by
and the weakest acidic phenol groups are only by solutions
15.74). By titration with all titers in the presence of suitable
indicators, it is therefore possible to quantitatively determine the content of acid
functionalities with the acid strength (expressed as in various acidity ranges.
Another method to access acidic groups on activated carbons is direct
metric titration, but it suffers from very slow exchange equilibria in solution
and the lack of definite endpoints. The latter fact reflects the spread of dissociation
constants of most surface functionalities on carbons caused by interactions between
neighboring groups and effects resulting from the limited size of the graphene
sheets. a continuous, rather than discrete, distribution of values,
showed that potentiometric titration data could be trans-
formed to proton affinity distribution (PAD) curves and that the results were in

excellent agreement with the quantitative estimate of surface site concentrations
obtained from the Boehm titration.
Heteropolyacids and their salts are formed by metal-oxygen

octahedra as the main structural units linked by corner sharing and edge sharing.
Of the wide variety of heteropoly compounds, those of the series are the
most important as catalysts. The compounds, build around a
central atom (X = or P) with tetrahedral coordination by oxygen. The unit
is surrounded by octahedra (M = Mo, V, Nb, Ta). They are arranged in
four groups of three edge-shared octahedra, which are linked by shared
corners with the central tetrahedron. This is the structure of polyoxoanions
exists in solutions and in the solid state as well. In the solid state a secondary
structure is formed by closepacking of the polyanions and is stabilized by
cations and water of hydration or, more precisely, protonated water dimers,
Several crystallographic arrangements may exist, depending on the
degree of hydration and the nature and size of the countercations.
Heteropolyacids are strong Bronsted acids both in solution and in the solid
state. Their strong acidity is related to the large size of the polyanion, which causes
a low density of negative charge at its periphery and therefore weak interaction
between adjacent polyanions and protons. The charge balance and hence the acid-
ity are strongly affected by both the elements constituting the polyoxoanions and
the nature of the countercations.
In the solid state, protons participate directly in the formation of the crystal
structure The lattice is stabilized by species hydrogen-bonded to
oxygen atoms at the most exposed positions of four neighboring anions (structure
XVII). Loss of water of crystallization gives rise to porosity (structure XVIII); thus
solids with considerable surface area can be obtained by moderate heat
The microporosity and the resulting surface area are largely controlled by the size
of the countercations. The most effective are the bulky and ions,
for which surface areas of about 180 have been obtained. By controlled sub-
stitution of it is possible to produce solids with homogeneous and
nized porosity
(XVII) (XVIII)
Solid-state heteropoly compounds have interesting catalytic properties and
have found applications in and acetylene hydration and alcohol dehydration
Adsorption of nonpolar molecules (hydrocarbons) reaches saturation at
amounts lower than monolayers ("surface-type" adsorption). However, it is char-
acteristic of heteropoly that adsorption of polar molecules (alcohols,
ethers, amines) takes place in amounts that exceed by many times

layer surface capacity. This property of "bulk absorption" is clearly result of
acid-base interactions between the polar molecules and the solid heteropoly com-
pound, since the amount absorbed depends on the polarity of and the
nature of the countercations.
The latter property is important for the catalytic activity of heteropoly com-
pounds in acid-catalyzed reactions. With certain heteropoly salts and polar
catalytic reactions such as alcohol dehydration occur in the bulk of the solid
catalyst, not only on its surface. The solid behaves like a very concentrated acid
solvation medium for the reactant. which allows these solid acids to be called
pseudoliquids
It has been clearly established that acid sites in heteropoly are of the
Bronsted type and that their strength, for dehydrated solids. is higher than that of
concentrated sulfuric acid. The acidity increases with increasing cation
For example, the partially exchanged is a superacid, and
its performance depends on the degree of exchange,
Because catalysis by solid acids and bases is central to a variety of industrial

processes, relationships solid acid-base behavior and catalytic properties
have been explored However, concepts, either qualitative or empiri-
cal, demand closer scrutiny.
A discussion of methods to quantify the surface acidity of solids, or
not, requires first the recognition of the fact that the term "acidity" is too vague.
For catalytic applications, one should be able to identify the nature, either Bronsted
or Lewis, of surface sites exposed by the surface of solids. On the other hand, even
for the purpose of correlation with another property (such as
catalytic activity), one should clearly distinguish the importance of measuring the
number and assessing the strength of acid sites. Both components coop-
erate to create the overall picture of "surface acidity," which is no more than a term
used for a badly defined macroscopic property of solids. If improperly used, this
term may generate misleading statements such as "catalyst A is more acidic than
catalyst B." It would be more appropriate to compare, specifically, the number of
acid sites in catalysts A and or their acid strength, assuming that it has been
previously found that the nature of the acid sites is the same in both catalysts, either
Bronsted or Lewis.
However, for several reasons it is difficult to compare the acidity of solids in
definitive terms as demanded above. They are related to the experimental methods
for characterizing surface acidity or to the theoretical definition of a common
acidity scale.
A variety of methods have been suggested to characterize the acidity or basicity of

solid surfaces in terms of the strength and number of surface sites they contain.

Some methods use the adsorption of acidic or basic probe molecules with the goals
to (1) the sites with opposite properties, (2) identify (and count, if possible)
the sites with either Lewis or Bronsted character, and (3) determine the sites'
strength in relation to a known property the acid or base strength) of the
probe A second group of methods uses standard catalytic reactions that
are known to occur on surface sites of certain strength as a probe of acid-base
characteristics.
These methods suffer from the lack of complementarity, and thus the signifi-
cance of results provided by any of them is limited. A standard practice to detect
the Bronsted or Lewis character of surface sites is pyridine adsorption combined
with FTIR measurements; the number of Lewis or Bronsted sites is more difficult to
count, however. Other titration methods use either color indicators and acid or
base in nonpolar solvents or the adsorption of gaseous acidic or basic
probes. They do not, in general, give consistent quantitative information about
the number of acid or base sites even when applied to the same sample. There
are several reasons the applicability of titration methods is limited: Either the
state of the surface is different for different methods or adsorption equilibrium is
not always achieved. Another more serious source of discrepancies between
methods is that probe molecules of different basicities "see" different surface
sites. The lack of a uniquely defined thermodynamic scale of acid strength of surface
sites makes difficult any correlation between results obtained with different probe
molecules. The use of standard catalytic tests for probing the so-called catalytic
acidity is not always a better approach, because the mechanistic assumptions
involved are neither straightforward nor subject to experimental proof.
1. Identifying the Type of Acid Sites: Bronsted or Lewis?

There are several probe molecules for which infrared spectroscopy can differentiate
between adsorption on Bronsted and Lewis acid sites and even estimate the
amounts adsorbed. Pyridine is the most widely used because it gives well-resolved
bands when protonated by Bronsted acid sites 1540 and 1640 or when
coordinated to Lewis acid sites (1450 and 1620 The values of extinction
coefficients are available in the literature for these bands, which makes pos-
sible semiquantitative measurements, separately, of Lewis and Bronsted sites.
Ammonia, with a smaller kinetic diameter that enables it to reach more easily
the acid sites in smaller pores, can also be used to distinguish Bronsted
and Lewis acid sites; however, the use of ammonia is less reliable, mainly because
the resulting IR bands overlap each other Another base that can distinguish
between Bronsted and Lewis acid sites is quinoline; because its size is greater than
that of pyridine quinoline can also be used to differentiate between acid sites at the
external surface and those in pores than its kinetic diameter (6 A). Bronsted
sites can be selectively measured with methods by using substituted pyridines as
probe molecules
However, owing to their strong basicity, it is expected that most of these bases
would adsorb even on the weakest acid sites. Weaker bases are
needed as probe molecules for spectroscopy, and studies have been reported
in which CO, acetone, benzene, acetonitrile, and even hydrogen have

been used Adsorption of CO at low temperatures is used extensively because
infrared spectra can differentiate between CO molecules in interaction (by hydro-
gen bonding) with surface hydroxyls with Bronsted acid character and CO coordi-
nated to unsaturated surface cations (Lewis sites) It appears that carbon
monoxide is a highly specific probe molecule that enables the of
the order of relative acidities when different types of or coordinately
unsaturated cations are present
N M R and MAS N M R methods can also be used to characterize the interaction
of amines with acid sites, and the shift of or rather than is conveni-
ently used to distinguish between Lewis and Bronsted sites.
2. Number and Strength of Acid-Base Sites

Counting the number (or measuring the concentration) of chemical species that
may participate in acid-base reactions is a where traditional chemistry
offers, in general, well-established solutions.
When the reaction medium is water (as it is in most analytical applications), the
relative strengths of acids and bases can be compared by using a thermodynami-
cally precise acidity scale, namely the scale with water as a reference. In this case
one can use titration methods. For example, an acid of known acid strength,
is added to the system and neutralizes all bases whose conjugate acids
are weaker than the Similarly, when the is a base
of known strength, all acids weaker than the will be neutralized:
The only experimental problem is recognition of the endpoint of
titration; the amount of base (or acid) neutralized is then calculated from the
amount of consumed. Recall that all the values just listed are acidity
constants and the strength of bases is compared using the strength of the
corresponding conjugate acids, =
In aqueous solutions the principles of analytical chemistry established the pro-
tocol to use strong acids (or bases) as they "see" all bases (or acids) in the
system, and the result of titration is a "total basicity" (or "acidity") number. If an
additional method to measure the concentration of "free" acid at different titration
points and under equilibrium conditions were available, then one could use
eons titration methods to isolate bases (or acids) of different strengths that might be
present in the system and evaluate their amount.
Extrapolation of the above principles to solid surfaces is not straightforward
because of the strong interplay of the two fundamentally different notions: the
strength and the number of acid-base sites. To accurately measure the number of
acid sites of a solid acid, one should be able to define their "strength" and have
available a "stronger" base to be used as a If the base is not strong
enough, some of the weakest acid sites will remain undetected. Titration in
phase systems, and especially when one is a microporous solid, may be influenced by
additional factors such as slow diffusion into the micropores and charge accumu-
lation at the interface or at the pore mouth. In other words, even though a strong
enough base was available, its access to the acidic sites on the internal surfaces might
be hindered by geometrical constraints or its intrinsic basicity might be perturbed by
intermolecular interactions in the confined space of the micropores.

3. Applications
The limitations of the experimental methods and potential methods to avoid mis-
interpretation of laboratory data are illustrated here with a few examples.
Benesi titration is one of the oldest methods used for counting the acid sites
of solid catalysts in different ranges of acid "strength."The use of quotation marks
is necessary because the method is based on the convention that the scale in
aqueous media can be extrapolated to very strong acid media by use of an appro-
priate series of color indicators. Thus, when a basic indicator (B) reacts with a
surface Bronsted acid site (HA) to form a surface base ( A ) and the conjugated
acid form of the indicator the acid strength is equated to the
acidity function in the following way:
Similarly, if interaction is with a Lewis acid site (A), the function is

expressed as
where [B] is the concentration of the unreacted base. The acid strength of surface
sites is estimated by comparing the color of several indicators with different
values adsorbed on the solid sample, while the number of sites in various
ranges is estimated from the of base (in most cases,
needed to block the sites. The suffers several limitations. The
most serious is the fact that solid samples are in nonaqueous media
benzene) but the values of indicators used for comparison are aqueous phase
values, and the medium-transfer corrections for are ignored. Also, equilibrium
may be achieved with difficulty and the method does not distinguish between
Bronsted and Lewis sites unless hindered bases, which are selective for
Bronsted sites, are used as
An alternative method proposed by Contescu et al. is aqueous titration
combined with a suitable algorithm for deconvolution of the measured proton
binding curve into a continuous spectrum of apparent values. Even neglecting
corrections for charge accumulation at the interface. the content of the
proton affinity distribution (PAD) function is complete with respect to both the
number and the apparent strength of proton-binding sites at the inter-
face. Agreement with the number of surface acidic sites was demonstrated by
comparing the PAD data for activated carbons with the results obtained indepen-
dently using an established analytical method (Boehm titration The validity
of the evaluation of values was demonstrated by comparing the experimental
results for oxides with "open" structures such as alumina with the acid strength of
surface predicted by the MUSIC model
Another method for quantitative determination of acid sites on solid catalysts
is calorimetric nitration with amines This method uses the very simple
concept that the enthalpy of adsorption for simple bases is a direct measure of
acid strength. If a variation of the heat of adsorption with surface coverage is

observed, it demonstrates that the distribution of surface sites is nonuniform.
However, with this method also there are difficulties with experimental procedures
and the interpretation of results. In addition to the uncertainties due to effects other
than acid--base interactions and more than one proton being actually trans-
ferred adsorption equilibrium may be slow because the mobility of
strong bases is low at low temperatures. For a microcalorimetric study
of the adsorption of various amines on several high silica zeolites by
and Gorte concluded that the structure of zeolites does not substantially affect their
ability to transfer a proton to the reactants. Other factors such as confinement
effects may explain the discrepancies between the results of calorimetric mea-
surements and catalytic activity data on zeolites.
Another methodology used to determine the number and strength of acid sites
on solid catalysts is based on temperature-programmed desorption of adsorbed
basic molecules (such as ammonia, pyridine, quinoline, or n-butylamine). In prin-
ciple, an excess of base is adsorbed and the physically adsorbed species are removed
by evacuation, in most cases combined with heat treatment. The amount that resists
evacuation is attributed to the total number of acid sites. If thermodesorption is
measured after heat treatment at several, progressively increasing temperatures, an
approximate acid strength distribution can be established A linear correlation
was found empirically between the strength of acid sites expressed by the
function and the reciprocal temperature for the highest rate of
thermodesorption
Alternatively, can be used to calculate the activation energy of adsorption for

the chemisorbed base, which is proportional to the heat of adsorption. However,
determination of heats of adsorption from ammonia temperature-programmed
desorption (TPD), although popular for its simplicity, may lead to inconsistent
data because the experimental protocols may greatly influence the results
Reactive amines other than ammonia have also been employed to distinguish
Bronsted acid sites in various zeolites using a combination of T P D and
gravimetric analysis (TGA) techniques The is based on the fact that
surface Bronsted sites may induce thermal decomposition of aliphatic amines to
alkenes and over a narrow temperature range. The number of
molecules reacted is equated to the number of strong Bronsted acid sites. By choos-
ing amines of appropriate sizes it is possible to discriminate between acid sites
located in pores of different diameters.
In general, when results obtained from titration methods are compared with the
catalytic activity, the correlation is not necessarily good. One of the reasons is that
only a small fraction of the total acid sites measured by titration are active for a
given catalytic reaction. The low temperature at which adsorption is mea-
sured with the use of color indicators favors adsorption on all sites, including the
weakest ones. Thermal desorption methods discriminate between sites of
different strength but are unable to distinguish between Lewis and Bronsted
sites. When coupled with vibrational spectroscopy data, the TPD technique may,
indeed, be adequate for analysis of surface acidity However, it is almost

always found that the use of a single technique is not sufficient for comprehensive
characterization of the surface acidity of catalysts and predictions of catalytic
activity.
The challenge is to identify the catalytically active sites for a specific
reaction, to count them, and then to correlate the catalytic activity (expressed
as reaction rate per unit catalyst surface) with the surface density of active sites.
This task is quite easily solved in the case of zeolites, where it has been demon-
strated that the number of Bronsted acid sites of a particular type (usually, those in
Al- OH- Si bridges) could be quantitatively related to the rates of catalytic
reactions
A more challenging issue is to correlate the strength of acid sites with the rates of
catalytic reactions homogeneous acid-base catalysis a relationship
between the acid strength of an acid in solution and the rate of a catalytic
reaction catalyzed by that acid was quantified first by Bronsted and Pedersen
The stronger the acid, the easier it is to activate the bond to form a transition state
complex and the faster is the catalytic reaction. The equation of Bronsted and
Pedersen relates the effectiveness of a general acid catalyst to its acid-base strength:
where is the specific rate constant for catalysis by an acid whose acid dissocia-
tion constant is G and are constants for a series of similar catalysts.
It was assumed by several authors that a similar correlation should exist for
heterogeneous catalysis but the difficulty of measuring the number and
strength of Bronsted acid sites on solid catalysts at the same time hindered any
experimental proof of a complementary relationship. Recently, the occurrence of a
Bronsted-type relationship was demonstrated experimentally by Contescu
for isomerization of 1-butene on a series of mixed oxide catalysts. The number
and strength of Bronsted acid sites were simultaneously determined from the pro-
ton affinity distribution of solid acid catalysts by potentiometric titration For
a homologous series of catalysts, such as or
the areal reaction rates on catalysts of various molar compositions were directly
correlated with the number of acid sites of a particular type (Fig. identified as
the bridging hydroxyl in Si- OH- M structures. The significant fact was that
active sites identified in each binary oxide system were characterized by nearly
constant acid strength values (expressed on the scale of water), which shows
that the acid strength in Si- OH- M structures is indeed determined by the
electronegativity of M. Moreover, when the logarithms of specific reaction rate
constants were plotted versus the values for the active sites found on catalysts
from various binary systems, the linear correlations shown in Figs. and were
found This is the equivalent in heterogeneous catalysis of the Bronsted rela-
tionship, first developed in homogeneous catalysis.
The fact that a relationship to Eq. (8) was experimentally found for solid
acid catalysts has both practical and theoretical significance. a practical point
of view, it demonstrates that at least one experimental exists, namely the
PAD method, that affords a direct and quantitative determination of both the
number and strength of Bronsted acid sites on surfaces of porous oxide catalysts.

Surface Density of Active Sites
Relationship between catalytic activity of 1-butene to 2-cis- and

on several mixed oxide catalysts and their surface acidity on the
aqueous scale. (a) Linear dependence between specific reaction rates and the
density of active sites from PAD data; (b) linear Bronsted relationship between the
reaction rate of formation and the acid strength of corresponding active
sites on different mixed oxides; (c) same relationship for formation of

The fact that the results from titration data for the
interface could be used for correlation with the catalytic activity of the
interface (in similar systems) reopens the discussion of the relevance of one or
another of possible acidity scales, either for gases or for water, in characterizing
the surface acidity of solids.
Concepts
I11 recent literature a considerable lack of agreement can be found as to which
acidity scale is more suitable as a reference for molecules probing the acidity of a
solid surface (gas-phase or aqueous-phase scale). This dichotomy presents a pro-
blem that is difficult to resolve and has interpretive implications for experimental
results.
The difficulty of establishing a unique acidity scale for solid surfaces has been
mentioned several times. Even the definition of "acid strength" of solid surface sites
is not thermodynamically consistent. It is appropriate to analyze some of the
theoretical concepts involved and to address some of the solutions advanced by
various authors who tried to solve this problem.
The definition of acidity in terms of the Lewis theory has a broad meaning that
encompasses the Bronsted definition as a particular case. In practice, especially in
the field of heterogeneous catalysis, it is important to differentiate between
Bronsted or Lewis acid sites on solid catalysts, since they may show completely
different catalytic behavior.
Limiting the discussion to acidity, there have been two opposing views
expressed regarding the question of which is the most appropriate acidity scale for
solid surfaces. Some authors advocate the use of the acidity scale of gaseous species;
others support the use of the scale of aqueous solutions.
Arguments for the Gas-Phase Acidity Scale

Acid-base properties determined for gas-phase molecular interactions are based on
experimental measurements of equilibrium constants in proton transfer reactions
between two gaseous bases:
+ +
The absolute proton affinity (PA) of a base molecule in the gas phase has a precise
thermodynamic definition in relation to the enthalpy variation of the hypothetical
reaction
in the absence of any interactions with the medium. The latter requirement is
difficult to satisfy for accurate measurements even in vacuum. However, if the
absolute proton affinity for a single base, could be independently
in the gas phase, then a relative scale of basicities could be established by
measuring the equilibrium constants of a series of gas-phase reactions with

the participation of Again, conditions of total isolation from medium effects
must be ensured, but this is less difficult for gas-phase reactions than for reactions
with condensed media.
On the other hand, quantitative scales of Bronsted acidity available for aqueous
solutions are based on measurements of equilibrium constants in proton transfer
reactions like The best-known scale of acidities, the scale, compares the
relative proton-donating capacities of various molecules with respect to a unique
base: water. The is that in this latter case the proton transfer cannot be
regarded as isolated, because the aqueous medium contributes to the energetic
stabilization (hydration) of all species, reactants and products, involved in the
equilibrium.
It is possible, of course, to combine gas-phase proton affinities and
measured equilibrium constants in order to gain insight into the relative
contribution of intrinsic (molecule-specific) versus medium-dependent effects
while still using a perfectly rigorous thermodynamic approach Calculations
showed that the solvation energy terms (medium effects), far from being a small
perturbation, may be as important as the intrinsic (molecule-specific effects).
When the reaction medium is water, the two types of effects are almost equally
important, and the details of interaction of water with all species involved be
accounted for in order to derive first-principles-based acidity scales. However, the
utility of the water-based scale has been demonstrated many times, and valu-
able insights into the molecule-specific effects have been gained by comparing
values between the members of a homologous series of compounds.
In principle, a similar approach may be extended to Bronsted sites on solid
surfaces. It is possible to design a thermochemical cycle that would permit an
estimate of the enthalpy variations during protonation of a reference base on
any solid surface site. All that is needed is a knowledge of gas-phase proton affinity
for the base molecule and an estimate of the enthalpy variations when two systems,
one with free base and the proton attached to the solid and the other with
nated base and the proton dissociated from the solid site, are brought from the
solid state to a (hypotheical) gas-phase state. This enthalpy change is, however,
very difficult to measure or calculate. The deprotonation energy of surface groups
on solids can be obtained from quantum calculations and NMR data The
real difficulty comes from the lack of any experimental data about transfer func-
tions, the energetic terms involved in the interaction of the protonated base
with the solid.
In other words, the medium-specific contribution is not easily measured or
quantified for probe molecules in interaction with solid surfaces. Moreover, in
the case of microporous solids, the short-distance interactions known as "confine-
ment effects" are even more difficult to evaluate. In all comparisons of experimental
data one should be aware that the reactivity of probe base molecules is largely
influenced by the size of adsorbates and micropore dimensions. As a result, the
acidity scales based on the free energy of proton transfer to a specific base are
expected to depend on the choice of reference base. This fact has been confirmed
experimentally, as calorimetric heats of adsorption of various bases on, zeo-
lites, depend on the base chosen. For example, a ZH zeolite may be a stronger acid

than a one on a scale based on pyridine but a weaker acid on a scale based on
ammonia
2. Arguments for the Water-Based Acidity Scale

The advocates of the use of the aqueous scale for comparing the acidity of
solids (mainly of oxides) invoke chemical, rather than thermodynamic, arguments.
Their rationale is as follows. Since the protonation of a base probe molecule cannot
be observed in total isolation from the environment, it is more logical to seek
similarities with other systems instead of neglecting the serious perturbations that
most solids induce on the state of the adsorbed base.
It is known that charged molecules are stabilized by different mechanisms in
various media. In the gas phase, stabilization of a charged protonated species, such
as from equilibrium (XIX), occurs by internal induction; in aqueous solu-
tions the protonated species is stabilized by interaction with water
dipoles. Consequently, the same molecule may exhibit different acid strengths in
different media. For example, in aqueous solutions ammonia is a stronger base
about 9) than pyridine about 5); however, in the gas phase the basicity of
pyridine is significantly higher.
It can be argued that the presence of oxygen ions (or of hydroxyl groups) on the
surfaces of oxides, in interlayers of layered hydroxides, and in zeolite micropores
creates a "waterlike" environment for adsorbed species. In other words,
the oxide may function as an aqueous solution, stabilizing the charge on the
adsorbed molecules A series of titration or displacement experiments have
shown, indeed, that the relative adsorption strength of various organic probes,
including alcohols. acids. and on various oxides reflects the order
of their aqueous values and does not follow their order of acidity in the gas
phase A conclusion was reached whereby the acidity and basicity of
an amphoteric oxide like was characterized by a combined
calorimetric technique with both basic and acidic probe molecules It was
shown that the number and strength of acidic sites were related to the aqueous
values of the adsorbed probe molecules.
A supplemental proof that metal oxide surfaces provide reaction environments
that mimic conventional solvents emerged from studies on the reactivity of several
organometallic compounds adsorbed on oxides. It is known from aqueous phase
chemistry that the synthesis of iridium carbonyls is sensitive to the basicity of the
solutions used. The same fact has been shown by Gates and coworkers for
metal carbonyl clusters on metal oxide surfaces (in the absence of solvents); their
chemistry parallels that occurring in solvents with properties (such as basicity)
analogous to those of the surface. For example, the environment the surface
of and in the supercages of zeolites is so highly basic that it allows the
synthesis of a series of anionic iridium (or rhodium) carbonyl clusters that is pos-
sible only in basic solutions. In contrast, alumina and zeolite are more neutral
surfaces on which neutral (or rhodium) carbonyl clusters form instead
The parallelism between the chemistry of carbonyls in solution and that
on oxide surfaces demonstrates the importance of surface hydroxyl groups or a
water association in the chemical environment offered by the oxide surface.

Comparing the nature of products synthesized on oxide surfaces with that of pro-
ducts formed in aqueous solutions allows an estimate of surface basicity. In the case
of alumina the comparison suggests that thermal treatment (at increases the
basicity of the surface.
These very new results are only preliminary and are sufficiently accurate
because neither the basicity of the organic solutions used for synthesis in
metallic chemistry nor that of solid surfaces are easily quantified. However, the
conclusion that the basicity of increases with decreasing surface
agrees with the conclusion that the Bronsted acidity of alumina and
several other oxides decreases under
similar conditions of thermal treatment
The latter result was obtained from several catalytic activity studies for butene
isomerization on mixed oxides dehydrated at various temperatures. Double-bond
migration of 1-butene requires weak Bronsted acid sites on the surface of solid acid
catalysts (with values in the range from about 4 to 8 depending on oxide
composition). Most oxides, especially oxides, are active for this simple
merization reaction if activated by mild dehydration. Infrared spectra show the
occurrence of vast populations of surface hydroxyls on the active catalyst surface,
which probably mediate the proton transfer Bronsted acid sites on the catalyst
surface to the reactant in the gas phase On the other hand, too high a
dehydroxylation temperature decreases the activity, while partially
restores the catalytic activity
T o bring to closure the plethora of possible examples on this subject we cite two
more. Each illustrates a specific domain of intrinsic practical importance in which
the acid-base properties of microporous solids play a dominant role in determining
a broad range of applications. The examples examine the and
interfaces, respectively. For the interface, the interdependence of
base properties and microporous structure is revealed by the catalytic properties of
zeolites, especially by their shape-selective properties. It is impossible to review all
the literature in this area, and only a few recent views are summarized. For the
interface, we focus on the role of acid-base chemistry in interlayers of
clay minerals.
Zeolites are widely used as molecular sieve adsorbents and solid acid catalysts in the
oil refining industry. The most important characteristic of these microporous solids
is their acidity, in particular its interplay with the chemical environ-
ment.
There are many differences between traditional acid-base chemistry in solutions
and acid functions on zeolites, but the general aspects of the chemistry between

proton acceptor reactants and proton donor sites on the solid's surface remain the
same as those that occur in the aqueous phase. Therefore, the catalytic acidity of H-
type zeolites is controlled by factors such as the acid strength and concentration
of zeolite hydroxyl groups, (2) the balance between media (solvation) and diffusion
(geometrical) effects; (3) the nature of reactant molecules, especially their base
strength or proton affinity; and (4) the stability and possible transformation routes
of carbocation-like intermediates In zeolite micropores, the interaction
between surface acid sites and reactant molecules involves not only the proton
donor site localized at a given hydroxyl position but also the contribution from
the surrounding crystal, or the solvent (medium) effect
Understanding the interaction mechanism of various reactants with acidic zeo-
lites is a prerequisite for understanding zeolite catalytic properties. Water is one of
the simpler reactants. However, earlier experimental work was not able to define
unambiguously the structures formed by that interaction. Quantum-mechanical
calculations suggested the formation of only hydrogen-bonded water, as shown
by structure (a) in Scheme 3, and not that of Some experimental data
agreed with the quantum-mechanical result, but others indicated the
formation of hydronium ions as well, as depicted by structure (b) in Scheme
or the coexistence of both sepcies, A study with powder neutron
diffraction and infrared spectroscopy confirmed that both species are formed
on hydrated silica-alumina phosphate (SAPO) microporous catalysts: the
ion is found in the eight-ring channel of the zeolite 3, structure and
the second water is hydrogen-bonded to an acid site on the six-member
ring (Scheme 3, structure a).
This insight is important in view of the large-scale use of these catalysts for
alkane to increase the octane rating of motor fuels. It may help
answer the question of which molecules become protonated by a particular catalyst
and which do not. For example, one may infer that if water becomes protonated,
stronger bases such as alcohols may also be protonated under similar conditions
Acidity centers in hydrated silica-alumina phosphates originating from

water molecules and ions. (Adapted from Ref. 151.)

An observation often made in the literature is that protonation of weak bases in
clay suspensions frequently proceeds to a greater extent than that predicted from
the p H of the bulk aqueous phase and the intrinsic constant of the
conjugated acid of the weak base This fact has important consequences for
the reactivity of, pesticides or other organic contaminants in soil and may
strongly modify their solubility, resistance to hydrolytic degradation, and long-
term accumulation.
For example, it was reported that simetone and prometryne, two
herbicides with weakly basic amino groups about 4) develop higher
apparent basicity when present at low concentration in dilute aqueous suspensions
of natural clays and Depending on the experi-
mental conditions (such as the presence of other electrolytes and their nature and
concentration), the increase in the effective of these herbicides in
suspensions amounted to up to two units. The effect has been accounted for by
the electrostatic effects of the interface. In view of the fact that clays
carry a net negative charge, their aqueous interface can significantly lower
the potential of the protonated form of the organic molecule relative to that
of the neutral (B). This results in shifting the B+ to
the left, which is equivalent to increasing the effective basicity of the herbicide in
suspensions. The role of other metal cations, water, or other organic
molecules in the aqueous phase is to control the accessibility and apparent number
of negatively charged sites on the clay. Interestingly, the effects were not observed
in water suspensions of a zeolite with pore size openings smaller than the
nated group of one of the herbicides, although the smaller
protonated group had access and underwent exchange.
Modification of the acid-base chemistry of exchanged species in the interlayer
spaces of minerals is supported by recent results in pillared clays, both anionic and
The acid-base properties of pillared clays and of their thermal degrada-
tion products were characterized by potentiometric titration data from which the
P A D functions were calculated It is generally accepted that the Bronsted and
Lewis acidity of pillared clays is related to properties of both the mineral layers and
the acidic sites introduced as pillars When pillars are composed of metal
oxides or hydroxides (such as hydroxyaluminum or hydroxyzirconum
the acidic properties of the clays are enhanced Similar changes were also
observed as a result of thermal treatment of intercalated clays and were
attributed to the rearrangements of pillars caused by However,
the of potentiometric titration data for nondehydrated samples revealed
that OH groups in various surface coordinations attached to hydroxyaluminum
pillars intercalated in clays underwent shifts in their
intrinsic values due to the influence of the clay sheets This result was
not reproduced for hydrotalcite-like anionic clays intercalated with molybdate and
anions, which interact weakly with the brucite-like layer during ion
exchange. In the latter case, the distribution of metalate-exchanged samples
was close to that of concentrated aqueous solutions of the same ions

This is an indication that in the starting materials the intercalated oxoanions
are monomeric and ions and undergo a mechanism in
the interlayers during acid titration similar to that in water. This is not a trivial
finding, because it further suggests that by adjustment of the pH of the water
suspension one can modify the form of the intercalated anion and thus obtain
pillared structures with different forms of the intercalated anions and thus with
different acid-base surface properties
The scope of this review was intentionally restricted to those and liquid/
solid interfaces where the acid-base properties at a molecular level can best be
explained and related to particularly important applications.
In the field of heterogeneous catalysis, in which most solid acids or bases used as
catalysts also have porous structure, the questions of primary interest are (1) how
to characterize in a consistent manner the number and strength of surface acid-base
couples, ( 2 ) how to distinguish between Lewis and Bronsted acid sites, and (3) how
to prepare solids with high surface area, selected micropore size, and stable surface
structures with desired acid properties. The progress made in this field is commen-
surate with its economic importance. Now it is possible to select and prepare
catalysts with an acid strength adequate for almost any specific acid-catalyzed
reaction. This has been made possible by the development of acid-base character-
ization methods that provided the basis for understanding the close relationship
between the atomic-scale structure and the properties of surface acid sites. The
progress in materials science and the development of molecularly engineering cat-
alysts have extended the spectrum of inorganic materials used as solid acid cata-
lysts. The new generation of zeolites and zeotypes with controlled molecular sieve
properties and an acid strength as high as 10 has induced a revolution in the
field of oil refining and petrochemistry. When catalysts with even stronger acidities
are required, the use of sulfated oxides is promising, although more work is needed
to stabilize these against premature deactivation.
Synthetic layered hydroxides with hydrotalcite-like structures are used as het-
erogeneous catalysts for base-catalyzed reactions. One of their major drawbacks for
technological application is their low stability during hydrothermal and/or thermal
treatments. Recent attempts to incorporate large, preferably multicharged anions
into their interlayer spaces introduced new routes for preparation of more stable
materials. The effect of intercalated polymetalates or transition metal complex
anions in the framework of layered hydroxides is twofold: They increase the gallery
heights and thermal stability on one hand and introduce additional Lewis or
Bronsted acidic sites in a basic framework on the other hand.
Oxides, especially those of Si, Fe, and Mg, are essential components of the
earth's minerals; their surface sites may have either Bronsted or Lewis acid-base
properties. The weathering of rocks and the formation of soils are processes at the
interface between minerals and natural waters or solutes from the
medium in which solid phases are either formed, altered, or dissolved.
Dissolution of solid minerals is controlled by slow surface reactions rather than

diffusion processes in the liquid phase and by acid-base reactions, either in the
Bronsted sense (proton-promoted dissolution) o r in the Lewis sense
dissolution). Moreover, the hydrophilicity of natural clays and minerals, as
well as their swelling and ion-exchange properties, originate in local acid-base
interactions in the interlayer space. Thus the chemistry of natural waters and the
quality of the environment are controlled by acid-base equilibria in multiphase
systems of natural waters, solute, and minerals.
Arrhenius, Phys, Chem. (1887).

Bronsted. Rec. Chim. Pays-Bas (1923).
Chem.
G. Lewis, Valence and the of Atoms and Molecules, The Chemical
Catalog Col, New York, 1923.
(a) Jensen, The Lewis Acid-Base Concepts: An Overview, Wiley. New York,
1980; (b) Jensen, Adhesion Tech. 5:1 (1991).
(a) R. Hard and Soft Acids and Bases, Hutchinson and
Ross, PA, 1973; (b) J. Am, Chem.
(1963).
(a) Drago and A. Matwiyoff, and Bases, Heath. Lexington,
MA, 1968; (b) Drago and J. Am. Chem. 87: 3571
(1965).
(a) V. Gutmann, The Donor Acceptor Approach to Molecular Interactions,
Plenum, New York, 1982; (b) V. Gutmann, A. Steininger, and Wychera,
Monatsch. Chem. 97: 460 (1966).
Hudson and Tetrahedron Lett. (1967).
Bowers, Aue, Webb. and Am.
(1971).
Aue and Bowers, in Gas Phase Ion Chemistry, Vol 2 Bowers,
ed.), Academic, New York, 1979, p. 2.
W. A. and Sauer, J. Am. Chem.
(1993).
L. Jorgensen and Am. Chem. (1989).
P. Stair, Am. (1982).
C. Berg, in Wettability Berg, Marcel Dekker, New York, 1993, p. 76.
(a) F. Finlayson and A. Shah, Adhesion Tech. (1990); (b)
Webster and J. P. Wightman, J. Adhesion Tech. (1991); (c) Oh,
and Kim, Adhesion Tech, (1990); (d) P. Papirer
and Adhesion Tech. (1990); (e) A. Larsson and W. E.
Johns, Adhesion (1988); Ma, W. Johns, A. Dunker, and A.
Adhesion Tech. 4-411 (1990).
A. Leon y Leon and Radovic, in Chemistry and of Carbon, Vol.
24 (P. A. Thrower, Marcel Dekker, New York, 1992, p.
Adhesion Tech. 1:7 (1987).

G. Fripiat, Berger-Andre, Andre, and G. Derouane, Zeolites
(1983).
Derouane. in Guidelines for Mustering the Properties of Molecular Sieves
Barthomeuf. Derouane. Plenum, New York, 1990.
p. 234.
W. Farneth and J. Gorte, Chem. Rev. (1995).
, and Pure Appl. (1993).
Lei. and W. H. Appl. Catal. A (1996).
A. van Santen and Chem. Rev. (1995).
S. in Approaches to Theory and Applications (M.
Schrader and New York. 1992, p. 279.
Salvador, Vandermeersche. and Fripiat, J. Chem.
(1968).
Sposito and Chem. Rev. (1982).
(a) R. and Clay. Miner. (1977); (b)
Hougardy, Stone, and J. Fripiat, (1977).
P. Stair, Am. (1982).
Conrad, and Latta, Surf. (1976).
(a) and Phys. Rev. Lett. 123 (1974); (b)
and Sci.
(198 1).
Soled, Catal. (1983).
V. Rep. Prog. Phys. (1985).
Kung, Transition Metal Oxides: Chemistry and (Stud. Surf.
Catal. Vol. 45) Elsevier, Amsterdam. 1989.
A. Zecchina and 0. Catal. Rev. - (1993).
A. Lokhov and A. A. (1980).
P. A. Plisken, and Low, Catal. (1962).
A. and Phys. B (1997).
A. A. Tsyganenko and V. Filimonov, Struct. (1973).
Contescu and J. A. unpublished results, 1995.
Knozinger and P. Ratnasamy. Catal. Rev, - (1978).
(a) P. A. Saad, Saur, and Lavalley,
Appl. Catal., (1990); (b) A. B. Mohammed V. A. and
Lavalley, Appl. Catal. A: General (1993).
(a) G. Busca. V. Lorenzelli. V. Sanchez and Catal.
(1991); (b) V. and R.
9:1492 (1993).
Alvarez, J. F. A. and J. A. Odriozola,
Faraday Trans. (1993).
(a) Catal. (1971); and
Catal. (1975).
V. Gunko, V. I. Zarko. V. V. Voronin, V. A. and A.
A. Langmuir (1995).
A. and D. Ellis, Am. Chem. (1992).

(a) Morrison The Chemical Physics of Surfaces, Plenum, New York, 1977;
(b) and V. Electronic Phenomena in Adsorption and
Catalysis on Semiconductors and Dielectrics, Nauka, Moscow, 1979.
49. (a) A. G. Oblad, and A . Mills, Am. Chem. (1955);
(b) I. Chen, and Weissman. Catal. (1987); (c)
Gonzales, Kobe, and A. Catal.
(1996); (d) H. Thesis, Syracuse University, 1997.
(a) and White, Adv. Colloid Interface (1978); (b)
James and A. Parks, in Surface and Colloid Science, 12
Plenum, New York, (1990).
5 H. Bolt and H. van Riemsdijk, in Soil Chemistry,
Models Bolt, Elsevier, Amsterdam, 1982, p. 459.
(a) Hiemstra, C. De Wit, and van Riemsdijk. Colloid Interface
(1989); (b) Hiemstra and van Riemsdijk, J. Colloid Interface
Sci. (1996); (c) Hiemstra. P. Venema, and W. H. van Riemsdijk,
Colloid Interface (1996).
53. Koopal, Electrochim, (1996).
54. (a) J. Jagiello, and A. Schwarz, Langmuir, (1993); (b)
Contescu, Hu, and A. Schwarz, Chem. Faraday Trans.
(1993).
Contescu, A. Contescu, and A. Schwarz, Phys. Chem. (1994).
56. Bleam, J. Colloid Interface (1993).
57. El Shafei, J. Phys. Chem. (1993).
58. (a) Contescu, A. Contescu, and A. Schwarz, J. Phys. Chem. (1994);
(b) C. V. T. and A. Schwarz, Colloid Interface
(1996).
59. (a) T. Morimoto, Yanai. and Nagao, Phys. Chem. (1976); (b)
Segawa and Hall, Catal. (1982); (c) Groff, Catal. (1984).
60. (a) Tanabe and A. Catal. (1967); (b) Shibata, Kiyoura, J.
Kitagawa, Sumiyoshi, and K. Tanabe, Bull. Chem. (1973).
61. (a) J. Take, Kikuchi, and Y. Yoneda, Catal. (1971): (b) T.
and Tanabe, Phys. Chem. (1976).
62. El Shafei, Colloid Interface Sci. 182: 249 (1996).
63. (a) A. Parks, Chem. Rev. (1965); (b) H. Salman. and
64. Knozinger, Jacob, Singh, and P. Hofmann, Surf. (1993).
65. H. Nakabayashi, Kakuta, and A. Ueno, Bull. Chem. (1991).
66. (a) Thomas. (1949); (b) J. Plank, Colloid Sci.
(1947).
67. K. Tanabe, Sumiyoshi, K. Shibata, Kiyoura, and Kitagawa. Bull. Chem.
Jpn. (1974).
68. Tanabe, and Hattori. New Solid Acids and Bases
(Stud. Surf. Catal. Vol. Kodansha, Tokyo, 1989.
69. H. Solid State Chem. (1984).
70. (a) B. Miller, and E. I. Catal. (1994); (b)
Miller, Johnston, and I. Catal. (1994); (c) Miller and
I. Catal. (1996).

(a) Connel and A. Dumesic, Catal. (1986); (b) Connel and
A. Dumesic, J. Catal. (1987).
P. 0. Scokart and P. Rouxhet, Colloid Interface (1982).
A. V. and Ortega, Catal. (1985).
A. Sarbak. and A. Somogyvary, Catal.
(1985).
(a) and N. Catal. (1969); (b) Sachett, P.
Da Lam, Dufaux. Frety, and Primet, Bull.
France (c) C. Narayanan, A. and
A, Appl. Catal. A: General (1995).
P. Berteau, A. and Chem. Faraday. Trans.
(1991).
(a) Arata, and N. Yamagata. Bull. Chem. Jpn. (1990);
(b) Chen, Courdurier, and Vedrine, Catal, (1993);
Riemer, Hunger, A. H. and
Chem. Chem. Commun. (d) J. W. de
Janchen, Lei, V. Schunemann, L. van de Ven, H.
Sachtler, and A , van Santen, Catal. (1995).
(a) Lunsford. H. Sang, S. Campbell, and Anthony,
Catal. Lett. (1994); (b) Courdurier, and Vedrine,
Catal. (1995); (c) A. V. Fornes, I. Juan-Rajadell, and
Lopez-Nieto, Appl. Catal. (1994); (d) Valyon,
Engelhardt, and Catal, (1996).
L. V. Kazansky, and Catal.
(1994).
(a) R. Miranda, and Davis, Catal. Lett. (1994); (b) A.
Ward and E. I. J. Catal. (1994); (c) Morterra. Cerato, and V.
Catal. Today (1993).
M. R. Gonzales, M. Kobe, and A. Dumesic, Catal.
(1996).
I. Wachs and K. in Characterization of Catalytic Materials ( I . E.
Wachs and L. Fitzpatrick, Butterworth-Heinemann, Boston, 1992, p. 69.
(a) Deo and I. Wachs, (1991); (b) Deo and I.
Wachs, Catal. (1991); (c) Kim, Soeya, and I.
Wachs, J. Catal. (1992).
J. A. L. and A. Lopez Agudo,
Catal. (1985).
and Jurczyk. Appl. Catal.
(a) Adachi, Contescu, and A. Schwarz, Catal. (1996); (b)
Adachi, C. Contescu, and A. Schwarz, Catal. (1996).
P. Courty and Preparation of Catalysis Amsterdam,
1983. 485.
(a) J. Sauer, Chem. Rev. (1989); (b) Kramer and A, van Santen,
Am. Chem. (1993); (c) P. P. Sherwood, and I.
Hillier, Phys. Chem. (1994).

(a) Grabowski, and Yoneda, J. Catal. (1980); (b)
Hawakami, S. Yoshida, and Yonezawa, Chem. Trans.
(1984); (c) Zhidomirov and V. Kazansky, Adv. Catal. (1986).
(a) A. Rabo (ed.), Zeolites Chemistry and Catalysis ACS Monograph No. 171,
American Chemical Society, Washington DC, 1976; (b) Mortier and A.
Schoonheydt, Solid State Cheni. 16: 1 (1985); Barthomeouf, Mater.
Chem. Phys. (1987); (d) P. 0. Fritz and Lunsford, Catal.
(1989); (e) P. A. Jacobs, Catal. Rev. Eng. (1982); (f) A. Corma,
Chem. Rev. (1995); (g) A. van Santen and Chem. Rev.
(1995).
Uytterhoeven, Christner, and Hall, Phys.
(1
(a) I. D. I. Senchenya, A. I. Lumpov, Zhidomirov, and V.
B. Kazansky, Kinet, (1979); (b) J. Mortier, Sauer, A.
Lercher, and Noller, Phys, Chem. (1984).
Smith, A. Morris, Marchese, Thomas, P. A.
Wright, and Chen, Science (1996).
Freude. Hunger, Pfeifer, and W. Schwieger, Chem. Phys. Lett.
(1986).
D. Barthomeuf and Beumont, J. Catal. (1973).
(a) Mortier, Catal. 55:138 (1978); (b) P. A, Jacobs, Catal. Rev. Sci.
(1982).
(a) T. Chu and Chang, Phys. Chem. (1988); (b) D.
J. Mortier, 0. H. Kenter, Janssen, and J. Verduijn,
Catal. (1991).
(a) Sauer, Chem. Rev. (1989); (b) Kassab, and Allavena,
Phys. Chem. (1988).
Derouanne, Catal. (1986).
A. Zicovich-Wilson. and P. Viruela. J. Phys. Chem.
(1994).
(a) J. Dwyer, Stud. Surf. Sci. Catal. (1988); M. Thomas and
Klinowski, Adv. Catal. (1985); (c) A. Martin, Wolf, S. Nowak, and
Zeolites (1991).
(a) M. Sonnemans, den and Phys. Chem.
(1993); (b) Coster, A. L. Blumenfeld, and J. Phys. Chem.
(1994).
Yin, A. Blumenfeld, and V. Phys. Chem. B 101-1824 (1997).
M. Huang, Kaliaguine, and A. Auroux, Phys. Chem. (1995).
(a) P. Boehm, Carbon (1994); (b) P. Boehm, High Temp. Press.
(1990).
0. Taushkanova, P. Smirnov, and A. A. Phys.
Chem. (1993).
(a) and P. Ehrburger, Carbon (1977); (b) and
Carbon (1987).
Tore, A. S. Arico, V. Alderucci, V. Antonucci, and Giordano, Appl. Catal.
A (1994).
V. A. and E. Weiss, Aust. J. (1957).

P. and M. Voll, Carbon (1970).
A. Leon y Leon. Solar. V. Calemma. and Carbon
(1992).
(a) P. Boehm, E. Heck, and R. Sappok, Chem. Int. Ed. Eng.
(1964); (b) Boehm, Adv. Catal. (1966); (c)
Carbon (1968) (d) in Chemistry and Physics of Carbon (P.
Walker, Vol. 6, Marcel Dekker, New York, p. 191.
(a) Matsumura, S. Hagwiara. and Takahashi, Carbon (1976); (b)
Neffe, Carbon (1987); (c) A. Antonucci, and
Giordano, Carbon (1989).
(a) A. Contescu, Contescu, Putyera, and J. A, Schwarz, Carbon
(1997); A. I. Contescu, Putyera, and A.
Carbon
(a) Brown, R. R. Bushing, and A. Levy.
Sect. B (1977); (b) I. V. Kozhevnikov, A. R . I.
and van Bekkum, Catal. Lett. (1994).
(a) B. and Moffat, J. Catal. (1991); (b) Bonardet,
Moffat, and Colloids Surf. A (1993).
I. V. Kozhenvnikov, Catal. Rev. (1995).
(a) Lee, Arai, S. Nakata, S. and J.
Am. Chem. (1992); (b) T. Okuhara, A. Ayakawa,
Yoneda, and Catal. (1983); (c) Catal. Rev.
Eng. (1987).
and Okuhara, CHEMTECH (1993).
(a) Solid and Bases, Academic, New York, 1970; (b) 0.
Catalysis b y Non-Metals, Academic, New York, 1970; in
Catalysis by Acids and Bases C. Naccache, G . Ben
Taarat, and Vedrine, Amsterdam, 1985 75;
Goldstein, in Experimental Methods in Catalytic Research Anderson and
P. Vol. 1, Academic, New York, 1968, p. 354.
(a) Highes and H. White, Phys. Chem (1967); (b) A.
Catal. (1993); (c) S. Khabtou, Chevreau, and Lavalley,
Microporous Mater. (1994); (d) A. Tutek, and
Catal. (1992).
and Forni, Faraday Trans.
92:1401 (1996).
(a) Knozinger, and P. Catal. (1977); (b)
A. and V. Fornes, Zeolites (1993); (c) A. Rodellas, and
V. Fornes, Catal. (1984).
(a) A. Lercher, H. and Ritter, Chem. Faraday Trans. I
(1981); (b) P. A. Jacobs, Catal. Rev. Eng. (1982); (c) and
V. Lorenzelli, Mater. Chem. (1982).
(a) E. A. Paukshtis and Yurchenko, React. Kinet. Catal. Lett.
(1982); Morterra, Cerrato, C. and V. Catal.
(1993); (c) M. A. C. and E.
Benazzi, Zeolites, (1993).

(a) A. Benesi, Am. Chem. (1950); (b) A. Benesi and
Adv. Catal. (1978).
(a) Take, Nomizo, and Yoneda, Bull. Chem. Jpn. (1973); (b)
Balikova, React. Kinet. Catal. Lett. (1975).
and N. C. Dispenziere, J. Catal. (1989).
(a) Cardona-Martinez and A. Dumesic. Adv. Catal. (1992); (b)
Sayed, A. Auroux. and J. Vedrine, Appl. Catal. (1986); (c) J. Parillo
and Gorte, Phys. Chem. (1993).
Parillo and Gorte, Catal. Lett. (1992).
Farcasiu and A. Ghenciu, Am. Chem. (1993).
Cardona-Martinez and A. Dumesic, Catal. (1990).
(a) Cardona-Martinez and A. Dumesic, Catal. (1991); (b) N.
Cardona-Martinez and A. Adv. Catal. (1992); (c) J.
Parillo and J. Gorte, Catal. Lett. (1992); (d) Karge, in Catalysis
and Adsorption by Zeolites Ohlman, and R. Fricke,
Amsterdam, 1991, p. 133.
J. L. Krasnogor, and A. A. Castro, React. Kinet. Catal. Lett.
(1986).
Pereira and Gorte. Appl. Catal. A. (1992).
P. J. Dallons. and A. Van Gysel, Appl. Catal.
(1 99 1).
(a) Barthomeuf, in Molecular Sieves, Vol. (J. R. Katzer, ACS Symp.
American Chemical Society, Washington. DC, 1977, p. 453; (b) Poncelet,
and A. Jacobs, in Molecular Sieves, Vol. Katzer,
ACS American Chemical Society. Washington, DC, 1977, p. 607.
N. Bronsted and Pederson, J. (1924).
Yoneda, J. Catal. (1967).
(a) Contescu, V. Miller. I. and A. Schwarz, Catal.
(1995); (b) Contescu, V. Miller, I. and A.
(1996).
(a) Arnett, in Proton Transfer Reactions Caldin and V. Gold,
Chapman and Hall. London, 1975, p. 88; (b) Wolf, P. Harch and W.
Taft, Am. Chem. (1975).
A. Bleiber, and Chem.
(1993).
Spitz. J. Barton. Barteau, R. Staley, and A. Sleight, Phys. Chem.
(1986).
A. Gervasini and A. Auroux, Phys. Chem. (1993).
(a) S. Kawi, Chang, and Gates, Catal. (1993); (b) Kawi,
Xu, and Gates. Chem. (1994); (c) Kawi, Chang,
and B. Gates, J. Phys. Chem. (1993); (d) A. Zhao and B. C. Gates,
Langmuir (1997).
5. A. Rabo and Catal. Rev. (1989-90).
A, Rabo, Catal, Rev. (1981).
A. R. A. van Santen, Janchen, and Phys.
Chem. (1993).

Marchese, P. A. Wright, Chen, and Thomas, J. Phys. Chem.
(1993).
Batamack. Dorieux-Morin. Vincent, and Phys. Chem.
(1993).
Smith. A. Cheetham, Morris, Marchese, Thomas, A.
Wright, and J. Chen, Science (1996).
(a) and J. White, Colloid Interface (1979); (b)
Harter and J. L. Soil. Sci. Am. Proc. (1969); W.
Karickoff and Bailey, Clays Clay Miner, (1974).
(a) Putyera, J. Jagiello, T. Bandosz, and A. Schwarz, Chem.
Faraday Trans. (1996); (b) Bandosz, Jagiello, Putyera, and
A. Schwarz, Chem. Faraday Trans. (1994).
(a) Pinnavaia, Science (1983); (b) and P. Tindwa,
Clays Clay Miner. (1983); (c) Chevalier, R. Franck, Suquet,
Lambert, and D. Barthomeuf, Chem. Faraday Trans, (1994); (d)
Lambert, S. Chevalier, Franck, Suquet, and Barthomeuf, Chem.
Xie, and Zhang, Appl. (1991).
Hiemstra, C. M. de Wit, and W. van Colloid Interface Sci.
105 (1989)
Ludwig, P. Schindler, Colloid Interface (1995).
Marshall, Ridgewell. and C. Rochester, Ind. (Lond.)
(1974).
Radioanal, Chem., Articles (1988).
C. Contescu, unpublished results.
Petro, Misak, and I. El-Naggar. Surf. (1990).
0. Spalla, and Colloid Interface (1993).

Institute of Physical Chemistry, Bulgarian Academy of
Sciences, Sofia, Bulgaria
Department of Applied Chemistry, Gifu University, Gifu,
Japan
I. Introduction
II. Electro-Optical Effects in Suspensions of Spherical Particles
A. Electro-optical methods
B. Characterization of the samples
C. Electro-optical responses of spherical particles
D. Discussion
I I I . Electro-Optical Effects in Suspensions of Anisometric Particles
A. Experimental results
B. Discussion
References
Colloidal electro-optical studies started with the discovery of the Kerr effect [I].
The theories of colloidal electro-optical phenomena were strongly influenced by the
classical molecular electro-optics theory [2], particularly by the part concerning
dipole molecules The latter explains the display of negative Kerr constants
by the existence of permanent dipole moments perpendicular to the particle sym-
metry axis. The first formal theory pretending to explain the extremely high
values of the Kerr constant for colloids attributed them to particle volume.
Experimental studies demonstrated that the processes causing electro-optical phe-
nomena in colloids are much more diverse than those in molecular systems
Deforming the ionic atmosphere surrounding the colloidal particle, the electric
field, causes, on the one hand, particle electrokinetic motion (hence flow-induced
particle orientation), while on the other hand it creates an induced dipole moment,
polarization-induced particle orientation, the Kerr effect.

Low-frequency "anomalies" in the electro-optical responses of aqueous colloidal
systems (sign reversal and accompanying deviations from Kerr behavior) have
attracted scientific attention since the very beginning of colloidal electro-optical
research. Starting in 1939 with the papers of Lauffer [9] on the tobacco mosaic
virus, the phenomena were studied over the years on an impressive number of rigid
colloids, micelles, etc. This tendency con-
tinues to this day. It is sufficient to mention the systematic investigations on
electrolytes of Oppermann and Kramer and Hoffmann the numerous
interesting results of and Hoffmann and coworkers on
micelles, the theoretical schemes on rod-shaped particles the
attempts to find a model sample for analysis of the electro-optical phenomena
the impressive pictures obtained on gels concerning gel electrophoresis as only
a part of this activity. Actually, the history of colloidal electro-optics is a history of
the investigations on the low-frequency electro-optical "anomalies."
The experimental observations gave birth to a series of theories competing to
explain the phenomena. Treating the peculiar concentration dependencies of the
low-frequency electro-optical responses, Lauffer was the first to indicate their
general character, citing the similar results on rod-shaped [9] and disk-shaped
particles and advancing the concept of "interaction forces." In the subse-
quent treatments of Sackmann and Rabinovich the colloidal surface
double layer was considered responsible for the sign-reversal phenomena.
Further, various concepts of particle properties and interaction forces were pro-
posed to explain their origin: permanent dipoles, either inherent or induced
by interactions asymmetry of surface charge field-induced deformation
optical interactions interaction-enhanced transverse conductivity
cluster orientation etc.
The sign-reversal phenomena in the semidilute regime were demonstrated
experimentally on a large variety of samples: rigid and flexible, of varying and
constant size and axis ratio. and of various from disks to rods. In our recent
papers we extended this list, reporting observations of similar effects on suspen-
sions of rigid isotropic spherical particles Evidence was presented for the
electroacoustic origin of the observed effects. Since all theoretical schemes con-
structed to explain the electro-optical sign reversal are based on the assumption
of transverse particle orientation (or deformation), our results concern an essential
problem in the understanding of the phenomenon. The observation of such effects
on a stable suspension of isotropic particles signifies that particle optical anisotropy
(or deformability) is not a necessary condition for the display of sign reversal.
Actually, the basic concept applied in the analysis of colloidal electro-optical phe-
nomena is their assignment to anisotropy fluctuations (Kerr effect or electrophore-
tic orientation), directed translation modes being ignored.
In the last decade the electroacoustic waves in colloids attracted consider-
able attention. Electrically induced concentration fluctuations were detected by
means of different experimental techniques. The Bragg diffraction technique was
used to analyze the concentration gradients induced in colloidal crystals by dc
electric pulses Density fluctuations in electric fields were observed with
time-resolved reflection spectroscopy and transmitted light spectroscopy
The conventional light scattering was also applied to detect the induced translation

Spectroscopy of
modes Our research is directed to the possibility that they could grow to be
large-scale and very significant in the hydrodynamic frequency range and could
influence the optical response of the system both directly and through acoustic
orientation of the particles. Thus for interacting systems self-orientation and col-
lective orientation modes are coupled with self-translation and collective transla-
tion modes, resulting in a frequency behavior of the electro-optical
responses. Although the precise mechanism of low-field electro-optical sign reversal
is not yet well understood and our studies are in the initial stage, we shall try to
demonstrate here the close similarity of the electro-optical behavior in the hertz
range of spherical and anisometric particles, hence the importance of
tic modes for the electro-optical anomalies in charged colloids.
An essential feature of this study is the choice of model samples. They are rigid
particles in highly stabilized suspensions (liquid and crystal phases), hence inter-
ference of field-induced deformation and aggregation of particles is avoided. The
highly monodisperse isotropic spheres are amorphous silica, thus neither form nor
intrinsic optical anisotropy can be related to the particles, and surface effects
attributed to latexes are also avoided. The anisometric particles are thin rods of
the clay mineral paltgorskite of axis ratio 20. The sample is and its
basic advantage is the great amount of electro-optical data accumulated in the
course of its application as a model for research into the origin of the electro-optical
phenomena. It displays a very large Kerr effect at high dilution and all features of
low-frequency anomalous behavior in the semidilute regime.
The electro-optical responses of colloidal systems were detected by reflection
spectroscopy and by the variations of scattered light intensity. Induced variations
in the reflection spectra of colloidal crystals turned out to be very sensitive to
electroacoustic effects, while the experimental setup of the second method permits
direct comparison with numerous data on anisometric colloids. Investigations into
reflection spectroscopy were carried out by Okubo and coworkers a t Gifu
University (Japan), and the light scattering studies by Stoimenova, Okubo, and
coworkers at the Bulgarian Academy of Sciences. The essential parts of the results
are due to our joint research.
General Aspects
Conventionally the electro-optical effects of interacting colloidal systems are
presented by the statistical average of the product of the optical effect of a single
particle and the orientation distribution function
a =< >
where the optical part is a function of the components of the particle-scattering
matrix. Thus, for field-induced birefringence and dichroism, the optical factor is
determined by the diagonal elements of at forward scattering:

An -
An' -
while the induced variations of scattered light intensity are followed by any pre-
ferred combination of the components of light polarization) as a function of
the observation angle:
For example, for light scattering observed from the side
and for backward scattering.
The above presentation illustrates the similar bases of the electro-optical meth-
ods y l ] . The components of the scattering matrix can differ in their sensitivity to
the various sources of particle anisotropy, but they yield similar information on the
induced anisotropy fluctuations, particularly on their relaxation. This scheme is
true in so far as noninteracting particles are concerned. Though the correlations
in particle orientation have been accounted for in recent electro-optical the-
ories, the influence of the applied field on the space distribution of the scattering
is generally neglected.
In interacting systems the optical and orientation factors in a are no longer
separable quantities. The induced optical effect is determined both by the single
particle scattering [form factor and by the pair distribution function [structure
factor the latter being direction-dependent Since in addition to orienta-
tion the electric field causes particle translation, even for spherical particles the
radial distribution function and the "static" structure factor attain
dependent induced anisotropy. The deformed surface potential also contributes to
this effect.
Formally introducing modified values for and (in the general case the
process involves quantities), we can write for the steady-state elec-
trical light scattering,
The field-free decay of the electro-optical response, displaying relaxation of the

induced anisotropy, can then be related to the structure factor,
In the ideal case of noninteracting particles = is the

dynamic structure factor that describes the single-particle motion [for spheres,
is constant and no electro-optical effect is observed]. In the presence of interactions
(including field-induced interactions), the decay process contains information on all
single and collective motions caused by the field. The latter can be enhanced by
appropriate choice of the applied field frequency and by the pulse duration. Basic

factors in this process are the area of observation (considerations with respect to
this factor are similar to those presented for the photon correlation spectroscopy)
and the pulse duration. The latter quantity actually determines the number of
experimentally detected dynamic modes.
2. Electrical Light Scattering

The light scattering observations of the electrically induced changes in the colloidal
systems were performed on a conventional apparatus used in numerous experi-
ments on the anisotropic colloidal systems
The measuring cell is shown schematically in Fig. Attached to a sled, the cell is
placed in a metal thermostated box. The movement of the sled in the box is
restricted in the direction of the incident light beam. Scattered light intensity is
measured at an observation angle of using a photomultiplier or oscilloscope
registration. The electric pulses are applied perpendicularly to the observation
plane. The electrodes are platinum plates of radius a t a distance of
apart. The volume of the cell is about the diameter is and
the distance between the front cell wall (of incidence) and the electrodes is about
5 mm.
The scattered intensity of the unperturbed system is compensated, and the
field-induced light-scattering variations = with time are observed
directly on the oscilloscope. The dc component of the effect. is determined
taking for the middle of the modulated steady-state signal, the ac component
by the amplitude of the steady-state oscillations. The results presented in this
chapter were obtained by the use of white light.
Besides the advantage of direct comparison with previously obtained results, the
construction of this cell permits the observation of propagating acoustic modes in
the crystal colloidal samples.
Scheme of the measuring cell for electrical light scattering.

3. Electrical Light Reflection
Field-induced variations of reflected light intensities and spectra were recorded by
the use of Y-type optical fiber cables. The measuring cell (rectangular,
248 x contains two parallel glass plates coated inside with electrically
conducting membranes (Sanyo Kogio Co., Tokyo). The optical pathlength
is 3mm. The incident beam, (from a tungsten is introduced into the cell
parallel to the electric field by a fiber cable, and reflected light is led in the same
way to a registration. The time-resolved reflection
spectra were recorded on a PMA-50 photonic multichannel analyzer
(Hamamatsu Photonics,
of the
Highly monodisperse spherical particles were the main object of our studies. The
colloidal silica spheres CS-81 and CS-91 used in the investigation are products of
Catalyst Chemicals Ind. Co. (Tokyo). By electron microscopy (TEM,
Type the spherical mean diameter and standard deviation were
determined to be 103 13.2 for CS-81 and 110 4.5 for The charge
density of strongly acidic groups on the spheres determined by
titration with a autobalance precision bridge were 0.38 and 0.48
c m , respectively. Before use the samples were purified with an ultrafiltration cell
and treated continuously on a bed of and anion-exchange resins
20-50 mesh]. The resins were used for the deionization of
water needed for the suspension preparation. The suspensions were obtained by
dilution of the stock solution with water and were further kept in contact with the
ion-exchange resins for a week in order to achieve the crystal state (conductivity
below ohm-' at Crystals in the dilute suspensions were very large
which is an advantage for light scattering observations of the dynamic
modes.
Turbidity of liquid like and crystal-like suspensions is very low compared to that
of systems due to the diminished intensity of multiple scattering for visible
light. Particularly low turbidity is observed for highly dilute and exhaustively deio-
nized samples This made it possible to follow the static and dynamic light
scattering behavior of the silica suspensions in gaseous, liquid, and crystalline states
Data analysis was performed with the and histogram
including Marquadt analysis. A single exponential decay was found only in the
colloidal gases, corresponding to free particle translation diffusion. The diffusion
coefficient is always lower than the value calculated with the Stokes-Einstein equa-
tion with the true diameter of spheres and correlates well with the effective diameter
including the atmosphere. The liquid samples display bimodal decay, corre-
sponding diffusion coefficients (differing by a factor of 3) being related to
and long-term self-diffusion of particles. static light-scattering variations for
the crystal state display very sharp peaks in their scattering vector dependence. The
analysis of crystal structure dynamics yields three apparent diffusion coefficients
that differ from each other by an order of magnitude: the basic one related to
particle (restricted) Brownian diffusion, a slow mode assigned to propagating

shear waves in the crystal lattice, and a fast possibly related to
electrolyte interactions.
Typical electro-optical responses of the suspensions of spherical particles

recorded by the scattered light intensity are presented in Fig. 2. They are detected
for the crystal state of the systems and for different intensities and frequencies of
the applied sine-wave electric pulses. The low-frequency responses are modulated;
they follow the field frequency at sufficiently low field intensity and exhibit a double
frequency modulation at higher field intensity. Two different time scales are
involved in the decay of the responses and 1 s), which can be both exponential
and oscillatory. At higher field intensity or frequency the effects cannot be distin-
guished by the responses of anisotropic colloids.
Figure 3 presents the frequency dependence of the electro-optical effects for
suspensions of different volume fractions of spherical particles. The stock solution
(8% sphere volume fraction) is highly deionized, and dilution is carried out with
deionized water. Thus the highly dilute samples are in the liquid-phase state.
Modulated responses are detected at low frequencies for highly dilute samples
. 2 Electrical light scattering (ELS) responses from highly deionized suspensions of

silica spheres CS8 1 of volume fraction 1 (dotted line, electric field). (a) 5 Hz, 15
(b) 0.5 Hz, 15 (c) 1 kHz, 400 (d) 10 Hz, 30 (outside the electrodes); (e)
10 Hz, 60 (f) 10 Hz, 120

3 Frequency dependence of the ELS responses from deionized suspensions of
(field intensity for different sphere volume fractions.
(0.1 %), below 1 Hz; for semidilute samples they cover the hertz interval, increasing
to higher frequencies with increasing volume fraction. In parallel a dc component is
observed, which, depending on volume fraction and ion content, can cover the
entire frequency range (up to 100 kHz), typically related to orientation effects. In
the semidilute regime the low-frequency sign reversal is observed, in close similarity
to the low-field responses of anisotropic systems. At higher sphere concentrations
(above 5% volume fraction), the responses are again positive.
The field intensity dependence of the ac and dc components of the responses in
the intermediate concentration range are complicated: linear at very low fields and
quadratic or rapidly saturating at higher fields On a larger scale they show
oscillatory behavior. In Fig. 4 they are shown for the transition concentration of
sphere volume fraction for moderate field intensities. The component
increases with field intensity and can be observed up to higher frequencies.
Modulated signals are obtained up to in the range of the negative dc
component.
Figure 5 shows typical responses for this particular sample (0.5% sphere volume
fraction). A basic difference in the electro-optical behavior of spherical compared
to anisometric particles is the linear response a t low field intensity. The modulated
signal follows the frequency of the applied field with a phase difference of
With an increase in frequency the ac component relaxes, changing in both ampli-
tude and phase and passing to double frequency, while the dc component changes

3
-2
0
0 1 2 3
log (Hz)
Frequency dependence of the ELS responses from deionized suspensions of
silica spheres (sphere volume fraction 0.5%) for different field intensities. 150
ELS responses from deionized (liquid-state) suspensions of of volume

fraction (dotted line, electric pulses).

112 Stoimenova and
from a negative value to a small positive value. The decay process displays two
basic relaxation times: one, associated with the positive kilohertz effect, on the
order of milliseconds and the second, related to the low-frequency modulations,
on the order of seconds. Additional slower modes can be observed upon applica-
tion of longer pulses, as can be seen in the kilohertz response. This figure concerns
the liquid state of the sample.
The electro-optical responses are strongly dependent on the phase state of the
colloidal system. The transition of the same sample the liquid to the crystal-
line state is accompanied by a considerable increase in the cutoff frequency of the ac
component and a displacement of the dc curve to more negative values (Fig. 6).
Similar variations can be observed by increasing of the sphere volume fraction, as
demonstrated in Fig. 3, which shows that the effective fraction determines
the process. In parallel, additional modes are observed to appear in the crystalline
state. They introduce resonance peaks at frequencies depending on the sphere
volume fraction (two fractions are presented in Fig. 6).
Typical responses of the crystal were shown earlier in Fig. 2. Below 1 Hz
they are similar to the responses of the liquid phase (Figs. 2a and 2b) and are
evidently related to induced longitudinal density waves. At low fields they follow
the field frequency (Figs. 2d and and at higher fields a harmonic frequency
appears (Fig. In the interval the responses show oscillatory decay.
Estimating the parameters involved, we identified this mode as related to shear
waves propagating in the crystal lattice Hence the registered peak frequencies
are a relative measure of the elastic modulus of the system. The kilohertz effect is of
reversed sign and displays distinctly two relaxation rates (Fig. The first one
0 1 2
log f (Hz)
Frequency dependence of the ELS responses deionized suspensions of
CS81 (field intensity for different sphere volume fractions at liquid and
crystal states of the sample.

(milliseconds) is most probably related to particle self-diffusion, and the second one
(on the order of seconds) concerns collective motion.
Detecting the scattered light while moving measuring cell in the direction of
the incident light beam permitted the observation of propagating shear waves in the
cell outside the electrodes for the crystal samples The phenomenon is illu-
strated in Fig. 7 , where the formation of standing waves upon application of lattice
self-resonance frequencies can be observed.
Systematic investigations of the induced density waves in suspensions of silica
spheres in the low-frequency range were performed by the use
of light reflection techniques X-Y type plots and Fourier transforms of the
reflection signals demonstrated the phase variations with both field frequency and
intensity for a different experimental setup (incident and scattered light along the
field direction). Detected phase changes seem to be relevant to inherent properties
of the samples. Despite differences in the phase variations of the responses
obtained by the two techniques, a basic feature of the effects is preserved. In the
linear low-field regime the electro-optical effects of aqueous systems display n o
harmonic components; the increase of field intensity changes the phase of the
response and induces a double frequency component. The picture is quite
to the responses obtained by electrical light scattering (Fig. 8).
Since time-resolved reflection spectroscopy allows for a more detailed study of
the reflected intensity, it was used to investigate the origin of the harmonic
.7 ELS responses from deionized (crystal-state) suspensions of of volume

fraction detected at different distances from the electrodes (dotted line, electric
pulses; denotes the center of the electrode space). (Adapted from Ref. 36.)

FIG.8 Electric reflection (ER) responses from deionized (crystal-state) suspensions of
CS91 of volume fraction 4.4% for pulses of 0.2 Hz and different field intensities (dotted
line, electric field).
component in the electro-optical response Detailed investigations of the spec-

tra of the responses of silica spheres in water and other solvents of different con-
ductivity demonstrated that the method is informative for the analysis of the
acoustic modes in colloids. Figure 9 shows schematically the variations of the
reflection spectra in the course of one cycle of the applied sine-wave electric field.
They are compared to variations in the integrated reflected intensity. From the
changes in the peak wavelength it is clear that the basic process at higher fields
remains the same: oscillation of interparticle distance, density waves. A differ-
ence is seen in the variation in the integrated reflected intensity. While the peak
wavelength variation follows the field frequency, the integrated intensity shows
double frequency modulation. The transition between the extreme wavelengths is
strongly asymmetric. The asymmetry is evidently due to different decay rates of the
densified and rarefied states to equilibrium. The line half-width increases (particu-
larly in the process of rarefaction), most probably due to reduced correlation
length. The results obtained by reflection spectroscope demonstrate that a double
frequency optical response can be obtained by the electrically induced translation
modes as well. The origin of the dc component (including sign reversal) can find
similar explanation.
Theoretical analyses predicted the propagation of low-frequency shear waves in

both semidilute polymer solutions and dilute colloidal crystals Different
experimental techniques were applied for their detection and for the determination
of the shear modules of colloidal samples The dispersion equation of the
transverse waves for the low-frequency regime (wavelengths much larger than the
interparticle distance)

time
0 1 2 3 4 5
FIG. 9 Time-resolved reflection spectra for one cycle of the applied field (0.2 Hz,
for a deionized (crystal-state) suspension of CS91 of fraction 4.4%. Upper
figure: Corresponding integrated intensities. (Adapted from Ref. 48.)
where 1 is the mass density, poise) the viscosity of the solvent

(water) and E the shear modulus of the sample.
For the volume fractions presented in Figs. (3-6) the shear modulus is on the
order of and the sound velocity V(= The micro-
scopic relaxation time and the attenuation length
= For frequencies below 1 kHz the dissipation
is small and the shear waves are propagating. The dimensions of the measuring
cell encourage the formation of standing waves.
The observed resonance-like behavior of our samples corresponds to the
expected parameters of induced shear waves, thus giving evidence for the origin
of the electro-optical effect. The decay rate of this mode corresponds to the slow
mode detected by dynamic light scattering in the unperturbed system In the
general case the analysis of light intensity scattered by electrically induced fluctua-
tions would be complicated. It depends strongly on the area of observation. If the
latter is sufficiently small, local density fluctuations are detected at low frequency
and scattered intensity follows directly the variations in the acoustic pressure. For
the gaseous state of the samples in the low-field regime it can be presented using
equation
AP) V

where is the system's complex dielectric constant, is the particle volume frac-
tion, is the particle relative density, and is the particle velocity in the
applied field. An additional term should be written to account for the entropy
fluctuations, which are evidently significant in the liquid state of the samples.
The introduction of Coulomb interactions will add to the above equation terms
accounting for correlations.
At higher frequencies, small-scale motions become predominant and light scat-
tering is determined by average fluctuations. Higher field intensities (conventionally
used to obtain electro-optical saturation) introduce nonlinear effects such as polar-
ization of the and interaction of acoustic waves. Thus the more general
picture can become highly complicated. Nevertheless, the above equation displays
several basic features of the process compared to Kerr electro-optical behavior.
First, it predicts the linear dependence of the process on particle velocity, hence on
field intensity and particle electrokinetic potential. It involves frequency- dependent
particle mobility, which enhances significant phase variations of the optical
response. Further, interparticle interactions could induce a considerable transverse
component of particle mobility, thus giving rise to transverse density waves. Next,
the equation predicts dependence on particle relative density that is much stronger
than that expected for anisotrcpy fluctuations. Particle volume fraction influences
the process both directly and through interactions. As can be seen from the figures,
the electro-optical responses of spherical particles display the basic features of the
low-frequency effects in colloids.
Several figures are presented in this section to illustrate the similarity of the
frequency behavior of spherical and anisometric particles. They were obtained by a
conventional electro-optical method (electric light scattering) on a typical and
studied colloidal sample The anisometric particles are rigid rods of the
clay mineral palygorskite (attapulgate). Their average length and diameter are
200 and respectively. Particle surface charge in the aqueous suspensions
is about A more detailed investigation of the sample is presented in
Refs. 58-60.
Figure 10 is a typical picture of field intensity and frequency variations of the
electro-optical responses of highly dilute aqueous colloids. The frequency curves of
the anisotropic particles (below the overlap concentration) are significantly differ-
ent from those of the spheres, due to the large effect in the kilohertz range. The
decay of the responses of the investigated sample is nearly exponential, the time
constant of several milliseconds (2-7 corresponding to particle rotational diffu-
sion. A tail on the order of 50 ms can still be detected. The high-frequency relaxa-
tion is proved to be related mainly to surface polarization of the particles, which
causes their orientation. The low-frequency relaxation interval coincides with the
cutoff frequency of the component (hydrodynamic range), where large variations

FIG. 10 Variation of the electro-optical responses of a highly dilute palygorskite
suspension (volume fraction 0.001%) with frequency and intensity of the applied
electric field.

in the effects (including sign reversal) are displayed in the semidilute regime or upon
modifications of the ionic content of the systems.
Deionization of the sample for 1 month induces changes in the frequency curves
as shown in Fig. 11. The decay of the kilohertz effect is biexponential. The short
time, corresponding to self-rotation, remains almost unchanged (several millise-
conds), and the long time increases drastically indicating collective rota-
tion (further deionization of the sample diminishes the amplitude of the effect and
further separates the time constants). The deionized suspension clearly shows a
twofold relaxation of the component. The cutoff frequency of the latter increases
similarly to the corresponding effect for spheres, and the width of the kilohertz line
is reduced, as is typical of correlated motions. This is a basic feature of the con-
sidered low-frequency anomalies that is difficult to explain adequately by a trans-
verse orientation scheme, since it is accompanied by a significant increase in the
decay time of the responses. O n the other hand, at low ionic strength, short-range
orientation ordering can be observed in such far below the overlap con-
centration. It is accessible to static light-scattering observation. as demonstrated by
the peaks in the static structure factor of tobacco mosaic virus under
similar conditions
Frequency dependence of the ELS responses of a dilute palygorskite

11
suspension fraction 0.01%) for different ionic strengths of the sample
dc compound; [ , ac compound; filled symbols, M open
symbols, about Upper diagram: Electro-optical responses of the
sample at field frequency 1 kHz (dotted line, deionized).

At low fields the quadratic and linear dependence of the effects in the two
frequency ranges are clearly distinguished (Fig. 12). Modulated effects are observed
up to and they are asymmetric. The amplitude of the dc component dis-
plays a deep hollow around the same frequency, reaching to sign reversal.
Approaching the overlap concentration of the rods, the amplitude of the kilo-
hertz response is reduced considerably and the sign reversal is observed at higher
electrolyte concentration. Figure 13 is for a sample of ionic strength
The same sign reversal for the same sample, obtained by increasing the particle
volume fraction, was reported in our previous papers Similarly to the
effects of spheres, the effective particle volume fraction determines the process.
Responses of the sample to dc pulses in the same range of field intensities are
shown in the same figure. The decay time constants are strongly dependent on
pulse duration. At pulse lengths below the time constants extracted by
biexponential fitting of the decay are approximately 1 and The increase in
pulse length leads to multiexponential decay involving additional slow modes.
Further deionization of the sample annihilates the Kerr line and turns the
frequency effects positive, stable, and single-frequency They are similar to the
log f (Hz)
. 12 Frequency dependence of the steady component of the ELS response from
suspension (volume fraction 0.01 for different intensities in the low-field
range.

I , . , .
13 Frequency dependence of the steady component of the ELS response from

suspension fraction 0.15%) for different intensities in the low-field
range. Upper figures: ELS responses to dc pulses.
effects shown by the more concentrated silica samples. Figure 14 shows several
typical responses of the anisometric particles.
Generally speaking, three frequency ranges are clearly distinguished in the electro-
optical dispersion curves of anisometric particles:
Below 1 H z the range of large-scale density fluctuations, which are not
averaged in the optical response and can be detected both directly and by
the induced flow orientation
2. The region around 100 Hz o n the order of particle reorientation frequency,
where translation motions reduce particle orientation
3. The Kerr line, which dominates at higher fields but only until the particle
rotation is not restricted by interaction forces
Hence, two basic factors are to be taken into account in the analysis of electro-
optical data. First is the pulse duration. A t certain values of the pulse length the
response shows steady-like behavior, which is due to the separation of the time
scales involved and the restricted sensitivity of the optical response to induced
changes in the system. Second is the choice of frequency range for the determina-
tion of the Kerr constant. The latter is related to the low-field regime, where the
acoustic modes contribute significantly. Adequate separation of orientation and

10 Hz,
20 Hz, 60
deionized
14 ELS response from palygorskite suspensions of volume fraction 0.15%

(dotted line. electric field).
translation effects requires analysis of the responses over a wide range of field
intensities and frequencies.
The results presented for spheres reject a general view lying a t the basis of all
schemes proposed for explanation for the low-frequency sign reversal in colloidal
electro-optics the necessary presence of anisotropic o r deformable species in the
disperse system as a condition for the detection of electro-optical responses. This is
not a necessary condition for the observation of low-frequency negative
optical responses either. The essential role of particle electrophoretic mobility (giv-
ing rise to electroacoustic effects) for the latter is evident. The results presented
concern a series of phenomena in different systems widely studied by researchers
involved in the field of colloids, etc. Below we
turn our attention to some of these phenomena.
It is to be noted, in the first place, that the acoustic modes in semidilute polymers
and colloids are quite similar hence the optoelectroacoustic phenomena can

also be observed polymer and polyelectrolyte solutions. Abundant experimental
data on the electro-optical behavior of micelles confirm the general similar-
ity of the electro-optical spectra. Our basic objection to the developing views in this
area concerns the interpretation of the volume fraction intervals (display of positive
effects, anomalous sign reversal, return to positive effects). We have demonstrated
similar observations on silica spheres, which could hardly be explained by the Odjik
theory. As pointed out by Cates several main features should be explained by a
theoretical scheme for the sign reversal effect: the anomalous sign and the concen-
tration dependence of the steady response; the disappearance of the anomaly at
high fields; the concentration dependence of the decay times; and the chain length
dependence of the steady response. We should add to the above several more.
First is the increase with volume fraction of the cutoff frequency of the alter-
nating response (closely related to the cutoff frequency of the sign reversal) (see
Figs. 3 and 11). The latter feature strongly contradicts the idea of cluster orienta-
tion, since the correlation length of particle rotation (and therefore the cluster size)
are expected to increase with volume fraction. The relation of this frequency to the
elastic modulus of the system is consistent with experimental data. It explains both
the universality of the phenomenon and its relation to interaction forces.
Second is the strong influence of the density ratio on the con-
centration dependence of the sign reversal This fact is always ignored in
proposed schemes and can simply be assigned to the influence of particle density
on the electro-acoustic pressure.
Third, the decay times of the electro-optical responses at dc pulses are system-
atically lower than those of the responses in the kilohertz range. This is true for
as shown by the detailed investigations of Oppermann it is
also true for the rigid particles discussed here Assignment of the
larger time constants to (collective) rotation is consistent with the fact that transla-
tion motions are restricted at higher frequency.
A considerable of papers have been and are being devoted to the
optical determination of particle permanent dipoles This is a topic of
everlasting discussions and contradictory views. Since the electro-acoustic effects
display many features of permanent dipole behavior, they are evidently relevant to
the problem. They should be taken into account even in the case of purple mem-
branes, where the relation of the negative electro-optical response to the permanent
dipole of the fragments seems generally accepted and beyond any doubt (See,
Ref 62 and papers cited therein). Having in view the universality of the phenom-
enon and the display of its basic characteristics (concentration dependence of
amplitude, frequency, and decay time of the negative effect) in the responses of
purple membranes their difference from the other colloidal systems might turn
out to be rather quantitative.
The acoustic modes deserve particular attention in studies of gels. The results
presented here for spheres are quite similar to the results reported for the
frequency electro-optical behavior of agarose gels particularly the single fre-
quency of the alternating component. Some of the figures presented in Ref. 24 for
bipolar pulses of unequal duration evidently indicate resonance amplifying of the
electro-optical signal. It is therefore not surprising that the molecular weight
separations observed in FIGE (field inversion gel electrophoresis) are highly

sensitive to pulse amplitude and duration Experiments directed to the deter-
mination of the self-resonance frequencies of the gel matrix might help to achieve
better control of gel electrophoresis.
The above examples constitute only a brief illustration of the very wide range of
phenomena and problems that may be relevant to electroacoustic effects in colloids.
Kerr, Phil. Mag. (1875).

P. Radium (1910).
Born, Ann. Phys. 55: 8).
A. Peterlin and Stuart Phys, (1930).
Heller, Rev. Mod. Phys. (1942).
V. Volkenstein, Molecular Optics, GITTL, Moscow, 1951 (in Russian).
C. O'Konski and Science (1950)
O'Konski, Phys. (1960).
M. A. Lauffer, Am. Chem. (1939).
O'Konski Molecular Electro-Optics, Marcel Dekker, New York, 1976.
B. Jennings Electro-Optics and Dielectrics of and Colloids,
Plenum, New York, 1979.
(ed.). Molecular Electro-Optics. Plenum, New York, 1981.
Watanabe (ed.), Macromolecules, Colloids, Liquid Crystals and Biological
Tokyo, 1988.
Stoylov and P. Stoylov (eds), Colloid and Molecular Electro-Optics 1991
IOP Publishing, London, 1992.
P. Stoylov, Colloid Ottevill and eds), series
'Colloid Science," Academic. New York, 1991.
Chem. (1988).
in Colloid and Molecular Electro-Optics 1991 Stoylov and
Stoylov, IOP Publishing, London, 1992, p. 93.
and Hoffman, Macromolecules (1991).
Jennings, M. and Toddy, Colloid Interface
(1989).
Angel, Hoffmann, Kramer, and Turn, Ber. Phys.
(1989).
Hoffman, U. Kramer. and H. Turn, Phys. (1990).
Grye and Phys. (1993).
F. Giardini. R. Piazza, and V. Degiorgio, J. Phys.:
Matter (1992).
S tellwagen and Stellwagen, in Colloidand Molecular Electro-Optics 1991 ( B .
Jennings and P. Stoylov, IOP London, 1992, pp.
Norton, Phys. Rev. (1939).
Mueller, Phys. Rev. 792 (1939).
Opt. Am. (1945).
Rabinovich, Phys. Rad. (1946).
J. Shah, C. Thompson, and C. Hart, Phys. Chem. (1963).

1 and
M. J. Shah, Phys. Chem. (1963).

and Watanabe, (1976).
Stoylov, Ferroelectrics (1988).
and Colloid Interface (1992).
Phys. France (1992).
and Okubo, Colloid Interface (1995).
M. Stoimenova. V. and Colloid Interface
(1996).
and C. Cannon, U.S. Patent 4,497,208 (1985).
and van de Opt. A: Opt. Image
(1984); Phys. Chem. (1985).
and Appl. Phys, (1979).
Faraday Trans. 1 (1988).
Stoimenova and Stoylov. Colloid Interface (1980).
and Faraday Discuss, Chem.
90: 1 (1990).
P. Stoylov. Adv. Colloid Interface 3-45 (1971).
Okubo, (1992).
Okubo, Colloid (1993).
Okubo. (1994).
and Colloid Polymer,
(1996).
Okubo, A. Okada. and Colloid Interface
(1998).
P. de and J. Phys. (1977).
F . Joanny, Colloid Interface (1979).
P. Pieranski. and Strzelecki, Phys. (Paris)
(1980).
and B. Russell, Colloid Interface (1981).
Lindsay and P. Chaikin. Chem. Phys. (1982).
A. N. A. Clark, R. Mockler, and Phys. Rev.
(1982).
Pieranski, (Paris) (1980).
J . Fluid (1988).
R. A. and Hunter. Faraday Discuss.
Chem. (1990).
V. Peikov, Radeva. and Stoimenova, J. Colloid Interface (1994).
Stoimenova and Radeva. Colloid Interface (1995).
V. A. and Stoimenova. J. Colloid Interface submitted.
and (1988).
S. and P. Stoilov, Colloid Interface (1994).
Todorov. and Biophys, (1982).
Carle and V. Olson, Nucleic Acids res. (1984).

. Catalysts Division, Petrobras Research and
Development Center, Rio Janeiro, Brazil
Chemistry Division, Petrobras

Research and Development Center, Rio de Janeiro, Brazil
I. Introduction
I I . Special Considerations
A. Nuclear properties
B. N M R signal narrowing effect in colloids
C. Relaxation rates in colloidal species
D. Free solution species in the colloidal system
E. Adsorption phenomena
III. N M R Studies on Colloidal Oxides and Hydroxide Systems
A. Colloidal silicas
B. Colloidal aluminas
C. Other colloidal oxides
IV. Conclusions
References
Colloids are of considerable industrial importance and have intrinsic scientific

interest. Generally they consist of a sol system, solid
particles dispersed in a liquid phase that can polymerize into a continuous three-
dimensional network called a gel. In particular, colloidal sols of oxides and
hydroxides, either as such or as precursor dispersions, have numerous and
ranging applications in catalysis, coatings, optics, microelectronics, and ceramics
and other composite materials. The importance of investigations on such systems
is unquestioned, and many articles on this field are issued every month in the
several periodicals on colloid science and materials science. Among these colloidal
systems, silicon and aluminum oxopolymers, which transfo m into an oxide

network upon calcination, have been the most investigated owing to the ready
availability of these elements in nature and thus to their practical importance.
and sols through multicomponent oxosystems such
as those of and Y-Ba-Cu [1 are important examples of other colloidal
oxides that have also been studied. Special attention has been given to the devel-
opment of the sol-gel process in the last 20 years
the several experimental techniques for studying colloid and sol-gel
chemistry, perhaps the most powerful one is nuclear magnetic resonance (NMR)
spectroscopy. allows investigation of these systems by a combination of
structural information, obtained through lineshape analysis and chemical shift
data, and indirect information from dynamic measurements of spin relaxation
diffusion.
magnetic resonance spectroscopy has been extensively used to
study aqueous solutions silicates. providing information on the
types of polymeric species in solution after partial hydrolysis. Indeed. high resolu-
tion N M R technique has been successfully used to follow the evolution of the
various chemical species in solution and has yielded detailed information on the
kinetics and mechanisms of the involved reactions These are certainly
important to understand the transformations that occur toward the resultant col-
loidal species formed in the later stages of polymerization. the other hand, the
magic angle spinning (MAS) technique in solid-state has supported a large
number of studies of gels, xerogels, and aerogels providing valuable but
indirect information on the sol precursors. In this case, the colloidal system is
inevitably disrupted with the evaporation or extraction of the solvent for the
solid-state analysis, and the original sol species may have undergone structural
changes. In contrast to the large use of N M R for analysis of chemical species in
solution and in the solid state. the use of liquid N M R for in situ investigation of
colloidal has not been extensively exploited.
Studies on adsorbed species and the surface chemistry of oxides in their normal
state have been carried out to a great extent by high resolution solid-state N M R
especially due to the widely growing area of heterogeneous catalysis. A rela-
tively small contribution is found for the use of NMR for probing adsorption
phenomena in colloidal oxides, and it is only briefly discussed here. Colloidal
systems such as organic polymers, proteins, and microemulsions are relatively
well covered in the literature and therefore are out of the scope of this chapter.
Metal colloids are not considered either.
The purpose of this chapter is to show the applicability of liquid
N M R spectroscopy to the in situ characterization of some colloidal inorganic
oxides and hydroxides. In the features and advantages of using the N M R
technique on such specific systems are summarized. In some recent applica-
tions are presented, focusing on the characterization of alumina sols. It is not our
intention to provide a review; rather we seek to highlight some
applications of N M R in this important field of study.

T h e nuclear properties of the nuclei usually found in colloidal oxides and/or hydro-
xides probed by NMR are shown in Table 1. NMR and NMR are often
instructive of structural modifications in solution o r colloidal precursors containing
organic such as the metal alkoxides during their hydrolysis in sol-gel
processes, as we briefly discuss in The properties of these two nuclei are
very well known and are not discussed here. T h e two investigated nuclei
in and are quite different in their properties. The natural
abundance and sensitivity of are considerably lower than those of
Furthermore, the spin-lattice relaxation time of silicon is much longer than
that of aluminum, which sometimes requires that either the sample be pulsed less
frequently o r a paramagnetic relaxation agent be employed. Despite these apparent
disadvantages for observation of the chemistry of silicon-based sol-gels and
aqueous silicates seems to be much better understood than that of similar systems
based on alumina. The main reason for this contrast is the strong quadrupolar
nature of F o r most nuclei, the dominant relaxation process is the
dipole interaction between the observed nuclei and other nearby magnetic nuclei. In
the case of nuclei, which have a (I and only a
moderately large ratio, the relaxation is dominated by the
polar term due to the interaction with the fluctuating electric field gradient experi-
enced by the nucleus. This phenomenon causes the resonance lines to be
considerably broadened o r even to vanish into the baseline when the nucleus is
not in a highly symmetric site. A specific approach of this relaxation mechanism for
colloidal species is summarized in
1 Comparison of Nuclear Spin Properties

Natural Comparative
abundance ratio moment sensitivity
Isotope Spin Q relative to
26.75
6.72
-3.63
6.98
-5.32
10.84
-1.51
7.04
1.12
5.54

The quadrupolar nucleus presents features similar to those of and
show properties closer to although the former is not easily
observed due to the low sensitivity. is the most appropriate nucleus to be
analyzed by of those listed in Table 1. However, the use of in
the subject under consideration is restricted to the study of some colloidal
xyphosphates or phosphorus-based molecules adsorbed on colloidal oxides
Oxygen-17 shows all the negative aspects of both and allied with a very
low natural abundance, which usually requires isotopic enrichment. Hence,
NMR has not been used very often in the characterization of colloidal oxides,
although there are notable exceptions described in the literature
Unfortunately, the properties of the magnetically active nuclei and are
very unfavorable for N M R measurements. These nuclei show a relatively low
natural abundance, low sensitivity, and a high quadrupole moment. fact, tita-
nium has received little attention from the N M R spectroscopist. Therefore, despite
the great practical importance of and associates in the colloid state, it is hardly
possible to find any reference on the use of or N M R for the characteriza-
tion of these oxides.
Effect in Colloids
Colloidal sols are arbitrarily regarded as particles falling anywhere in the range of
in diameter. However, a smaller upper limit will probably be required if
the colloidal oxide species to be observed and interpreted in a spectrum by
means of conventional Fourier transform N M R in the liquid state. First, the par-
ticles must form a stable colloidal suspension ensuring the homogeneity of the
medium. This is a basic condition to avoid a gradient of concentration in the
volume being analyzed and guarantee equal excitation of the nuclei within the
spectral range. Furthermore, the size of the particles should be sufficiently small
to allow their incoherent isotropic tumbling motions so that both dipolar couplings
and chemical shift anisotropy can be eliminated. Therefore, Brownian motion is
expected to sharpen the signals of the colloidal particle and also of related
adsorbed species. In fact, the use of colloidal suspensions for N M R analysis in
specific cases has been claimed as an alternative tool for N M R analysis of solids
replacing the magic angle sample spinning method where rapid mechanical
rotation of the solid sample is applied.
For a basic of N M R relaxation processes, which is needed for compre-

hension of the following paragraphs. the reader is referred to the basic
literature including a recent article by The two main relaxa-
tion processes related to the nuclei discussed in this chapter are to the
dipole and interactions. They are briefly reviewed as follows.

Dipole-Dipole Relaxation
As the only property of a nucleus of spin 1/2 that depends on orientation is its
magnetic moment, transitions between nuclear spin levels can be induced only by
magnetic fields. A nucleus in a liquid will experience a fluctuating field due to the
magnetic moments of nuclei in other molecules as they execute Brownian motion.
This chaotic random motion has a time scale that depends o n a of factors
and is characterized by a somewhat loosely defined term. the rotational correlation
time This is the time interval after which the molecular motion contains n o
vestige of its earlier angular momentum. The relaxation field intensity due to the
fluctuations of the magnetic fields in a liquid is related to the resonance
frequency and to the correlation time (Fig. 1). The intensity of the relaxation
field is constant over the usual N M R frequency range (flat portion of the curves)
and increases with increasing This condition is ensured when the correlation
time is much shorter than one Larmor period of the nuclei. that is, when
This is known as the region of extreme narrowing, and small molecules that
rapidly in solution (very short will be in this region. However. when is
sufficiently long, the relaxation field (and so the relaxation rate) starts to decrease
with increasing NMR frequencies. Hence, for relatively large particles, where the
rate of molecular reorientation is lower than the spectrometer frequency, the
extreme narrowing condition is not fulfilled.
The effect of varying and is also shown in Fig. 2a, where the behavior of the
intramolecular dipole-dipole longitudinal transverse relaxation
times is depicted for the homonuclear case. The main feature to note is the
minimum, which marks the limit of extreme narrowing region
<< 1 and way this minimum moves with spectrometer
frequency. The higher this frequency, the shorter becomes the maximum allowed
T o the right of the minimum. the motion is too slow to cause a n effective
relaxation, and the contribution of the transverse relaxation to the global
relaxation process is enhanced. Moreover, becomes shorter, resulting in
increased linewidths =
F o r the heteronuclear case (Fig. where Si atoms are relaxed by neighboring
protons, the relaxation behavior is similar to that shown in Fig. 2a. It can be
inferred Fig. 2b that both the longitudinal and transverse relaxation rates
are two orders of magnitude faster than in the case. Similarly, the
moves left with increasing magnetic field strength. In comparison, with
the homonuclear case, the relaxation of Si caused by protons is the dominant
relaxation process and has a allowed slightly greater than that in
the case.
The correlation time can be estimated from the particle size on the basis of the
theory of electrical dispersion
where is the viscosity of the liquid, T is the temperature, k is the

constant, and a is the radius of the particle. Assuming colloidal particles in a n
aqueous system temperature) and with diameters greater than 1
values larger than s are calculated. F r o m Figs. 1 and 2, this means that

100 1000 10000
v (MHz)
. Intensity of fluctuations. in magnetic fields in a liquid sample due to

Brownian motion as a function of the frequency, for different correlation times.
spin 1/2 nuclei of such particles will often be in the threshold region of the extreme
narrowing o r far from that. Furthermore, within the usual range of NMR fre-
quency (60-750 MHz),the magnitude of may be of crucial importance for the
relaxation behavior, especially in the range of between and
2. Quadrupole Relaxation
F o r the case of quadrupolar nuclei in molecules in the extreme narrowing region,
the quadrupole coupling constant influences NMR spectra through very efficient

.2 Plots of dipole-dipole relaxation times and versus correlation time
at spectrometer frequencies of 200 MHz (dashed line) and 500 MHz. (a)
case: two silicon nuclei separated by in a substructure;
case: a silicon and a proton separated by 258 in a -0 H -
substructure. (Bond lengths from Ref. 70; calculations according to equations
provided in Ref. 71.)

relaxation processes and is related to the equivalent longitudinal and transverse
quadrupole relaxation by equation
the asymmetry parameter and = = electric field

tensor: = nuclear quadrupole moment; = Planck's constant). In this case, the
of a resonance corresponding to a site that experiences an electric
field gradient is proportional to the rotational correlation time for molecular
motion
The simple exponential relaxation behavior dictated by Eq. (2) for molecules in
the "extreme narrowing" conditions cannot be applied to molecules of limited
mobility. Outside the region of extreme narrowing the behavior is reminiscent of
that of spin 1/2 dipole-dipole relaxation (Fig. 2), but it is complicated by the fact
that several nuclear transitions are possible. Thus for a nucleus like with
= there are five transitions, 3/2
and and the relaxation rates for nuclei undergoing
external transitions differ from those of nuclei undergoing the central
transition. This produces what is called nonexponential relaxation. Bull et a!.
presented numerical solutions for the nonexponential quadrupole relation of spin
2 and spin 7 2 nuclei. They found that the longitudinal relaxation of nuclei
with both and 7/2 spin may be described approximately as a single exponential
for all correlation However, the transverse relaxation is governed by
three and four contributions for the 5/2 and cases, respectively, and the weights
of the different contributions all have similar magnitudes for longer correlation
times. This can be seen in Fig. 3, which shows the transverse relaxation behavior of
the spin nuclei as a function of The limiting values of the coefficients of the
exponentially decaying contributions for large corresponds to
and for components I. and respectively, in the spin case and
similarly for nuclei of spin Curve I corresponds to the central transition
F o r values of less than 1, the transverse relaxation is approxi-
mately a single exponential. F o r 1, a of exponentials is observed.
but the very rapid decay (short and large relaxation rate) of the spin
component may allow the analysis in of just two exponentials. F o r
1. single exponential analysis is allowed because the contribution of com-
ponent decays very rapidly too. In this case, the signal due to the central transi-
tion of the quadrupolar nucleus can give rise to a relatively narrow resonance, while
the remaining transitions broaden dramatically, corresponding to a loss of the
NMR signal u p to (cf. decaying contribution of the components). Care
must be taken to avoid confusion between loss of intensity at chemical shift
frequency through rapid relaxation of the remaining components. indicating long
correlation times, and the loss of intensity through splitting caused by,
for example. anisotropic molecular motion,
If the quadrupole nucleus is situated in a colloidal particle of limited mobility, it
may be sensitive to second-order effects, and the central transition
signal is thought to be shifted to lower frequency dynamic shift").

. 3 Transverse relaxation of spin 5/2 nuclei in the absence of chemical exchange.
(a) The normalized amplitudes of the three exponentially decaying components. and
their relaxation rates, both as a function of Note that for the longer correlation
times, all three components are of amplitude. (From Ref. 27.)
The magnitude of this shift is related to the coupling constant and

decreases with increasing field. F o r I nuclei it is given by
where the asymmetry parameter, is usually assumed to be zero as a reasonable

approximation.

The of the signal is also dependent on the resonance frequency of the
nucleus and has been derived by and coworkers for the same I =
nuclei as
It is worth noting that the linewidth of this resonance decreases with increasing
This behavior is in sharp contrast to the situation described by Eq. (2) and to the
dipolar relaxation process, where the linewidth increases with increasing From
the field dependence of the chemical shift and described in (3) and
the values of 7 and for spin 5/2 nuclei in the slow motion limit can be
obtained. As the parameter provides a measure of the coupling between the
nuclear quadrupole and the electric field gradient at the nucleus, it reflects the
of the local environment of the nucleus and can give information
on the nature of the colloidal species, is a measure of the mobility of the species
involved and, in principle, can provide information on the magnitude of the particle
size when (1) is used.
Species in the
The colloidal state inevitably brings about difficulties for the experimentalist when
separation of the disperse phase from the dispersion medium is needed. This is the
case when the speciation and concentration of only the free soluble species have to
be determined. Separation of the ionic solution from the small colloidal particles
for conventional analysis is nontrivial, although separation techniques
such as ultracentrifugation. dialysis, and field-flow fractionation have been success-
fully used. If the soluble species of interest have an active nuclear spin, the liquid
NMR technique will constitute an alternative and simpler way to characterize and
quantify those species without being affected by the disperse phase. An exception is
the case where the colloidal species gives a signal that fully overlaps the sharp
resonance of the solution entity. As is quantitative, the absolute concentra-
tion of the species can be estimated based an internal reference of known
concentration but different chemical shift relative to the sample signals.
Alternatively, a calibration curve can be established from a set of external standard
solutions (preferably the same substance found in the sample) measured under the
same experimental NMR conditions as those applied to the sample.
Care should be taken when the soluble species show a significant interaction
with the surface of the disperse phase. In this case, a modification in the NMR
spectrum profile may reflect an adsorption phenomenon, which should be properly
interpreted.
Adsorption of solution species in colloidal oxides and associates will probably

involve attractions between hydroxyl groups at the surface and a specific group
of the adsorbed molecule. If the latter contains a magnetically active nucleus, the

perturbation of its electron density can induce a shift in the signal and
qualitative information can be obtained. However, the molecule becomes less
mobile and the relaxation rates increase. so N M R will also provide further infor-
mation about the structure of the adsorbed layers, which allows one to investigate
the interfacial properties of these colloidal systems. The interactions of the colloidal
particles with either the dispersion medium (solvent) or other small molecules,
surfactants, or polymers present in the dispersion medium can be investigated by
NMR; a brief review focusing on this area has been published by and
Obey As mentioned before, this aspect is not stressed here. Instead, we show a
small number of applications in which colloidal oxides are involved. As these
colloids are usually in aqueous dispersions, information about the dynamics of
water molecules is quite important and can be obtained through water spin relaxa-
tion.
The particles of colloidal size are ideally suited to provide a high surface area per
unit mass of substrate, which in turn is a highly desirable condition for obtaining
NMR spectra of interfacial species. Another factor determining the sensitivity of
the measurement is the existence of dynamic exchange of the dispersion medium
(water) between the bulk and the surface. In the fast exchange limit, the average
relaxation rate of the solvent will depend on the accessible area and its residence
mechanism at the particle surface. The residence time of the sol-
vent is also sensitive to competitive solutes. The measured relaxation rates, in
the fast exchange limit may be simplified by the equation
where and are the fractions of free and bound water and i 1, 2 stands for
longitudinal and transverse relaxations, respectively. The average relaxation rate of
water molecules in the corresponding colloidal dispersions will give information
about the particle structure and affinity of the surface for the solvent molecule. This
is illustrated in Fig. 4, where the specific water proton spin-spin relaxation rate is
plotted as a function of surface area (different particle size) of silica and alumina
sols [The specific relaxation rate - where is the relaxa-
tion rate of pure water and is related to the overall colloidal dispersion given by
The fact that the data fall on the same line is a clear demonstration that
N M R can be used to monitor the total available surface area in a dispersion in
which is not straightforwardly achieved by other means. The straight line depen-
dencies indicate that the samples satisfy the fast exchange limit. Besides, the higher
slope of the alumina line indicates a stronger solvent-surface interaction than that
for silica. More detailed studies of the water layer can be made by using and
relaxation measurements, which are sensitive to changes in motion
Moreover, splitting of and N M R signals in surface-bound
water molecules can elucidate structural modifications on oriented particle systems
as in sol-gel transition N M R relaxation studies of water protons using para-
magnetic probes as relaxation rate enhancers also provide a convenient way to
investigate ion adsorption onto the surface of colloidal oxides
When surface species are not in fast exchange, direct observation of the spectra
can be used to probe both the surface orientation and the dynamics of the adsorbed
species.

0 0.5 1.5 2 2.5 3
Surface area
Specific solvent (water) relaxation rates for oxide surfaces as a function of
4
surface area. Alumina; silica. (From Ref. 30.)
Since the introduction of pulsed Fourier transform techniques,

has become a powerful and routine tool for gathering knowledge of the complex
nature of aqueous silicate solutions. This is due to the characteristic and mostly
well-separated signals for the groups in different structural surroundings and
the possibility of estimating the relative concentrations of the various structural
entities A successive shift of the resonance is observed to lower frequencies
o n replacement of Si OH bonds in silicate species by siloxane bonds 0
Si). F o r every siloxane bond less, there is a shift of about to higher fre-
quency. T h e notation introduced by Engelhardt for the assignment of
the resultant building units has been widely used. In this notation, Q (quarternary)
represents a silicate atom bound to four oxygen forming a tetrahedron and
the superscript indicates the number of other Q units attached to the tetra-
hedron under study (Fig. 5). is the fully condensed tetrafunctional species that
lies around ppm, and is the fully hydrolyzed monomer with 5
- 101 (4 x + The same notation has been used for the characterization of the
degree of condensation of silicon species in the colloidal and solid states.
The studies with liquid NMR spectroscopy for investigation of the chem-
istry and kinetics of the hydrolysis and condensation reactions either in alkali metal
silicate solutions o r in alkoxide sol-gel precursors are abundant and have been very
well covered in reviews It is intended here to point to a few recent
results and draw attention to the application of N M R o n colloidal silicas as
stable particulate sols, which has rarely been exploited.

OR
- Monomer - End Units
(Fully hydrolyzed)
OR OR
-Si - Si -
I
OR
Q2 Middle Units Branching Units
-Si -
- Fully Condensed
Tetrafunctional Units
FIG. 5 building units in silicon environments. can denote either atoms or
The condensation mechanisms elucidated by are particularly impor-

tant for understanding the formation of silica colloids. A common condensation
reaction sequence of acid in aqueous systems proceeds via monomer,
dimer. trimer, and cyclic di-. tri-, and species (rings) to polymeric pro-
ducts. In silicate solutions containing initially di- or acid, partial hydrolysis
to monomer and respectively, occurs simultaneously with the condensation
reactions The N M R spectra of formed by hydrolysis of
tetraalkoxysilanes can also be complex, as hydrolysis and condensation occur con-
currently In general it is accepted that intramolecular reactions leading to the
formation of rings cages play a significant role in the polymerization
sequence. Moreover, the condensation reaction pathway (conversion of
to 0 Si) is quite dependent on the synthesis conditions particularly the pH
of the medium.
In a recent paper and reported the prepara-
tion of silica sols by neutralization of sodium silicate solution with They used
N M R to monitor the polycondensation reaction under acidic and basic pH. It
was found that polycondensation sets in immediately after the reagents are mixed,
as evidenced by the disappearance of the and a progressive con-
tinuous increase of the fraction of 4-connected silicate tetrahedra at the
expense of and species (Fig. 6). However, for base-catalyzed reactions

.6 spectra showing the polycondensation reaction in a silica (pH 1,
produced from neutralization of sodium silicate solutions with
hydrochloric acid. Spectra taken at different moments after preparation. The last
spectrum was obtained immediately after gelation; the first = 0) is given for
comparison, referring to the starting solution = pH 11). (From Ref. 38.)

and still exist in significant amounts throughout the polycondensation. Besides,
there was an indication that polycondensation (towards species) continued
beyond the point for acidic p H but not for the basic pH condition.
Similar pH dependence effects have been observed in the time evolution
N M R analysis of tetraethylorthosilicate (TEOS) hydrolyzed systems
although at more basic pH, hydrolysis was still found to be at
with a significant amount of species remaining. The results reported in both
works suggest that acidic conditions promote hydrolysis but inhibit full condensa-
tion In basic conditions the reverse effect occurs, and hydrolysis is the
determining step. For strongly acidic media, where rapid hydrolysis takes place,
liquid-state has made it possible to follow the silicon site distribution
79 . .
past the gel point without any apparent discontinuity or loss of signal, indicat-
ing that even after segments are still quite At strongly
basic conditions, however, significant loss of N M R signal is observed
probably due to the insufficient rotational mobility (large of the silicon in highly
condensed forms (cf.
Complementary N M R measurements such as can provide
information the hydrolysis that is not immediately visible from NMR
spectra and can be used to monitor the behavior of water and organic ligand
in sol-gel systems For instance, N M R studies on the effects of pH in a TEOS
sol-gel showed that in an acid-catalyzed process all the ethoxy groups were
hydrolyzed long before the sample gels, whereas in a base-catalyzed process some
ethoxy groups remained attached to the silicon at the point. It is evident
that this observation supports the above-mentioned mechanism according to
An indirect estimation of the remaining ethoxy groups on colloidal silica by
N M R has also been reported based on the relative intensities of the
species in the spectra and the respective ratio (see Fig. 5).
By comparison, N M R spectra of four commercial silica sols of different
particle sizes are shown in Fig. 7. Commercially produced colloidal silicas are
usually alkali-stabilized and consist of amorphous silica particles suspended in
water that are kept in suspension due to the negatively ionized surface. The spectra
of these particulate silica sols show the presence of the silicon species and
with the clear prevalence of the fully condensed species (Table 2). In the
interpretation of the resonance corresponds to the silica core
of the sol particles, and the and components to polyhydroxy units, either
bound to the silica surface or existing as free or associated at the sol
surface. However, this last hypothesis seems to be more consistent because the
proportion of these hydroxylated species (32% in the case of the sol with the
smallest particle size) is far in excess of that corresponding to a simple hydroxylated
plane surface. The decrease of the relative proportion of silica with the
size of the sol particles might arise from the condensation of the available oligomers
at the particle surface. This observation is in agreement with the base-catalyzed
polycondensation in neutralized silicates, where the species was claimed
to be formed at the expense of near In the case of the
catalyzed alkoxides. the conditions of growth appear to be dominated by the addi-
tion of the monomer to the highly condensed particles In both cases, the
mechanism of particle growth proposed by is valid.

Chemical Shift (ppm)
FIG. 7 N M R spectra at 39 MHz of commercial colloidal silicas (see also Table 2).
(a) (b) S2; (c) S3; (d) (From Ref. 42.)
Despite assignment of species for the core silicon, the observation

of nuclei buried in the colloidal particles was questioned by Siedle a n d
They pointed out that the dipole-dipole relaxation time of the
"internal" should be very long (Fig. 2a). Thus, the only nuclei observable
in NMR would be those o n the outside of the silica sphere, which can be relaxed by
protons from the solvent o r which participate in chemical reactions such as
exchange of silanol protons o r equilibria involving linkages. This indeed
may be true for particulate silica sols. Alternatively, the proximity of protons from
and species might enhance the relaxation rate of Si the buried species
2 Proportions of Species in Commercial Silica Sols

Article Concentration Particle
Sol code diameter
1 115 S 16.1 8 9 23 68
Ludox S2 32.8 12 4 20 76
Ludox HS 3 40.3 16 5 17 78
Ludox S4 47.2 30 1 13 86
Source: Ref.

(Fig. However. the efficiency of this heteronuclear process is depen-
dent o n the angles and bond lengths involved. O n the other hand, there are ways of
suppressing the excessively long spin-lattice relaxation times of such as the use
of a paramagnetic agent which allows data acquisition at a reasonable
pulse repetition time. This will also depend o n the efficiency of the paramagnetic
agent in reaching the core silicon.
Alkali-stabilized aqueous silica sols have also been studied by "0 and NMR
to determine the water spin relaxation in these colloidal systems. In the excellent
articles of and Halle the dynamics of the water molecules were ana-
lyzed as functions of particle diameter, alkali content, and particle concentration.
They found differences between the longitudinal and transverse relaxation
rates demonstrating contributions from slow motions to the relaxations (see
and estimated that the slow motion occurred on a time scale of more than
based on the frequency dependence of the relaxation rates. The fast and slow
dynamics were independent of particle concentration. T h e longitudinal excess
relaxation rates (where excess relaxation rate is defined as = with
= 1, 2 standing for longitudinal and transverse rates, respectively; is the trans-
verse rate of the free water) were independent of particle size. supporting the
assumption that the surface structure of the silica particles is independent of
their size. In contrast, the transverse rates which contain contributions from
slow dynamics, increased with particle size. It is worth noting that the relaxation
rates were divided by a factor including surface area. so that the net effect of the
bound water per unit surface area could be seen (see, in contrast. the
ized case in Fig. 4). However, the contribution most important to those interested
in characterizing the colloidal silica was the different responses of the relaxation
rates of the two nuclei, and after addition of alkali. As expected of a n
increase in the of charged groups on the surface, relaxation rates
increased o n addition of alkali. but the transverse relaxation of deceased
tonically. A reasonable explanation was that the transverse relaxation was not
dominantly coming from the water molecules the surface silanol groups,
exchanging deuterons very rapidly with bulk water. The reduction of on
addition of alkali led to the mechanism of deprotonation of groups, con-
trasting with the mechanism usually accepted in the literature of adsorption of
hydroxide ions.
The NMR investigation of adsorbed compounds on colloidal silica can be illu-
strated with examples. Siedle and studied the adsorption of
in a silica sol with a mean diameter of 35 nm dispersed in
ethanol (EEO) by means of NMR. In addition to a sharp peak at
whose position and width were the same as those of in there was
a broad resonance at in the colloid-EEO but not in the pure solvent. This
higher frequency shift could be related to a formed between the fluorinated
adsorbate and basic sites o n the surface of the colloidal particles. Besides. the
observation of separate peaks was indicative of a slow exchange rate between the
free and bound o n the N M R time scale. The longitudinal relaxa-
tion times in the free and bound states were 1 . O 1 and 0.26 respectively. Assuming
a n intramolecular dipole-dipole relaxation, the correlation time, was calculated
as 4 through appropriate equations. This was much shorter the

3x overall correlation time for tumbling of a particle of radius 17.5 [Eq.
indicating a fast reorientation of the axially symmetrical groups that
would provide a relaxation mechanism independent of motion of the small particles
bearing the surface compound. characterization of adsorbed polymer layers
over colloidal silica (Ludox) was reported that focuses on of the
adsorbed layer through a criterion based on the solvent as perceived by
pulsed field gradient nuclear magnetic resonance (PFG NMR) By using this
technique, based on the diffusion of the solvent, Cosgrove and
able to estimate the average concentration of adsorbed polymer layers in concen-
trated silica sols various conditions. et used as an
technique to study the adsorption of
tants colloidal silica. From their relaxation they found a
more pronounced enhancement of the specific relaxation rate of the water as a
function of the indicates a large reduc-
tion of water mobility compared to of related polyethylene
oxide so that the conformation of the adsorbed surfactants could be pro-
posed.
Stable particulate alumina sols are precursors for many materials such as controlled
porosity catalysts, ceramics, and coatings. are commonly formed by
of by dilute monoprotic acids. They can be
prepared either by the process developed by Yoldas [4] from hydrolyzed aluminum
alkoxides or by acidic peptization of the precipitates derived from hydrolyzed
aluminum salts As will be seen these alumina sols are usually
component fluids that also contain soluble aluminum species. colloi-
dal can also be by the of hydrolyzed
cations. Therefore, a brief review the characterization of aluminum species in
solution by N M R is presented in studies on
are then discussed
1. Studies on Hydrolyzed Aluminum Species
Aluminum exists aqueous solutions as the monomeric ion
below pH 3 or as ion above At intermediate pH values, for
an molar ratio 0 and 2.5, is hydrolyzed removal of
the coordinated water followed by condensation of the O H groups
to polynuclear ions. this process leads to the growth of large
colloidal species and the formation of sol-gels, N M R spectroscopic studies
have contributed remarkably to establish speciation in hydrolyzed aluminum solu-
tions. These studies were pioneered by et and notable
have been made by Bottero et al. Bertsch et al. and more recently
Kloprogge coworkers Many species with different degrees of polymeriza-
tion and different internal hydrolysis ratios have been reported in this complex
However, only the monomer and the
have been unequivocally assigned. These species
show sharp resonances in the spectrum at 0 ppm (octahedral region) and

(tetrahedral region), respectively. The dimeric cation species
was thought to be the source of a broadened N M R resonance
at about in hydrolyzed aluminum solutions but was later assigned
to octahedral oligomeric species possibly a precursor of the tridecameric-
ion The resonance at Oppm is in fact a weight-average of the
hexaaquo ion and its at a fast exchange rate,
which induces small changes in linewidth depending the pH of the medium

The tridecameric complex is usually expressed as is composed
of a central highly symmetric tetrahedrally coordinated aluminum atom (respon-
sible for the resonance at 63 surrounded by 12 octahedrally coordinated
sites These latter are presumed not to have a symmetrical environment, thus
causing a great broadening of the peak due to the quadrupolar interaction
(fig. so that the signal is not readily or completely detectable at room tempera-
ture, except with increasing temperature
The relative concentrations of these species, their interconversion mechanisms,
and the rate at which they can convert to solid phases are largely dependent on
the synthesis and aging conditions, as numerous workers have noted by means of
The degree of base hydrolysis, denoted by the molar ratio
is particularly important the speciation, as in Fig. The
of condensation of the transient species, which eventually leads to
crystalline aluminum hydroxides and have
also been investigated in detail Besides the transient monomeric,
oligomeric and tridecameric species cited above, octahedral and tetrahe-
dral broad resonances have been observed in some specific hydrolyzed aluminum
by liquid-state These new species have been
assigned to or mixtures of oiigomers. Fu et showed evidence
a broad resonance at due to dimerization of the species
during its treatment. In general, however, the species in the
later stages of condensation are not detected by resolution in
liquid state, since an appropriate colloidal state is likely not obtained obser-
vation Alternatively, the corresponding solid phase can
be separated from the solvent and analyzed by MAS NMR as performed by
Bottero and coworkers and Morgado et Despite the of
solvent, the presence of in the amorphous gel suggested the
mechanism of aggregation of the tridecameric ion the latter present
in the precursor solution.
Although hydrolysis and condensation proceed differently in systems
than in aqueous similar N M R speciation and sensitivity to the
synthesis conditions are found in the hydrolyzed solutions that can be
used as precursors to sol-gels. In to silica the rates of
hydrolysis of precursors are very fast, and
gomers are scarce in the resultant alumina sols, Therefore, the influence of organic

spectra at of hydrolyzed base as a
of degree of (a) = Ref.
versus inorganic precursors on evolving appears to be

Nazar Klein investigated alumina sols derived from hydrolysis of
followed by acidic peptization The data
that the clear sols processed at room temperature the is
predominantly in form of the a minor of
the monomer at (Fig. Conversely, in the sols processed at 363
species the (Fig. Furthermore, heating the low

. at 104.2 MHz of sols, 0.5 M (a) hydrolyzed
aged at 293 K for 8 (b) hydrolyzed and aged at 363 K for hydrolyzed
aged at 293 K 14 followed by aging a t 363 K for (From Ref.
temperature sols a t 363 K does not transform them into the form
directly but results the of tetrahedral and octahedral species (Fig.
probably resulting from condensation, A feature of all spectra in
Fig. 9 is a small broad peak a t 7 p p m which been attributed by the to
overlapping peaks: one a t 4.2 tentatively assigned to the "dimeric" species
and the other a t assigned to colloidal species.

2. Studies on Systems
Among the hydrated oxides only the oxyhydroxide in its
form is expected to be converted into a clear or sol
by means of acidic In general, peptizable are known to be
obtained the Yoldas process aithough they also be produced from the
hydrolysis of salt in aqueous solution The Yoldas process consists of
hydrolyzing an aluminum alkoxide in a large excess of water at 353-373 result-
ing in the precipitation of a boehmite gel, which is with a monoprotic acid
yields a stable sol. Dos and found that sols prepared
by the Yoldas method consist of crystalline boehmite particles with dimensions on
order of nanometers of amorphous or imperfectly crystallized clusters. In
fact, the boehinite phase is thermodynamically favored at temperatures of 353 K or
higher. studies by Olson and Nazar and Fig.
and et alumina sols prepared by the Yoldas procedure at 373,
363, 353 respectively, showed the presence of only coordinated
species. In sharp contrast, alumina sols obtained at similar conditions
showed an additional very broad tetrahedral resonance in the
N M R spectra as obtained by Assih et and more recently by et
(Fig. 10). Assih assigned this new resonance to the
adsorbed at the surface of boehmite particles being responsible for the
stabilization of the sol. Although no direct evidence of the mentioned adsorption
demonstrated, qualitative broadening of the tetrahedral resonance is indeed
suggestive of adsorption (see Unfortunately. no detailed on
the of the observed resonances was provided. However, it is remarkable that
both Assih and Janosovits groups combined spectro-
scopy, presence of tetrahedral could not be probed by the latter
technique. The speciation discrepancy previous results the
same basic method of be explained by differences in the
although such contrast cannot be easily explained. extre-
mely large resonance observed in the tetrahedral region by Assih and Janosovits is
very similar to a broad resonance 7 KHz in width that found by
Akitt and and Morgado and to a background
signal arising from in probe head. This background signal varies
intensity depending the experimental conditions and sometimes to be sub-
tracted from the spectra. If not considered, it may cause erroneous inter-
pretation of the spectra and affect the of overlapping
peaks, In principle, the relative decrease in intensity of the band
acidic hydrolysis time observed by and
(Fig. does not this
A detailed characterization of boehmite by N M R has been
reported by Morgado et Alumina sols prepared a commercial
mite powder (produced by hydroiysis of aluminum alkoxides) were at
different ratios. The resulting sol containing
were susceptible to sol-gel transformation following the addition of nitric
acid 0.06) and were by N M R spectra recorded at
periodic intervals to the gel point (Fig. 1 l), The spectra of the sol recorded at

spectra of the alumina sol synthesized by hydrolysis of
under molar ratios for 3 days,
followed by distillation for solvent removal (2-butanol). (From Ref. 60.)
130.3 M H z after 60 min (Fig. 1 la) showed a sharp peak a t assigned to the
monomer and its hydrolysates (6) and In addition, a rela-
tively broad peak (linewidth at slightly overlapped with the
monomer resonance. This latter, referred to as decreased in intensity with
the process, concomitant with the growth of another extremely broad
octahedral resonance, referred to as SB2, centered a t approximately 8 p p m (Figs.

FIG. 11 spectra at 130.3 MHz of the boehmite SBS dispersion (10%
with = 0.10 for (a) 60 (b) 240 (c) 420 (d) 780
1200 Scale in axis valid only for spectrum (a): spectra were equally
displaced in the and directions for better viewing of the broad resonance
transformation during sol-gel transition. (From Ref. 61.)
1 The is relatively unaffected by gelation. Very similar

spectral changes had already been observed by Olson and in their
= although their extremely broad peak (only
visible by deconvolution) was not so well resolved as in the higher resolution
spectra obtained by Morgado et Another interesting observation was the
thermal reversibility of the sol-gel as monitored by This can be seen in
Fig. 12, where the original spectral profile of the sol (without the very broad
resonance) is recovered after the semisolid gel is heated at 353 (A decrease of
about 20% in the total resonance signal was found for spectrum of Fig. 12 acquired
a t 353 K relative to spectrum a. This is explained by the known inverse proportion-
ality of the NMR signal to the temperature due to the decrease in the population
difference between spins states The SB2 species responsible for the very broad
resonance in the gel state is likely a larger structure derived from the species in
a more anisotropic environment. However, due its thermal reversibility, weak
attractive bonds are probably involved van der forces), in contrast

. N M R spectra at 130.3 MHz of the boehmite SBS dispersion (10%
with = 0.10. (a) Sol after 60 of peptization (298 K); gel
after 1200 of peptization (c) gel transformed to sol after 1260 and
recorded at 353 (From Ref. 61.)
with other rigid structures such as those found in silica systems. It is also
worth noting the presence of monomer even in the semisolid gel. probably moving
freely in the interstitial water enclosed in the continuous solid structure.
It is obvious a t this point that NMR can be a powerful tool for examining
the chemistry of alumina sols, but the most interesting result obtained by
Morgado al. was the unequivocal assignment of resonance to boehmite
colloidal particles. This was possible only because of the field dependence shown by
this species. The spectrum of the sol in Figs. 1 l a and 12a are seen to be quite similar
to that in Fig. 8a (except for the very signal at and it would be
tempting to assign species as the solution species discussed in
the previous section. However, a remarkable shift to lower frequency and broad-
ening of the resonance in the boehmite sol were observed at lower
magnetic fields, as depicted in Fig. 13. Morgado could demonstrate that this
field dependence was due to a second-order quadrupole-induced shift, which is
observed in solids but only for species of limited mobility in the liquid
state, far from the extreme narrowing condition (cf. The chemical shifts

. 13Field dependence of the colloid resonance in an aqueous
= (a) 52.1 MHz, 78.2 MHz. (c) (d) 130.3 MHz.
(From Ref. 6 1
and linewidths determined a t four magnetic fields were plotted against the inverse
of the square of the nuclear resonance frequency as shown in Fig. 14 (data denoted
by solid squares). A quadrupolar coupling constant of = 22. MHz and an iso-
tropic chemical shift of were calculated respectively the slope of the
line and from the intercept in the plot of 5 in accordance with (3)
(assuming = 0). A comparison of the value with those of other aluminum
described in the literature indicated that it was intermediate between
those of sites of very high symmetry (such as the tetrahedral site of with
7 0.65 MHz) and those of sites of low symmetry (such as the octahedral site of
with = but very close to 7 values for sites of reasonably high
symmetry such as in Similarly, a value of = 3.5 x was calcu-
lated from the slope of the line in the plot of according to by
using the previously estimated value of 7. This value of correlation time was much
higher than those observed for the soluble complexes = 4.2

Chemical Shift Linewidth (Hz)
800
600
400
0 1 2 3 4 200
1 2 3
nuclear field (x nuclear field (x
FIG. 14 Dependence of (a) chemical shift and (b) linewidth of the colloid
octahedral resonance as a function of the nuclear magnetic field (see also Table 4).
and line) = 0.10; = 0.08; -
= 0.06. (From Ref,
and = 1.3 s) and of the same order of magnitude as that of an

aluminum ion bound to large proteins Thus, the authors could confirm that
species was not a molecule in solution. In fact, the product is 4.5
(130.3MHz) and deviates from the extreme narrowing condition. From Fig. 3 this
means that the contribution of the central transition should correspond to 26%
of the total signal. In their quantitative measurements (Fig. 15), Morgado and
coworkers found that the signal of the resonance amounted to about 30% of
the total aluminum, which is consistent with the complete contribution of the
central transition (26%) plus a minor contribution from the transitions
= (in Ref. 61, there is a mistake in expressing such transitions
of the nuclei), which has a considerably higher relaxation rate at = 4.5
(curve in Fig. 3). This conclusion was reinforced by the fact that the correspond-
ing contribution of the signal measured at was slightly higher
35% of the total than that measured at Indeed, is 2.7 for
the former frequency, and at this point the relaxation rate of transition is closer
to that of the central transition I, becoming more significant. Moreover, upon
gelling, the contribution of the + SB2) signal dropped slightly to 26% of
the total This is easily explained by the likely lower mobility of the SB2 species
(larger bringing the product to higher values at the right-hand end of the
plot in Fig. 3, where only the central transition accounts for the visible signal with a
contribution of 25.7%. Therefore, the species along with the monomer
accounted for all the aluminum in the sol system. All this evidence pointed to
the assignment of resonance to the colloidal boehmite particles in the
system.

% of total A!
20 50 100 200 500 1,000

Time,
. 15 Absolute and relative concentrations of of the monomer and the SB
species in systems (SBS - 10% =
determined by N M R spectral integration and comparison with standard
solutions. Solid lines denote results at 130.3 MHz, and dashed lines those at
78.2 MHz. Relative concentration is referred to the total present in the (From
Ref. 6 1 .)
The behavior and quantification of the monomeric species in their

boehmite systems have also been studied. Significant changes in linewidth of the
monomer resonance were noted and were found to be dependent on the pH of the
medium. This behavior was attributed to the differences in the composition of the
ion and its as shown by Eqs. (6) and (7). The dependence of
the monomer line broadening on the p H and other experimental conditions fol-
lowed the same relationship found in earlier studies of the hydrolysis of
in aqueous solutions This allowed Morgado et al. to conclude
that the monomeric species freely in the dispersion medium of their alumina
sols without significant interaction with the surface of colloidal boehmite particles,
which are probably stabilized by charging of the potential-determining ion
As for the quantification of the monomeric species, those species were always
present in their systems, accounting for less than 10% of the total
indicating the partial dissolution of the boehmitic material. This was confirmed by
the proportionality of the monomer concentration to the content of boehmite in the

slurry and to the ratio. Besides, the amount of released as mono-
mers in solution increased continuously after the sol formation (Fig. 15), indicating
that the dissolution process proceeds slowly after the first breakdown of the par-
ticles into the colloidal system. The monitoring of the content by NMR
was certainly useful for the elucidation of the peptization mechanisms. This in situ
measurement is particularly important considering that isolation of the dispersion
medium for conventional determination of in solution would not be trivial
(see
Morgado and Pacheco recently applied the same approach to
investigate other types of colloidal aluminas. Three other commercial
with different degrees of crystallinity were submitted to conditions of nitric
similar to those of the earlier study Table 3 shows the relative
phous content and crystallite sizes determined by X-ray diffraction of the new
boehmites SBF, Disperal, and Pural 200 including the sample SBS used
in the previous study. The aqueous suspensions with 10 were acidified
with nitric acid, and the resulting colloidal systems were analyzed with NMR
after 30 min of peptization. Different ratios were used to achieve
similar p H values, according to the type of alumina and their distinct acid uptake
(Table 4). The samples SBS, SBF, and Disperal provided translucid sols, whereas a
white dispersion was obtained with the highly crystalline boehmite P-200.
This latter gave rise to a very ill-defined and broad peak in the N M R spectrum
7 KHz. fig. 16) centered between 6 and 12 irrespective of the magnetic
field employed. It is obvious that in this case a homogeneous system with suffi-
ciently particles could not be achieved for observation in liquid N M R (cf.
However, an interesting aspect of the spectra was the detection of a very
small sharp signal at due to a monomeric species in solution that was
estimated as less than 0.05% of the total in the system. Indeed. the expectation
Properties of the
XRD data
Code amor p hous c ) (nm)
SBS 36
SBF 27
18
Pural 200 P-200 0
B 36
B2 39
B
supplied by Condea GmbH (synthesis from hydrolysis of
gels synthesized according to Ref. 5 by neutralization of
relative to the highly crystalline boehmite P-200, assumed as 100% crystalline (0%
amorphous).
= crystallite size.

4 Characteristics of the Colloidal
Monomer Magnetic properties
Total percent
ratio (nm) of
26 0.048 2.3
SBS 24 0.052 2.5
22 0.056 2.7
16 0.033 1.6
13 0.038 1.9
16 0.010 0.5
Disperal 15 0.019 0.9
P-200 100 0.001 0.05
B 19 0.029 3.6
B2 12 0.036 4.5
B3 13 0.039 4.9
nd not determined; spectrum acquired at only one magnetic field.
of commercial boehmites and of boehmite gels characterized, respectively, 30 min and 24 h after addition.
= mean particle radius determined dynamic light scattering (hydrodynamic radius corrected by classical
of peak in the spectra acquired at 130.3 MHz. Concentration calculated from an external calibration curve. Integral vs.
of a series of solutions measured with the same N M R conditions.
from field-dependent chemical shifts and linewidths (see text).
to the total aluminum present in sol system.

this highly crystalline alumina was that practically no would be available
for dissolution in the peptization. This fact illustrates the sensitivity of the
technique.
The other alumina sols showed the same spectrum profile as that shown in Fig.
12a for the SBS sample, except for different relative intensities of the monomer
signal to the colloidal resonance. Indeed, the concentration of monomeric
species was clearly dependent on the type of boehmite, as shown in Table 4. There
is a direct relationship between the concentration of the monomer and the amount
of amorphous phase in the original boehmite material (Table 3). It indicates that
the monomeric species forms from dissolution of the amorphous phase associated
with the dominating nanocrystalline part, which in turn is related to the colloidal
resonance For the new boehmite systems, the colloidal signal showed the
same second-order shift as before. The results of the
dependent chemical shifts and linewidths were also plotted against as shown
in Fig. 14 (open symbols), along with the previous data reported by Morgado et
with the boehinite SBS. It is clear from Fig. 14a that the chemical shifts
observed for all three types of lay on the same line, so that the same
coupling constant and isostropic chemical shift were calculated
for the different systems, as shown in Table 4. This fact may provide
some insight into the controversial crystal chemistry of the pseudoboehmites.
Considering that the value reflects the local symmetry around the nucleus, it
can be concluded that the chemical environment around the sites in the different
boehmite colloids is not significantly different. It is in agreement with the theory of
Tettenhorst and Hofmann in which both poorly and well crystallized
mites are continuous in their structure, they consist of the same or similar
octahedral layers in the plane, although there is a lack of three-dimensional
order for the materials of smaller crystallite sizes because of the restricted number
of unit cells along the axis. In Fig. 14b. the new data follow roughly the same
trend as previously. The of the colloid resonance in the most crystalline
(Disperal) were systematically lower than those of the other two. In all
cases, the correlation times were of the same order of magnitude (Table
4). As is related to the of the particle, the small differences in particle
sizes (MPR in Table 4) are consistent with the lack of discrimination among the
values, although Eq. (1) indicates a cubic dependence of on particle radius. In
principle, an independent estimation of is possible from that equation. Using
= 0.89 x Pa-s (298 K) and a particle radius of 15 as determined by
dynamic light scattering, a value of = 3.1 x was obtained, which is two
orders of magnitude higher than the values calculated through (4) (see Table 4).
This deviation is partly explained by the fact that (1) assumes the rotational
diffusion of a perfect sphere, whereas boehmite particles are known to be prolate or
oblate spheroids whose rotational motion is therefore anisotropic. However, exist-
ing models for correcting such an assumption predict a maximum deviation of one
order of magnitude from the classical theory Other sources of error may be
derived from the determination of particle radius (MPR a) by dynamic light
scattering. Small errors in the estimation of particle size are enhanced in the calcu-
lation of through Eq. since = f(a 3 ). The appropriate use of N M R for

obtaining information on the size of the particles and molecules is still a cloudy
subject.
Alumina sols from peptization of boehmites obtained by an inorganic route
have also been analyzed by N M R at 130.3 M H z Boehmite precipitates
were prepared by neutralization of solution with according to
the procedure described in Ref. (5). They were diluted with water to 4% and
with nitric acid, providing clear sols that did not gel, even 24 h after acid
addition, when NMR analysis was carried out. Their spectra are shown in Fig. 17,
and the concentration in monomeric species is shown in Table 4 The
spectra presented the same characteristic peaks as those of alumina sols from
boehmite of hydrolyzed alkoxides (organic route). The colloid showed a resonance
peak at 8.5 ppm and 470 Hz, similar to the spectrum of Fig. 12a. The mono-
mer linewidth showed the same p H dependence of in aqueous solution as
observed for the alumina sols of commercial boehmites. The magnitude of the
relative monomer concentration (percent of the total) is consistent with their
high content of amorphous phase (Table 3), but it is larger than that shown by
SBS of similar amorphous content by XRD. This can be explained either by the
longer peptization time to which these sols were submitted or by a
. 17 NMR spectra at 130.3 MHz of sols (4% after peptization

with = 0.11 of boehmite precipitates derived inorganic route (see
Tables 3 and 4). (From Ref. 65.)

ference in the nature of their amorphous phase, which might be more susceptible to
dissolution by
Another alumina sol from the inorganic route was studied by Kurokawa
Fresh precipitate produced by neutralization of solution with
was aged for 12 h and separated from the mother liquor to be with acetic
acid and by heating at 353 K. These researchers monitored the peptization process
of the aluminum hydroxide precipitate toward a final alumina sol as a function of
period (Fig. 18). Although they used (MAS) N M R with the sample
rotating at only soluble and colloidal species were observed, so that
static N M R would likely give the spectra. It was found that the starting
precipitate (before was associated with disappearance of the
N M R signal. Considering the amorphous nature of the material at this point (high
quadrupole interaction with the electric field gradient), this fact is quite reasonable.
It is also evident that the species that was present in the precursor hydrolyzed
solution is still present with the precipitate (Fig. The tetrahedral species
was decomposed upon acidic peptization at 353 while the sharp monomer peak
and a broader resonance at 3.1-3.3 increased with time of peptization. It was
assumed that acetic acid is consumed by both peptization of the hydroxide and
of the tetrahedral structure. It was not considered, however, that at
their conditions (353 K), colloidal boehmite is likely to be formed from the
amorphous hydroxide along with peptization. The broader resonance at
3.3 ppm appearing after 1 h of was assigned to oligomers of hydrous alumi-
num. However, from our previous discussion, such resonance is most probably due
to colloidal boehmite. The spectral profile of their alumina sol (Fig. is very
similar to the spectrum of the boehmite system acquired at 78.4 MHz (Fig.
In fact, their alumina sol showed particles in the range of a few tens of
nanometers, and after solvent removal it showed an X-ray diffraction pattern
with ill-defined reflections close to those of poorly crystallized boehmite.
However, their N M R measurements were performed at = and
from Fig. 14 an apparent chemical shift of (instead of 3.1 ppm) would
be expected for the colloidal boehmite. Either their assignment of soluble oligomers
is correct or another colloidal species with sufficiently large, but different from
that in commercial boehmites (Table 4), is involved. It should be stressed that the
rotation at the magic angle is not enough to remove the second-order
quadrupolar effects.
3. Studies of Adsorbed Species
The use of NMR spectroscopy for probing molecules adsorbed on alumina disper-
sions is briefly cited. and used deuterium N M R to probe the
dynamics and structure of a surfactant (an alkylbenzenesulfonate, SHBS) adsorbed
onto alumina. The ring was deuterium-labeled, and from the analysis of the
lineshapes of spectra, valuable information on the degree of motional
in the adsorbed layer was obtained. As deuterium possesses a quadrupole moment,
it is sensitive to the orientation of the molecule with respect to the external magnetic
field. This sensitivity generates a quadrupolar splitting, which is defined as the
separation between the two allowed transitions for the deuterium nucleus. In liquid
systems, the quadrupole interaction is averaged out, but for static systems powder

before he
NMR spectra at of hydroxide precipitate [from

neutralization of with with acetic acid = 0.15)
at 353 K as a function of period. (From Ref. 48.)

or Pake patterns are obtained. Molecular that are comparable to or faster
than the quadrupole coupling constant will affect the lineshape and reduce the
splitting from its rigid value.
Phosphorus-31 N M R was used by Burton et to analyze
zenes (dispersants) bound to the surface of alumina ceramic particles. The differ-
ences between the spectra of the inorganic polymer in solution and in colloidal
suspensions were used to probe the chemical interaction with the alumina. Their
N M R results showed a small lower frequency shift associated with phosphazenes
bound to alumina. Solution samples consistently displayed a broadening of the
resonance upon cumulative addition of alumina to the suspension, indicating some
degree of chemical adsorption However, Burton et pointed out that
this lineshape difference might be mistaken as an artifact caused by the amount of
solids present in the N M R sample.
Nuclear magnetic resonance studies have proved valuable for the characterization
of other colloidal oxides and hydroxides. Some selected examples of such studies
are shown in the following paragraphs.
Pozarnsky and used N M R to propose for
the synthesis of aqueous sols. In contrast with the colloids discussed in the
previous section, the N M R signal due to colloidal species (vanadate polymers)
has mostly broadened beyond the detection limit, probably due to its very low
symmetry, which enhances the problem of quadrupole interaction. Nevertheless,
based on the solution chemistry associated with the sol, the process could
be rationalized in terms of the chain polymerization of the dioxovanadium cation,
from This polymerization could be rate-limited by the
release of the cations from decavanadate anions (-426, -512, -531 if
these anions were initially present at high concentration. Using an external refer-
ence spectrum, Pozarnsky and found different starting concentrations
of the decavanadate anion that were dependent on the synthesis method. In the
slow synthesis method, the decavanadate ion accounted for 80% of the
vanadium present, whereas in the fast synthesis method it accounted for only 20%.
Therefore, a reaction scheme that explains the N M R data (after a brief initial
transient period) could be obtained, thus rationalizing the polymerization mechan-
ism.
Judeinstein and used and N M R to analyze molecular precur-
sors formed during the sol-gel synthesis of thin films but were not very
successful in using N M R for their unequivocal characterization. Three
broad resonance peaks appeared, but a definite conclusion was not possible
because the chemical shifts might be related either to bonding distortions in a single
oligomeric molecule or to different molecular species. In both cases the broadening
of the N M R signals could be explained by short relaxation times arising from
formation of oligomeric species and exchange reactions between different chemical
species.
King et applied and N M R to provide structural information
on HTMA (hexamethylenetetramine) used as a agent in the sol-gel

of microspherical nuclear fuel Furthermore, they used N M R to
monitor uranyl speciation during the hydrolysis and processes. et
[6] used N M R to study the early stages in the preparation of colloids.
Most advanced ceramics are multicomponent materials having two o r more
cations in the oxide network such as and cordierite
In particular, the sol-gel route may be applied in their synthesis to
allow the homogeneous mixing of precursors at a molecular level. However, a
major problem in forming homogeneous multicomponent gels is the unequal
hydrolysis and condensation rates, as observed between silicon and aluminum
precursor cases Compiexed precursors with
reduced reactivity have been used to overcome this drawback, and the effectiveness
of this route has been studied by N M R .By using this technique, and
coworkers have followed the hydrolysis of TEOS in the presence of an aqu-
eous (or ethanolic) solution of aluminum and magnesium nitrates and compared it
with that in the presence of complexed by ethylacetoacetate (etac) and
an aqueous solution of magnesium acetate. The N M R spectra of the these two
precursor sols for cordierite are shown in Figs. 19a and respectively. For the
first procedure (Fig. only a sharp peak at Oppm is noticed,
indicating that no Si- 0- Al bonds are formed at this stage. In contrast, in case
(b) when the aluminum precursor was complexed, the N M R spectra showed
the formation of transient species characterized by a sharp peak at 51 pprn
FIG. 19 N M R spectra of two precursor solutions from cordierite. (a) TEOS

hydrolyzed with an aqueous solution of aluminum and magnesium nitrates; (b)
TEOS and (etac) hydrolyzed with an aqueous solution of magnesium
acetate. Ref. 9.)

(similar to but not the same), which assigned to bonds
probably associated with Indeed, the corresponding
spectrum showed a peak caused by suggesting a structure.
Upon sol-gel transformation, the sharp tetrahedral signal in the spectrum of Fig.
19b progressively disappears, giving rise to a broad resonance around 55 ppm along
with the remaining broad peak at The main point is that the chemical
shift of is typical of that found for cordierite, thus demonstrating the
importance of the formation of bonds in the precursor sols as probed
by and Another approach to ensure a good level of mixing of
and silicate species was used by and Douy in the synthesis of
They used urea decomposition to provide slow in generation of ammo-
nia to hydrolyze and the aluminum species while promoting
condensation of the acid due to the slow increase in pH. In this process, a gel
was formed within a few hours but was slowly disaggregated into a clear sol after 3
days, and finally a second gel was obtained after a week. The intermediate colloidal
phase was formed at 0.8 2.7. During this period of hydrolysis, static
and N M R spectra were obtained and compared to the
solid-state MAS N M R of the first and second gel phases. spectra of
the sol phase showed the remaining resonance peaks at - 110 and 102
predominant in the first gel, progressively disappeared, being replaced by lines at
-90 and These lines in turn disappeared, giving rise to a single resonance
at -78 ppm for a degree of hydrolysis slightly higher than 2.4, remaining
in the second gel. Surprisingly, N M R studies of the sol phase showed only the
peak at Oppm (like that of Fig. in contrast with a small additional tetrahedral
signal in the first and second dried gels observed by MAS The
authors attributed the unexpected absence of in the intermediate sol phase
to species highly distorted symmetry> probably not observable by
N M R under static conditions. However, there is no reason to believe that the
rotation of the sample at the magic angle make visible such an asymmetric
species. The MAS tends to eliminate anisotropy effects (CSA). Moreover,
the particles of 2nm radius reported for their phase indicates that the
colloidal species was probably moving rapidly enough to account for isotropic
behavior so that could be observed even in a static experiment (cf.
species the previous section). The authors attributed the predominant
acoordinated aluminum and condensed silicon species in the later stages of this
sol-gel process to a chemical environment very close to that found in natural
In these minerals, each silicon atom is linked to
atoms bridges and to hydroxyl groups, showing a unique
resonance at NMR) and only octahedral aluminum. This work and
many others show the adequacy of the combination of and solid-
state N M R to elucidate the sequence in processes.
The same combination of techniques may be helpful gaining a better under-
standing of adsorption phenomena in colloidal oxides
Lead-207 N M R was employed to probe a sol-gel system A
precursor was prepared by mixing lead acetate and isopropylorthotitanate
tions in methoxyethanol and Only one signal appeared, around
-883 (from immediately on mixing, remaining unchanged with time

even during the process. However, compared with the simple lead acetate
the solvent a signal at -1248 there was a significantly higher fre-
quency shift of the N M R signal in the two-component precursor. indi-
cated that interaction between the compounds happened very quickly and
a new chemical enviroment with respect to the lead atom was established. the
other hand, and spectra of the freshly prepared precursor showed
that neither the - OPr group of titanium compound nor the group
of the lead compound were changed. So the chemical shift at -883 ppm could not
be explained by a Pb- 0- Tibonding at this stage. Based on additional infor-
mation gained from infrared spectra, Lan et al. suggested that the two
compounds rapidly interacting, forming a linkage or association through
the bidentate acetate bridging, unchanged during solution reactions and gelation.
Pb- 0- Tibonds formed only upon calcination with the decomposition of the
groups.
Carbon-13 N M R was used for examination of precursor sols by
Kramer et interesting aspect of this work was immediate broad-
ening effect (without changing the chemical shift) in the reso-
nance when the ethoxide toluene was added to the barium alkoxide
solution. Because the initial copper solution colloidal, it was suggested that the
line broadening was due to the interaction of the quadrupolar copper nucleus with
alkoxide. However, as hydrolysis proceeded, the signals became progressively nar-
rower until the original aspect was recovered. This narrowing effect was coincident
with a decrease in the particle size of the sol, indicating that yttrium and barium
alkoxides react with copper ethoxide in such a way as to break up the colloidal
particles. Duboudin et used and to study the kinetics and
structure of an material prepared by the sol-gel process.
An interesting example of characterization of a colloidal hydro-
xide and surface-adsorbed species was studied by Yesinowski Calcium
xyapatite, in colloidal was analyzed by N M R in
the liquid state. A stable colloidal suspension of this solid could be formed only
with the addition of sodium diphosphonate acting as a peptizing
agent. The spectrum of the colloidal suspension is shown Fig. 20b.
The peak at 19.1 ppm (from arises from EHDP. In the phosphate region
there are other two peaks. A sharp peak at 0.9 ppm arises the orthophosphate
ions liberated from hydroxyapatite into solution when EHDP was added. The
broader peak at was assigned to the hydroxyapatite.
Ultracentrifugation of the system and resuspension of the residue in
pure water resulted in the nearly complete loss of the solution phosphate signal
but not the colloidal signal (Fig. In addition, there was a reduction in size and
a remarkable increase in the linewidth of the EHDP resonance, whereas the
tion phosphate peak had no broadening. Yesinowski therefore concluded that the
small broadened peak arose from EHDP on the hydroxyapatite sur-
face. A closer look at the N M R spectrum of this sample before
and resuspension (Fig. 20b) showed a sharp component arising from free
EHDP over a broader one the chemisorbed EHDP that is not in rapid
exchange with free molecules in solution.

Solution
Phosphate
. NMR spectra at of (a) solid hydroxyapatite powder

(undecoupled); (b) colloidal suspension of of hydroxyapatite in 2 m L of
+ 0.075 g of 1.4 h data accumulation and (c) colloidal
suspension by (b) and pellet in 2 of
scale as spectrum (From Ref.

All the studies described illustrate ability of to investigate the
of a variety of colloidal oxides and associates in the liquid state, D u e to
the recent growth and of sol-gel products derived from metal
oxides, NMR spectroscopy has been extensively to study the
composition and kinetics of such systems from their initial formulation through
their final processing stages. Nevertheless, similar results can be
obtained in colloidal dispersions by inorganic routes. simultaneous analysis
of the chemistry in both the dispersion medium and the disperse phase is a great
advantage of liquid-state N M R of such colloidal systems. However, the disperse
phase is investigated conveniently only the particulate species meet a
narrowing condition, broadening interactions (typical of the
solid phase) are It seems that particles with effective diameters smaller
than are appropriate. although other parameters related to particle
motion, as solvent viscosity and temperature? have to be considered as
In the case of colloids containing nuclei such as studies of the
field-dependent phenomena are particularly for qualitative quantitative
examination. The spectra of adsorbed compounds o n suitable colloidal oxi-
des can easily be obtained and give structural about the surface of the
particle and adsorbate. I n some cases, colloids will constitute useful models for bulk
oxides. O n the other hand, the of NMR spectroscopy in fluid furnishes a
powerful tool in with MAS N M R t o analyze the complete chemical
of sol-gel transformations.
New York. 1979.

Bergna The Colloid Silica. American
Society, Washington,
C. Bricker and Sol-Gel Science:
Sol-Gel Processing. Academic. San CA.
E. Am. Bull. (1975).
Lam, and Colloid Interface
(1997).
A. Leaustic, and Mater. (1989).
A. Pozarnsky and A. Mater. Lett. (1996).
Judeinstein and Livage. J. Mater. (1991).
Livage, Babonneau, M. Chatry, and Coury, (1997).
I. Jaymes and A. Douy, Ceram. (1996).
6. Moore, A. and Knight, Mater. Res.
(1988).
C. (ed.), Sol-Gel for Electronics
and Ridge. 1988.
A. and Kay, Rev, Mater. (1991).
L. and A. Mater. (1991).
Joseph, and W. T. Fored,

J. D. Tchoubar, Cases, J. Fipriat,
J. Colloid Interface
D. Gay, in Encyclopedia of Nuclear Magnetic 2 M. Grant
and Harris, Wiley, Chichestes, 1996, pp.
and and Carbon-13 An
Integrated Approach, London,
Yesinowski, Chem. (1981).
D. Burton, G. and Mater.. Res.
Symp. (1991)
(a) C. King, A. Thompson, and R. Buchanan,
Mater, Symp, (1990); (b) R. King, King, and A.
ibid, p. 1083,
A. Pozarnsky, Thesis, University of
and Lett,
J. NMR An to
Hall, London, 1992,
R. Harris, Nuclear A
Longmans, 1986.
in Magnetic 6
Grant and R. Wiley, Chichestes, 1996,
Bull. and D. Turnmer, Chem. 701306
A. C. Kunwar, A. R. Thompson, Gutowsky. and E. J.
Aramini, and J. Am.
T. Cosgrove and Obey, in of Resonance

Grant and 2, Chichestes, 1996,
(a) Piculell, J. Chem. Trans. 1 (1986); (b) and
41
L. and Sparks, Colloids A 78157(1993).
Roose, and A
Engelhardt and D. Solid-state and

Wiley, Chichestes, 1987, Chapter
G . Engelhardt, Jancke, and 141109 (1974).
and L. Kelts, in The Colloid of Silica
Am. Chem. Washington, Chapter 19.
J. and A. Today (1997).
and Gregoskiewitz, Colloid Intesface
(1997).
Kelts. J. and Solids 831375
Keefer, W. Shaefes, Kay, and

J. Solids (1986).
R. J. and 61792
(1990).

F. S. A. Matsumoto, and Goberdhan, in The
Colloid Chemistry o f Silica E. Bergna, Adv. Chem. Am. Chem.
Washington, DC. 1994, Chapter 3.
J. Brinker, in The Colloid Chemistry of Silica Bergna, Adv. Chem.
Ser. Am. Chem. Washington, DC, 1994, Chapter 18.
A, R. Siedle and A. Am. Chem. (1981).
Cosgrove and Griffiths, Colloids Surf. A (1994).
(a) E. Hurd, Magn. (1990); (b) R. Hurd. in Encyclopedia
Nuclear Magnetic Resonance M. Grant and R. Harris, 3, Wiley,
Chichester, UK, 1996, pp.
R. Bohmer, and (1992).
Kurokawa, and Nakata, Chem. Rev.
(1994).
(a) Akitt, Greenwood, and Lester, J. A
(1969); (b) W. Akitt and A , J. Magn. (1978); (c) W.
Akitt and A. Farthing, Chem. Dalton Trans. (1981); (d) W.
Akitt and A. Farthing, Chem. Dalton Trans. (1981); (e) W.
Akitt, and B. Mann, Magn, (1981); Akitt, and M.
Elders, Chem. Faraday Trans. (1985); (g) Akitt and M .
Elders, Chem. Dalton Trans. (1988); (h) Akitt, J.
Elders, X. Fontaine, and A. K. Kundu, J. Chem. Dalton Trans.
7989:1889 989).
Y. Bottero, Cases, Fiessinger, and Poirier, Phys. Chem.
(1980); (b) Bottero, J. Cases, and Fiessinger, J. Phys.
Chem.
(a) P. Bertch, and I. Barnhisel.. Soil Am.
(1981); (b) P. M. Bertch, (1987).
(a) Kloprogge, Seykens, Jansen, and Geus. Non-Cryst.
Solids (1992); (b) Kloprogge, Seykens, Jansen. and J.
Geus, Non-Cryst. Solids (1993); (c) Kloprogge, Seykens,
Jansen, and Geus, Non-Cryst, Solids (1993).
A. Thompson, A. C. Kunwar, H. S. and Oldfield, Chem.
Dalton Trans, 17 (1987).
Wood, A. Siedle, Hill, P. and C. J. Goodbrake, Mater.
Symp. Proc. (1990).
J. Bottero, Tchoubar, Cases. J. and Fiessinger. in
Interfacial Phenomena in and Materials Processing (Y. A. Attia,
and Chander, Elsevier, New York. 1988, pp.
Nazar, and J. Am. (1988).
Doss and R. Zahlen, Phys. Rev. (1993).
L. Olson and Bauer, Mater, Symp. Proc. (1986).
T. A. Ayral, and Phalippou. Mater. (1988).
Scharf, and A. Wokaum, Non-Cryst. Solids
(1997).
Morgado, Lam. Menezes, and Nazar, Colloid
Interface (1995).
Morgado, Jr. and Pacheco, unpublished results, 1997.

R. Tettenhorst, and A. Clays Clay Miner. (1980).
Boere and R. G. Kidd, Rep. N M R Spectrosc. (1982).
Morgado Thesis, Federal Univ. Rio de Janeiro, 1995.
and D. Colloid Interface (1993).
and Mater. (1993).
Lan, A. Montenero, Gnappi. and Dradi, Mater. Sci. (1995).
F. Douboudin, J. Dunogues, Puyoo-Castaings, and J.
Mater. (1990).
J. in Modelling of Structure and Reactivity in Zeolites R. A.
Academic, San 1992, pp.
R. Carper and J. (1997).

Department of Polymer Science and Engineering,
Gyeongsang National University, Korea
I. Introduction
I I . Experimental
A. Materials
B. Methods
I I I . Results and Discussion
A. Surface density and wettability on alkyl-inimobilized glass
substrates
B. Effect of various alkyi chain lengths on contact angle and
surface energies
IV. Conclusions
References
Polymer surface dynamics or mobility at interfaces is an interesting and practically

applicable subject Solid surfaces are assumed to be rigid and immobile due to
surface chemistry. However, polymer molecules in the near surface or interfacial
regions are really expected to exhibit motion and relaxation. Of course, these are
not identical to the motion observed in the bulk because of the different interfacial
environment. In general, given sufficient mobility, surfaces and interfaces
can be mobile and the surface can rearrange or reorient at interfaces to minimize
the interfacial free energy with the surrounding phase.
For biomedical devices and chromatographic supports the nature of
such surface motion can be applied. We know that polymer surfaces in aqueous
solution can have different properties than in air or vacuum due to their different
mobilities. The chromatographic properties of a reverse phase are dependent on the

17 ark
conditions of the bonded phases. The motion of such chains is the key
to understanding their chromatographic properties Many researchers, includ-
ing and Squires have investigated dual models of monomeric alkyl
surfaces ranging from a folded or "liquid-like" orientation to a rigid or bristly
type in an aqueous environment.
Electron spectroscopy for chemical analysis (ESCA) [9] and Fourier transform
infrared (FTIR) spectroscopy with attenuated total reflectance (ATR) can also
be used for routine surface studies. FTIR spectroscopy is known to have the
sensitivity to determine the average orientation and reorientation of interfacial
chains. But it does not directly provide information on the itself. The
mobility of a solute in the neighborhood of an alkyl chain can be measured by
fluorescence spectroscopy
Direct measurement of molecular motion and dynamics is offered by NMR
studies using the spin-lattice and spin-spin relaxation times In a
general sense, the changes in spin-lattice relaxation time were consistent with
increased segmental motion as a function of distance from the surface. Sindorf and
investigated the molecular motion of n-alkylsilane bonded to silica
particles using N M R . Alkyl chain mobility increase with chain length
up to = 8) until the interactions between chains occur at longer chain length
= 18, for example). However, the main disadvantage of methods is
that they are limited to fine particles having high surface areas and are not applic-
able to plate-type surfaces.
The plate contact angle method might provide a sequential scanning
curve or hysteresis loop that can be interpreted in terms of surface mobility, reor-
ientation, solvent penetration, and intrinsic under both water and air.
Water is known to be a difficult liquid for use in contact angle studies due to its
small molecular resulting in penetration and local swelling of the solid
surfaces
According to Van et the receding contact angles on
surfaces decrease in the range of = 6 to = 12, then increase to
= 18. Because of the increasing mobility due to the decreasing polymer sur-
faces with moderate-length alkyl side chains may reorganize, which may lead to a
decreased number of hydrophobic segments exposed to water. In addition, as a
minor factor, the hydrophilic ether groups may also reorient toward water surfaces.
This result may be closely related to another study of the brittle temperature
for acrylates and methacrylates. They showed a minimum in the brittle
temperatures in the ranges of = 8-12 due to increased side chain mobility.
Judging by these cases, alkyl side chain bonded surfaces may have different motions
with different chain mobilities and lengths.
It is not easy to measure such chain length effects directly on polymer
substrates because of the packing density problem polymer swelling effects
etc. Therefore, immobilized on glass substrates can be used as a
good model system for the study of surface dynamics. Because rigid glass
surfaces are nondeformable, unlike polymer substrates, a rigid glass substrate may
have a contact angle very different from that of a polymer substrate. Pure clean
glass substrates contain groups high surface energy), and the contact
angles are zero in a pure aqueous environment. Hence, we may be sure that the

contact angles obtained are indeed due to the alkyl side chains attached to the glass
substrates.
In this work, we applied dynamic contact angle methods to the study of wetting
and interfacial phenomena. The reaction of monofunctional reagents with
sites at the silica surface can result in a single monolayer, whereas di- and
trifunctional silanes, especially if water is not completely excluded, hydrolyze and
also polymerize, forming a polymeric bonded phase Therefore, a series of n-
alkyl monochlorosilanes of differing chain lengths = 8, and
available commercially, were bonded on the glass surfaces. We studied the
effects of different chain lengths on borosilicate glass surfaces by varying
the surface concentrations and using different solvents. In addition, using a non-
linear programming method, surface energy terms, and were obtained from
the data under both pure water solvent and mixed solvents of water and methanol.
II.
. Materials
Trimethyl chlorosilane, chlorosilane, n-octyldimethyl chlorosilane,
and chlorosilane were purchased from Petrarch Systems, Inc.
and used without further purification (all 99.5%). The chemical structures of
these silanes are shown in Table 1. present as minor constituents in
reagents for the n = 1 case], contain no reactive groups.
Hence, they were removed during the rinsing procedure.
Prepurified water was further purified by passing through a reagent
water system Co.). Toluene was dried with molecular sieve particles
for 24 h before use. Methanol and dehydrated ethyl alcohol were used as received.
Corning borosilicate cover glass (2940, No. 1 24 x 50 mm, 0.16-0.19 mm
thickness) was used as substrate. As determined by ESCA, surface elemental com-
positions of this borosilicate glass were B 2.3, 28, 0 = 63, C = 4.4 atomic
percentages, respectively. Minor components below 1 were Na, and
Ti.
Cleaning Glasses
Glass slides were immersed in chromic-sulfuric acid solution at for 40
which supplied active oxygen atoms that oxidize carbonaceous materials on glass
surfaces, rinsed extensively with purified water, and then dried overnight in a clean
oven at in air. All other glassware used was also carefully cleaned by the
same process.
We examined the cleanliness of the glass slides by measuring the surface tension
of purified water(72.6 0.5 at room temperature and by verifying
that there was no water contact angle hysteresis and that the slides were perfectly
wetted.

1 Structures of Silane Coupling Agents
Designationa Chemical name Chemical structure
Trimethyl chlorosilane
Me
Me
-
Me
Me
chlorosilane Si
Me
system designation
2. Preparation of Coated Surfaces

After verifying that the dried slide showed no contact angle hysteresis, clean glass
slides were immersed in a suitable concentration of
silane in dried toluene and allowed to equilibrate for 15 min. reaction
time was 30 min for the formation of monolayers at room temperature
The coupling chemical reaction of silanes with glass surfaces is shown in
Fig. 1. Hydrolyzed silanes in aqueous solution can react with the
groups in the glass surfaces to produce immobilized chains.
The slides were then rinsed thoroughly four or five times with pure ethanol to
remove unreacted silane materials. The treated slides were then heated in a vacuum
oven a t for 3 h under nitrogen to remove and ethanol. The stability of the
silane coating on the glass slides was tested by remeasuring the contact angle in
purified water after drying in a vacuum oven at room temperature and by
verifying the reproducibility of the contact angle.
3. Dynamic Contact Angle Measurements

As soon as the coated slides were removed from the oven and cooled to
temperature, dynamic contact angles were measured by the Wilhelmy plate tech-
nique We obtained the wetting curve with an (Cahn Model

(Chemical reaction)
FIG. 1 Chemical reactions of coupling agents with glass surfaces for

= 3, 7, or
RM-2) as a function of the immersion depth. A motor drives the shelf to advance or
withdraw the water container at a controlled speed. The chamber is insulated and
maintained at room temperature and constant humidity (35% RH).
The output from the electrobalance measurement is fed to an X-Y recorder, with
the balance output (force) feeding the axis and immersion depth feeding the
axis. To measure the contact angle, the silane-treated glass slides were immersed
into or withdrawn from the purified water or mixed solvent at a constant speed of
Dipping velocity was chosen to be fast enough for convenience of
measurement but slow enough to avoid speed effects (such as hysteresis on the
clean glasses) as shown in Fig. 2. Advancing contact angles were used for this
study because they can show the surface energy condition directly, whereas reced-
ing contact angles can show the surface energy condition of the wetted surfaces.
Water-methanol solvents were made with increasing fraction. Each
solution was kept for 15 min at room temperature to equilibrate; then its surface
tension was measured. The contact angle and surface tension of each sample were
measured three times, and their error ranges were within and
respectively.
4. Surface Energy Measurement

Young's equation, which describes the force balance of a liquid drop on a solid
surface, is used with the usual symbols,
The total free energy is the sum of contributions from the different
forces at the surface,

.2 Dipping velocity effect at various speeds using cleaned glass plates in the
plate technique: (a) 10 (b) 30 (c) 40 (d) 60
(e) 120
where the superscripts d and p refer to the dispersion and polar force components,
respectively.
The surface free energy of a interface can be described as a function
of individual force components of the solid and liquid. F o r the relationship, the
following equation, which was assumed by Owens a n d Wendt is most reliable
for polymeric solids:
Combining (1) and (3) yields the well-known geometric mean equation,
By measuring the contact angle 9 of two different liquids against a solid, simul-
taneous equations are obtained that can be solved for and if the values of
and are known. With some organic liquids, the components of free energy
were evaluated using reference solids. The result shows that the components
deviated by over 10% from the mean, and they cannot be accepted as constant.

When more than two kinds of liquids are used, a useful technique introduced by
Erbil and can be used. The method, based on a parameter identification
problem, uses only the total surface free energy of liquids, which can be obtained
precisely by experiment. For this method, an equation that shows the interaction
between the solid and liquid surfaces is needed.
The work required to pull the liquid away from the surface the work of
adhesion) was given by Dupre as
= +
The work of adhesion was described as follows by and Good using
the individual component of free energy,
=
where is the interaction parameter. Combining (4)-(6) yields
+
When = 1, the network of adhesion is maximized for a given value of
d d
a d This
value of corresponds to and
= Thus = 1 only when the fractional molecular force contri-
butions from dispersion and polar components in both the liquid and the solid to
the total surface tension are equal. This idealized condition is difficult to achieve
during contact angle determination for most polymers due to the difference
between the surface free energy properties of probe liquids and polymers. In this
work, a new concept that involves the virtual liquid surface free energy components
is introduced. With the known value of total surface free energy of liquid and
experimental data of the contact angle one can estimate the individual surface
energy components of solids and virtual components of liquids with a parameter
estimation algorithm using a nonlinear programming method. The method forces
the interaction parameter values close to 1 for each polymer-liquid pair under
the constraint of the geometric mean relation, Eq. (4).
Because of some inevitable errors in measuring the contact angle, the geometric
equation might show some deviation from the exact equality. With a value of
acceptable error in Eq. 3% or for example, one can optimize the free
energy components to maximize the interaction parameter. We used a multivariable
nonlinear regression algorithm developed by Marquadt (see Ref. 24) with some
modification for constrained optimization. The result shows less than 3 % error in
the geometric mean relation and a mean value of over 0.95 for the interaction
parameter in all cases.
The samples of different alkyl chain lengths (n 4, 8, and 18) and different alkyl
surface concentrations obtained by varying the silane concentration

toluene) in the treatment solution were characterized by contact angle
measurements via the Wilhelmy plate method. Five different solvents were made
up by mixing water and methanol.
Advancing contact angles measured for these modified glass substrates are
shown in Table 2. For the monofunctional bonded phases = 1, 4, 8, and 18)
in various solution environments, contact angles increase with both increasing alkyl
concentration and increasing water fraction in the mixed solutions. With
increases in the fraction, for 4 and 8, contact angles decrease steadily
to around whereas for the 1 and 18 cases, even with a pure envir-
onment, contact angles approach due to the high hydrophobicity of silane on
glass substrates. In an aqueous environment, the chain is "folded" and shows
2 Advancing Contact Angles as a Function of Alkyl Chain Length and Log

Concentration
Concentration toluene)
(surface
0.5
indicates no tested
error range within 0.5.

hydrophobicity, whereas in aqueous methanol solvent the chain is "brushlike" or
"extended" as stated by and Sebestian Due to the lower surface ten-
sions of mixed solvents, contact angles for glass surfaces modified with silane
are low.
Using the above contact angle data, corresponding surface energies were
obtained by using the nonlinear programming method as shown in Fig. 3.
Figures 3a-3d show the surface energies of four different silane immobilized
glass surfaces as a function of concentration. For all chain lengths the general
trends for concentration are similar. As solution concentration decreases, contact
angle values generally increase. Hence, total surface energies and the polar term
increase, whereas the dispersion term gently decreases.
In this model system, we may change the relative surface concentration by
preparing the samples with different silane solution concentrations. At low con-
centration, there may be less surface density and the surface may be nonuniform of
an "island type" The predominance of the hydrophilic phase causes low
4 -3 -2 -1 0
LOG CONC (MO
-5 4 -3 -2 -1 0 -5 4 -3 -2 -1 0
LOG CONC (MOLE'S00 LOG CONC TOLUENE)
Contact angles for four different alkyl lengths as a function of log

concentration in pure water. (a) = 1; (b) = 4; (c) = 8; (d) = 18.

17 ark
contact angles and high polar surface energy. Also the disperse surface energy term
showed a lower value. However, the total surface energy exhibited higher values.
At high concentration the bonded alkyl chains are more prevalent and are
probably relatively densely packed at the higher silane solution concentrations.
Therefore, modified surfaces with alkyi chains can show a larger contact angle.
Therefore a lower polar term of surface energy, like paraffin surfaces. However, a
higher disperse term of surface energy was shown as expected, causing the total
surface energy to decrease.
Water can usually be used as a unique solvent for the study of biomedical
materials. In other applications such as reverse-phase liquid chromatography, how-
ever, different solvents such as water-methanol mixed solvents are used. As the
fraction increases, the surface tension of the solution decreases. In mixed
solvents, the trend of contact angles as a function of concentration was generally
similar to that of the aqueous case. As the proportion of methanol increases, the
contact angle decreases dramatically as in the case of decreased concentration.
ffect of Various on Contact
Several factors govern the exact nature of alkyi chain surfaces. First, the function-
ality of silane materials may affect their monomeric or nature.
Second, the degree of and the surfce concentration may influence the
accessibility of water molecules to the unreacted hydrophilic groups remain-
ing on the underlying glass surfaces. Third, the length of the alkyl chain is a main
factor in this work. It is expected that for moderate chain lengths, the chains are
relatively mobile There are interactions between the chains van der
forces) at long chain lengths These different chain types, folded or extended,
may affect contact due to the different outermost groups exposed
and According to I R spectroscopic and ellipsometric studies by
Porter et the long (about = 18) form a densely packed,
"crystal-like" assembly with fully extended alkyl chains. As alkyl chain length
decreases, the surface structure becomes increasingly disordered and liquidlike with
lower packing density and coverage. Fourth, the effect of an aqueous environment
is different from that of the aqueous methanol solvent environment. Although we
usually use water for biomedical studies, for chromatographic studies the mixed
solvent is very important. In a mixed solvent environment, the bonded alkyi chain
can be 'bristly" or
Figure 4 show total surface energies as a function of concentration with four
different chain lengths. Except for the n = 18 case, there are transitions in surface
energies just below This results in two contact angles for = 4
and 8 in two different relatively high (0.1 and and low and mol)
concentration ranges, and the possible reason is described below.
Figure 5 shows the advancing contact angles as a function of alkyl chain
length at four different high and Sow surface concentrations mol alkyi
toluene). With increasing carbon number the advancing contact
angles decrease until = 8 at relatively high concentrations (0.5 and 0.1
or decrease until 4 at relatively low concentrations and

-5 -4 -3 -2 -1 0
LOG CONC 100 TOLUENE)
FIG. 4 Total surface energies as a function of concentration with four different chain
lengths.
.5 Advancing contact angles for four different concentrations (0.5, 0.1, and
as a function of carbon number in the alkyl silanes.

then all cases increase until = 18. The = 4 and 8 cases,
which are in the intermediate chain length ranges, showed lower surface energies
than either the = I or 18 cases.
Contact angle data are not presented for chain lengths in the range between
= 8 and = because silanes o f these lengths were not available commercially.
However, the van der Waals forces at intermediate chain lengths = 12, 14,
16) may be expected to increase continuously up to = 1 8 , well known to be
capable o f self-assembly. According to Johnson and Dettre the advancing
contact angle is associated with the low surface energy region high contact
angle).W e can tell that the advancing angle may depend on the total hydrophobi-
city o f surfaces,which also depends on the number and chain length o f hydropho-
bic alkyl groups and in the silane molecules adsorbed on the
surfaces. N M R studies show that the o f terminal methyl carbons o f
monochlorosilanes on silica particles increases until = 8 ; then at
= 18 it becomes almost constant.
Now we analyze the trends in Fig. 5 in two steps: (1) the silanization procedure
in toluene solution and ( 2 ) contact angle measurement in an aqueous environment
and their surface energies. In the = 1 case, silane molecules can bind to reactive
sites on the surfaces in a densely packed state but probably patchy style, due to the
three bulky methyl groups per silane molecule.
For the longer = 4 and 8 cases their chain lengths are 0.71 and 1.20 nm,
respectively), the adsorbed amount may decrease slightly until = 8 , probably due
to an increased excluded volume effect.Therefore, the mean distances between
chains increase with In the = 18 case chain length is the
adsorbed surfaces may be more densely packed due to van der Waals forces
between the long chains. The reason for changing minimum contact angle at
with a decrease in concentration may be the contribution from the combi-
nation o f van der Waals interaction due to different alkyl lengths and concentra-
tions.
During measurement o f contact angles, the expected behavior o f hydrophobic
alkyl chains in an aqueous environment is as follows. In the case o f = 1 , because
water molecules contact their rigid, hydrophobic, methyl groups on the surfaces,
they exhibit a high advancing angle, which causes lower total and polar surface
energies. the = 4 case, the advancing angle i s lower than that o f = 1 . Due to
the folded style in an aqueous environment, it may expose the many methylene
groups on the surface.W e know that the methyl group has a lower critical surface
tension 24 than that o f the methylene group, 31
Therefore water molecules may touch methylene groups in chains, resulting in a
lower contact angle. the n = 8 case, the general trend is similar to that o f the
n 4 case. However, the greater methylene numbers in a chain cause the advancing
to decrease slightly. In addition, slightly decreased amounts o f attached silane
may contribute to the decreased contact angles until n = 8 .
For moderately longer alkyl chains = 4 and due to the low degree
o f van der Waals interactions possible between the attached intermediate length
alkyl chains, there may be an excluded volume for the folded chains to be
relatively more mobile, disordered, and "liquid-like" than the self-assembling,crys-
tal-like 18 case. Therefore, there may be a shielding effect o f the folded chains

= 4 and 8) against water molecules approaching glass surfaces. For such
intermediate alkyl lengths, total surface energies are relatively higher for the above
reasons.
In the case of = 18, the contact angle increases again. Due to alkyl chain
interaction and packing, the water molecules may touch only the outermost alkyl
chain methyl groups on the top of the surfaces, meaning that the water molecules
find it difficult to penetrate into the bundles, giving less chance to contact
methylene groups inside bundles. Hence, the advancing angle is high for n =
Therefore, surface energies become low again.
At lower concentration, the general shapes are similar to those
trends at higher concentration, 0.5. Minima near = 4 and 8 are again observed at
high concentrations; hence, we believe that at low surface coverages, the chain may
be mobile and adopt folded styles at = 4 and = 8. At the long chain length,
n = 18, these insufficiently covered surfaces from partially ordered "islands" rather
than homogeneous layer Water molecules may contact most methyl groups on
the top of these islands rather than methylene groups. Therefore, as in the high
concentration case, the contact angle increases again at = 18. The simplified
model systems for these surfaces for four different alkyl chain lengths are shown
in Fig. 6, based on the above descriptions.
Contact angles obtained with the Wilhelmy plate method and near surface energies,
d
and obtained with nonlinear programming methods were varied with chain
length surface concentrations for monochlorosilanes 1, 4,
8, and 18) in various solvents.
As solution concentration decreases, total surface energy and the polar term
increase, whereas the disperse term decreases, due to surface density with
18
Simplified model for four different chain lengths.

In the cases of chain lengths n = 1 and = 18, contact angle and surface energy
held high and low values due to patch-type terminal methyl groups and relatively
high van der Waals interactions between long chains, respectively.
F o r both and low surface concentrations, monomeric silanes in the
range of = 4 8 exhibit the minimum contact angle and higher surface energies,
This suggests that these chains may be relatively disordered o r
then the self-assembling, crystal-like, = 18 case. This is probably because of the
excluded-volume effect due to the low degree o f van der Waals interactions possible
between the attached intermediate length chains.
Andrade, D.E. Gregonis, and Smith, in Surface Aspects

of Biomedical Polymers Andrade, Plenum, New York, 1985, p. 15.
Yasuda, A. K. Sharma. and Yasuda, Polym. Phys.
(1981).
Ward, Org. Coat. Chem. (1981).
D. Andrade, in Polymers in Medicine, Vol. ( E . et Plenum,
New York, 1986, p. 29.
Kessaissia, Papirer. and J. Colloid Interface Sci.
(198 1).
Sander and A. Wise, Anal. Chem. (1984).
L. C. Sander, J . B. and Field, Anal. Chem. (1983).
and A. Squires. Chromatogr. (1981).
Andrade, in and Interfacial Aspects Biomedical Polymers (J.
Andrade, Plenum, New York, 1985, 105.
L. and Lo, (1984).
A. and Harris, Am.
Chem. (1984).
L. Faccini and A. P. Macromolecules (1984).
G . Levy, Dechter. and Am.
Chem. (1984).
and Maciel, Am. Chem. (1983).
0. and A. Colloid Interface (1966).
Van A. Hogt, and Colloid Interface Sci.
(1986).
Rehbreg and C. Fisher, Chem. (1948).
Chromatographia (1975).
Gun and J. Colloid Interface (1966).
Andrade. Ma. King, and D. Colloid Interface
(1979).
Owens and R. Wendt, Sci. (1969).
Y. and R. A. Colloid Surf. (1988).
A. Girifalco and R. Good. J. Chem. (1957).
Kuester and J. Optimization Techniques with FORTRAN,
Hill, New York, 1973.

25. I. and Chem. Int. Ed, (1969).
26. Deckinan, and Alvarez. Proc. Electrochem.
(1985).
27. 0. Levine and A. Chem. (1957).
A. and Gould, Adv. Chem. Ser. (1964).
29. Porter, Bright, and Am. (1988).
R. Johnson and Dettre, Surf. Colloid (1969).

Department of Pharmaceutical Sciences, University, Osaka,
Japan
I. Introduction
I I . Hierarchical Structure of Collagen Fiber
A. Hierarchical structure of collagen fiber
B. Surface structure of collagen fiber
III. Pretreatment and Inner Structure of Collagen Fiber
A. Pore size distribution determined from adsorption isotherm
of water vapor on collagen fiber
B. Pore size distribution determined from adsorption isotherm
of nitrogen gas on collagen fiber
IV. Experimental Procedures
A. SEM, TEM. and AFM
B. Adsorption method
V. Summary
References
Collagen, the essential component of the extracellular matrix, is present in the skin,
bone, tendons, and connective tissues of the higher animals. The collagen fibril is
functionally related to the tenacity in bone, the tensile strength in tendons, and the
preservation of strength in skin or connective tissues. In the native state, collagen
exists in a form. Water is essentially related to the maintenance of collagen
molecules in the native conformation. That is, the inner spaces inside the collagen
fibers functionally maintain water in the native state.
Observations of the collagen fibril surface structure were performed using
atomic force microscopy (AFM) scanning electron microscopy (SEM)

186
and transmission electron microscopy These observations mainly give

visual information about the of the collagen fibril surface. the
other hand, the microporous structure, that is, the pore size distribution and pore
volume. has been investigated by the adsorption method
I n this chapter, the structure of the collagen fiber, the surface and
structure of the collagen fiber. and experimental methods for the investigation of
the are discussed.
To understand the structure of the collagen fiber, the structures of the

collagen molecule, collagen fibril, and collagen fiber must be recog-
nized.
1. Collagen Molecule
Collagen molecules have a triple-helix and conformation. Their length and
form depend on the type of collagen. The type I collagen molecule, tropocollagen,
is approximately 300 nm long and 1.5 nm in diameter. It is composed of a
helix atelocollagen and N-terminal and C-terminal nonhelical telopeptides (Fig. 1).
the results of calorimetric and mechanical experiments of hydration on rat
tail tendons, i t was proposed that water molecules are fixed by three hydrogen
bonds inside the triple helix or by double hydrogen bonds in the available sites
inside the triple helix This hypothesis indicates the existence of microspacing
inside the tropocollagen structure that water molecules can enter.
2. Microfibrils
Five collagen are aggregated to pack a microfibril
scheme is shown in Fig. 2. A five-stranded inicrofibril is approximately 3.5 nm
or 3.8 [7] in diameter. Each tropocollagen is axially staggered by a
period with respect to its nearest neighbors. The period includes one gap
(hole) zone and one overlap (head and tail overlap) zone. The D period has
been reported to be 67nm [9] or According to
N-terminal Atelocollagen C-terminal

(triple-helix)
(nonhelical)
.1 Scheme of type I collagen molecule.

Fibers
period
crosslink
collagen nonhelical telopeptide
nonhelical telopeptide
ridge
FIG. 2 Scheme of microfibril. (a) Cross section; (b) latitudinal section
et water molecules are fixed in the microfibrils and in the hole zones
adjacent to the end of the tropocollagen. It was reported that intermolecular
space enclosed by the five tropocollagen in a unit is
This indicates that there is microspacing inside the
3. Collagen Fibril and Collagen Fiber
Microfibrils are packed on a tetragonal lattice o r quasi-hexagonal lattice to
form the collagen fibril. The collagen fibril is the fibrous entity observed by A F M
or The diameters of the collagen fibrils are close to a multiple of
The collagen fibrils form a fiber, and the fibers are bundled together
with elastic fibers and fibrocytes to form fibrous connective tissue.
1. Electron Microscopic Surface Observation of Collagen Fibril

et [2] reported field emission electron gun surface obser-
vations of collagen fibrils dehydrated by different procedures. Figures 3-6 are
micrographs of air-dried (AD) fibrils, lyophilized (L) fibrils, critical point-dried
(CPD) fibrils, and freeze-dried ( F D ) fibrils in tert-butyl alcohol, respectively. The
A D fibrils appear flattened and distorted, and the D period is at times evidently
altered and disrupted (Fig. 3). The L. C P D . and fibrils appear rather similar
(Figs. 4-6). Their surface features show high porosity (Figs. 3-6). However. the
micropore sizes o r shapes are not easily observed in these pictures. Three
ridges, and lie within the D period The ridge is
located in the gap zone (Fig. 2). The and ridges are located on the
overlap transition and correspond in position t o the N-terminal nonhelical
tide and nonhelical telopeptide, respectively (Figs. 2 and 4). The L fibrils
exhibit a sharp D banding and two evident ridges. The ridge is recognizable by
its larger size (arrows in Fig. 4). The CPD fibrils show a gap- overlap alternation
marked by two transverse ridges (Fig. 5). The FD fibrils show a D period, and
ridges, and a sort of flaky and filamentous coating laid upon the fibril surface

Boki
FIG. 3 Field emission electron gun-scanning electron microscope (FEG-SEM) picture

of air-dried collagen fibrils ( x 160,000). (From Ref. 2.)
FIG. 4 FEG-SEM picture of lyophilized fibrils ( x 160,000). (From Ref. 2.)

Collagen Fibers
FIG. 5 FEG-SEM picture of critical point-dried fibrils ( x 160,000). (From Ref. 2.)
FIG. 6 FEG-SEM picture of freeze-dried fibrils in tert-butyl alcohol

(From Ref. 2.)

(arrows) (Fig. 6). fibril ultrastructure observed using transmission electron
microscopy distinct ridges,
align regular from right to left, respectively (Fig. 7) the
appear as a distinct row of globular subunits.
2. Observation of Collagen Fibril
the AD rat collagen by AFM. is operated
by lightly touching a tip to the sample and producing a topographic by
measuring the sample height as the tip scans over the sample. cross sections
of the images are graphed in Figs. 8 and 9. The cross section shows the
periodicity, and groove widths by dividing lateral distance by
groove values are 63 nm (Fig. 8) or 60 nm (Fig.
From Figs. 8 and 9, we observe the difference in height between the grooves
and the ridges of the collagen fibril surface. The difference varies between 5 nm
on the 15 nm fibrils (Fig. 8).
difference between overlap (elevated) zone and (depressed) zone
4nm Two grooves per D period are observed on the smaller diameter sheath
fibrils (Fig. 9). The deep grooves measure 6 nm in depth, and minor grooves
measure nm in grooves show hints of the presence of
pores surface.
The D period of the rat tail tendon collagen was reduced by dehydration from 68
to 64nm The micropore volume of the collagen fiber was increased by
acidification Therefore, the inner structure of the collagen fiber was also
Transmission electron micrograph of freeze-etched collagen fibrils;

of 40,000 Ref. 2.)

position
FIG. Differences in height grooves and ridges of D periodicity of
in cross section (native collagen fibril), Fibril height is 300
(the zero points on both axes are arbitrary). (From Ref.
. 9 Differences in height between grooves ridges of D periodicity of

microfibrils cross rat tail sheath fibril), Fibril is 22 (the
zero points on both axes are arbitrary). (From Ref.

effectively varied by its The microspaces in triple helix,
fibril, and collagen fibril are decreased or increased by chemical
drying methods
The adsorption method has an advantage of evaluating the
structure of the collagen fiber. That is, the inner pore volume of the collagen fiber is
determined by using adsorption method. case, it is that the
pores are cylindrical and capillary condensation occurs in the pores according
to Kelvin's equation,
where is the surface tension, is molar volume of the condensing vapor, is

the angle of the liquid with the is Kelvin's radius, and is the relative
pressure. The pores with widths of less than about in radius are
termed whereas pores widths from 2.0 to are referred
to as or "transitional pores."
Distribution Determined fro

of Water Vapor on
The differential size distributions obtained by et [3] are shown
Figs, 10 and 1 The raw collagen fibers prepared from a portion of bovine belly
hide have radii of 3.9-7.7 nm and micropores radii
0.55 0.45 They suggested that a of 0.9 is
Pore size distribution of and enzyme-treated collagen fibers.

calculated using the method. Ref.

Fibers
Control collagen
Enzyme treated 11
collagen
Tanned collagen
. 11 Micropore size distribution of raw, enzyme-treated, and tanned collagen

fibers. Micropore volume was calculated using the MP method. (From Ref. 5.)
near the intermolecular spacing (1 inside the microfibril obtained by

angle X-ray diffraction When the components are removed by
the enzyme treatment, the 3.9-7.7 radii pores are widened to have radii of
(Fig. 10). When the raw collagen fiber is tanned, the 3.9-7.7 n m radii pores
are absent (Fig. 10). However, neither the enzyme treatment nor the tanning causes
any change in the of radii 0.35-0.55 n m (Fig. 11).
et al. reported the structural analysis of alkali-treated [4] collagen fibers

followed by preparation in acidic p H solution The cumulative pore size dis-
tributions are shown in Fig. 12. No micropores with radii less than are
found o n the raw collagen fibers prepared from steer hide. However, micropores
with radii of 1.2 o r larger appear after the alkali treatment. These results are not
in agreement with the data obtained by the moisture adsorption method [3] o r
angle X-ray diffraction
The change in pore volume brought about by acidifying the alkali-treated col-
lagen fibers was investigated The pore size distributions of the collagen fibers
are shown in Fig. 13. Micropores with radii of 1.2 o r larger appear after acid-
ifying the raw collagen fiber with pores of radii larger than The increase in
pore volume is greater with radii of 1.2-1.55 o r 1.55-2.45 n m rather than with radii
of 2.45-9.5 nm. It was suggested that the interspaces of the triple helix o r microfibril
are increased due to the formation of carboxylic acid moieties during the acid
preparation

oki
FIG. 12 Cumulative pore volume of alkali-treated collagen fibers. C-10, C-20,

and C-30 are collagen fibers alkali-treated for 0, 10. 20, and 30 days, respectively.
Micropore volume 1.55 nm) and macropore (8.5 1.55 were
calculated using the method and the DH respectively. (From Ref. 4.)
. 13 Pore size distribution of collagen fibers. The labels C-0, CS-0, C-10, CS-10. C-
20, CS-20, and CS-30 refer to the number of days the raw collagen fibers were
alkali-treated. C denotes the alkali-treated and collagens. denot
alkali-treated, freeze-dried, and acidified collagens. 1.2
1.55 2.45 (From Ref.

The air-dry preparation technique is inherently prone to artifacts Other
dehydration procedures, including Iyophilization, critical point drying, and
freeze-drying, provide better results on biological for S E M and
observation
I a m not experienced in the techniques of preparing a specimen, instrumenta-
tion, methododoly, resolution. etc. References and 18-21 may be useful for
additional reading.
The pore size distribution of collagen fiber is determined by using adsorption iso-
therms of water vapor o r nitrogen gas. The preparation of the sample. adsorption
procedure, and calculation method are mentioned below.
1. Preparation of Sample
Collagen fibers are from a n organic matrix. They are thoroughly washed
in distilled water and prepared through a process of several steps. The sample
should be dried. The fiber is dried in a vacuum desiccator over phosphorus
oxide for 3 months to completely remove water from the collagen fiber The
water inside the triple helix [5] be removed. If the water inside the collagen
fibers is not completely removed, then the amounts of nitrogen gas adsorbed will
vary widely. This wide variation in the amount adsorbed may be caused by the
phase change of the residual water inside the collagen fibers due to an extremely low
adsorption temperature of
2. Adsorption Procedure
The water vapor and nitrogen gas adsorption isotherms determined by
using automatic adsorption apparatus have been reported. O n other hand. the
manual adsorption method for the adsorption of water vapor and nitrogen gas is
discussed in Refs. 25 and 4, respectively. The adsorbed is determined by
using a Shibata moisture and surface area apparatus, Model P-850
Scientific Technology, Ltd., Tokyo, Japan) The of water vapor
and nitrogen gas adsorbed are determined by the gravimetric method and
volumetric method respectively.
The pore size distribution is calculated based on the water vapor or nitrogen
gas adsorption isotherm. Before the calculation, it is essential to obtain a n accu-
rate adsorption isotherm. The amount of nitrogen gas adsorbed on the collagen
fiber is extremely low in comparison to the amount of water vapor adsorbed on
it. Approximately. 1 x kg of collagen fiber. therefore. is put in the adsorp-
tion glass bulb. The amount adsorbed on the collagen fiber is measured at equili-
brium pressures of 6.7, 13.3, 20.0, 26.7, 33.3, 40.0, 53.3, 66.7, 80.0, 93.3, and
101.3 in order to get an accurate quantity. T h e adsorption isotherm for each

1
collagen fiber is performed in triplicate; therefore the reported amounts adsorbed

are mean values
3. Calculation of Pore Size Distribution

The pore size distributions are calculated using the (MP) method
and the Dollimore-Heal (DH) method based o n the adsorption isotherm data.
Prior to calculating the micropore distribution by the MP method, it is necessary to
convert the adsorption isotherms into plots The amount adsorbed at
each relative pressure on the basis of the adsorption isotherms in Fig, 14 are
converted into and t values for the corresponding relative pressures calculated
from equation. The plots. as shown in Fig. 15, begin to deviate
Relative pressure
. 14 Adsorption and desorption isotherms of nitrogen gas on collagen fibers. C-0,
C-20, and C-30 are collagen fibers alkali-treated for 0. 10. 20, and 30 days,
respectively. Experimental data of adsorption and desorption. (From Ref. 4.)

. 15Volume of nitrogen gas adsorbed versus statistical thickness. C-0. C-20,
and C-30 are collagen fibers alkali-treated for 0, 10. 20, and 30 days, respectively.
amount adsorbed converted into liquid; t = adsorbed thickness. (From Ref. 4.)
downward from straight lines. The downward deviations are used for the calcula-
tion of the pore and the pore size distribution of the
u p to 1.55 The pore size distributions for radii of 1.55 n m or larger are
calculated using the D H The cumulative pore size distributions cal-
culated using the MP and D H methods are shown in Fig. 12.
The surface structure of collagen fibril is visually described using the SEM,
and TEM observations. The of collagen fiber is quantitatively
described using the pore size distribution calculated from adsorption isotherms.
The depends on the achievements of other researchers. The latter is mainly
my work. It is regrettable that the SEM o r AFM observation cannot be compared
with the pore size distribution o n the same collagen sample. The functions of
micropores inside the collagen in the native state must be elucidated.
I. Tran Garrone, and Franc, Cell.

(1994).
2. Raspanti, A, Alessandrini. P. and A. Microsc. Res. Tech.
(1996).

N. Colloid Interface
(1976).
K. Colloid Interface
(1993).
and G.
Nature
A. Miller, and Nature (1971).
A. and A. Bailey, in of
Connective Skeletal Fitton-Jackson, S.
Partridge, and London. 1965,
A. 2151931(1967).
and D. (1993).
A. and S. Bear. Leather
A. Parry,
A. Miller, Nature
A. Parry, and A. Craig: (1979).
A. and L. Torsi
(1969).
R. and in
Academic, 1973,
Boki, Kawasaki, and Colloid Interface
Yang, A. and
Inoue. and Arch.
Arch. (1992).
A. Ottani. and A.
(1997).
A. and Gamble.
Naono and Colloid Interface (1991).
Naono and Colloid Interface 1281146(1989).
Boki. and Food
Colloid Interface 26145
and Heal, Colloid Interface (1970).

of Civil and Jersey
Institute of Technology, Jersey
Engineering, Illinois of Chicago,
I. Introduction
II. A Conceptual Model for Sorption of Inorganic Substances at
the Oxide/Water Interface
III. The Diffusion Model
IV. Implications from the Conceptual and mathematical Models
V. Critical Review of Relevant Literature
VI. Conclusions
References
In natural aquatic systems and in engineered processes,

mineral surfaces can play an important role determining the transport and
fate of inorganic compounds. These interactions can influence nutrient
processes, and in disturbed
faces caii the of chemicals. surfaces,
oxides probably play a dominant role because of their large surface areas
strong affinity for trace In this chapter are especially interested
in poorly crystallized porous oxides of and manganese. These
can exist as discrete particles and as coatings on other mineral
surfaces. not stable a therinodynamic sense, and transformation
of these to more stable forms are often limited because of
environmental conditions or the presence of other inhibiting
The chapter a conceptual model describing sorption of inorganic
materials onto oxides and a brief discussion about possible for the
often observed two-step process, cases intraparticle is

the rate-limiting mechanism, and includes expression for that
Together, these conceptual mathematical models suggest
should be observed in studies of these systems; review implications and
then look into the literature for examples. Although there is confirmation of
these ideas, few of the existing studies were designed to test hypotheses. Lastly,
we present our conclusions.
Sorption of to oxides can be divided into a number of sequential

possibly parallel steps. (In this chapter follow the suggestion of and
use the term sorption to represent adsorption, absorption. and surface precipita-
tion.) The adsorbate must be transported from the bulk solution through
hydrodynamic boundary layer to the oxide external Once at the sur-
face, the adsorbate will try to find a desirable and/or available surface site to react
Initially, sites are presumably readily available on or near the external surface.
Eventually, the adsorbate must leave the surface and to into
the particle interior in search of appropriate binding site. This intraparticle
diffusion step is our subject. Both of these first two steps involve some
form of to maintain the electroneutrality condition the appropriate
of of counterions must diffuse in concert with the adsorbate.
Once an appropriate site is found, assuming the formation of an
sphere complex, the adsorbate lose its of hydration and then bind to
that surface site. Most conceptual and mathematical of that surface com-
reaction describe it as exchange reaction. that no previous
surface complexation reactions have occurred, sorption of or anionic
adsorbates that hydronium or hydroxide ions, respectively, will be released
from the surface. These exchanged can either diffuse back into the bulk solu-
tion or find a desirable and/or available site to react Once again,
must be maintained. so appropriate counterions will redistribute to balance
the surface charge diffuse the hydronium or hydroxide ions. In sum-
mary> overall sorption process can be divided into four steps:
Diffusion through the hydrodynamic boundary layer to the external surface
Diffusion to a available surface site
that surface site
of ions released from the surface
Keep in mind, however, that each of these steps is more than the
foregoing summary statements imply. of the factors to consider
include waters of hydration diffusing the adsorbate and the that
diffuse with the adsorbate to maintain electroneutrality. Furthermore, adsorbate
reaction with a specific surface site does not guarantee a permanent surface com-
plex. It is possible that although reaction the first site was favored,
there is another, more energetically favorable site available to which the adsorbate
could migrate.

Most studies of inorganic sorption by oxides report a two-step process
Initial uptake, which is rapid and typically complete within hours, most likely
involves sorption to external sites. However, initial uptake, second-
ary changes continue for o r months o r longer. A variety of mechanisms have
been suggested to explain the long, slow uptake process, including the formation of
surface complexes, the of solid solutions [3] or
precipitates particle-particle interactions slow chemical reaction rates
and adsorbate diffusion into the interior of porous particles
A of the literature suggests that different may play impor-
tant roles in different systems. For example, results from pressure-jump studies of
sorption reactions at nonporous surfaces have that the chemical reactions
themseleve are very fast so that in cases a chemical reaction rate
is not likely a rate-limiting step. Spectroscopic evidence indicates that in some
systems surface precipitation or solid solution formation does occur even at
high adsorbate loadings We that intraparticle diffusion
can be an important rate-limiting mechanism for all porous oxides.
Transport of an adsorbate through interior pores of an amorphous oxide involves

intraparticle diffusion For macropores, occurs by diffusion.
Although the International Union of Pure and Applied Chemistry (IUPAC) defines
micropores as pores of mesopores as having and
macropores as having d A this categorization is based on gas-solid
average pore diameters for amorphous and Mn oxides are 20A
[1 and 35-143 A respectively. In the aqueous phase, amorphous oxides
are hygroscopic, results in layers of adsorbed on the particle
both external and internal surfaces These layers of adsorbed on
that are typically characterized as mesopores result micropores, For the
porous amorphous oxides. flux into the particle is due to the concentration gradient
this includes both surface and pore diffusion, In surface diffusion the adsor-
bate desorbs from one site and to an adjacent vacant site. Concurrently, the
adsorbate will diffuse in the aqueous phase of the pore. The overall rate of these
parallel diffusion be determined by the fastest process.
The rate of metal removal in the porous oxide can be described a
film transfer process and either surface pore models. T o
the transfer of adsorbate from bulk solution to the adsorbent surface. some
studies assume a linear concentration gradient existing in a hypothetical film sur-
rounding the adsorbent particle When film transfer limits the rate, which, for
example, is likely nonporous particles, the equation can be used to
simulate the transfer in a batch reactor:
contaminant concentration in the aqueous phase

contaminant concentration in aqueous phase at external surface of particle

film transfer coefficient
of adsorbent
R particle radius
solution
porosity in reactor
apparent of adsorbent
However, for oxides,
microposous structure
intraparticle diffusion For oxides, Axe
Anderson showed transfer resistance is to
particle diffusion
a spherical particle, the species mass balance yields the
this expression, is the the sorbed phase is a

function of process, is the porosity of particle, is
surface diffusion coefficient particle, is oxide
density, is the pore coefficient the and is the radial
position the from the center. aqueous environments,
where are dilute, diffusivities are constant Also,
the is rapid or for adsorption to oxides; the
reduces to
where adsorption is described by the fitting When

the contaminant is upon oxide precipitation. the is
0) = 0
At center of particle. bondary condition is given by
and at the external oxide surface, of two conditions

due to a turbulent hydraulic the concentration at the surface is in
with the aqueous phase o r by the flux of adsorbate to
the surface:
t)
Distinguishing between pore and diffusion. however, be difficult if

the surface boundary is However, diffusion in

micropores is negligible there is a strong affinity between the adsorbate and
the surface, resulting in high adsorbate loadings The sorption of
metal contaminants to amorphous oxides is one of a strong
bent- adsorbate system where surface diffusion dominates Diffusion obeys
a random walk model where
In this expression, defines the mean square hopping distance, and the probability
that the adsorbate has the energy required to h o p to the next site is the
factor with representing the activation energy. The adsorbate vibrational fre-
quency. is a function of the molecular mass and the force constant, which is
dependent o n the potential energy on the pore surface.
In this chapter, we d o not distinguish between surface and pore diffusion but
instead talk about an overall effective diffusivity. In this way, we can include all the
different approaches in our review. In summary, the mass balance model is based
on the following assumptions: The adsorbent is composed of uniform, porous
spheres with uniformly distributed sorption capacity, sorption is reversible, and
diffusion is not a function of the particle dimension. These conceptual and math-
ematical models provide information about how sorption kinetics and the general
system behavior can be influenced by changes in environmental conditions.
If inorganic sorption onto porous oxides follows the conceptual and mathematical
models presented above, certain types of behavior should be observed. F o r exam-
ple, in many aquatic environments, amorphous Fe, and oxides often pre-
cipitate in the presence of other dissolved ions. Under these conditions, adsorbates
that have a n affinity for oxide binding sites should have access to those sites during
the formation process before they become sequestered in internal pores. Therefore,
when coprecipitation occurs, intraparticle diffusion is probably not a rate-limiting
step in the sorption process. Coprecipitation conditions, however, usually are not
equilibrium conditions, because the amorphous oxides continue to age and trans-
form into more stable mineral forms. During this aging process, adsorbates that are
associated with the changing solid could become involved in several different reac-
tions. If it does not fit well in the evolving mineral lattice, the adsorbate could be
released to solution. Under these conditions, initial removal during coprecipitation
would exceed the equilibrium conditions, and aging would cause a n increase in
adsorbate concentration in solution. In contrast, if it readily substitutes for
major metal ions in the lattice, the adsorbate could become fixed in the oxide. I n
all cases, the rates of aging and of the oxide, and possible release of
adsorbate, would be controlled by intraparticle diffusion.
Adsorbate size relative to the size of pores in the oxides is also an important
factor to consider. the IUPAC classification, it seems reasonable that inor-
ganic adsorbates should have access to binding sites in pores of the size

and larger. Micropores, however, may not be accessible to all adsorbates, especially
the larger oxyanions such as phosphate, and selenite. Furthermore, as
mentioned earlier, these pore size categories were developed for gas-solid systems
and may not be appropriate for oxides in aqueous systems. For example, oxides in
solution have hydrated surfaces, and, depending on surface charge, an electrical
double layer may be present. The form and extent of these surface phenomena
within pores are not clear.
Another complicating factor involves the exchange reaction at the surface
wherein or anionic adsorbates form a surface and hydronium
or hydroxide ions, respectively, are released. These released ions could diffuse
back into bulk solution, but, depending on initial conditions, there may be
sites deeper within the particle. If so, during at least part of the
sorption reaction there should be a p H front preceding the adsorbate front diffus-
ing into the particle. In effect, availble sites deeper within the particle would be
by the hydronium or hydroxide ions released during sorption. Eventually,
the pore and quantity of available sites become limiting and the
released hydronium or hydroxide ions must diffuse out of the particle.
These surface charge effects could also influence other aspects of the
chemical behavior of the suspension. For example, depending on the solution ionic
strength and the pH versus ZPC relationship, particles could aggregate, creating
additional pore volume. Furthermore, surface charge and ionic strength must influ-
ence the counterion diffusion that is required to maintain
In summary, although intraparticle diffusion is a rate-limiting mechanism for
sorption in porous minerals, implications in using the diffusion model include
accounting for such effects as coprecipitation, adsorbate type, potential exchange
reactions, sorbent and solution chemistry, and the stability of the particle size
distribution.
Although the subsurface system is composed of a number of amorphous

oxides are an important phase in controlling metal contaminant distribution
because they exist not only as discrete particles but also as coatings on other
minerals. Sorption reactions between the bulk aqueous phase and organic matter
and/or clay are generally readily reversible; as a result, adsorbates are eventually
redistributed to porous amorphous oxides, where, in the slow sorption process of
surface diffusion, the particle acts as a sink for the contaminant. In laboratory
studies, Axe and Anderson found that surface diffusivities range between
and indicating that it may take years to reach equilibrium.
Likewise, when the contaminant a constant source of contaminant exists,
which is presently not included in hydrogeochemical transport models and not
accounted for during site remediation.
There are few studies that have examined configurational diffusion of metal
contaminants diffusing in aqueous mineral systems. Furthermore. experimental
methods and the conceptual and mathematical used to represent diffusion
vary widely. The diffusivities presented in Tables 1 and 2 are composite diffusivity

Adsorption onto
A Summary of Diffusivity Values from Studies Involving Single-Phase

Adsorbents with Significant Porosity
Adsorbate Adsorbent Aporox. D (cm2 Ref
Cd H F O (pH 7 )
Cd H F O (pH 6)
Cd
Alumina
Sr HFO
'Effective diffusivities summarized here represent an average from each study; actual values were
typically within an order of magnitude of these averages.
= hydrous ferric oxide. also known as amorphous iron oxide or ferrihydrite.
values, which in some cases incorporate the effects of porosity, tortuosity, and
adsorption. Also, some diffusivity values are estimates o r averages from the infor-
mation presented in the original paper. It is clear that pore size plays a significant
role in diffusion. In general, values for the adsorbents with what the
I U P A C characterizes as micropores and (Table 1) are around
In contrast, values for adsorbents with (Table 2) are
typically u p to six orders of magnitude larger. T o better understand the effect of
diffusion. ranging from to were used in the integrated
analytical solution to the mass balance using a constant boundary condition at the
particle surface and a particle diameter of 40 (Fig. In systems with
pores, intraparticle diffusion approaches the rates expected for bulk diffusion. As
pore size increases from micropores to mesopores to macropores, diffusion rates
2 A of Values from Studies Involving Single-Phase

Adsorbents with Limited Porosity and Modified or
Adsorbate Adsorbent b Approx. (cm 2/s) Ref.
Cd Proprietary a-FeOOH 1 x 29
Cd Fe oxide-coated 2x 30
Cd Mn oxide-coated GAC 31
Proprietary 1 x 29
Fe oxide-coated G A C 1x 30
Mn oxide-coated GAC 9x 31
Pb Proprietary 5 29
Cd, Zn a-FeOOH with 0.86% 2x 32
diffusivities summarized here represent an average from each study; actual values
typically within an order of magnitude of these averages.
" G AC = granular activated carbon.

IG. 1 Theoretical mass fraction sorbed versus the square root of using a particle
diameter of and the integrated solution to the mass balance with a constant
boundary condition.
can increase by from one to three orders of magnitude. Furthermore, although

few data are available, it seems that these effects are more important for large
The adsorbate's charge and solution composition will determine its size, which
affects the resulting type of intraparticle diffusion. Specifically, when pores are
large relative to the diameter of the ion, bulk diffusion is possible. F o r pores
that are small relative to the size of the ion, transport is limited to solid-state
diffusion. When pore and ion are about the same size, the ion interacts with the
pore wall as a result of physical forces. such as electrostatic and van der
forces Diffusion under these conditions is known as configurational diffusion
In aqueous environments, layers of water adsorb in what is typically defined
as mesopores, and interaction of the adsorbate with the pore wall results in con-
figurational or surface diffusion. The size differences among anions compared to
those among cations would suggest that anionic diffusion may fall into the

Adsorption onto 207
state regime for certain cases, while hydrated cation can move by both surface
and pore diffusion.
Of the studies reviewed here, only the work by al. directly exam-
ined the effects of pore size. Under similar experimental conditions. they reported
that when adsorbents with micropores and were compared. diffusivity
values for Cd and increased by factors of about 2 and 33, respectively. As
noted by those authors, the fact that is more sensitive to changes in the pore
size is likely due to the fact that hydrated selenite anion is large relative to C d
(diameters of about 6.5 A and 4.5 A, respectively). The larger selenite anion be
significantly hindered while diffusing through smaller diameter pores.
et noted that in contrast to postprecipitation adsorption
tests, uptake of onto amorphous iron oxide was significantly greater and
not diffusion-limited. Furthermore, subsequent aging of the system, they
observed desorption of presumably because it did not fit well in the structure
of the aging oxide. In contrast, there apparently are some where the
adsorbate does fit well into the structure of the aging oxide. F o r example,
et reported increased uptake of on gels aged for 205
days. The behavior could not be due to slow diffusion into the solid, however,
because in the same systems the amount of Sr, and Ba decreased with aging.
Trade-offs between diffusion and incorporation into the solid can have interesting
implications for the transformation pathways of these unstable solids.
reported a correlation between the relative affinities of anions for
and enhanced formation of hematite rather then goethite.
Because amorphous oxides are metastable minerals, over time they will crystal-
lize into more stable forms. F o r example, amorphous oxide, also known as
ferrihydrite, will restructure to hematite. However, upon dissolution of the amor-
phous Fe oxide, it would reprecipitate to goethite. These structural changes in the
oxide, which are a function of time and solution chemistry, will affect the
particle diffusion process. However, Axe studied amorphous oxide aging
with experiments that included spectroscopic analyses of the before, during,
and a t the conclusion of sorption experiments. X-ray diffraction was used to
observe the oxide aging, with and without contaminant present. The diffraction
pattern of freshly precipitated amorphous Fe oxide (Fig. 2) showed little o r no
cyrstallization; however, as it aged, more structure could be seen. Over the
patterns were analyzed for the peak widths a t half-height. After 7 weeks, there is
evidence of both hematite and goethite minerals, but the oxide without Sr had more
goethite present than the one with This crystallization increased with time; after
14 weeks diffraction pattern peaks narrowed and increased in height (less noise). At
14 weeks, there were n o significant differences between the Fe oxide with and
without The fresh, 7-week, and 14-week patterns were from one set of samples,
but in the final time series, 22 weeks. a different sample was used. F o r the 22-week
samples, the diffraction pattern of the Fe oxide without has stronger, narrower
peaks than does the pattern of oxide with An analysis of the peak widths
indicates that for the oxide without Sr there is evidence of more crystal-
lization. As a result, the 5.5-month-old samples suggest that the presence of con-
taminant inhibited the crystallization of the amorphous oxide. Interestingly. the
week-old samples showed less crystallization than the samples a t 14 weeks. Overall,

14 week old
X-ray diffractograms of amorphous Fe oxide with and without contaminant

and as a function of = hematite; goethite. (From Refs. 11 and 36.)
however, all diffraction patterns have substantial noise, indicative of an

solid. While noise decreases, revealing crystallization, with time, there is still very
little crystallization.
Amorphous Fe, and Mn oxides are generally found with adsorbed inorganic
ions and organic matter. Although amorphous oxides are metastable minerals, their
aging to crystalline stable minerals can be retarded by the sorption of organic and
inorganic species As a result, amorphous oxide minerals that have high
surface areas, porous structure, and a high affinity for inorganic species are effec-
tive sorbents for inorganic contaminants. Further studies are needed to investigate
the (or not) of amorphous oxides as a function of contaminant
loading and time for better understanding the oxide stability as well as for poten-
tially modeling the effect of the contaminant's presence.

T o accurately model contaminant transport when Fe, and oxide minerals
are present, intraparticle diffusivities are needed. Additionally, as we tried to point
out in this chapter, there are a number of implications in using the diffusion model
with oxides. Some of these implications of intraparticle diffusion have
been observed by researchers in macroscopic studies of both model and real sys-
tems. However, as only a small number of studies have been conducted o n metal
contaminant diffusion in aqueous oxide systems, many implications need yet to be
addressed such as the long-term effect of contaminants sorbed in micopores of
metastable minerals and desorption of contaminants both coprecipitated oxi-
des and oxides exposed to contaminants over long periods of time. Therefore,
future studies are needed to study and improve our understanding of this slow
sorption process, intraparticle diffusion.
A. in Processes at Mineral Surfaces A. Davis and

Hayes. Am. Chem. Symp. 323. Chem. Washington,
DC, 1986, pp.
A. Davis and Kent, in Mineral-Water Interface Geochemistry F.
and A. White, Mineralogical Society of America,
Washington. DC, 1990, pp.
and A. Stone. Environ. Sci. Technol. (1995).
A. and Morel., Colloid Interface
(1986).
A. Davis, C. Fuller, and A. Cook,
1490 (1987).
A. Anderson. I. and R. Environ.
Technol. (1985).
and Santschi, Environ. Sci. Technol. (1991).
and Environ. Sci. Technol.
(1988).
Stollenwerk and Kipp, in Chemical Modeling Aqueous Systems
C . Melchior and Bassett. A. Symp. Ser. 323, Am. Chem.
Washington. DC, pp.
V. Roberts, and 0.Leckie, Environ. Technol.
(1995).
Axe and Anderson, Colloid Interface (1995).
L. Axe and P. Anderson, Colloid Interface (1997).
K. Hachiya, M . Sasaki, and Chem.
(1984).
K . Hayes and J. 0. Leckie, in Geochemical Processes at Mineral Surfaces A.
Davis and Hayes, A. Symp. 323, Am. Chem.
Washington, DC, 1986, pp. 1
P. Zhang and Sparks, Environ. Sci. Technol. (1990).
P. R. and L. Sparks, Geoderma (1995).

Hayes. A. Roe, Brown. 0. 0. Leckie. and A.
Parks. Science (1987).
G. Brown, in Mineral-Water Interface Geochemistry, Review in
F . Hochella and A. White, Vol. 23, Mineralogical Society of
Washington, DC. 1990, pp. 309--363.
5. Combes, Brown, and A. Parks, Environ.
(1992).
A. A. Rea. C . Fuller, and A , Davis. Geochim.
Acta. (1993).
Elements of Chemical Reaction Engineering, 2nd Prentice-Hall,
Englewood Cliffs. NJ. 1992.
P. K. Satapathy. A. and Das, Colloid Interface
(1995).
and Diffusion in Zeolites and Other Solids,
Wiley. New York. 1992.
Axe and P. Anderson. in Adsorption of Variables,
Mechanisms, and Model Jenne, ed.). Academic Press, New
York, 1998. pp.
and Water Res. (1989).
Fuller, J. A. Davis. and A. Geochim.
(1993).
L. Trowbridge, Isotopic selectivity of surface diffusion: an activated diffusion
model, DOE, Oak Ridge, 1989.
and K. Chemical Reactor and Design, 2nd
New York, 1990.
and Ellis, Environ. (1994).
and removal and recovery by iron oxide-coated
granular activated carbon. Dissertation. Illinois Institute of Technology,
Chicago, 1995.
Fan, Removal and recovery of and by oxide-coated
posite adsorbent, Dissertation, Illinois Institute of Technology. Chicago,
1996.
and Tiller, Soil (1988).
J. and Wei, (1992).
Jackson, and Syers. Soil
(1976).
A. C . Burns, and A. Lawrence. Environ.
(1996).
Axe, Sr and Cd diffusion and reaction within Fe oxides. Dissertation,
Illinois Institute of Technology, Chicago, 1995.
A. in Trace in Water. Ad. 73 A. Baker, Am.
Cheni. Washington, DC. 1968, Chapter 21.
and R. (1973).

Milieux Poreux, Institut de
Physique du Globe de Paris, Paris, France
de Combustion et de Detonique.
Centre National de la Recherche Scientifique. France
MICHAEL Department of Mechanical Engineering, Israel
Institute of Technology, Haifa, Israel
I. Introduction
I I . Electrokinetic Equations
A . Flux of the ionic species
B. Flow velocity field
C. Electric potential
D . Boundary conditions
E. Equilibrium ion density distribution
F. Equilibrium ion density distribution in spatially periodic porous
media
III . Linearized Equations
A. Perturbations
B. Simplified equations
C. Dimensionless formulation
IV. Examples
A. Porous medium submitted to external pressure and potential
gradients
B. Semi-infinite void space
C. Plane Poiseuille configuration
D. Circular Poiseuille configuration
V. Numerical Simulation of Electro-Osmotic Phenomena
VI. Porous Media
A. Regular packings of spheres and ellipsoids
B. Random packings of spheres and ellipsoids
C. Reconstructed porous media
D . Discussion
VII. Fractal Aggregates

21 ler et
A. Cluster generation 245

B Macroscopic conductivity of suspensions of uncharged aggregates 247
C. Macroscopic conductivity of suspensions of charged aggregates 250
D . Electrophoretic mobility 252
VIII. Concluding Remarks 253
References 255
Consider an dilute electrolyte filling the interstices of a porous med-

ium. This medium is composed of electrically charged smooth solid particles
arranged in a spatially periodic lattice structure. F a r from the particles the solute
may be considered neutral. However, in the vicinity of the particles' surfaces the
distribution of ions within the electrolyte is affected by the particles' surface charge.
For example, positively charged particle surfaces attract negative ions from the
solution and repel positive ions. In the absence of external electric fields, pressure
gradients, or body forces, the electrolyte rests in equilibrium. In this state there
exists a region close to the particle surfaces with a nonzero volumetric charge
distribution within the electrolyte. This region is called a sheath layer
its thickness may vary hundreds of nanometers to several micrometers
When the porous saturated with the electrolyte is embodied in an
external electric field there appears a nonzero volumetric body force within
the layer, which sets the ions in that region into motion. Far from the
particle surfaces, this volumetric force is zero, since the solute there is neutral.
However, the electrolyte is brought into motion also in the latter region as a result
of the solute's viscosity. These processes lead to the appearance of an interstitial
flow velocity field ). This velocity field, when integrated over a representative
volume of the porous medium, yields a nonzero seepage velocity in the absence o f
any macroscopic pressure gradient applied to the porous This process is
called electro-osmosis, and the velocity is called the electro-osmotic velocity.
Hence, application of an external electric field results in two processes: the flow
of electric current I (which is typical of any continuum possessing free electric
charges), and the electro-osmotic velocity Both are proportional to the electric
field intensity if the latter is sufficiently weak.
O n the other hand, when a macroscopic pressure gradient V P is applied to the
porous medium, the fluid percolates through it with a Darcy velocity
Additionally, the electrolyte flowing in the interstices affects the equilibrium ion
distribution within the layer, so that these ions are also set into motion. This
results in a macroscopic electric current density I flowing through the porous
medium in the absence o f any external electric field.
When the two "forcing factors" (or generalized forces) E and V P are applied to
the porous medium, there ensue corresponding responses (or generalized fluxes ) I
and which are related by

where is the electric conductivity tensor, is the permeability tensor, and a. are
the electro-osmotic tensor coefficients.
One can imagine limiting circumstances for which the latter equations are
decoupled. Explicitly, for an almost ion-free (almost nonconductive) solution,
one can see that 1 = 0, while Eq. ( l b ) reproduces law. In the other extreme
case of a solid (infinitely viscous) electrolyte, one has = 0, and Eq. (la) reduces to
the Ohm's law. In general. e can show that a = by virtue
thermodynamic theorem denotes the transpose of the tensor
These coupled charge and mass transports have given rise to a vast literature
that started with Huckel Henry and Booth
who considered limit cases (very thin or very thick double layers) in simple geome-
tries (along a flat wall or around a single sphere) for small surface potentials. The
relevant literature has been analyzed in detail by et and it will be
simply summarized here by giving some of the major landmarks.
There is, of course, an analogy between the study of a fixed and regular suspen-
sion and that of a spatially periodic medium. Dilute suspensions of particles have
been intensively investigated by and coworkers (cf., for instance, Ref. 10
for a recent review of these works). Only a few contributions deal with nondilute
suspensions; an alternative and efficient approach is to use a cell model as Levine
and did in order to take into account the effect of the finite solid
void fraction.
Electrokinetic effects in porous media have been investigated by using either
capillary tube models or, as mentioned before, concentrated suspensions or
solidated of particles.
A first important result was obtained by Overbeek who extended
[4] expression for the electrophoretic velocity beyond the double
layer along a flat plate,
where is the fluid permittivity, the viscosity, the surface potential, and the
electric field parallel to the wall. Overbeek has shown that Eq. (2) applies locally
everywhere in the pore space provided that E represents the local electric field. If
the double layer volume is negligible, this implies the following relation between the
transport coefficients in (I):
where is the electrolyte conductivity.

The transport properties of circular capillaries were initiated by Osterle and
coauthors (Ref. 2 and references therein). Various limitations were successively sup-
pressed in these papers; the model was extended to account for nonvanishing double
layer thickness, for concentration gradients, and for strong surface potentials.
Packed beds or concentrated suspensions of particles were mostly addressed in
the limit of thin double layers. In this situation, the convected charge flux is

confined within the double layer and is determined by the local electrochemical
potential gradients. Thus, it is necessary to solve for the potential distribution,
which can be done by matching outer and inner solutions. F o r the inner problem,
the particle surface can be viewed as locally planar. In framework, and
and determined the conductivity and electro-osmosis coef-
ficients, respectively. in three types of cubic close packings of equal spheres. The
equation for the potential was solved with a expansion.
Aside from these theoretical studies, a large set of data has also been accumu-
lated since the first observation of electrokinetic phenomena by which is
only very briefly reviewed here. Attention was focused on various problems, like the
of suspensions of charged particles liquid transfers in soils
membrane separation and various model situations of theoretical interest
like suspensions of spheres o r streaming potential in specific geometries
Recently, electrokinetic phenomena have been considered for interpreting
the signals just before earthquakes Measurements of streaming
potential and electro-osmosis coefficients have also been proposed as a n easy and
powerful alterntive to direct permeability determination
This review is organized as follows. T h e general electrokinetic equations that
govern the flow of a n electrolyte are presented in along with a description of
the infinite spatially periodic porous medium.
The influence of an external electric field and/or of an external pressure is
studied in The equations that govern the perturbations of the ion distribu-
tion, the electric potential, the pressure, and the velocity are derived. They can be
further simplified and linearized when the external electric field is small with respect
to the local field induced by the zeta potentials and when the ionic distribution is
not disturbed by the flow The linearized equations are made dimensionless
by a n adequate choice of basic units.
Some simple and useful examples are worked out in IV. First, the simulta-
neous motions of fluid and charges through a spatially periodic porous medium are
analyzed: linearization enables us to summarize the results by means of four
tro-osmotic tensors that relate the fluxes of charge and matter t o the potential and
pressure gradients. cases of a void space limited by a plane solid wall
and of plane o r circular channels are briefly addressed; the symmetry properties
deduced the general theorem are verified; these simple configurations
are further used to check the numerical routines.
The general numerical method of solution is described in V. In the
approximation, all the equations are linear. First. the potential
is solved; second. the ionic potential and the velocity field are alternately obtained
in a n iteration loop. These three systems are solved by of classical iteration
schemes such as the conjugate gradient for the potentials and a technique
for the velocity. General estimates of the expected precision are given.
Results relative to porous media are presented and discussed in Regular
arrays of spheres and ellipsoids are addressed first. The influence of the surface
potentials and that of the double layer thickness are studied for a
given choice of these two electrokinetic quantities.
Random packings of spheres and ellipsoids are simulated by a sequential deposi-
tion of particles under gravity. The influence of the surface potential and that of the

Phenomena 215
double layer have been systematically studied and discussed by comparison with
the previously discussed ordered packings.
Reconstructed media based o n Fontainebleau sandstones were generated and
studied for three double layer thicknesses. These important data are discussed and
compared with existing results
Section VI ends with a general relation in the limit of large double layer thick-
ness that, to the best of our knowledge, is novel. This yields a single
reduced representation of the electro-osmotic coefficients for all the studied con-
figurations as functions of a dimensionless double layer thickness.
Section VII with suspensions made of fractal aggregates, because of the
practical importance of such structures that may be present in circumstances.
The conductivity of such suspensions is determined for charged and uncharged
clusters; the electrophoretic mobility is calculated in the dilute limit.
The chapter ends with some concluding remarks.
All the relevant microscopic equations that govern electro-osmotic phenomena are
given in this section, as well as a general description of spatially periodic porous
media.
Consider a n A'-component Newtonian electrolyte of density dynamic viscosity

= constant, and dielectric constant flowing with velocity t ) in interstices of
a porous material. Let , be the electric potential prevailing within the solute.
The flux of each ith ion species composing the solute is given by the constitutive
equation
= - + = ...
, is the ion's concentration in molecules per unit volume, is
the electric potential, and and are the ion's diffusivity and electric mobility
respectively, related by the Stokes-Einstein equation
=
Furthermore is the ion's algebraic valence (greater or less than zero), is the
electron charge (e and is the flow velocity field. The flux and the concen-
tration obey the continuity (species conservation) equation
We will be concerned with dense porous materials wherein the percolation flow
velocity is normally small and the characteristic Reynolds number is much less than

unity. In these circumstances the (generally nonsteady) flow velocity field is gov-
erned by the Stokes equations
where is the electrolyte's density, is the pressure, is the electrolyte's constant

dynamic viscosity, and is the electric volumetric force density,
with
being the electric charge density.
Potential
The electric potential is given by the Poisson equation
where is the dielectric constant of the electrolyte. It is assumed that for any
nonsteady process of ion transport, transient phenomena associated with the elec-
tric potential occur so fast that one can use (4) in a quasi-steady approximation,
assuming that = , t ) satisfies (10) with time being a parameter. This
that the electromagnetic wave propagation characteristic time is
smaller than the characteristic time associated with the diffusive species
transport , and being the characteristic interstitial scale and velocity of light,
Equations (6)-(10) have to be solved subject to the boundary conditions

v la)
where v is the outer normal to the particle bed surface its surface charge
density, and zeta potential. The two boundary conditions, namely the Dirichlet
or Neuman condition on the electric field, are found in the literature.

Suppose that neither an external electric field nor a pressure (or concentration)
gradient is imposed on the porous medium. Then one can rewrite (6) in the
form
with the obvious zero-flux solution, which, with the help of (4) and be
written as
In the above, the superscript 0, appearing on and refers to equilibrium con-

ditions. From one obtains the familiar Boltzmann distribution for
where is a certain value of (at = 0). T o obtain the equilibrium potential

distribution the introduction of Eq. (14) into (10) yields the
Boltzmann equation,
This equation is to be solved subject to the equilibrium boundary condition
where is the zeta potential of the particle surfaces. These two conditions are
related by the condition that the sum of all the charges within the unit cell is
equal to zero:
This condition will be called the macroscopic condition. the

most usual case, the surface potential is taken to be constant on
The flow velocity field equation (7) becomes, with the help of (8) and
To complete the formulation, one should specify the porous

ture of the material. Several types of ordered and disordered microstructures are
considered here that are all characterized by the spatial periodicity property
The porous medium is assumed to consist of a multitude of identical unit cells,

21 et
indefinitely reproduced within an infinite space (see Fig. 1). Therefore, the solution
of (6)-(11) is to be sought in the three-dimensional space with
,
where {I,. are the three basic vectors characterizing the unit cell of the porous
medium and the trio of integers belongs to is the
position vector within the unit cell.
At equilibrium. all properties appearing in (14) and (15) are spatially per-
iodic functions. depend only on the local position vector r :
i= ..., = = (18)
and may be considered only within the unit cell o n the external faces of which the
periodicity conditions should be imposed. These will be formulated in the form
=0 =0
where . . . denotes the difference between the values of function at the opposite
points r lying at the corresponding unit cell boundaries.
The general system of equations, (6) and is nonlinear. which greatly

complicates its solution in the general case, where the external fields
.1 Schematic view of a spatially periodic medium

are applied to the medium. One technique, generally accepted in most studies. is to
assume that these external fields are as specified by appropriate
criteria. Accordingly, the ion concentration distributions and other quanti-
ties d o not differ significantly with respect to their equilibrium values
uO = This makes it possible t o formulate the transport problem by
linearized equations written in terms of corresponding small perturbation quanti-
ties t), , I ), t):
u= = 0
In the above, the j subscript on jump conditions signifies that the component of the
jump condition along the unit vedtor I, is written for any value of
Introduce the ionic potentials by means of the relationship (cf.
Because of the combination of Eq. (24) with Eqs. (4) and (13) for the flux
of ionic species yields
Since the ionic flux across a particle surface is nil as well as the fluid velocity, ionic
potentials are subject to the boundary conditions
This condition is independent of the precise boundary condition lc).

Equations (10) and (6) may be rewritten as
Development of the velocity field equation leads to
Differentiating (24) gives

20
and using the solution for the flow velocity field at equilibrium, one can
show that
Thus, 5 perturbation fields are given by the

21V + 5 equations and along with the jump conditions in
(21)- (23).
We will further assume that all the perturbing quantities are small, which is valid
only when is small with respect to the electric field prevailing in the double layer.
A condition for V P will be established below. In addition, only sta-
tionary situations are considered. Accordingly, the time derivatives and the
order perturbation terms are dropped in (24)-(31).
Then the conservation equation for the ionic species, is reduced to a
linear differential equation.
In the absence of any macroscopic gradient of ionic concentration

the ionic potentials are subject to jump conditions.
+ = 0 implies
Similarly, the equations governing the fluid motion become
. = 0, = 0, u = 0 on
If there is no macroscopic gradient of ionic concentration +
The 4 equations and allow the determination of the N + 4

quantities and Equation (25) can be linearized by drop-
ping second-order perturbation terms; this yields

Hence, the solvent and solute fluxes can be deduced from and only.
Therefore it is not necessary to solve for and
In this subsection, the electro-osmosis is written in a

T o this end, dimensionless variables are introduced as follows.
The average ionic strength is used as a characteristic concentration,
From the Poisson-Boltzmann equation, potentials are scaled with
Distances are normalized by the characteristic length of potential variations.
This length is called the length. Therefore the electrostatic field is

scaled
The pressure is scaled in accordance with the Hoff formula
F r o m the equations governing the flow velocity field (34). one can devise a char-
acteristic
Finally, the diffusion coefficients are scaled by
These scales are used to define the reduced quantities, which are denoted by a
prime:

Using the above scaling relationships, we write the following set of reduced
equations:
This system (44)-(47) is to be solved subject to the boundary conditions inside the
unit cell,
The has still to be completed by the boundary conditions at the surface of

the unit cell:
In addition, under the assumption that either the surface zeta potential o r the
surface charge is small, the approximation can be used to
linearize (45) and with respect to , This was done in the numerical
implementation. This approximation is generally of the second order with respect

to In the particular case of two symmetric species with = the order of
approximation is 3.

Consider first an external electric field of intensity applied to a porous material.
In its dimensionless form, the flux of ionic species is given by
The electric current density flowing through the porous medium, resulting from
the application of is
The electro-osmotic velocity can be calculated by integrating the velocity

throughout the unit cell volume:
Because of the linearity of the previous system of equations and boundary condi-
tions, all solutions should be proportional to E. The conductivity tensor of the
is defined by
and the electro-osmosis tensor by
the following, the conductivity is sometimes normalized by the conductivity

of the undisturbed fluid, instead of as is customary for porous media. These
two quantities are related by
Let us consider now a n external pressure gradient V P applied to the same

porous material. Again, the solvent and charge fluxes are proportional to the
driving force VP. Thus, the permeability tensor K and the electro-osmosis tensor
a are defined by

Since both problems considered in this section are linear, a superposition of the
two generalized forces, simultaneous application of the pressure gradient and
the electric field, leads to relationship (1) with the electro-osmotic tensors a,
and K). This is true only under the assumption that the ion distribution is slightly
distorted, by appliction of either V P or E. The condition imposed on the latter
quantity is obviously
E'
The comparable condition to be imposed on V P can also be obtained.
Application of V P results in a force that must be much smaller than electric forces
near the walls Therefore, the restriction on V P is
A last physical situation may be of general interest. Previously, the jump con-
dition for the ionic potentials was calculated from (24) assuming that the
ionic concentrations have no macroscopic gradient =
= =
If this condition is not fulfilled, for example with
+ =
then the ionic concentrations are subject to a jump condition.
This expression is equivalent to a macroscopic gradient of unperturbed concentra-

tion and, under the restriction that it may be reduced to
In the presence of an externally imposed concentration gradient, the modified

pressure field ,
in the Stokes equation, Eq. has a jump con-
dition that depends on
In a semi-infinite space bounded by a plane charged surface (cf. Fig. 2a)

equation for the undisturbed potential can be solved analytically It may be
useful to reproduce this derivation here. For a symmetric electrolyte, = = 1,
= the equation for the undisturbed potential reads as when is the
direction normal to the plane,
Integrating twice while accounting for the boundary conditions a t the walls
= and far from it (uniform we get finally

(a) Semi-infinite void space, (b) plane, and circular Poiseuille flows.
=
1 + , tanh
1-
Note that for small potentials can be linearized into
which yields the solution
This expression could also be obtained by a first-order expansion of the exact

solution The surface charge can be obtained from and or as
Consider now the application of an external electric field parallel to the

tion along the wall without any global pressure gradient. Equations (46)-(49) yield
The tangential velocity increases from 0 at the wall to the value far from it,
which is given by formula [4]
or dimensionally

226
If the height of a plane channel (see Fig. 2b) is assumed to be very large
compared to the thickness then ( 5 8 ) is a solution for the
undisturbed potential distribution Under this condition, the conductivity
and electrokinetic tensors can be calculated analytically.
On the other hand. if the channel is not very thick compared to the layer,
Poisson's equation must be linearized to be solved analytically when the surface
potential is assumed to be low <<
Two species are present in the fluid with = = 1 a t the same initial con-
centration = Therefore, (45) becomes
The solution of this system is
The surface charge on the walls deduced from and (62) is
The average velocity induced by an axial external electric field can be calculated
analytically. as well as the electric flux induced by a pressure gradient,
In the above, the subscript denots the components of the vectors along the
direction of the channel. we deduce the coefficients a' and
which are equal, in agreement with theorem:
These coupling coefficients are very small = 3' in a narrow channel and
tend towards in a wide one 1).
The electric flux under a given electric field gives the conductivity coefficient,

In (67). the first term corresponds to the fluid conductivity with uncharged
walls. The second term is due to the difference between the concentrations of the
ionic species, which have different mobilities. The third one results the con-
vection of the local charge disequilibrium; since this disequilibrium and the solvent
velocity field are proportional to this is a second-order contribution. However,
since expression (36) for is only first-order accurate, the last term in (67) accounts
only partly for the contribution.
Assume that the porous medium is composed of a series of parallel tubes (bundle of
tubes model) each of diameter (Fig. 2c) The unit cell is now formed by the
region 0 a, 0 where L is a n arbitrary length. The external field and
the pressure gradient V P are applied parallel to the tube axis. T h e capillary axis is
and the distance to this axis is
Owing to the cylindrical symmetry, (44) for reads, for two
species with 2, 1, and iff
The solution of the linearized equation involves the modified function of the
first kind
It can be given in a power series representation,
Note that
Therefore, for wide capillaries the potential along the tube axis is
The surface charge on the walls deduced and (69) is

The cylindrical also imposes that reduces to its axial component
Equation (46) and the jump condition (49) for yield
Thus, from and 9 components of (47) we have
Finally. is obtained the z' component of (47):
Integration of (75) yields
and by averaging over the capillary cross section,
Finally, the four global transport coefficients are
Again. is verified.
The expression for is truncated to order since our expression for is
only first-order accurate as in the previous case of a plane channel. F o r wide
capillaries >> 1).

The simplified equations that govern electro-osmotic phenomena are
(49). Apart from some simple situations such as flow. isolated
spheres, these equations cannot be solved analytically. Hence, it was found useful
to start a systematic numerical analysis, which is summarized below.
In this first numerical approach to electro-osmotic phenomena, (45) and
were linearized by making the classical assumption that the potential is
it is equivalent to assuming that either the surface zeta potential or the
surface charge is small. Hence, (45) and are replaced by
Note that these approximations are generally second-order. They are third-order
if there are two symmetric species with = This linearization is generally
called the approximation
The geometry of the porous medium is discretized into a n array of
elementary cubes of size a filled with fluid o r solid. These cubes are by
mesh points. The equations are solved in a finite difference formulation with a
second-order discretization, which yields linear problems of the generic form
The vector is the unknown field or the matrix results from the operator
in the transport equation and from the local boundary conditions and the
right-hand term B results from the driving force, from the global boundary
conditions (49) and the current value of u' when solving for or the current
value of when solving for
The fields are are determined by a program that solves in any geometry
any linear differential equation of the general form
with the boundary conditions
by the use of a conjugate gradient method The convergence criterion is based

on the global residue
with generally taken smaller than

Equations (47) are solved by means of a routine derived from the one used for
solving the Stokes equation The so-called artificial compressibility is applied

with a staggered marker-and-cell (MAC) mesh. Thus, a given pressure field is used
to calculate velocity field. The resulting velocity field does not fit the incom-
pressibility equation and the pressure field must be modified to minimize the
local velocity divergence. These steps are repeated until Eq. is satisfied. This
corresponds to the equality of the total fluxes through all the cross sections of the
sample, within the same accuracy as in Eq. (84).
The general iterative is illustrated by the flowchart in Fig. 3. Note the
iteration loop on the determination of the coupled fields and u. The equilibrium
potential is determined first. Then Eqs. (46) and (47) are solved alternately until
convergence of the global electric and fluxes I and An estimate of the
velocity is given (zero velocity when the routine enters the loop); a n estimate of
is determined by solving (46); the velocity field is then updated by solving
(47). The stabilization criterion for the fluxes is generally set to
These alternate usually converge in a few iterations. In the pre-
sence of a macroscopic electric field E without a macroscopic pressure gradient, one
solves first for assuming u ' 0. Then u ' is determined from and reinjected
into (46). However, the corrections to resulting from the presence of are very
small, as it is a secondary effect. Similarly, if a macroscopic pressure gradient is
imposed without an electric field, one first solves (47) for the flow field assuming
Calculation of characteristic values and

dimensionless equations
not
converged
calculation of U
Printing
End
Flowchart of the computing program.

constant Then is injected into (46) to compute The potential in (46)
yields only a very small correction to
A large number of tests have been conducted to check these complex routines.
They are detailed in Ref. (9) and can be as follows.
Several non physical parameters have been introduced when implementing the
numerical solution. The first one is the arbitrary choice of the unit cell to generate
the periodic medium. It has been checked that is has absolutely no influence on the
results, by performing similar runs with shifted unit cell boundaries. The same
holds for rotations of the geometry by permutation of the coordinate axes.
There are also two discretization parameters. The elementary cube size a for
discretizing the geometry is to be compared to a characteristic length scale L of the
porous medium. Large or small ratios correspond to coarse or fine descriptions
of the geometry, respectively. The grid size is the spatial resolution for the
discretization of the transport equations, and it must be compared to the
thickness. Roughly speaking, the ratio is the number of grid points in the
electrical double layer.
All the following computations are conducted for an aqueous solution of
at where species 1 is the with 9.31 and species 2 is
with = 1.79 this
,
corresponds to a ratio = 5.2. Plane
and circular flows have been obtained with good precision, as well as the
conductivity and the coefficient of suspensions of spheres, which were compared
to the results found in the literature.
The results of all these tests can be summarized in a few statements. For a
given geometry, with a fixed ratio or the accuracy of the predictions
critically depends upon the ratio K G (or of the grid spacing to the double
layer thickness. For 1, the errors o n all the perturbation quantities
. . d o no exceed a few percent of When discretization is too coarse
(KO the predictions become grossly inaccurate, as could be expected since the
double layers are overlooked by the discretization grid.
Moreover, the ratio or of the characteristic length scale of the
to the discretization step determines the quality of the description of the geometry.
Therefore, it affects equally the transport coefficients for uncharged solid walls and
their disturbance by the electrokinetic effects. For example. the precision on is
bout 2-3 times irrespective of the surface potential.
Porous media can be classified as unconsolidated and consolidated materials. Soils,

clays, and packings of various sorts belong to the first class, while rocks such as
sandstones and limestones belong to the second. Both classes are investigated in
this section. The first two subsections are devoted to ordered or random systems of
particles, and the third to a different model of consolidated rocks, based on the
classical reconstruction technique Finally, a formula valid in the thick
double layer limit is used to gather all the numerical results and also some experi-
mental results found in the literature.

Computations were conducted for simple cubic arrays of spheres and ellipsoids for
various solid concentrations in an aqueous solution of Of course, the
conductivity and permeability of cubic sphere packings are isotropic, as may be
reasoned from linear superposition. The external electric field is applied along one
of the main directions of the cubic lattice. The double layer thickness was kept
equal to = 1.65. The ratio of the sphere radius to the particle separation
varied = 0.2 to = 0.5, which correspond to a dilute suspension
= 0.034) and close packing = respectively. The size a of the elemen-
tary cube was set to obtain a good discretization of the particle geometry and of the
double layer, with 4 and K O
The dimensionless conductivity and coupling coefficients depend linearly
upon The conductivity and electro-osmotic coefficients 6- = and
are plotted in Figs 4a and 4b. respectively, versus the solid volume concen-
trations for the three reduced surface potentials = For = 0 , one
can see that 6- tends to as 0. Moreover, for uncharged particles, the
values of are very close to in the whole range of The corresponding
analytical results of and Acrivos are also plotted in Fig. 4a. For
uncharged particles, the reduced conductivity significantly depends upon the pack-
ing density, which is explained by the effect of interactions between the double
layers of adjacent particles.
Basically, the effect of increasing is to reduce the relative volume of the
conducting matter and to diminish the absolute value of as may be observed
for uncharged particles. This trend is also observed for positively charged particles.
For negative values, one can note the existence of a critical concentration at which
0 (a= where the above trend is counterbalanced by the excess of
mobile positive ions within the double layer. Due to this effect, the electrical con-
ductivity of suspensions of negatively charged particles is always greater than for
positively charged particles.
For any given the electro-osmotic coefficient is a decreasing function of the
solid volume fraction, as be seen in Fig, 4b. In the dilute limit, the fluid in the
bulk of the pore space is driven by the slip velocity at short distances of the
particles. For lower porosities, the double layers overlap. and the free streaming
zone shrinks and disappears.
The data for are compared in Fig. 4b with similar results of and
Neal computed with a cell model. The values of differ by about 0.05
throughout the whole range of solid fractions. This that for dilute suspen-
sions the agreement is satisfactory 10%) where the cell model is applicable.
et showed that for 0.01, these two results agree to within a
few percent. However. the relative error is large for close packing. This is because
the cell ignores the contacts between particles and assumes that the double
layers surrounding different grains d o not overlap. Accordingly. the departure
worsens when increases.
Computations were also performed for simple cubic arrays of oblate ellipsoids of
revolution with aspect ratio 5/1 and with solid concentrations varying from 0.02 up
to 0.524. The double layer thickness was kept equal to = 2.81. The geometry is

-0.8
0 0.1 0.2 0.3 0.4 0.5 0.6
(a) Conductivity coefficient and coupling coefficient as functions of

the solid volume fraction in a cubic array of spheres (solid lines) or ellipsoids (dashed
lines). Data are for - 1.65, = and The dotted line in (a) was
obtained for uncharged particles and the dash-dotted line for uncharged ellipsoids
the dotted line in (b) results a cell model
discretized with = = 22.5 and = 4.5 where are the

axes of the ellipsoids. The double layer thickness ranged = 1.3 u p to 13,
where R is the equivalent radius,
The electrical conductivity and the electro-osmotic coefficient evalu-

ated along the direction were found to be within about 15% of the comparable
data for spheres calculated for the porosity. This does not quite agree with the
approximate prediction of Tinga et for the conductivity at = 0 from a cell
model, which differs more significantly from the conductivity for spheres (see Fig.
4). In the dilute limit, Tinga et al. obtained -1.14 for uncharged aligned
5 x 5 x 1 ellipsoids versus 1.5 for spheres.

The results for both geometries a t maximal solid fraction are recast in terms of
the slopes of the reduced conductivity and of the coupling
coefficient as functions of in 5 , where data for the vertical direction in
packings of ellipsoids are also included. Identical trends are observed for the con-
ductivities for both types of grains. The slopes for ellipsoids along the
and directions are smaller and larger, respectively, than for spheres. F o r the
thinner double layers our data for spheres converge with the results
obtained by and or and using
asymptotic expansions in the of vanishing length. The
tro-osmotic coefficients for ellipsoids are different along the two directions,
since they are related to the correspnding permeabilities, as shown by Fig. 5b. The
comparison with the results of asymptotic expansions for spheres with
(a) Derivative of the reduced conductivity and (b) coupling

coefficient in a cubic array of contacting spheres (solid lines) or oblate ellipsoids
(dashed lines) along the horizontal and vertical directions as functions of the
inverse dimensionless double layer thickness The dash-dotted line in (a) is the result
for spheres of a matched asymptotic expansions technique

thin double layers is bad in this range of Ratios of about 3 and 2 are
found for = 5 and 10, respectively. Thus, matched asymptotic expansion
results converge more rapidly toward the limit -0.345 than predicted by
the numerical resolution of the full system of equations.
Finally, et [9] provided detailed results for orthorhombic arrays of
ellipsoids of revolution with an aspect ratio of in an attempt to describe the
porous geometric structure of clays that are composed of platelike particles.
The packings were built by simulating the sequential deposition of particles

in a parallelelepiped cell with periodicity conditions along the two horizontal direc-
tions. After a particle is introduced at a random location from above the cell, it
settles vertically along the axis until it touches a grain already in place. Then it
rolls and glides until i t reaches an equilibrium position, where it sits on at least three
supporting points. While reaching its equilibrium position, it undergoes rotations
and translations (without bouncing) along the bed surface until it reaches the lowest
position. The numerical procedure is described in detail in Ref. 38. An example of
such a packing of ellipsoids with an aspect ratio 5 x 5 x 1 is displayed in Fig. 6.
The transport properties of these random beds were investigated along the two
horizontal and directions. The packings obtained in the above procedure are
statistically isotropic in a horizontal plane. Accordingly, the results obtained for the
two directions differed by at most 1 % . In contrast with this. differences (up to
between the values of the two coupling parameters a ' and were observed
stemming from the insufficiently fine discretization. In such cases, the arithmetic
averages between these coefficients are presented.
Flat ellipsoids tend to settle with their largest section parallel to the plane (see
Fig. 6), as in the orthorhombic arrangements of the previous paragraph. The
average solid fraction is equal to 0.595 for spheres and 0.596 for ellipsoids. Since
the unit cell used for the computations contained numerous particles, in order to be
statistically representative. discretizations coarser than those for the previous reg-
ular packings were used. The unit cell contained about 30 spheres with = 4.27,
or 90 ellipsoids with = = 6.4 and = 1.28. Note that discretizations
were performed a posteriori, when the packing had been built in a continuous
description.
The conductivity and permeability for uncharged particles are
Spheres:
Ellipsoids:
= 0.208. = 2.5 x
The above permeabilities roughly constitute the averages of the values obtained
along the various directions considered in the orthorhombic packings, while the
conductivities are closer to the smallest conduction.

Example of a bed of 5 x 5 x 1 ellipsoids obtained by sequential deposition.
The dimensionless conductivity and coupling coefficients depend linearly

on similarly to the comparable dependence of the ordered arrays (cf. Ref. 9). For
the initial slopes for the conductivity,
-0.88 (spheres) or -0.12 (ellipsoids)
are somewhat smaller than for the due to a larger overlap of

the double layers. The ratios are within the range
-0.790 (spheres) or -0.860 (ellipsoids)

' K '
The influence of the double layer thickness in these random packings was also
investigated by and follows the same trends as those calculated for
orthorhombic

Finally, electro-osmotic transport coefficients for reconstructed porous media,
based on Fontainebleau sandstones, were calculated. Since these materials contain
only quartz mineral. no interaction occurs during percolation of
water. Therefore, the permeabilities measured in water and in air are almost iden-
tical. The Fontainebleau sandstones have only open (interstitial) porosity (no pore
space inside the grains). These media are generated randomly, according to statis-
tical geometrical characteristics measured on real materials. The reconstruction
procedure has been described elsewhere and only the main lines are given
here. First, measurements were performed on a thin section of a given porous
A phase function is introduced:
if x belongs to the pore space
1
otherwise
where x denotes the spatial position with respect to an arbitrary origin.
The porosity and the correlation function can be defined by the statistical
averages (denoted below by an
or equivalently, because =
On a given sample, these measurements were performed using analysis

in a single arbitrarily chosen plane, since the considered materials were
isotropic. Hence, can be replaced by its absolute value
The process of reconstruction of a three-dimensional homogeneous and isotro-
pic random medium of a given porosity and a given correlation function is
equivalent to generating a random discrete function that satisfies the proper-
ties specified in Eqs. (90). It can be shown that such can be devised from a
Gaussian field when the latter is passed successively through a linear and a
nonlinear filter.
For the present investigation, porous media have been reconstructed using the
data measured on two samples of sandstones with porosities
= 0.21 and E = 0.31. respectively. Geometries of the reconstructed samples were
discretized using solid or fluid elementary cubes of size a using = 12 and
= 1.58 for = 0.21 or = 1.89 for 0.31. The length scale L is the
correlation length deduced from the correlation function by
An example of a reconstructed sample with = 0.31 is displayed in Fig. 7 .

The correlation length scale L as measured by image analysis was about 20
which exceeds the layer thickness by at least two orders of magnitude.
Nevertheless, we performed calculations with the dimensionless parameter

FIG. 7 Example of a reconstructed sample with 0.31.
= 1. and This corresponds to the situation where the geometry of

the reconstructed samples was scaled down about 100 times.
F o r the the elementary were subdivided into mesh cubes
with o r 4. We started with the coarsest grid, 1 , and passed to finer
subdivisions until the results were reproduced within about 5 % . F o r it
to taken 11 2. For the smallest double layers = o r for
sample 0.21, whose primary transport coefficients are the smallest, 11 = 4
necessary. Even under these conditions, the results could be reproduced only to
within about 10% in a few cases, to the least conducting directions.
As a general data are believed t o be accurate t o within a few percent
multiplied by
The properties of two reconstructed samples for each sandstone were
along three directions and One of the samples for 0.31 did not
percolate along the axis. The results are presented in Table 1, including the
derivatives of a between 0 and the conductivity was
indeed checked to be a linear function o f in the range = 0 to = - 1. The
primary coefficients and A" are very scattered, and each individual sample is
strongly anisotropic. However. when normalized by the medium effective conduc-
tivity along the same direction a significantly lower scatter is observed. In
particular, for thick double layers, a remarkably constant value is obtained:

Length Scale A for the Various Models
Model'' Direction
spheres
ellipsoids
0-1. spheres
0-1, ellipsoids
ellipsoids
spheres
RP, ellipsoids
RM.
2A3
Plane channel
Circular channel
=simple cubic; 0-1, = types I a n d random
packings; reconstructed media.
The data for = 1 are slightly more scattered:
F o r thin double layers = the data are still more scattered. Three values
are far below the others and correspond to the least accurate computations. When
these are excluded, one obtains
It is interesting to note that the variations in the relative conductivity for the
reconstructed media are comparable to those in the granular media. F o r all the
arrangements of spherical o r ellipsoidal grains considered above. including random
packings, we obtained for = 2.46
These numerical results agree with the theoretical value of obtained

for a constant potential applied over all the pore space.

The data for the coupling coefficient are also easier to analyze in comparison with
the corresponding permeabilities. Again, the ratio tends to -1 for thick
double layers. These results are discussed in the next section.
Since most sandstones have a relatively coarse microstructure with typical pore
sizes of about 10 the experimental data in the literature concerning electroki-
netic effects in sandstones are beyond the range of double layer thicknesses inves-
tigated in these numerical computations. and coworkers
measured streaming potentials in various samples of sandstones
saturated with brines. From their values for the brine resistivity, is likely
to range from to This corresponds to Accordingly, the
streaming potential obeys and the ratio is very small
They also report the value of the surface potential -0.88. Similar results
were obtained by Li et who analyzed sandstones, limestones, and beds of
fused glass beads, since the brine was more concentrated (0.1 which
yields K 1
A new important result of these computations is the relation between the coupling
coefficients and the permeability for thick double layers [cf. Eq. The case of
relatively thick double layers (fine porous media) was largely overlooked in the
literature. Attention was generally focused on the opposite limit of vanishing
length, where the matching technique of inner and outer developments was exten-
sively used (see I), which yielded the general relation (3) of Overbeek
The analytical treatment for thick double layers is a formidable task because of
the layers' overlap, which requires full resolution of the governing equations.
However, for plane and circular channels, one can use formulas (66) and (78) to
obtain
for a plane channel
for a circular channel
The thick double layer limits can be easily deduced:
This result can be generalized to any porous It is more convenient to

consider a macroscopic pressure gradient V P applied to the porous thus
aiming to obtain the comparable limit of the coupling coefficient a instead of the
equivalent coefficient Suppose that V P induces a flow with velocity The con-
vected electric current is

Using the same hypotheses and approximations as before, namely that the equili-
brium ion concentrations are slightly distorted by the flow, that the electrolyte is
symmetric, and that the surface potential is small enough to linearize the
Boltzmann equation, ( 9 6 ) can be recast into
Moreover, for thick double layers. with larger than the characteristic pore size,
is nearly uniform and approximately equal to the surface potential
Therefore, replacing by and using the definition of permeability, one gets
Snce V P is arbitrary. one obtains
Note. however, that the convergence in (98) is not necessarily uniform along
different directions, as illustrated below.
The limit ( 9 5 ) has already been pointed out in several cases in the previous
section. Since in most practical situations the length is unlikely to exceed
widely the typical pore size, it is essential to study the rate of convergence of the
coupling coefficients toward the limit (95) to determine how far off this limit is from
the actual value in a n intermediate range, and possibly to introduce a correction
factor. T o this end, all the results of the previous section for all the investigated
directions in all the geometric models were gathered.
In order to plot ratios against a normalized double layer thickness, a
length scale applicable to all the directions and geometries has to be defined. As
already pointed out above and also shown by the criterion for conver-
gence toward ( 9 8 ) corresponds to the covering of the pore space by the electrical
double layer rather than its thickness relative to a typical grain size. Thus, an
adequate length scale should be a measure of the characteristic pore size. Such a
length scale A was introduced in Ref 42. It is essentially a pore volume to pore
surface ratio, with a measure weighted by the local value of the electric field in
the conduction process:
This length scale A is derived from the correction to due to the interfacial
conductivity in thin layers.
An electrokinetic method was proposed to measure A , as a n alternative way
to determine the permeability via

This relation was shown to provide excellent predictions for various models
or real porous Note that for a plane channel, (99) and (100) yield
slightly different results, namely, A = and A = respectively. For a
circular channel, both expressions give A R. For all the media considered here, A
was evaluated by solving the equation in the pore space of the unit cell with
insulating boundary conditions. For the ordered and random A ranges
from A = up to A = (Table 2). Our data are compared to (100) in
Fig. 8. Note that points that are significantly above the dotted line correspond
to the orthorhombic and simple cubic packings of ellipsoids.
All the electro-osmotic data from our simulations are thus gathered in Fig. 9 and
compared with the theoretical results, for channels. For all anisotropic
A was evaluated along each direction. All the data cluster around a
single curve, with very little dispersion considering variety of geometric config-
urations. As expected tends to - 1 for vanishing For intermediate
thicknesses, is well approximated by valid for a
circular channel.
It is tempting to (99) to obtain an even better agreement with the
numerical data. A least squares fit yields a modified expression for A ,
The prediction with Eq. with R replaced by A in (101) is also plotted in

Fig. 9. For an easier evaluation, expression for a plane channel could be
substituted into
Alternatively, can be deduced from the primary coefficients and mea-
sured along the direction:
The constant in (103) results from a least squares fit of our data (Fig. 10) Note that
(102) can be recast into
which underestimates formula (3) for large by a factor of
this model was compared with experimental data for a porous

material when the double layer thickness is comparable to the typical pore size
Coupled charge and mass transport were measured in a plug of

FIG. 8 Conductivity and permeability for regular close simple cubic or
packings, random packings and reconstructed sandstones
The slope of the least squares fit (dashed line) is 0.41.
.9 Reduced coupling coefficients and as functions of for all

the configurations studied in VI. Data are for regular packings orthorhombic
packings random packings and reconstructed sandstones Theoretical
results with A = (solid line), with A = R (dashed
line). and (100) (dotted line). (+) Experimental results for from
Refs. 17 and 44-46.

.10 The reduced coupling coefficients and as functions of ' for
all the configurations studied in VI. Data are for regular packings,
packings, (*) random packings, and (+) reconstructed sandstones. The
dotted line is the least squares fit. (101) and (102).
montmorillonite saturated with brine. The measurements were interpreted by

regarding the pore space between the clay platelets as plane channels. The effective
half thickness b of the water layer is also deduced from the measurements. These
data were interpreted by et [9] and are displayed in Fig. 9, where they are
seen to compare well to the numerical predictions.
The transport properties of aggregates are of interest in a number of processes such

as sedimentation, filtration, and electrophoresis whenever they are charged and
submitted to a n external electric field. Industrial applications include the determi-
nation of the settling velocity of sludge particle aggregates
S o far most efforts have been devoted to the determination of the drag force
exerted by the surrounding fluid on a translating aggregate. The literature that has
been devoted to this topic is already quite impressive and has been recently sum-
marized by Coelho et al. T h e purpose of this section is to extend the analysis to
electrophoretic mobility, to isolated particles to which an electric field is
applied. As previously stated, the same formalism as for porous media can be
applied to suspensions.
A cluster can be characterized by its gyration radius, which may be defined as

where is the total number of particles of the cluster, and denote the numbers
of any two particles, and is position vector of a particle. Each elementary
particle is cubic of size a.
fractal dimension may be introduced by the relation
where is the of the cluster.

Many procedures have been proposed in the literature for generating
clusters with various fractal dimensions. Three of them have been selected for brief
discussion here.
A modified Witten-Sander scheme was used to accelerate the convergence t o the
asymptotic regime The random particle does not stick to the cluster when it is
o n a site adjacent to occupied sites. but it is until it is o n a n occupied site;
then its final position is assumed to be the previous one. Such a scheme yields a
fractal dimension of 2.6, which is slightly larger than the 2.5 obtained with the
original Witten- Sander scheme.
The hierarchical model with linear trajectories was applied without any modifi-
cation two clusters are to stick together when they occupy adjacent
sites. Such a model is much less time-consuming than aggregation
with Brownian trajectories. The resulting fractal dimension is difficult to measure
with precision because of the large statistical fluctuations due to the itself: it
was found to be close to 1.9.
The last model is a hierarchical model where the fractal dimension is tuned by
selecting the orientation and sticking point of clusters of size where p
is the iteration F o u r values of the fractal dimension were studied: 1.6, 1.9.
2.2. and 2.5. Two of these values were chosen close to the values of the
Sander aggregate and to tlie standard hierarchical model. for reasons that will be
clear in the next section.
The numerical results relative to these aggregates are summarized in Table 3.
T o compute the transport properties of these aggregates, a spatially periodic
pattern of such particles was created. F o r the sake of convenience, one can say that
the center of gravity of each aggregate is located a t the center of each unit cell of
size however, because of the periodicity condition, the precise position of the
cluster inside the unit cell does not matter. was chosen equal to 32, 48, 64, and
96; usually, is large with respect to the gyration radius, and the resulting solid
concentration is smaller than 0.02. The structure of the suspension is represented
in Fig. 1 l a .
T o determine the hydrodynamic radius of the clusters. one has to solve the
Stokes equations that govern the low Reynolds number flow of Newtonian fluids.
By taking two opposite limits (namely, the solid concentration goes to zero and the
size of the aggregates goes to infinity), it could be shown that the hydrodynamic
drag F o n an aggregate scale is

Phenomena 247
Fractal Dimension and of the Various Types of Clusters

and
D 2.64 1.90 1.6 1.9 2.22.5
64 4.13 0.22 3.80 0.07 3.83 4.03 4.40
A' 256 4.34 0.14 3.71 0.12 3.82 4.19 3.96 4.26
= 1024 4.48 0.13 3.66 0.26 3.82 4.18 3.94
= Witten- Sander model.
= hierarchical model with linear trajectories.
where is the translational velocity of the aggregate and its gyration radius.
This formula is in agreement with the classical Stokes formula. since for a solid
sphere = 3 and =
T o have a complete overview of conductivity, it was found useful to determine the

macroscopic conductivity of a suspension of uncharged aggregates, which
requires solving a simple equation in the spatially periodic geometry
described in
An analogous situation was considered in the literature. The equation
was solved between an aggregate a t a given potential and infinity with potential
zero. This is actually equivalent to determining the growth rate and the distribution
of absorbing sites in the cluster during the aggregation process This was done
with walk methods by et al. and et and with
renormalization methods by Cates and The incident flux has a scaling
structure charcterized by an infinite family of independent scaling exponents. The
equation was also solved in two dimensions between two clusters a t con-
stant potentials by and Mandelbrot
Here, the boundary condition at the surface of the aggregate is a
condition since it is considered perfectly insulating.
The results are represented by using the coefficient
According to Maxwell this coefficient is proportional to the solid concentra-

tion; for solid spheres, it is given by
3
2
The proportionality is well verified in Fig. 12. Moreover, all the results relative to
fractal aggregates are gathered around a single curve, whatever their fractal dimen-
sions and their gyration radius. The equation of this curve is

. 11 Spatially periodic suspensions of fractal aggregates. The aggregate in (a)
contains 1024 cubic particles of size a'; it was built with the hierarchical model with
linear trajectories. The deterministic flake in (b) is at the third-generation
stage with b = 5.
These unexpected results are in a sense confirmed by the fact that is

indeed well verified by the numerical data for solid spheres. Similar calculations
were conducted with spheres with a uniform porosity. Their solid fraction was
set to equal to 0.25, 0.50, and 0.75, yielding = 2.6, 2.5, and 2.05, respec-
tively. In this situation, can be estimated by replacing the porous spheres with
solid homogeneous ones with conductivity Their volume fraction in

Phenomena 24
FIG. 12 The coefficient for uncharged aggregates as a function of the solid fraction
Each point represents an average over 10 realizations. Data are for
Sander (modified: see text), D 2.5; hierarchical with linear trajectories. D = 1.9;
Thouy and D 1.6; (+) Thouy and Jullien, D = 1.9; ( x ) Thouy and Jullien,
D 2.2: Thouy and Jullien. D 2.5; spheres, D = 3.0; (0)porous spheres,
D = 3.0; and Acrivos D = 3.0.
the suspension would be c = . The conductivity is then given by the general

form of Maxwell's formula for the conductivity of a random dilute suspension
of spheres with conductivity
a for random porous solids was determined independently by It

is equal to 0.55. 0.14, and 0 for = 0.25, 0.50, and 0.75. respectively, which yields
= 2.12, 2.41, and 2.0. The predictions are in good agreement with the numer-
ical data, except for the looser porous spheres = 0.25); since their internal
correlation length is comparable to their radii, such porous spheres cannot be
represented by an equivalent effective material; note that the correlation length is
defined as usual as average size of the connected solid clusters.
It be worth noting, although there is n o physical substantiation, that in
view of and (1 all the aggregates considered here behave like porous

spheres with identical solid content and = They may also be viewed as solid
spheres of identical volume and a pseudo relative conductivity =
Complete systematic cannot be achieved here because they are extre-

mely long. Hence, a complete picture cannot be presented as for the hydrodynamic
radius.
Let us restrict ourselves in the following to a two-component electrolyte with
valences = 1. The conductivity of the solution at equilibrium is [see
where is the electrolyte concentration and the diffusion coefficient of the

positive and ions. When the local potential of the solution is equal to
the ionic concentrations are derived from the Boltzmann equation as
The fluid is denied as
This relation suggests an adequate representation of the conductivity of a sus-

pension of charged aggregates. The dimensionless coefficient 6 is defined as
In the two extreme cases of vanishing or infinite double layer thickness, the poten-
tial uniform throughout the fluid, equal to 0 or respectively. The fluid
conductivity , (1 is then also uniform. Accordingly, the coefficient in
(114) tends to 0 and 1, respectively, these two limits.
The coefficient 6 defined by (1 14) is expected to depend on the four dimension-
less quantities K O , and We are interested in the limiting behavior when
and tend to 0.
Let us restrict the presentation of our numerical results to the only case that
could be completely worked out, 1. The first simplifying feature is that the
coefficient is found to be proportional to the solid concentration, as can be seen
in Fig. 13.

FIG. 13 The coefficient for charged aggregates as a function of the solid fraction
Same conventions as in Fig. 12. Data are for = 1, = 256.
Then, for a fixed value = 2.315 x is represented as a function of

the gyration radius. F o r all the aggregates, the data are very well fitted by a relation
of the form
+
a correlation coefficient 0.99.
As tends to infinity, becomes negligible in (115) and 6/41 tends to a
constant value for each type of aggregate. This could be expected from a dimen-
sional analysis argument. Since is the only dimensionless group in the pro-
blem involving a length scale, with the exception of which is kept constant here,
its influence can vanish only when it tends to infinity.
Finally, a linear fit of the coefficient a ' in (115) for the various types of aggre-
gates yields
7.86 - Re,
with a correlation coefficient = 0.991. The decrease of with D could also be
expected on a physical basis. F r o m its definition in Eq. (1 is a measure of the
fraction of the suspending fluid whose conductivity is disturbed by the surface
potential of the particles, which is approximately a layer of thickness on the
order of around the solid. F o r the same solid volume, aggregates with a
lower fractal dimension develop a larger surface, and thus the volume of disturbed
fluid is larger.

T o summarize, for the aggregates considered in this paper and for = 1,
(1 and 16) combine into
Of course, the numerical coefficients in (116) are relative to = and another

double layer thickness would yield different
Relation (117) has the same form as the prediction of for solid
spheres,
where 2.1 and L ' increases when decreases. T h e asymptotic analysis

leading to (116) was not repeated for other values of K G . However, for a moderate
aggregate size = 128, for and 1/8 was found to be larger than
for K G = 1 by a ratio of about 2.4. 6, and 17, respectively. The corresponding
successive values of = 2.9, 9.6, and 34 are of comparable magnitudes.
. Electrophoretic Mobility
Let us apply to a suspension a n electric field it will induce electrophoretic motion
of the particles with velocity This is used to define the reduced electrophoretic
mobility as
When the particle concentration is sufficiently small, v,, is equal and opposite to
the fluid seepage velocity.
Since computations are very long, the investigations on were also restricted to
Furthermore, the extrapolation to infinite cluster sizes was also impaired by
this limitation.
The reduced electrokinetic mobility defined by (1 19) is again expected to
depend o n the set of dimensionless parameters KG , and Its dependence
o n can be analyzed with the help of the calculations of and Neale
By use of a cell model, these authors generalized Henry's [7] prediction of the
mobility of a single spherical particle with radius a to concentrated systems.
Their result is cast in the form
f
Although Eq. (120) is not expected to apply directly to suspensions of fractal
aggregates, the function is a natural candidate to describe the dependence of
o n the concentration with a replaced by some characteristic radius
Thus the following form was tried:
where would be the aggregate mobility in the dilute limit.

Electrokinetic Phenomena 253
The analytical expression for f is too cumbersome to allow a determination of

by a least squares fit. Various simplified expressions have been proposed
but none is applicable in the present range, since is always close to 1.
Therefore, was determined for each type of aggregate and for 256 from
the ratio of for = and = 0.0185:
It turns out that is close to a in all cases. even though ranges from
about 5 to 14. F o r a constant size 256, is even a slightly decreasing function
of This indicates that the particle interaction effects on mobility result mostly
from interactions between the of the aggregates and that the cluster pair
interactions are negligible. The value of R depends on the relative arrangement of
the particle within the aggregates. F o r 256, can be represented by the
linear fit
with a correlation coefficient 0.989. no asymptotic analysis in

the limit of infinite could be performed; however. since the relevant interactions
are short-range within the aggregates, (123) is probably a good approximation
for any large cluster.
It appears also that the variations in this range of K O for small are very
close to linear. The adjusted form, (121). is plotted in Fig. 14 together with the
numerical data.
Moreover. it was found that the mobility in the dilute limit has very small
variations with compared to the error bars of the statistical averages (see Fig.
11). Hence, the electropheric mobility of fractal aggregates for = 1 is given
within by the model (121) with (123) and 0.55.
A few additional computations were run with A' 128, = 3.9 and
KO 1 . The data compare within with predictions of (123). and
0.55, which supports the assumption that A' 256 is close to the asymptotic
regime. In addition, was set to and with = 128 and
= 3.9 for the six types of aggregates. The data are compared to the
model. in Fig. 15. The deviations are significant and systematic.
This suggests that a proper analysis should allow the extension of (123) for #
with the same general form, (121).
The electrokinetic properties of porous media. suspensions, and isolated particles

can be determined within the same theoretical framework. which essentially
assumes that the basic governing equations can be linearized. The resulting system
can be solved numerically when the double layer thickness is not too small com-
pared to a characteristic dimension of the medium under consideration. Hence,
these results are applicable to finely divided media of dimensions.

IG.14 Mobility of fractal aggregates as a function of the solid fraction Same
conventions as in Fig. 12. Data are for 256, 1. The dotted lines are the fit,
(121).
Mobility of fractal aggregates divided by the prediction of (122) and

(123) as a function of the fractal dimension Data are for = 128 and
and 1 . 8 dotted a r e least fits.
This approach can be extended to many situations as polyelectrolytes and
porous particles.
The present set of investigations also needs to be completed with another numer-
ical code able to address opposite limit where the double layers are thin with
respect to a characteristic length of the porous medium.
Most computations were performed a t whose support is gratefully

acknowledged.
V. Prentice-Hall. Englewood Cliffs,

1962.
Fair and Osterle, J. Phys. (1971)
de Groot and P. Nan-Equilibrium Thermodynamics, North-Holland,
Amsterdam. 1969.
93: 8).
E. (1924).
P. Polar Molecules, Reinhold, New York, 1929,
Henry. Proc. Roy. A (1931).
Booth, Proc. Roy. A (1950).
and Colloid Interface
(1996).
and D. N. Ward, Colloid Inteface (1988).
J. in Colloid Science ed.).
Amsterdam, 1952.
O'Brien., Colloid Interface (1986).
Moscou (1809).
and P. Soil.
Am. (1976).
I. V. and Peck. Colloid Interface Sci.
(1993).
A. Van der Put and J. Colloid Interface (1983).
and O'Brien, Colloid Interface (1988).
A. Van and I. Colloid Interface (1980).
Effets et circulations de dans
les roches Thesis, Paris, 1994.

Pengra, Li, S. X. Li. and Wong, in in
Systems Drake, Klafter. R. and Mater.
Res. Proc. 366, 1995.
R. W. Colloid Interface (1981).
Russel, Fluid. (1978).
P. Adler. Porous Media: and Transport,
MA, 1992.
Brenner. Phil. Trans. Roy. (1980).
W. and White, Faraday Trans. 7411607 (1978).
Fluid (1980).
Rice and R . Whitehead, (1965).
Hunter, Foundations of Colloid Science, 1 , Clarendon Press, Oxford,
UK, 1987.
Thovert. Wary, and Adler, Phys. (1990).
and P. Adler. Porous Media (1990).
and I. A. Int. Flow
(1990).
A. and A. Int. Flow (1982).
A. Voss, and Appl. Phys.
Kang and A. Sangani. Colloid Interface (1994).
and Adler. Phys. Rev. (1997).
Jouniaux and Res.
P. Pozzi, and L. Paris (1994).
X. Li, B. Pengra, and P. Phys. Rev. E (1995).
L. Johnson, Koplik. and Phys. Rev. Lett. (1986).
and Garboczi. Phys. Rev. B (1992).
P. Groenevelt. and Soil. Am. J.
(1978).
Groenevelt. and Soil. Am.
(1980).
A. Williams. and P. Williams, Colloid Interface Sci. (1978).
and J . Ganczarczyk. Water Environ. Res. (1992).
Logan and Wilkinson, (1991).
J. and Ganczarczyk. Water Res. Can. (1994).
Jullien and Aggregation and Fractal Aggregates, Workd Scientific,
Singapore, 1987,
A. and L. Sander. Rev. Lett. (1981).
Sutherland. Nature (1970).
and (1994).
A. Random kinetic aggregation. in Physics of Finely Divided Matter
Boccara and Springer. New York, 1985.
P. Meakin. Stanley, A. and A. Rev. A
2369 (1985).
P. Meakin, and I. Phys. Rev. Lett. (1986).
and T. A. Phys. Rev. Lett. (1986).
and B. Mandelbrot, A (1991).
Maxwell, Electricity and Clarendon, Oxford, 1873.

60. Wary. and Adler. (1990).
61. Stauffer and A. Aharony, Introduction Percolation Theory, Taylor and
Francis. London, 1991.
62. Levine and Neale, Colloid Interface (1974).

Center for Surface Science, Department of Chemistry,
Jadavpur University, Calcutta, India
Geological Studies Unit, Indian Statistical Institute, Calcutta,
India
I. Introduction 259
I I . Transport Processes 26 1
A. Self-diffusion in niicroemulsions 26 1
B. Conductance of microemulsions 263
C. Transport across niicroemulsion acting as liquid membrane 272
References 275
are a unique class of systems in which either

water or oil disperses in a niicrofine state, being surface coated with organized
in an otherwise incompatible continuum of either oil or water. Here
a liquid that is immiscible with water is considered an oil and the are
surfactants present alone or mixed with cosurfactants (normally lower
amines, etc.) In nonaqueous microemulsions, polar liquids such as glycol, glycerol,
formamide, and dimethylformamide replace water. Microemulsions are of two
categories: water-in-oil and oil-in-water They are isotropic,
of low viscosity, and thermodynamically stable, having particle dimensions falling
in the range of 10-200nm. The microemulsions are grossly differentiated from
reverse micelles in terms of content of water in the pool. If the water content
exceeds the solvation requirement of the head groups, the "reverse
is replaced by A pictorial representation of different
types of is given in Fig. 1.
The formation of a microemulsion is fundamentally controlled by the
lowered interfacial tension (of the order of 0.001 o r less) by the

Core
Layers
rigidly held water molecules held
to the head groups
Cosurfactant
Layer
FIG. 1 Pictorial representations of micelles and
action of the interfacially adsorbed amphiphiles. The interfacial tension is not

lowered to such a n extent in they are thus
unstable" and exist in much larger dimensions (200-3000 Appreciable work
(positive free energy) is required for their formation. The work for microemulsion
formulation is insignificant, and they are said to be thermodynamically stable.
Much depends on the types of amphiphiles used and their combination. For easy
formation, the chemistry of formulation should be right.
The internal structure of microemulsions can be of three different types:
1. Isolated dispersed droplets of water in oil or oil in water.
2. Clusters of droplets of the above types in a continuum.
3. Both and dispersions simultaneously present to form a
state in solution
These structures can be located in the phase diagrams essentially constructed to
study the composition-dependent phase behaviors of ternary or quaternary systems.
For a general understanding of books and review articles on
the subject may be consulted
The overall physical state and function of a liquid can be understood by exam-
ining transport processes in it. Microemulsions are compartmentalized liquids;
transport in them can help to reveal their internal consistency, interparticle inter-
action, overall particle geometry and stability, etc. In line with the scope of this
review, concise descriptions and discussions of investigations related to

in Microemulsions 261
diffusion of components in (2) transport of ions (conductance)

in them. and (3) mass transport across employed as liquid
are herein presented.
The components of a constantly The process is inti-

mately related to the internal structural state of the system. The surfactant and
molecules diffuse back and forth the interfacial layer and the
bulk, the water and oil molecules self-diffuse in the medium. and the droplets
diffuse in the continuum. This dynamic process is influenced by the self-association
o r clustering of the droplets. Self-diffusion studies in are therefore
of considerable importance.
Measurements by NMR Method
The method of NMR based on physical properties of molecular spin provides an
elegant way to investigate molecular degrees of freedom It has been used to
determine characteristics such as critical concentration (CMC);
aggregation number; counterion association; shape, size, and hydration of aggre-
gates (micelles): structure of solution; and monomer-micelle equilibria.
The self-organized system of a can be intrinsically understood by
determining the self-diffusion coefficients of the constituent species. water, oil. and
as well as the dispersed droplets. The two o r three phase of the
system can also be understood by this method along with the degree of anisotropy
and the presence of long-range discontinuities and continuities. In a
the self-diffusion characteristics (manifested as the diffusion constant can have
the following distinctions:
1.
2.
I system: and
and
-D
3. (bicontinuous) system: D is
lower because the amphiphiles exist in large aggregates)
The self-diffusion coefficients of the constituents can thus provide information
about their domain-controlled states of aggregation and Typical
self-diffusion behavior of the different species in is illustrated on
a n arbitary scale in Fig. 2,
The self-diffusion process of the constituent species in a can
speak in favor of their degrees of freedom and hence their residential status in
different domains. F o r highlighting information on the process by the NMR
method. the articles of Zana and Lang and Lindman Lindman
and and others may be consulted.
For systems, the Fourier transform nuclear magnetic resonance
( F T N M R ) method has been used extensively for self-diffusion studies of the com-
ponents The nature of the composition-dependent self-diffusion coefficients of
the constituents of the system saline was determined by

Relative
diffusion
D
D
P h a s e domain
FIG. 2 characteristics of different species in drawn on an

arbitrary scale.
Guering and Lindman The results support I

transformations with increasing salinity of the system. and Kaler
used in the study of the bicontinuous structure of
where glycol, propylene glycol, and glycerol were used as polar solvents, with
pentaethylene glycol mono-n-ether and straight-chain alkane as the amphiphile and
oil. respectively. Dilute formulations in oil showed the presence of microstructures.
The system ammonium bromide in con-
centrated form gave evidence of the presence of spherical aggregates
studies of three different polar systems were made by
Jonstromer the surfactants used were DDAB (didodecyl
ammonium bromide). and (tetraethylene glycol dodecyl ether). In the
derived systems, isolated reverse micelles closely aggregate at lower temperature;
the structure changes toward the bicontinuous state at higher temperature. The
DDAB- and systems did not provide evidence in favor of aggregate
formation. The aggregates did not show a dominating diffusion phenomenon.
diffusion studies by the N M R technique indicated that the
system is structureless
A number of other NMR-probed microemulsions have appeared in recent
literature. The diffusion coefficients in and ammonium
microemulsions investigated by Olsson established
that replacement of water by ammonium hydroxide destabilizes the liquid crystal-
line phase and reduces the size of the colloidal association structure in the isotropic
liquid region, Olsson and worked on nonionic
prepared from pentaethylene glycol dodecyl ether and Discrete
swollen micelles have been evidenced by N M R self-diffusion measurements; the
preparations are in conformity with the hard-sphere model. The N M R
sion measurements on a microemulsion sys-
tem advocated a progressive decrease in the mean curvature of the surfactant film
with water addition at a constant level of the oil It was concluded that the

microstructure changes to via a bicontinuous state. From self-diffu-
sion measurements, et observed that in a glycol
dodecyl system prior to the bicontinuous state, limited growth occurs
spherical to oil-swollen prolate niicelles. By N M R measurements Hendrikx
explored the local properties of the A O T monolayer for
alkane systems. A lamellar phase was observed for alkanes with carbon
11; for longer chains, depending on salinity, the bicontinuous state appears.
There can be chain penetration in the monolaver for lower alkanes: the
is inconclusive for higher alkanes. Microemulsions have the prospect of use as
resistant hydraulic fluids. A model study on glycol
dodecyl was studied by the N M R method by
Waysbort et A n inversion to type was found to occur with
water in the range of 55-60% as corroborated from microviscosity measurements.
A composition-dependent distinction between free and aggregated water was estab-
lished. The Fourier transform pulse gradient spin echo H technique was
used by in a dynamic study of glycol dodecyl
oil (cyclohexane + 1:1 Rapid fusion-fission of the
was observed at low oil content and a t low temperature;
emulsion dominates in the system. A slight increase in temperature results in sig-
nificant growth in the size.
Self-diffusion measurements using the N M R technique have also elucidated the
matter of water structure in the pool; u p to 10--15 water molecules per
group (when A O T is the surfactant) were shown to be structurally perturbed
From self-diffusion N M R studies of AOT-induced
proposed several regions of bound water in the water pool domain The
state of water in the pool for
bromide o r
o r cyclohexane) was examined by NMR self-diffusion studies
at different ratios o r values. Three types of water molecules--
bound, aggregated, and free-in various proportions have been envisaged
The account given above illustrates that the transport phenomenon of
sion is useful for the study of the internal states and dynamics of
neous microemulsion systems.
The ion conductance in a depends on its type. In an I)

the conductance is almost like that of an aqueous in a
it is very low, whereas in the bicontinuous
condition, the conductance can be conspicuously large. Depending on the
sition and temperature, a dramatic increase in conductance may occur; this
phenomenon is called percolation.
The very low conductance in a dilute system has been
explained on the basis of migration of statistically charged droplets with charge
fluctuation features In the ternary systems of
benzyl alcohol, polypropylene glycol, etc., the compositions toward the oil
corner in the ternary phase diagram (called the L-phase), show a n increase in

conductance increasing alcohol concentration that passes through a
and then decreases. The increase and decrease have been attributed to the ioniza-
tion of the surfactant by incorporation of alcohol in the micelle and the formation
of mixed reverse micelles as well as a change in the shape of the aggregates
In this context, the conductance behaviors of perfluoropolyether is
noteworthy.
The conductance of the salt of perfluoropolyether carboxylic
system as a function of water content a t different
mass ratios has shown a well-shaped maximum in the curve The initial
rise in conductance has been considered to be due to the increased dissociation of
counterions from the groups by their solvation with water. The formation of
water droplets (reverse micelles) begins at the conductance which has
been supported by scattering measurements. A working model and quantitative
analysis for this distinct system have been attempted.
Unexpected conductance behaviors have been shown by microemulsions
obtained from mixed surfactants and mixed cosurfactants [56-621. T h e transport
property depends on the mixing ratio. This is discussed in Section II. B.5.
1. Percolation of Conductance
Binary inclusions containing conductors and insulators after a threshold volume
fraction of the conductors can exhibit a rapid rise in conductance followed by its
leveling off. This phenomenon is called static percolation and has been critically
studied by a number of researchers In colloidal solution as in micro-
emulsion, droplets can approach their neighbors by diffusion. and after reaching a
threshold distance they can transfer charge to augment a sharp rise in conductance;
this is called dynamic percolation Such a conductance manifestation can
occur after a threshold dispersant concentration at a constant temperature (volume
percolation) is reached as well as after a threshold temperature is reached a t a
constant dispersant concentration (temperature percolation). Conductance perco-
lation can also occur in microemulsions with nonionic containing
soluble salts in the aqueous core The origin, nature, and mechanistic behavior
of various percolating and quantification of the results can be found in the
literature The mole ratio plays a n important role
in the process of percolation. The threshold fraction of the disperse phase
depends on below which the change in conductance with volume fraction of
the dispersed phase is only In the case of temperature percolation, the
threshold temperature for percolation depends o n which is clearly demonstrated
in the works of et and et
investigated the structure and percolation of
employing sodium (AOT) and
phosphate (SDEHP). The sulfosuccinate-derived micro-
emulsions have produced spherical aggregates, whereas those derived from the
phosphate are composed of cylindrical aggregates. But the preparations have
shown percolation in conductance; the conductance increase with the phosphate
system is 10-100 times more than that with the succinate system. Moreover, for the
former, 0.06. and for the latter In the region of 0.2, the
electrical conductivity of the phosphate system has been found to be 1000-fold

recesses in 265
greater than that of the AOT system. The higher in electrical conductivity
have been considered to be a consequence of the presence of elongated aggregates
in the system, inside which the charge can for a long distance without
hindrance.
2. Mechanism of Percolation
The threshold volume fraction of percolation is guided by the shell
length and the overall volume fraction of the dispersed phase. F o r zero shell length
and no interparticle attractive interaction, according to the close-packed
hard sphere 4, = 0.65; systems with strong attractive interactions end
with being lowered from 0.65 to 0.10.
During percolation, the transfer of charge occur in two ways: (1) by the
mechanism of "hopping" of surfactant ions from one droplet to another
and (2) by the mechanism of transient fusion, ion transfer, and fission
While Maitra et are in favor of the hopping mechanism, et
support transient fusion and the mass transfer mechanism. The dynamics of
the fusion model have been amply treated by use of the time-resolved fluorescence
quenching (TRFQ) method The formation of the bicontinuous state by
droplet association has also been suggested for augmenting percolation The
evidence for the existence of the globular form of droplets in concentrated water/
microemulsions given by the dynamic light scattering (DLS) method does
not support the necessity of a bicontinuous structure for the occurrence of percola-
tion Spherical droplets with strong short-range attractive interaction can
produce percolation. The hopping, transient fusion, and transfer mechanisms
are modeled in Fig. 3.
A thorough recent study on conductance percolation of water
and dodecane) systems with regard to and tempera-
ture is due to their observations on percola-
tion are in agreement with the reports of al.
According to Feldman et the conductivity and dielectric polarization
below the percolation threshold arise the motion of through
water channels and are not due to the hopping of ions. Above the threshold, the
same mechanism still dominates.
3. Theory of Percolation
The with ionic o r added electrolyte in the case of
can be considered equivalent to mixtures of conductor and
insulator. A t a constant temperature in the percolation range, the conductance
of the system on the whole is guided by the scaling law
= -
where k and are constants and is related to the conductance of the dispersed
phase.
According to a report of et the exponent is 1.9 for both
static and dynamic percolations. The scaling law for temperature percolation has
the form
= -

Random dispersion charge Association
transfer restricted b y the augmenting charge
distant of separation transfer b y hopping
Fusion Fission
Transfer of ions Net transfer of
m ions
FIG. 3 (A) Hopping Ions hop in the direction indicated by the curved
arrows. (For simplicity, one ion is shown.) Ion transfer by fusion
and fission. cations are in each droplet; cations are transferred.
where is the conductance of the solution a t temperature 0, 0, is the threshold

temperature for transition, and and are constants.
The plots of and versus depicted in Fig. 4 are used to detect the transi-
tion values and respectively.
The expected value of like that of is 1.9 But the values usually
realized in actual experiments differ from 1.9, especially in the presence of additives.
The conductance of can be examined in the light of effective
theories of binary inclusions The E M T is not ade-
quate to deal with microheterogeneous dispersions of metals and metal oxides in a
continuous medium, and and Hunderi proposed an effective med-
ium theory considering dipole-dipole interaction among the dispersed entities
The dipolar interactions give rise to the formation of chains and
ters of spheroids. A percolation equation for the dispersion of
solids in a suitable was also proposed by and
Under practical conditions, for a nonconducting dispersion medium the
equation of Bottcher transforms into the scaling form with the exponent
= 1. Thus,
Bottcher:

recesses in
4 Schematic representations of and temperature-induced percolation.

and are indicated by arrowheads.
Scaling form:
= 1.5 - 0.333)
where is the conductance of the dispersed phase and the other terms have
already been defined. 0.333 is the threshold volume fraction for
dispersions 0.333. From mass balance and the rationale put forward by
and Fang and et structural information (particle
size and population, aggregation number of surfactant and cosurfactant per dro-
plet, etc.) on microemulsions can be realized. A set of typical results is pre-
sented in Table 1. The influence of surfactant, and cosurfactant in determining
the droplet size and amphiphile aggregation number is apparent from table.
Pal et simplified the E M T D D thoery of Granqvist and Hunderi for
dispersed particles of different states of aggregation:
For spheres,
=1 - 0.333)
For chains,
For clusters,
The E M T D D (cluster) and equations have been found

not to be valid for all 29 systems examined by Pal et The E M T and
E M T D D (chain) formulations have been found to be applicable to most of the
systems examined. It can be mentioned that the threshold volume fraction
= 0.157 proposed by Bernasconi and (considering clustering of dro-
plets) is in exact agreement with the for clusters when the conductance of
the continuous is zero.

Estimated Water Pool Size Overall Droplet
Dimension Aggregation Number (A,). and Aggregation
Number ( A , , ) per Droplet tor a of 0 at 293 K at Constant
Weight Fraction of Water 0.2 and Ratio =
System' (11111)
CTAB Hp
CTAB HA
Watei Dc
Hp
Bu
SDS Bu
Dc H 4
Water AOT Hp
AOT Dc
AOT Dc HA
'Hp HA =
Source: Adapted from Ref. 114.
The equation is valid u p to 0.75; for higher values of 4) it assumes the

of the E M T . Pal also examined the possible correlation between the
performance of and EMTDD (chain) in terms of (the structural ratio
at the droplet interface) and They further observed that
systems that give 1 satisfy both E M T and E M T D D (chain)
They further noticed that E M T and E M T D D (chain) propositions are obeyed in
the presence of without cosurfactant, the E M T D D (cluster) and
propositions are valid.
Effects of Additives on Percolation

The phenomenon of percolation has been found to be affected by the presence of
additives; both enhancement and retardation of the percolation threshold have
been witnessed. This phenomenon is related to the consistency of the
interface of the water Softening of the interface by butanol
and its rigidification by cholesterol and its derivatives have been demon-
strated. According to one report toluene blocks the transfer of charge carriers
(ions). suppressing the conductance, prevents percolation. Mukhopadhyay et
made a detailed study of the additive effect using alkanols, cholesterol,
esters of cholesterol, benzyl alcohol, and a crown ether in a
Except for benzyl alcohol and the crown ether. other additives
increased and also lowered the energy of activation for percolation. increased
with the increasing chain length of the The mechanism of transient fusion
followed by mass transfer and fission as demonstrated by the method
has been considered to be valid; the rate constant for fruitful droplet
collision in the absence of additive has been shown to be of the same order as the

recesses in 26
exchange rate constant realized by the T R F Q method. The

system of (here xylene itself is a
blocker, like toluene can be rendered percolating by the progressive addition
of In the presence of butanol, cholesterol reduces the percolating efficacy
Pileni al. reported that c can induce significant tempera-
ture percolation in a sys-
tem. Ray made an elaborate study of the effects of 29 additives of
different kinds (electrolytes, and surfactants) on the temperature
as well as volume percolation of the system. Bile salts. sodium
and were observed to lower both and 9,; the other addi-
nent of -
tives either remained indifferent or increased the percolation threshold. The expo-
(1) was found to be 1.0, less than the expected value of 1.9. A special
droplet fusion model was proposed to explain the percolation-assisting effects of
the bile salts. In a subsequent study. Ray extended the additive effects on
the percolation of the water AOT heptane system, a unique effect of
sodium the conveision of a o
(water to a percolating one In addition to this. the scaling laws
were tested, and activation energies for percolation were evaluated. Like toluene
and xylene, benzene and naphthalene increased the percolation threshold; the aro-
matic rings are considered to block the droplet fusion process, thus hindering the
phenomenon. A representation of typical results is given in Table 2. It is seen from
the table that at 33 the threshold percolation temperature of ((without
additive) is lowered to by the presence of 0.1 whereas
of same strength does not affect the threshold temperature.
iodide (TEAI) at 0.1 dm 3 increases 9, to which decreases to
in the presence of of and 0.02 TEAI. toluene,
benzene, and naphthalene at comparable weight percents (11.5) increase to 43.0,
46.5, 46.5, and respectively. With 0.10 9, for 11%
xylene diminishes to From the analysis of the results, the values of and
of (1) and ( 2 ) were found to be much lower than the expected value of 1.9 in
the presence of additives. The modes of action of the bile salts and aro-
matics (blocker) on the droplet fusion-fission process are presented in Fig. 5. For
establishing the proposed model on a firm basis, further investigations are needed.
The dependence of the percolation temperature of a microemuision
on the concentration of triblock copolymers
and water droplets was straightforwardly suggested by Eicke on the
basis of a thermodynamic model. Two distinct ranges of the
droplet] ratio were observed to correlate the percolation temperature.
5. Miscellaneous Conductance Studies
et reported the unusual conductance behavior of 35 n-
propanol alkane and water SDS in the
presence of The conductance was found to decrease with increasing tem-
perature At lower the conductance decline has been to the difficult)
encountered by ions in hopping from droplet to dioplet At higher the con-
ductance increases due to the clustering of droplets in the
The binary mixtures of lower

2 Temperature Percolation Parameters for Microemulsion
in the Presence of Additives
Additive 'dm 3 )
0 33 35 2.0 1.80 1480

(0.10) 33 9.0 1.70 2.50 1451
(0.08) (0.02) 33 14.0 0.43 3.01 882
(0.10) 33 43.0 2.10 1.90 379
(0.08) + (0.02) 15 19.5 -2.31 2.78 295
(0.10) 15 35.5 -5.75 3.60 222
SDS (0.20) 33 33.5 1.70 3.45 608
33 34.0 1.20 2.57 1028
Xy 11% in H p 33 43.0 790
Xy 1 1 in Hp 33 17.5 867
(0.10)
11.23% in Hp 33 46.5 532
11.37% in H p 33 46.5 532
Np 12.1 in H p 33 51 538
= sodium = sodium TEAI iodide;
= Tl toluene; Bz = benzene; Np = naphthalene; Hp = heptane.
Source: Ref, 107.
(cosurfactants) were observed to behave like of carbon chain lengths

intermediate between those of the two individual components. The system also
showed a fair degree of percolation.
Structural and conductivity studies of surfactants of anionic-nonanionic,
and cationic-anionic types were also examined by and
and Li et The former authors observed that the threshold
volume fraction of percolation for and
TX depends o n the mixing ratio; it increases with increasing propor-
tion of the ionic component. The formation of phases for
100, and N a combinations together with lower
alkanols and was reported by Li et
with diesel oil and heptane as the continuous media. The systems showed conspic-
uous conductance
Eicke and Meier studied the interfacial charge transport in
sions with mixed monododecyi ether) and
n-octane and observed unusual reductions in conductance producing a percolation
type of pattern in the conductance versus temperature course. The diffuse double
layer a t the interface of droplets was considered to be highly
compressed, which accounts for reduced mobility and surface conductivity.

Processes in 71
Ill
Y
Surfactant 0 Counter i o n Na C
anion
of events in percolation. Fusion-fission process for mass transfer

(A) without additive: (B) in the presence of bile salt favored mass transfer or
percolation; (C) in the presence of unfavored mass transfer or percolation.
The area of microemulsion research employing mixed amphiphiles and mixed

oils has been only sparsely studied. For practical applications of
this area needs elaborate exploration.
6. Pressure Effect on Conductance Percolation

On the whole. the phenomenon of droplet association augments percolation, which
is manifested in the physical properties of conductance, viscosity, and permittivity.
Like thermal energy at a constant pressure and composition, variable pressure at a
constant temperature and composition can also initiate percolation. The droplet
density required for pressure percolation is normally lower than that required for
temperature percolation. droplet attraction with increasing pressure that
does not affect the droplet geometry was reported by Eastoe et The

pressure-independent droplet size was also reported by et al. the
increasing pressure enhances the formation of aggregates without altering the
dimension of the contributing droplets
The work of Boned on pressure-induced conductance percolation of
different systems and
tane) is of fundamental importance. They varied pressure to 1000 bar and
explained their results on viscosity and conductance in terms of percolation theory.
The percolation threshold decreased with pressure for the
cane suggesting increasing interdroplet attraction; for the
isooctane system, varied insignificantly. The forms of scaling equations consid-
ered for the solution conductance ( a ) are
In and are prefactors, and are conductances of the dispersed

phase, is the threshold pressure, K is the pressure derivative of and +
S'P)
is the "crossover regime" or the transitional interval, which is of order of
The analysis of the dependence of on P is simple: if K is indepen-
dent of and if and are also independent of the equation then
takes the simplified
In =f -
Tingey et reported an "antipercolation" feature of a

propane microemulsion in the pressure
range of 80-400 bar; the conductance decreased by three orders of magnitude.
The interconnected channels of the studied system were anticipated
to break down into dispersed droplets.
The pressure effect on percolation has been only limitedly studied. More thor-
ough and works on different categories of systems under varied
environmental conditions are wanted.
In the recent past liquid membranes were employed for the separation and extrac-
tion of materials, and they can be conveniently employed for separating biological
materials Microemulsions of I and types
are considered dispersed liquid membranes that can the transfer of
soluble and water-soluble compounds, respectively, across them by trapping them
in microdroplets for convenient uptake and subsequent release. The
sions I and are called bulk liquid membranes. They are recent additions
in the field of separation science and technology. This field has been fundamentally
explored and advanced by Tondre and coworkers who worked out the
fundamentals of the transport process by studying the transfer of alkali metal
and other compounds across the They
also studied the transport of lipophilic compounds and
cene) across liquid membranes

in 273
Their investigations were elaborate and extensive in of cell design, solute

choice, amphiphile selection. additive effects, and data analysis. The transfer of
and as well as the nonionic compounds perylene. and
anthracene rapidly increases with Wiencek and also
reported sharp separation of acetic acid from water. In Table 3, results of transport
of by a liquid membrane are presented.
Tondre and coworkers proposed elaborate mechanisms of transport of both
hydrophilic and compounds across liquid membranes
This is not presented here due to lack of space.
O n the basis of ion transport across a the flux (F) of
the transported solute is given by the general equation
In this equation, 1 and 2 for neutral solute and alkali metal

the subscript stands for the operating concentration of S in the source
phase at steady-state conditions; is the droplet (carrier) concentration, D the
diffusion coefficient of the droplets, and L the cumulative thickness of the diffusion
layer; and is either the extraction constant o r the product of a partition constant
times the equilibrium constant for inside the droplets.
The usefulness of (8) under varied experimental conditions has been exam-
ined; the measurement of flux compared with the classical procedure of direct
contact has shown that membrane diameters of about d o no affect
the interfacial transfer mechanism that occurs in systems.
The transport flux for lipophilic substances through a n has
been shown by and Tondre in the light of reasonings of Ward
and Cussler to be
where D is the diffusion coefficient of the droplet, is the equilibrium constant

for binding of the solute with the droplet, is the partition coefficient of solute
3 Co-Ion-Dependent t of by lene
Glycol Dodecyl Ether Decane W 0
Transport
Anion
Acetate
Source: Adapted from Ref. 143

between the oil and is total droplet concentration, is the con-
centration of the solute outside the membrane, and L is the thickness of diffu-
sion layer.
Equation (9) has been experimentally verified; information regarding structural
organization and of the process of
the solutes in a microemulsion dro let have been also obtained.
In the case of separation of ions, 8-hydroxyquinoline and the industrial
extractant Kelex 100 were used in or micro-
The kinetic and mechanistic in the extraction process were
worked out by and coworkers T o understand the coupled trans-
port, crown ether o r Kelex 100 were used by and Tondre In several
instances of ion transport, synergistic effects were observed: the results were
explained on the basis of carrier diffusion in the stagnant layers.
investigated the transport of amino acids
nine and tryptophan) through reverse micelles, and the rates of
transport were compared and analyzed in relation to partition coefficients between
water and the AOT palisade layer. The rate of uptake of amino acids by the reverse
micelles and the influence of the ionization states of the acids on the extent of
transport were also presented. transport using chiral AOT was
also attempted, but the rigidity of the interfacial film according to them is not
sufficient for chiral recognition for enantiomeric enrichment.
The ways to extract copper ions in compartments, association
through ion exchange and complexation with hydrophobic
extractants, were reported by Tondre et From the electrochemical beha-
vior of ions they concluded that the 100 complex is too stable to
recover by electrolysis from droplets.
The overall efficacy of extraction of heavy metals (parti-
cularly mercury) from contaminated water involving oleic acid was reported and
successfully modeled by and coworkers who used experimen-
tally determined equilibrium extraction, stripping, and the initial reaction kinetics.
This model accurately predicts both the initial extraction kinetics and final mercury
extraction Good agreement between theory and experiment on the
of extraction using a to that of coarse emulsions has
been found. Electrostatic coalescence and addition were evaluated as
potential techniques for recovery of the components from mer-
cury-rich
The works on microemulsion liquid membranes have come mainly from the
laboratory of Tondre and coworkers. Only a limited number of studies have
appeared from other laboratories Further elaborate investigations in
this potent field are needed for versatile applications in the area of separation,
purification, and extraction.
1. Advantages and Limitations of Microemulsion Liquid Membranes

The advantages of liquid membranes are the following:
1. Lower interfacial tension produces smaller droplets with large surface area
per unit volume.

recesses in Microemulsions
2. They are thermodynamically stable and can be conveniently handled,

3 . Easy emulsification and with change of temperature aids in
transfer and recovery of solute.
4. The rates of transfer are faster than with synthetic membranes.
The following features constitute the limitations of liquid mem-
branes:
F o r high efficiency of transport, the receiving end should contain a signifi-
cant concentration of the separating component.
2. F o r versatile separation, the membrane should withstand a wide range of
p H , which can be a driving force for the process.
3. The microemulsion-forming constituents should have minimum solubility in
the feed phase (this limits the use of cosurfactants).
4. The extraction is by the attainment of equilibrium as in all passive
transport processes.
5 . Detailed knowledge about the phase behavior, stability, and so on of
emulsions is needed prior to putting them to use.
Support from the Department of Science and Technology, New Delhi (GOI) and
the Indian Statistical Institute, Calcutta in the preparation of the manuscript is
gratefully acknowledged.
Hoar and Nature (London) (1943).

Stoekenius, and Prince, Phys. Chem. (1959).
Shinoda and Friberg. Adv. Colloid Interface (1975).
Science (1988).
V. Schubert and Ber. (1996).
and Bock, J. Colloid Interface 486 (1988).
Verhoeckx, P. de Bruyn, and G . Overbeek. J. Colloid Interface
(1987).
P. de Bruyn. Overbeek. and Verhoeckx, J. Colloid Interface Sci.
(1989).
Shinoda and (1987).
and
Paul and P. Dispersion (1997).
P. A. Solvent Properties of Compounds,
London, 1954. p. 68.
Prince Theory and Practice. Academic, New York,
1977.
Gillberg. in Emulsions and Emulsion Marcel
Dekker. New York, 1984, pp. 1-43.
I. Robb Plenum, New York. 1977.

recesses in 277
I . Molinero, Sierra, and Colloid Interface

(1997).
M. Sierra and (1990); (1996).
L. Sierra and (1993).
and Colloid Interface Colloid
Sci. (1993).
and Perez-Benito, Phys. 9496 (1991).
A. Chittofrati. A. Sanguineti. M. and Colloids Surf. A
(1992).
Ajith, A. John, and A. Rakshit, Pure (1994).
A. John and A. Rakshit, (1994).
A. John and A. Rakshit, Colloid Surf. A (1995).
and A. Rakshit. Surf. Sci. Technol. (1992).
and Manik, Surf. Sci. Technol. (1993).
X. Li, Lin, and Qin, Dispersion (1996).
X. Li. Lin, Zhao, and Xiao. Colloid Interface (1996) and
references
and Zallen, Phys. (1970).
Blanc and E. in Percolation, Structures and of the
Israel Physical Society Deutcher, Zallen, and Alder. Israel
Physical Society, Jerusalem. 5. 1983, 229.
A. Burggeman, Ann. Phys. (Leipz.) 665 (1935).
Kirkpatrick, Phys. Rev. Lett. (1971); Rev. Mod. Phys. (1973).
Phys. Rev. B (1973); (1974).
Elliott. A. and Leath. Rev. Mod. Phys. (1974).
Phys, Rev. (1975).
Eicke. and I , in Solution
and Lindman, eds.). New York. 1984. p. 1533.
A. Cazabat. F. Guering. Langevin,
and Pailette. in in Solution and
Plenum, New York, 1984. 1737.
A, Cazabat, Langevin, J. Mennier. and A. J. Phys. Lett. 43:L 89
(1982).
and Phys. Rev, A (1986).
A. Bug. A. Safran, Grest, and Rev. Lett.
(1986).
Bhattacharya. Stokes. Kim. and Phys. Rev. Lett. 55:1884
(1985).
M. A. Van Phys. Rev. Lett. (1985).
A. Safran. Grest, and A. Bug, in Systems
Rosano and Marcel Dekker, New York, 1987. 235.
H. Eicke, Meier, and (1990).
Phys. Lett. (Paris) (1979).
Eicke. Geiger. Hilfiker, A . Saner, and in Time
Dependent Effects Disordered Systems and Plenum,
New York. 1987. p. 219.
I. Lapeyzynska, and Gillberg, Colloid Interface (1976).

Borkovec, Eicke, and Phys. Chem.
(1988).
P. P. Tartaglia, Chen. and Phys. Rev.
A 45:R 5358 (1992).
Mukhopadhyay, Bhattacharya, and Colloids Surf. A
(1990).
P. and A. J. Phys. Chem.
(1995).
K. A. Shioi, and Phys. (1994).
Grest, I. A. Safran, and A. I. Bug, Phys. Rev. A
(1986).
Huang, A. Safran, Kim, Grest, Kotlarchyk, and
Phys. Rev. Lett (1984).
Hilfiker, Eicke, Geiger, and G. Colloid Interface Sci.
(1985).
Peyrelasse, M. Moha-Quchane, and Boned. Phys. Rev. A (1988).
A. Jada, Lang, and Zana, Phys. Chem. (1989); (1990).
Dutkiewicz and B. Robinson, Chem. (1988).
A. C. and Varshney, Phys, Chem. (1990).
and J. Phys. Chem. (1988).
P. P. and J. K. Thomas, Chem. (1963).
Chem. Phys. Lett, (1976).
Lang in The Structure, Dynamics and Equilibrium Properties Colloidal
Bloor and Wyn-Jones, Academic,
Amsterdam, The Netherlands, 1990, p. 1 and references therein.
Lang, and Zana, Colloids Surf. (1972).
Urbach. A , Cazabat, and Langevin, Rev. Lett.
(1985).
M. Kotlarchyk, S. H. Chen, Huang, and Phys. Rev. Lett.
(1984).
N. I. V. Archipov,
and V. J. Phys. Chem, (1996).
Peyrelasse, Boned, and Clausse, Nature (London)
(1979).
M. and Phys. Lett. (Paris) (1978).
Peyrelasse, Boned, Nicoles, and A. in
Solution Properties of Plenum, New York, 1983.
Kozlovich. I. Nir, and Phys. Rev. E (1995).
and Phys. Lett. (Paris) 44: L679 (1983).
Ray, Paul, and P. Moulik, Colloid Interface (1996).
Ray, and S. Moulik, Proc. Conf. Physical and Chemical
Aspects of Organized Biological Assemblies, Ind. Surf. and Techno].,
University (Calcutta) 1991, 85.
Phys. (1952).
Bottcher, (1945).
and 0. Phys. Rev. B (1976).
Bernasconi and Phys. Rev. (1976).

Venable and Fang, Colloid Interface (1987).
Bisal, P. Bhattacharya. and P. Moulik, Chem. (1990).
Pal, Bisal. and Moulik, Phys. Chem.
Patanjali, A. Nabi, and A. N. Colloids Surf.
(1988).
in Solutions: New Methods o f Investigation Zana,
Marcel Dekker, New York, 1987, p. 241.
Mukhopadhyay, P. K. Bhattacharya, and S. P. Moulik, J. Chem.
(1993).
M. J. P. and C. Petit, in The Structure, Dynamics and
Equilibrium Properties of Colloidal Systems Bloor and Wyn-Jones,
Academic. Amsterdam, The Netherlands, 1990. 355.
Ray. S. Bisal. and P. Moulik, Chem. Faraday Trans.
(1993).
Eicke, Gauthier, and J. Phys. France (1993).
Eicke and W. Meier, Chem. (1996).
Eastoe, B. Robinson, Steytler. and Thron-Lesson, Adv. Colloid
Interface 36:1 (1991).
E. J. Fulton, and R. D . Chem.
(1991).
Eastoe, Young, H. Robinson, and Steytler,
P. Blitz, and Fulton in Super Critical Fluid Science and
Technology, ACS Symp. Ser. 406, Am. Chem. Washington. DC. 1989.
165.
Boned, P. and Phys. Rev. E (1994).
Fulton, and Smith, J. Chem.
Dubey, and Phys. (1961).

and Steinrauf, Biol. (1970).
I. A. and A. B. Fed, Fed. Am.
Biol. 29: 1327 (1970).
Yagi. and Membr. Biol. (1974).
J. Bedke, and
Am. Chem. (1980).
and (1973).
Hanji, Horiguchi, and
Am. Chem. (1976).
Kirch and Lehn, Chem. Int. Ed. (1975).
and Varoqui. J. Colloid Interface (1975).
C. Tondre and A, Xenakis. Colloid Sci. (1982).
A. Xenakis and Tondre, Phys. (1983).
Tondre and A. Xenakis, Faraday Discuss. Chem. (1984).
A. Xenakis and Tondre. Colloid Interface Sci. (1987);
(1987).
A. and C. Tondre, Chem. Faraday Trans. 1 (1989).
A. Derouiche and Tondre, Colloids Surf. (1990).

Tondre and Phys. (1989).
A. Derouiche. and Tondre, Colloid Interface Sci.
(1989).
and Tondre, Sep.
Technol. (1989).
and Tondre. Sci. (1992).
Wiencek and Sep. Technol. (1992).
Ward. (1970).
Diffusion, New 1976.
Hebrant, P. Mettelin. Tondre, P. and
Colloids Surf. A (1993).
C. Tondre. and (1997).
Larson and Wiencek, Environ. Prog. (1994).
Larson. B. Raghuraman, and Wiencek, Res.
(1994).
Larson, Raghuraman, and Wiencek, Sci. (1994).
Paatero, Sjoblom, and Datta, Colloid Interface Sci. (1990).
Bauer and Komornicki, Int. Solvent Extraction Conf. 315, (1983).
U.S. Patent 3641181 (1972).
Flaim and S. Patent 16, 467 (1981).
K. Colloids Surf. (1988); Sep. Technol. (1988).
A. Gaonkar and Colloid Interface (1987).
and Anal. (1983).
P. and Ion (1983).

Engineering and Materials Science, University of Minnesota, Minneapolis,
Minnesota
Department of Chemistry, University of Minnesota,
Minneapolis. Minnesota
I. Introduction
I I . Chromatographic Materials
A. Chromatographic zirconia
B. Colloidal aggregates
C. Other materials
I I I . Diffusion
A. Definitions of diffusion
B. Porous media
C. Large solutes
IV. Conclusion
References
The successful design of structured materials for separations depends on an under-

standing of the structural and chemical properties that are necessary and desirable
for a given process. Reproducibility of manufacture and performance depend on
adequate characterization of the chemical, physical. and process characteristics of
the materials. An ideal chromatographic support would have a high solute load
capacity, be highly selective between solutes. and allow complete solute recovery.
These place specific requirements on chemical and physical properties of any
support.
The outcome of a separation depends on fluid dynamics, mass transfer, and
chemical kinetics within the system, with the relative importance of each changing
with experimental conditions Porous chromatographic supports are commonly
used because they present surface area per bed volume than nonporous

supports. increasing the retention and the bed capacity for a given solute without
the costs associated with a longer column (higher pressure drops, greater capital
investment, longer separation times). However, resistance to mass transfer between
the extraparticle and fluid and intraparticle diffusion rates, which may
be significantly reduced those in the surrounding extraparticle fluid, can
reduce the efficiency of a separation. In this chapter we focus on intraparticle
diffusion.
This review is intended to familiarize the reader with an important class of
chromatographic materials that are made by aggregating colloids. We review the
literature relevant to predicting the diffusivity of large molecules within these mate-
rials. For design of material pore structures, rigorous theories and calculations
incorporating explicit solute-support hydrodynamic and electrostatic interactions
are desired. A rigorous level of understanding does exist for hindered transport in
straight cylindrical pores such as are found in track-etched membranes.
Researchers have explored the fundamental solute-pore interactions and have
incorporated them into models that are successful in predicting both partition
and transport coefficients However, an important limitation is the difficulty
in relating fundamental geometric properties of solutes in a generalized porous
medium solute size and shape, pore geometry and topology) to those of
solutes in cylindrical pores.
This discussion focuses on chromatographic materials made from aggregated
spherical colloids. The interstitial space between the packed colloids forms a pore
network within the chromatographic particle. Colloid aggregation is not the only
way to produce porous materials; however, technologies for generating porous
ceramics by colloid aggregation are accessible in of equipment, time, and
starting materials, and they produce particles of appropriate pore size, surface area,
particle size, and mechanical strength One example of a chromatographic
material made by aggregating colloids is porous zirconia spheres made in our
laboratories over the last few years.
Rigorously alkaline conditions-up to 0.5 M with alcohol and elevated

temperature-are commonly used to certify sterilization in large-scale commer-
cial bioseparations Zirconia is orders of magnitude less soluble than most
other metal oxides, especially at elevated pH only harsh acidic conditions
have caused detectable dissolution of zirconia, likely assisted by
plexation with a solution anion The chemical stability of zirconia has moti-
vated the development of zirconia-based supports using a variety of
chromatographic modes. including size-exclusion, affinity, ion-exchange, and
reverse-phase f9-
Zirconia is the only pure metal oxide that possesses four different chemical
properties on its surface: acidic and basic as well as oxidizing and reducing
The surface concentration of is higher than 20

Reports of the surface concentrations of acidic and basic sites vary. The highest
results indicate about 5 pmol of Lewis acid sites per square meter and about 4 pmol
of Lewis base sites per square meter Thus for zirconia to be useful as a chro-
matographic support, surface treatment is necessary to prevent the binding of any
available molecule to compete the surface coordination, in particular to block the
strong adsorption of carboxylate groups on proteins and Zirconia
faces have been modified dynamically within chromatographic columns by the
addition of competing Lewis bases to the mobile phase or by covering the zirconia
surface with polymers o r elemental carbon
In spray-drying processes, a stable suspension of colloids, which may contain a

binder, is forced through a nozzle The resulting droplets are dried by hot air to
yield particles as they sediment. The resulting structure is assumed to be a random
packing of spheres. although in some cases gross heterogeneity has been observed,
notably hollow or doughnut-shaped particles at rapid drying rates An oil
emulsion process is conceptually similar to spray drying, with droplets formed by
mixing or agitation and drying accomplished by extraction into hot oil
In polymerization-induced colloidal aggregation (PICA) processes, a reactive

monomer, generally urea formaldehyde, is mixed with a stable, submicrometer
diameter oxide sol and undergoes an acid-catalyzed polymerization that
results in porous. uniformly sized polymer-oxide composite microspheres
PICA has been applied to a variety of metal oxide systems, primarily silica, but also
alumina, titania, zirconia, ferric oxide, and antimony pentoxide The process
is affected strongly by solution acidity At lower pH, polymerization is more
rapid and a porous but mechanically weaker particle is formed.
Pa ticles made by oil emulsion drying and by PICA are significantly different
Each produces particles of distinct size ranges. Recently PICA has been
used to produce zirconia aggregates in the range of in diameter (final
sintered size) In oil emulsion processing, particle size and are easier
to control at larger sizes, 20 and greater. Particles made by PICA processing are
more uniformly spherical and monodisperse, while those made by oil emulsion
processing tend to be more irregularly shaped (Fig. 1). The microscopic pore
structure of particles made by the PICA process is more resilient to high tempera-
ture than particles made by oil emulsion
Microscopic examination of particle interiors reveals a gross difference in inter-
nal structure between particles made by oil emulsion processing and those by
PICA processing Materials by PICA processing exhibit characteristic
void spaces within the particles (Fig. 2). The particles appear to contain a spherical
core around which is an approximately concentric void space; the appearance is
reminiscent of an overripe peach in which the flesh has begun to pull away from the
pit. These voids are not observed in materials by oil emulsion processing and
an interesting and poorly understood feature of the mechanisms of particle
formation and growth PICA.

eeder et al.
FIG. 1 Scanning electron microscopic images of zirconia colloidal aggregates made by

(a) oil emulsion processing (batch OE-10) and (b) polymerization-induced colloid
aggregation (batch (From Ref. 28.)

FIG. 2 Scanning electron microscopic images of particle interiors for particles made by
colloid aggregation. Sections were made by dispersing particles
in epoxy resin and slicing with a microtome. Damage to the resin from the electron
beam is visible in the images. (a) Zirconia, batch PICA-7; (b) zirconia, batch PICA-4.
(From Ref. 28.)

Inorganic sol-gel polymerization processes are yet a feasible alternative to
colloid aggregation. These typically produce highly porous gels with small pores
and low mechanical strength materials have also been pro-
duced by inorganic polymerization in the aqueous phase of an
emulsion. The phase is then extracted with an appropriate solvent. The
most widely reported work has been on materials with unconnected cylindrical
pores but bicontinuous materials have also been reported These mate-
rials also generally lack sufficient strength for use in high pressure
applications and have not been widely used, although a few reports exist
Highly porous polymer gels are widely used in laboratory-scale purification of
proteins, but tend to under high pressure gradients and to shrink and
swell upon solvent changes and may be stable over only a narrow range of
features that make them undesirable for use in large-scale or high-speed processes.
Advances polymer chemistry have improved the chemical and mechanical prop-
erties of these materials, but the syntheses do not result in gel structures. These
gel materials are made either by hardening a biphasic emulsion, where the porogen
is later removed to form the pore space, or by aggregating smaller, rigid polymer
particles
We show in that the pore space of colloidal aggregates can be analyzed
to yield diffusivity predictions for large molecules. It is hoped that similar models
can be produced for the wider variety of materials just discussed.
According to Fick's first law, the rate at which a species diffuses in a given direction
is proportional to the concentration gradient in that direction.
In Eq. is the diffusive flux of species A (mass or per unit area per unit
time), a vector quantity; is the binary or mutual diffusion tensor describing the
diffusion of A in a of A and B; and is the spatial concentration
gradient of A. If diffusion is isotropic, then it may be characterized by a scalar
value law is a phenomenological description of diffusion on a macro-
scopic scale. It is useful in the design and analysis of processes like chromatography
that involve nonequilibrium mass transport, as mathematical models of chromato-
graphy concern with how fast a solute penetrates into the stationary
phase, the flux.
Another definition of diffusion is given by the intradiffusion coefficient, which
quantitatively describes how far an average molecule A will travel with time

In is the intradiffusion coefficient of species A, a scalar quantity
obtained from the trace of the dyadic made from the vector displacement of
A. The brackets with subscript A indicate that this represents an average over a
large number of A molecules. Here there is n o concentration gradient. Movement
occurs due to random transfer during the frequent interactions between
molecules. The intradiffusion coefficient is thus a thermodynamic quantity deter-
mined by the specific state of the system. Binary diffusion involves primarily sol-
vent-solute interactions. Intradiffusion is the result not only of solute-solvent
interaction but also of solute-solute interactions; thus in the of an infinitely
dilute solution, where essentially only solute-solvent interactions occur, the binary
diffusion coefficient and intradiffusion coefficient become identical
The term "tracer diffusion" is often used t o refer to measurements in which two
forms of the species of interest diffuse into each other. The difference between the
two forms is typically in isotopic composition o r optical rotation: it should have a
negligible effect on the chemical properties but allow for detection of a concentra-
tion gradient between the two forms. If the two of the solute are sufficiently
similar, then the measured value is the as the intradiffusivity. In actuality,
though, this is a nonequilibrium ternary system
Phenomenological descriptions of Fickian diffusion in porous media that

there is degree of homogeneity to the diffusivity within the system. In order to
write the differential equations describing the mass transport, i t is necessary to
assume that the quantities involved can be treated as smoothly varying point func-
tions. This is done by defining appropriate statistical averages over some distance
that is large relative to the level of the pore structure but small relative to that of the
entire system Thus a critical issue in any study of transport within a porous
medium is the degree of microscopic detail that is necessary o r useful to predict the
desired macroscopic properties of the pore space.
Because a diffusivity (including the tracer diffusivity) is defined
by a flux, it is important on a scale to consider how much of the plane
through which the flux is being measured actually the passage of the
diffusing species. The simplest description of diffusion in a porous medium
(assumed here to be isotropic) relates the flux within the porous medium to that
in the bulk solution with which it would be in equilibrium by introducing the
fractional porosity and the tortuosity
The fractional porosity accounts for the reduction in the flux due solely to the
reduced area available for diffusion. It is customary to use the fractional volume
porosity, but it has been argued tht the fractional area porosity at the material's
surface is more appropriate The definition given by (3) uses the tortuosity
to account for all other effects that alter the binary diffusivity, and it lacks any
inherent mechanistic o r predictive detail.

The porosity is not a separate factor for diffusion measurements using tech-
niques such as pulsed field gradient nuclear magnetic resonance spectroscopy and
some modes of quasielastic light scattering in which the diffusivity is determined
from the mean-square displacement of solutes:
Here diffusion is measured only within regions that are permeable to diffusion.
the liquid-filled pore space, rather than spatially averaged over permeable and
regions as with (3). The tortuosity in is not identical
to in Eq. (3). As will be discussed below, partitioning and steric hindrance affect
the two differently.
When diffusing molecules are of a size comparable to that of a confining liquid-
filled pore, the measured diffusivity may be significantly less than that measured in
free solution due to frictional drag, steric hindrance, and differences between the
solute concentration within the pore and in the free solution (partitioning)
There are many quantitative predictions for rigid and flexible particles in variously
shaped pores Partitioning affects both equilibrium and
measures of diffusion. Because the average concentration within the pore is differ-
ent from that outside, the state of solutes within the pore differs
from that of the bulk solution in important thermodynamic quantities like activity
and partial molar volume measures of diffusion are further
reduced by steric hindrance. In addition to the area through which there is no
diffusive flux because it is occupied by the solid, there is a portion of the area
near the pore surface that is inaccessible to the center of mass of the solute because
of the finite size of the solute.
models also exist for the contribution of hydrodynamic o r frictional
drag in straight cylindrical pores. These are well reviewed in Ref. 2. If a solute
adsorbs onto the pore surface and spends some portion of its time immobile. its net
rate of transport will be reduced. Additionally, in some cases adsorption of the
solute can significantly diminish the size of the pore and may even block the pore.
preventing passage of another solute However, less is known about how to
model the hydrodynamic drag in the convex, multiply connected pores provided by
colloidal assemblies.
There is also the effect of the structure of the porous material. F o r a
equilibrium measurement of diffusion, one can consider that there is no straight
path for solutes to travel in the direction of the flux. In a n equilibrium measurement
of intradiffusion, this represents the fact that solutes are not longer subject to a
purely random walk. When a solute is near a pore surface, the probabilities for
moving in each direction are no longer uniform; certain directions are prohibited by
the pore wall. F o r technical precision, then, one should differentiate between a
structure factor and a tortuosity. A tortuosity, quantitatively describes experi-
mental results in which multiple interactions affect the diffusion. A structure factor,
q, quantitatively describes only the effect of pore space geometry and topology on
diffusion. Note that for limited conditions- when studying diffusion of small
and a passive pore surface- this allows for

The geometry and topology of real porous media tend to be very complex,
making a complete description impractical. The challenge then is to determine
the amount and type of information necessary to adequately and concisely describe
transport within the material. Network models have demonstrated that diffusion
within a pore network depends strongly on the mean pore coordination number
and the size distribution Luckily, irregular networks have transport
properties identical or nearly identical to those of regular networks as long as they
have equal average coordination numbers and pore volume distribution For
any regular lattice of identical pores, q = 3 this result is also obtained for
randomly oriented uniform pores If the pore size is distributed, values of
q 3 are predicted. These predictions are much higher than experiments indicate
for diffusion within colloidal aggregates These errors arise because these
models erroneously assume that diffusion is unidimensional in a direction parallel
to the pore axis and that the nodes connecting the pores have no
that are evidently invalid for highly or moderately porous materials. Efforts
have been made to correct for the nonzero aspect ratio of the pores, with partial
success Superlattice models have the potential to account for the capacity
of pore bodies but would be computation-intensive and would require the devel-
opment of a method to generate the appropriate conductivity distributions from
experimental data.
In the cell the macroscopic system is divided into smaller cells
whose properties are assumed to be the same as those of the macroscopic system.
This model gives
where (fa is the fraction of solids =1 Equation (5) is valid for low
volume fractions of solids placed periodically the dispersed solid is treated as a
perturbation to the continuous liquid phase. Other than the required periodicity,
the structure is not specified.
Exact predicition of the diffusivity requires solution of the continuity equations
For an arbitrary structure, this is not feasible. However, for randomly placed,
overlapping spheres, a statistical argument provides a closed-form analytical solu-
tion as a limiting case to a more general series solution
For periodic structures, the continuity equations may also be solved explicitly by
expansion to yield tractable series solutions T o a first approx-
imation, these treatments are equivalent at low volume fractions of solids.
Experimental studies of diffusion of small nonadsorbing solutes within ordered
assemblies of colloids is has been well predicted by these treatments
However, these theories d o not consider solutes of nonzero size, and another
method is necessary to examine this.

Travel under a random walk is desribed by where ( r 2 ) is the mean square
displacement of the walker, is the length of the steps, and n is the number of steps
taken.
can be mathematically described as a random walk
where, when stepping is at a constant rate, time is proportional to the number of

steps. In a porous medium, the distance traveled will be reduced due to obstruction
by the solid,
A random walk simulation thus provides the structure factor for any geometry that
one defines. This technique has been used to examine the structure factor for an
infinitely solute within random arrays of freely overlapping cylinders with
various orientation distributions to study the effect of the distribution of
reactive sites within a porous catalyst and to study the effect of solute size
on the structure factor within ordered assemblies of spherical colloids
While the random walk simulation is straightforward in concept and in imple-
mentation, it is limited in application. The simulation does not contain time or
energy; thus forces cannot be incorporated, and solutes and pore surfaces cannot
interact other than to prevent overlap. For hydrodynamically interacting Brownian
particles, the apparent that would be measured at long times is
expressible as the product of a hydrodynamically determined short-time diffusivity
and a contribution that accounts for the distortion of the equilibrium dispersion
structure caused by the diffusing particle The short-time diffusivity, is an
ensemble-averaged quantity that accounts for hydrodynamic forces between part-
icles and can be evaluated from the hydrodynamic diffusion tensor using Monte
simulation It has subsequently been postulated that diffusion in a
porous medium can be described analogously as a product of a factor accounting
for structure, a factor accounting for hydrodynamic forces between diffusing
solutes and the pore and the diffusivity of the solute in a solution
The friction coefficient f has been determined either from the hydrodynamic
diffusion tensor using Monte simulation or by using one of several
simple analytical expressions for a sphere translating through a right cylindrical
pore 68-70]. The most commonly used is (11) where is the
characteristic ratio of solute size to pore size

This and alternative expressions are well reviewed in Ref. 2.
N M R and light-scattering techniques as well as computer simulation have been
used to measure the diffusion of large molecules within porous media. Both
simulation and experimentation support the general form of Eq. (10) in the
absence of solute adsorption or reaction for a variety of solutes and pore net-
works, including proteins within randomly oriented fiber systems polystyrene
in glasses proteins in polymer-based chromatographic
supports and polystyrene within colloidal aggregates I n most investiga-
tions, has been assumed to be an empirical constant, independent of solute size
though for colloidal aggregates it has been predicted by computer simu-
lation to be a function of solute size Values (or functions) for q reported in
the literature vary between 1 and approximtely 0.3 depending on the type of
material and its porosity. Of the materials reported, fiber arrays generally have
the highest porosities and the highest values for followed by aggregated cera-
mic colloids and aggregated polymer colloids, with cont olled-pore glasses gen-
erally having the lowest values of q There is. however, general
agreement that should be defined using the hydrodynamic radius of the solute
and the mean pore constriction radius, such as one would measure by mercury
porosimetry
T o aid in the design of the aggregates we make in our laboratories, we have sought
to understand the diffusion of large molecules within these porous materials as well
as other materials. N M R and light-scattering techniques have allowed researchers
to directly measure diffusion in porous media that were previously inaccessible to
experimental measurement due to small length and time scales. We have summar-
ized here the framework providing a general relation between the measurement of a
diffusivity and its use in a design equation. Hydrodynamic theories
and computer simulations have led to significant advances in quantitative predic-
tion of large molecules within porous materials that possess a periodic or statistical
regularity, which we have applied to our own materials made of aggregated sphe-
rical colloids; however, these results have underscored the necessity of having at
least a qualitative knowledge of the pore geometry-generally related to the
r
sythesis method.
This work has been supported by grants G M 45988 the National Institutes
of Health and CHE 917029 from the National Science Foundation. Additional
support has come from the Training Program of the National
Institutes of General Medical Sciences the Institute for
Advanced Studies in Technology at the University of Minnesota,
and the Regional Instrumentation Facility for Surface Analysis at the
University of Minnesota.

and A (1994).
(1987).
Nawrocki, Rigney. A. V. McCormick, and P. A
(1993).
A. R. Barry and (1994).
Inc.. Downstream News find Process N o . 2,
1986.
Porous Silica Chromatogr. Silica Library.
Amsterdam. 1977.
Weber, and Carr, Chromatogr. (1989).
P. Carr, A. Blackwell, Weber, A. Schafer, and M. P. Rigney, in
in and American
Society. Washington. DC. 1993.
H. and J. (1990).
A. and Carr, Chromatogr. (1991).
and P. Carr, Anal. Chein. (1994).
and P. Cars. Anal. Chein. (1995).
Sun and Carr, Anal. (1995).
P. Carr, and A. V. (1996).
A. A. Maroto. and
Mater. Sci. (1985).
A. A. Lecomte. and A. Danger,
Mater. 25:1268 (1990).
and J . Mater. (1991).
and Tanabe.
Faraday Trans. (198
J. A. Guizard, Cot, M . and S.
Colloids Surf.
Wax and Patent 0 490 226 (1991).
Masters, Spray Drying Handbook, 3rd ed., Wiley, New York. 1977.
Roth and Kobrich. Aerosol
P. Carr. Funkenbusch, M. P. Rigney. Coleman. A. and
W. A. Schafer, U.S. Patent to Regents of the University of Minnesota
1991).
K. Her and Patent 4,010,242, to de
(1977).
Sun. Annen. Lorenzano-Porras, W. Carr. and A. McCormick.
Annen, P. Carr. and A. McCormick, Mater.
(1994).
Lorenzano-Porras, Annen, C. Flickinger, W. Carr, and A.
J. Colloid Interface (1995).
Cars. C. Flickinger, and A. V. McCormick, presented by
A. V. American Chemical Society Spring National Meeting, San
Francisco. 1997.

Lorenzano-Porras, P. and A. Colloid Interface
(1994).
Lorenzano-Porras, Reeder, P. Carr, and A. V.
Cheni. Res. (1995).
H. Reeder, Pore structure and mass transport in colloidal aggregates for liquid
Thesis, Minnesota, Minneapolis, MN, 1997.
Q. Xu and A. Anderson, (1993).
Miller. and I, Catal. (1994).
Kresge, W. Roth, and Beck, Nature
M. I. and
Unger, Adv. (1997).
Kresge, Roth. Leonowicz,
S. Hellring. Beck. Schenker, Olson, and
Sheppard, Chem. (1994).
A. Kurganov, Unger. and J. A (1996).
F. and Science (1996).
A. Einstein,
Leipzig. 1922.
Diffusion: Mass Transfer in Systems, Cambridge Press,
Cambridge, UK, 1984.
Chem. (1992).
L. and N. A. Sci. 49-41 (1994).
Foster, J. Butt, and Bliss, Catal. (1967).
E. P. Russell, and N.
(1968).
and (1969).
(1981).
Nitsche and K . W. Res. (1994).
Anderson and J. H. Brannon. (1981).
J. A. I. P. Petrou, and
Bioseparation (1990).
Flow and transport in chaotic media, Thesis,
Minnesota, Minneapolis, MN, 1985.
and Chem. (1990).
Johnson and Stewart, Catal. (1965).
P. Friedinan and A. Chem. Eng. (1995).
D. Keffer, A. and Davis, (1996).
Maxwell. A Treatise on Electricity and Magnetism, 1. Clarendon Press,
Oxford, 1873.
B. and P. Colloid
(1986).
Bird, E. Stewart, and E. Lightfoot. Transport Phenomena. New
York.
(1963).
P. R. Leyte, and Colloid Interface Sci.
(1991).
and Leyte, Colloid Interface (1994).

M. and V. J. Chem. Phys. (1993).
P. and A, (1995).
Brady. J. Fluid Mech. (1994).
and Brady, Rev. Fluid (1988).
Phillips, J. Brady. and Phys. Fluids (1988).
J. Phillips, J. Brady. and Phys. Fluids (1988).
Claque and J. Phillips. Phys. Fluids (1996).
Zhou, I. K. and
(1 994).
Bishop, and Macromolecules (1989).
J. E. and T. Root, Phys. Chem.
(1997).
L. Anderson and A. Quinn. (1974).
Smith, (1986)

Micromeritics Instrument Corporation, Inc., Norcross,
Georgia
I. Introduction
A. Real and ideal surfaces
B. Origin of the adsorptive potential
C . Thermodynamic meaning of the adsorptive potential
I I . Theoretical Background
A. A short history of the integral equation of adsorption
B. Solving and using the integral equation of adsorption
III. Application of Density Functional Theory
A. The deconvolution method
B. Parameters of the model matrix
C . Results
IV. Discussion
V. Conclusions
References
A heterogeneous surface is one described by a distribution of potentials
In the formative years of adsorption theory. it was usual to assume that a solid
surface presented a set of equivalent sites for the attachment of a n adsorbate
molecule. Such was the idealized picture used by Langmuir [2] in developing his
classic equation. Even when the model was extended to include
adsorbate interaction as was done by d e Boer [3] in developing his nonideal

two-dimensional gas model for mobile adsorption and by Fowler and Guggenheim
in their statistical thermodynamic derivation [4] for the case of localized adsorption,
the assumption of an energetically homogeneous, or surface was
retained, as was the assumption of adsorption,
Today, there is probably universal agreement that all real surfaces are
significantly heterogeneous. The importance of physical adsorption in chemical
engineering processes involving separations and catalysis has led to increased inter-
est in this subject and to the need to explore ways to characterize the surfaces of
engineering adsorbents in terms of their adsorptive potential distributions.
In addition to providing a measure of the specific extent of a surface, the adsorp-
tion isotherm conveys a great deal of information about the energetic heterogeneity
and geometric topology or porosity of a solid substrate. As a source of information
for the characterization of materials, the gas adsorption isotherm has much to
recommend it: It is quite easily measured, it is highly reproducible from laboratory
to laboratory, and it provides a very sensitive indicator of surface properties. Using
graphite at cryogenic temperature, for example, a 5% variation in adsorptive
potential would correspond to a twofold change in the pressure required to
tain the same of argon adsorbed
. Origin of the Adsorptive Potential

The adsorptive potential is a measure of the net attractive force between a solid
surface and an adsorbed For physical adsorption, these forces arise
chiefly London-type dispersion interactions (van der forces) resulting
from induced dipole-induced dipole and higher multipolar attractions, which in
turn depend on the size, and magnetic susceptibility of the interact-
ing particles (atoms or molecules). Additional contributions may come
interactions or from induction forces brought about by the operation
of a surface electric field on induced or permanent dipoles of resident molecules.
The net potential due to these forces acting on an adsorbed molecule is generally
short-range and is the summed effect all nearby surface atoms.
The contribution of the dispersion forces to the total potential can be well
approximated by the Lennard-Jones equation for interaction,
where is the potential. the depth of the potential well. the molecular separa-
tion. and the separation at which = 0. For an adsorptive molecule at the
surface, the total dispersion potential, is the of its interactions
with each atom of the solid:
is the distance the adsorbed to the solid atom j

having energy parameter, It is readily shown that of the value of
is provided by the surface atoms within three to four molecular diameters of
the adsorbed molecule. It is clear that any irregularity in the local chemical

composition, density, o r geometry of the surface will cause a variation in the
adsorptive potential a t that point. If the surface topography is locally reentrant,
so as to constitute a fine pore of dimensions. then the adsorptive poten-
tial can be more than doubled. Since the effects of surface energetic heterogeneity
and the effects of porosity on the adsorption isotherm are both expressed through
this same adsorptive potential, we can anticipate some difficulty in separating the
two. W e shall see that such a separation is possible when the pores in the
material to be characterized are large enough to have negligible effect on
formation.
the preceding section, we presented the idea that real surfaces can be, and usually
are, "energetically heterogeneous" in that they provide regions of varying attractive
force o r potential to physically adsorbed molecules. In describing a surface energy
characteristic. a n energy function that is independent of temperature and concen-
tration is desirable. Such a quantity is conveniently defined by the potential energy
difference between the lowest energy state of the adsorptive molecule in the gas
phase and its lowest energy state in the adsorbed phase. both at infinite dilution. I n
this way we obtain a quantity that is independent of the kinetic states of the
molecule in either phase and that most directly the adsorptive potential
of the system. The adsorptive potential, is therefore defined for a homotattic
surface p. as
where is the potential energy per mole lost by an isolated in moving

from a distant point to its equilibrium adsorbed position. The second term o n the
right is the zero-point vibrational energy of the adsorbed molecule with respect to
the surface. Note that may differ if mechanisms other than disper-
sion forces are at work.
When a molecule is adsorbed, the process is by a liberation of heat
that can be measured calorimetrically. The experimentally measured heat can be
related to a thermodynamic quantity, the differential heat of adsorption, by rela-
tionships that depend specifics of the calorimeter used The
differential heat of adsorption, o n a homotattic surface at any isotherm
point is related to by
where the term - is the thermal vibrational energy per mole of adsor-
bate in excess of the zero-point energy, and are the kinetic energy
changes on adsorption due to loss in translational and rotational degrees o f free-
dom, and the energy of interaction with all neighboring adsorbed molecules.
This latter term is clearly dependent on surface concentration and could be esti-
mated by a n equation analogous to
Another thermodynamic quantity, the heat of adsorption, can be
calculated from the temperature dependence of the adsorption isotherm by obtain-
ing the slope of the adsorption and is defined by

29
-
where is the moles adsorbed and V and V are the molar of the gas and
adsorbed phases, respectively. Neglecting the molar volume of the adsorbed phase
and assuming that the gas phase is ideal gives the more familiar expression
It can be shown that the differential and isosteric heats are related by
The importance of (4) is that it expresses the differential heat of adsorption in

terms of a number of concepts that have a readily visualized physical basis; it
reasserts and emphasizes that the differential heat of adsorption contains inter
alia separate expressions for the adsorbate-adsorbent interaction and the adsor-
bate-adsorbate interaction; and since all the other experimentally determined heats
of adsorption are related to the same conclusion holds true for them. The
quantity expressing as it does the adsorbate-adsorbent interaction stripped of
all other incidental energy changes such as lateral interaction, work terms, and
kinetic and vibrational energy changes, is more suitable than any of the experimen-
tally measured heats as an index of the fundamental "affinity" of a solid surface for
adsorbing a particular gas
The relation of the various heats of adsorption to the adsorptive potential is
shown schematically in Fig. 1 . It should be noted that although heat lost from a
system is a negative quantity, it is a custom of long standing in
adsorption science to employ a sign. This is frequently confusing to new-
comers to adsorption studies.
For a more complete and detailed discussion of the thermodynamic quantities of
interest in physical adsorption, the reader is referred to Chapter of Ref. 1.
Although the concepts are somewhat older, the widely used model for describ-
ing adsorption on an energetically heterogeneous surface was first explicitly stated
by Ross and The postulates that the surface of a real solid is
composed of small patches of different adsorptive potential that adsorb indepen-
dently of one another. The distribution of adsorptive energies, among these
patches can be represented by a continuous distribution function:
is the patch (or site) frequency per unit energy interval on a surface of area
A. The distribution function must normalize to unity, as was pointed out by Hill

Distance from Surface (molecular diameters)
FIG. 1 Potential energy well for adsorbate-adsorbent interaction, showing the relation
of the various heats of adsorption.
since we are dealing with a surface of finite extent; that is, = 1 over the
range of energies considered significant,
At any equilibrium pressure under isothermal conditions, the quantity
adsorbed per unit area, on a given surface patch will depend only on the adsorp-
tive potential of that patch according to the function
or, more generally,
The observed total amount adsorbed. a t pressure is then the sum of contribu-
tions from each patch of surface,
Equation (9) is therefore the general form for any adsorption isotherm and corre-
sponds to (IV-4) of Ref. The function is called the kernel function o r
the local isotherm. Equation (9) is now often referred to as "the integral equation of
adsorption".

Analytical Methods
Referring to we see that in any treatment of surface heterogeneity we have
to deal three functions, any two of which, if known, assumed, o r determined,
can be used in theory to obtain the third. Equation (9) represents a Fredholm
integral of the first kind. The solution of equations of this type is well known to
present an ill-posed o r ill-conditioned problem. F o r o u r purposes, this means that
the data, can be "well represented" by many function pairs in the integrand;
therefore, simply fitting the data does not guarantee that the kernel function o r the
distribution is individually "correct." In addition. the mathematical difficulties of
handling (9) analytically has severely restricted the number of possible varia-
tions that have hitherto been published. In Table 1 are given descrip-
tions of some models of adsorption o n homotattic surfaces. We note that the
function be a n explicit expression for the quantity adsorbed in
terms of and This restriction alone means that of all the common theoretical
monolayer isotherm equations, only the two-dimensional ideal gas (Henry's law)
equation and the Langmuir equation are usable. The Hill-de Boer equation and the
Fowler equation cannot be solved explicitly for quantity adsorbed.
1 Mathematical Description of Models of Adsorption on

Substrates
Description of adsorbed Isotherm equation
Mobile film, two-dimensional
ideal gas, Henry's law
Mobile film, hard sphere

particle, equation
Mobile van der gas

interaction. Hill-de Boer
equation
Localized no interaction.
equation
Localized film. interaction,

Fowler equation
Mobile film. gas,

interaction, equation

The Henry's law equation can be written = where a function of
and T only. is given by
Substituting into (9) gives the adsorption isotherm equation
where is a constant corresponding to a n average surface potential This result

shows that regardless of the form of the distribution function, Henry's law yields a n
adsorption isotherm that is a straight line through the origin. All theoretical equa-
tions for adsorption surfaces must reduce to Henry's law at low
surface concentration; therefore all experimental isotherms must reduce to Eq. (17)
a t sufficiently low uptake. The zero pressure limiting slope of any isotherm allows
us to calculate the differential o r isosteric heat of adsorption through Eqs.
and (16). In practice, the pressure at which this occurs may be immeasur-
ably small, as the requirement is for a low surface concentration on even the highest
energy patches.
Langmuir himself was the first to propose the form that the Langmuir equation
[Eq. would take for a set of sites of different energies:
where is the fraction of the surface characterized by and is the

fraction of the sruface occupied at pressure p.
[9] and later Roginsky and and Taylor [I still using
the Langmuir equation, replaced the concept of a set of sites with a continuous
distribution of energies, thus permitting the summation to be replaced with an
integral. It was found that by selecting an exponential energy distribution of the
form
over the whole range of both positive and negative. the integral yielded the
equation in the
,
In these early efforts, the goal was to find an analytical description of using
a specific theoretical model for adsorption on a surface and a n assumed
distribution function. An alternative way to use Eq. (9) is to assume the of
both and and thence derive the distribution by inver-
sion. That this can be done and that a unique solution exists for was demon-
strated rigorously by Sips T o describe the experimental isotherm, Sips chose
for an empirical equation of the form
= +

and for took once more the Langmuir equation. He inverted the integral
equation by means of a transform and obtained for a unique dis-
tribution function that he showed to be nearly equivalent to a Gaussian distribution
of adsorptive energies.
An effort to evade the Langmuir equation and model mobile adsorption was
made by and The results were marred by the
needed to their model function explicitly for the quantity adsorbed.
Another way to avoid introducing the equation with its mechanism of
localized adsorption and no lateral interaction was pointed out by Morrison et
By differentiating (9) one obtains
For the solution of this equation, a new set of experimental results are demanded,
namely the differential heat of adsorption, as a function of Q. Equation (18)
still contains the unknown function This function can, however, be elimi-
nated when we know as a function of Q at absolute zero. At that temperature,
every surface patch is filled in the serial order of its adsorptive potential, beginning
with the greatest. A surface patch is therefore either completely filled or completely
The function can be replaced with the Dirac delta function,
times called in this context the condensation isotherm. This results in
The relation between and is given in (4)-(6). In order to use this

method, the heat adsorption and the heat capacity of the adsorbed film must be
measured at a series of low temperatures for various surface concentrations and the
zero degree heats calculated by extrapolation. In spite of this tremendous experi-
mental effort, the results are flawed, as we see (4) that at absolute zero,
= + The interaction term is of the same magnitude as and not
easily determined. The distribution obtained, while of the correct shape, is dis-
placed by the molar value of the adsorbate-adsorbate interaction at K.
The replacement of the kernel function with the condensation isotherm
has been used by several authors, invoking the so-called condensation approxima-
tion at experimental temperatures far from absolute zero. These efforts generally
use (18) without observing the temperature limitation. The condensation
approximation can also be applied in (9) directly:
where represents the condensation isotherm. This can be shown to give
where is the slope of the adsorption isotherm and is obtained from

the relation between the condensation pressure and the adsorptive potential.

Applications of condensation approximation have been well reviewed by
Rudzinski and Everett In general we can say that ignoring the equilibrium
thermal distribution of the adsorbate among the surface patches of different values
of can result in a seriously distorted picture of the adsorptive potential distribu-
tion.
No analytic solution of Eq. (9) has yet been made based on reasonable models of
adsorption incorporating adsorbate- adsorbate interaction. Such a solution may
not be possible.
2. Early Nonanalytic Solutions of the Integral Equation of Adsorption

The period of renewed adsorption research activity in the decade of the 1950s
happened at a time when high speed electronic computing was just becoming
available to researchers in this field. This the numerical solution of (9)
a feasible undertaking. F o r the first time, it was possible to a t least calculate the
numerical values of from the integral equation of adsorption using more
theoretically sophisticated kernel functions that incorporated adsorbate- adsorbate
interaction, together with a reasonable distribution function.
The first such solutions were carried out by Ross and 129;
Using Gaussian distributions of adsorptive potential of varying width, they com-
puted tables of model isotherms using kernel functions based on the Hill-de Boer
equation for a mobile, nonideal two-dimensional gas and o n the
Guggenheim equation [Eq. for localized adsorption with lateral interaction.
The fact that these functions are implicit for quantity adsorbed was no longer a
problem since they could be solved iteratively in the numerical integration.
It was not actually possible to fit data to the computed models numerical
methods in 1957; while computer speeds were astonishing a t the time, they were a t
least 1 x slower than today's desktop personal computers and had very limited
memory. Ross and Olivier developed a technique to find the best-fitting model for
a n experimental isotherm data set by means of graphical overlays. They found that
excellent fits to the experiment data could be obtained provided that the degree of
heterogeneity was not too great. As pointed out in a good fit to the data
does not in itself verify a kernel function o r the distribution. However, as the
adsorbent becomes more and more homotattic, its isotherms should approach
agreement with the kernel function itself. This was shown to be the case for adsorp-
tion measurements on a series of carbons graphitized at increasing temperatures,
culminating in the highly graphitized carbon black P-33, whose isotherms of argon
and nitrogen a t 77 and 9 0 K are closely fitted by the Hill-de Boer equation in the
monolayer region. By also correctly describing the heat of adsorption as a function
of quantity adsorbed for heterogeneous surfaces, this work confirmed (9) as a
powerful tool for investigating surface heterogeneity and the validity of the
nonideal gas model for the kernel function.
In later work, Ross and Morrison and and Morrison were
able to make several advances. The van der equation of state for real gases,
which is the basis of the Hill-de Boer equation, is known to be rather inaccurate.
Morrison and coworkers based their kernel function o n a two-dimensional form of
the better equation of state [Eq. But importantly, advances in

computing resources made i t possible to solve Eq. (9) for the unknown distribution
function rather than assuming a form a priori.
Again. it was found to be difficult to fit uniquely isotherm data for very hetero-
geneous surfaces. A limitation lies in the fact that the kernel functions used were
only models for monolayer adsorption, yet it is well known that adsorption pro-
ceeds to multilayers as pressure is increased. T o ensure that the more strongly
adsorbing portions of the surface remained in the monolayer range, only the lowest
pressure portion of the isotherm could be used. This means that the low adsorptive
energy portions of the surface are contributing little information in the total
amount adsorbed, making their estimation uncertain. If higher pressure data are
included in then multilayers exist on some surface patches; hence these are
not correctly modeled by the kernel function.
While good descriptions of adsorption on uniform surfaces in the submonolayer

region are available, only recently has accurate calculation of the whole isotherm.
including the region, been demonstrated These calculations use a
modified nonlocal density functional theory (MDFT). The first use of multilayer
local isotherms calculated by M D F T in obtaining a measure of surface energetic
heterogeneity for several solid adsorbents was reported in 1996
The formalism of density functional theory has received considerable
attention as a way to describe the adsorption process at the interface.
The older approach was to treat the adsorbate as a separate, two-dimensional
phase existing in equilibrium with the bulk gas phase. This model works well in
the monolayer region, but at higher surface concentrations the formation of
layers requires the adoption of some sort of three-dimensional model to account for
increasing adsorbate-adsorbate interaction and diminishing adsorption potential.
Using density functional theory, the adsorbtive can be treated as a single,
mogeneous fluid phase. The fluid varies in density from that of the bulk gas to a
much higher value at the adsorbent surface in response to the strength and config-
uration of the surface forces. In this paradigm. there is no separate
adsorbed phase; indeed, the concept of a monolayer capacity, fundamental to the
two-phase paradigm, has disappeared as well. The benefit of this approach is that
the isotherm can be modeled from the Henry's law region through to saturation, or
even above the adsorptive's critical temperature.
In particular, the ability with D F T to model physical adsorption in a pore space
of or cylindrical geometry has led to potentially useful methods for extract-
ing surface area and pore size distribution information experimental adsorp-
tion isotherms The predictions of density function theory have been reported
to compare well with the results of simulations using Monte or
molecular dynamics methods. We have made stringent comparisons to real data
for the adsorption of nitrogen and argon on the near-homotattic surface of a
highly graphitized carbon, Sterling FT-G(2700). performing such comparisons,
the only unknown intensive parameter is the Lennard-Jones interaction
energy between the adsorbate and adsorbent atoms. Using the customary

prescription (with corrected weight functions) for free energy
density functional, we found that experimental data in the monolayer
region of coverage can be moderately well described by D F T calculations; however,
in the multilayer region of the isotherm, the quantity adsorbed per unit area is
significantly Recent work showed that a modification to the
mean field approximation used t o calculate the attractive component of the con-
figurational chemical potential leads to the theoretical isotherms that agree closely
with experiment over a six-decade range of pressure. A n example is shown in Fig. 2.
The integral equation of adsorption, can be rewritten in specific units as
where is the total quantity of adsorbate per gram of adsorbent at pressure

the kernel function (the local isotherm). describes the adsorption isotherm
for a n ideally homotattic surface characterized by an interaction energy as quan-
tity of adsorbate per square meter of surface; and is the surface area distribu-
tion function with respect to The [Eq. as we shall see, is closely
1e-6 1e-5 1e-4 1e-3 1e-2

Relative Pressure
. 2 Experimental data for nitrogen at 77 K on Sterling compared to
calculation.

related to the adsorptive potential and can be related to the quantity in Eq. (3)
and in Fig. 1.
While DFT allows us to calculate values for it of course provides no
analytic form for the function, and in general the form is also unknown.
However, by using carefully designed numerical methods, model isotherms calcu-
lated by M D F T can be used in carrying out the inversion of the discretized form of
the integral equation of adsorption. In this way one can determine the effective
adsorptive potential distribution of the adsorbent from the experimental adsorp-
tion isotherm. The method used can be expressed by
where is the experimental adsorption isotherm interpolated onto the vector p

of pressure points, is a matrix of quantity adsorbed per square meter, each
row calculated by M D F T for a value of a t pressures and is the vector of
positive o r null values whose terms represent the area of surface in the sample
characterized by energy The total surface area of the sample is given by
The solution values desired are those positive numbers that most
nearly, in a least squares sense, solve (23). Additional constraints o n the solu-
tion may be required to stabilize the deconvolution process
Equation (23) differs from previous work in a n important way. previous
attempts, was the amount adsorbed at the experimental pressures p. This
required that be calculated for that specific set of pressures and that the
size of be no greater than the number of experimental points. N o t only does
this result in a large computing task, it also causes the evaluation of t o be
subject to a varying bias depending o n the of isotherm points a n d where on
the pressure scale they were measured. The automatic adsorption equipment avail-
able today permits a large number of experimental points to be measured, and the
isotherm can be interpolated accurately onto a predetermined optimized set of
pressures. Hence the vector can be chosen to best represent the kernel function
the wide pressure range required by the set of
If we consider of set of and a vector of length it is clear that
in must hold. If = the solution vector can "noisy'" because of even
small imperfections in the data o r For the solution is smoother
because of the additional data constraints. Various other regularization techniques
have been proposed to mitigate the inversion in this work we use the
method of of the second derivative of together with an over-
determined matrix for which >
F o r use with (23). a model matrix was calculated by the M D F T method

using the suitable for nitrogen a t 77.3 F o r convenience, the values of
were specifically the values of used to calculate the wall potential of
that reference, and ranged from 20 to in steps of 2.0 K. Relative pressure
points were chosen in geometric progression from 1 to 0.6 with 40 points per
decade. Model isotherms were normalized to 1 of surface.

The data reported here were obtained using a Micromeritics ASAP 2010 equipped
with an optional 1 pressure transducer. Low pressure data were corrected for
effusion.
1. Graphitic Carbons
T o confirm the algorithm. we show in Figs. 3a and 3b the result of
applying Eq. (23) to the experimental data of Fig. 2. Since this data set was used in
developing the M D F T we would expect to recover a energy
distribution with = 57.0 as used in the fit shown in Fig. 2. The best fit
contained contributions from the classes representing = 56 and 58, with an
area-weighted mean of 56.7 which is satisfactory agreement. The total surface
area obtained is The BET area of this certified reference material is
Because the M D F T model ignores the slight corrugation of the wall
potential, the commensurate transition seen at 0.008 relative pressure is not
reproduced.
Figures 4a and 4b illustrate the application of Eq. (23) to the nitrogen isotherm
obtained with Vulcan 3-G(2700). While graphitized at the same temperature as the
Sterling FT, Vulcan has previously been reported as less uniform than the Sterling
material As additional evidence, note that the commensurate film transition
near 0.008 seen in Figs. 2 and 3 is not experimentally detected on the Vulcan
surface. The area-weighted mean value of is 56.1 and the total surface area
is 80.78 The surface area by BET is
2. Carbon
An example of a much more heterogeneous surface is shown in Fig. 5. The adsor-
bent in this case is a nonporous carbon black designated C4, used by ASTM
Committee D24 as a reference reinforcing black.
The total surface area by the present method is the BET method
gives The weighted mean value of is 53.61 The central mode of
the distribution is lower than that for the graphites.
3. Noncarbon Adsorbents
Candidate adsorbents for the present method must not be that is. the
adsorption isotherm be of a free surface at least through near completion
of the Zeolitic adsorbents and activated carbons are excluded by this
requirement, but materials are usually appropriate.
Figure 6 shows the interaction parameter distribution and the goodness of fit for
a of controlled pore glass, CPG-120. Again we see a distribu-
tion with rather more of the surface having a low interaction value compared to
carbon. The mean value of the interaction parameter is 51.54, and the indicated
surface area is The BET area is calculated to be Fitting
the data up to a relative pressure of 0.6 was possible because pore filling is negli-
gible below about 0.8 relative pressure with this sample.
The isotherm for a yet more complex surface is shown in Fig. 7a.The adsorbent
is an extruded silica-alumina catalyst support used by Micromeritics Instrument

Fitted
(a) Expected of a surface from of

(20) to the experimental of nitrogen adsorbed at 77 K on Sterling

4 (a) Distribution of for from the nitrogen isotherm at
77 (b) Nitrogen isotherm data showing excellent DFT fitted result,

. (a) of for carbon C4 the nitrogen isotherm at
77 K. (b) data the fitted result.

(a) of from the nitrogen at 77K
isotherm data D F T fitted result

Pressure
Diameter
. (a) Adsorption of nitrogen a t 77 K on a catalyst reference
(b) area by using Jura
thickness curve.

Corp. as a surface area and porosity reference material. The sample is quite
obviously mesoporous. The pore area distribution shown in Fig. 7b obtained
with the BJH method, based o n the Kelvin equation, with the and
thickness curve shows porosity extending upward just below 3.0 T o avoid
the effect of pore capillary condensation in this size range, the surface parameter
analysis was performed with only the isotherm data below 0.4 relative pressure. The
surface parameter distribution shown in Fig.8a displays at least four modes that
are resistant to high levels of regularization; there may indeed be two maxima in the
range between 45 and 65 Figure shows the goodness of fit obtained from
this distribution.
The mean value of the interaction parameter for the catalyst surface is 55.58 K,
and the surface area is calculated t o be 225.68 The BET surface area calcu-
2
lated from data is 223.26m /g; is obtained from the BJH calcula-
tion.
The magnitude of the surface adsorptive potential for an adsorbed nitrogen

depends on two major factors: the local density o r proximity of the con-
stituent atoms of the adsorbent a t the site of adsorption and ( 2 ) the chemical nature
of those atoms; for instance, their size, polarizability, susceptibility, and ionic
charge. We have extracted the energy parameter distribution in the preceding
examples in of however, we must realize that it represents the net
adsorptive energy variation arising all factors. Even for a chemically pure
graphite surface, a step dislocation will create a region along its base where the
adsorptive potential will approximate 1.5 times the value for a plane and a region at
its top edge with a potential about 0.5 the flat surface value. The thermodynamic
meaning of the magnitude of the interaction parameter value is determined in the
following way.
Although D F T yields n o general analytic form for the adsorption isotherm, we
know that in the limit of low amount adsorbed it must converge to Henry's law,
is the capacity." The limiting slope, of the can

readily be calculated from the numerical results a t very low pressure at each
value. The definition of is by (16):
T o determine in terms of we plot vs. as shown in Fig. 9.

From the slope of the regression line we obtain
where is the adsorptive potential on a characterized by F r o m this

we can obtain the limiting isosteric heat of adsorption for the patch, since by Eq. (4)
= + The zero pressure limit of the isosteric heat of adsorption for

Fitted
100
Relative Pressure
(a) Distribution of for a silica-alumina catalyst support from the nitrogen
isotherm at 77 (b) data showing the M D F T fitted result.

Plot showing the logarithmic relation between the law constant and
the whole surface can be calculated from the reported distributions of using
(24) and (25) together with (16):
where a, is the area of the patch. The results of this calculation are shown in
Table 2, together with total surface areas determined with the present method. The
BET values are included for reference.
The value of for Sterling from this work, corresponding to
2.15 compares very well with the best estimate of Ross and Olivier
p. of 2.16 We see from (25) that the surface parameter intervals
of correspond to about
Most previous attempts to employ (9) to extract a surface interaction dis-

tribution used a local isotherm function that describes only formation.
Some, such as the equation, are manifestly inappropriate, as they d o
not include adsorbate- adsorbate interaction, hence cannot describe an

2 Calculated Heats at Zero Coverage with Surface Area Comparisons
Adsorbent No A' (BET
Sterling FT- 3 9 9.96 12.4 11.1
3- 10.02 10.98 80.8 73.5
- 5 14.12 15.09 38.7 129.6

120 6 14.9 15.18 124.8
Silica-alumina 8 14.9 15.86 225.7 223.2
catalyst
with a shape, as in Fig. 2. Others, such as the Hill-de Boer equation,

and the equation, can provide good descriptions of
layer formation o n moderately uniform surfaces yet have difficulty when applied to
surfaces with a wide distribution of adsorptive potential.
A reason for the failure of previous methods was discussed in II.B.2. It can
be better seen by inspection of the set of model isotherms shown in Fig. 10. As a
reference, a n approximate value of is indicated a t If range of
heterogeneity of the sample surface is narrow, all surface patches can be sampled
adequately without danger of multilayer formation o n the higher energy portions.
We can see, however, that if a significant part of the surface were characterized by a
maximum of of 70, one would need to obtain data a t pressures no
greater than. say, 1 torr. At this pressure, patches with 50 would barely
be contributing information to the total isotherm, and areas with 70 would
be building multilayers. thus making a monolayer kernel function invalid. It is also
clear from Fig. 10 that a multilayer model that incorrectly predicts the pressure
range for the formation of a second layer relative to the will also fail to
yield a good fit. The unmodified nonlocal DFT is such a n example
The modifications to the D F T calculations that bring about the agreement seen
in Fig. 2, while empirical in magnitude, have a plausible theoretical basis. F o r the
present, the modifications assume that the surface of Sterling deviates
from uniformity. T o the extent that this is not true, we must point
that the parameter distributions obtained through Eq. (9) and reported here should
be viewed as relative to that graphite surface.
The patchwise model underlying the integral equation of adsorption
that each patch adsorbs independently of others- that is, there are no "edge
effectsu- and the local isotherm function is assumed to account for all lateral, o r
adsorbate- adsorbate, interaction. If the patches are large, this assumption is valid.
However, if adjacent patches are small and have quite different adsorptive poten-
tials, then their occupancy a t a given pressure will be different and also many
adsorbed molecules will be near patch edges. A molecule at the boundary between
patches of high and low occupancy will experience a kind of average lateral inter-
action different from that given by its local isotherm function, hence not be truly

0.0001 0.001 0.1 1 100
Pressure
Representative isotherms for nitrogen on surfaces cliaracterized by the
indicated values of
independent of adsorptive energy is distributed. This limitation of the patch-

model is so remains as to it work so well
reproducing data. A tentative hypothesis is that distributions of real
but to be correlated to the extent that
patches differ but little in energy greatly edge effects.
Ross and On
123.
2. Langmuir. 38: 2267 (1916); Phys. Rev.
de Boer, The Press,

Oxford,
Fowler A. Statistical Cambridge
Press, Cambridge, 1949, 431.
Sanford and Ross,
Ross and Olivier,
7. P. Olivier and Ross, Roy. (Lond.) (1962).

Hill, Phys.
URSS
A. Rend. Acad. URSS 194 (1944): Bull, Acad,
(1945); URSS (1947).
and Taylor. Phys.
R. J. Phys. (1950).
and
A. E. Trans. (1951); L.
Drain and A.
Rudzinski and Everett, of
London,
Ross and Surface (1975).
Morrison and Ross. Surface (1973).
and Morrison. Colloid Interface (1979).
Porous Mates,
in tire
1996,
A. R. Walton, and Quirke, (1989).
and (1993).
and Quirke, (1993).
Conklin. and in Science
87, of Porous Solids et
Tarazona, A (1985);
and Mo!. (1987).
and Solving

Department of Chemistry, The Pennsylvania State
University, University Park,
I.Introduction
II.Algorithms
III.Modeling the Interactions
IV. Specific Examples
A. Elementary systems
B. Lennard-Jones atoms o n various surfaces
C. A d s o r p t i o n the critical region of the bulk fluid
D. Freezing of adsorbed layers
E. Energies of adsorption
F. Wetting and drying
V. Discussion
References
Beginning about 40 years allow to use computers to

simulate thermodynamic, and dynamic properties of bulk molecular
began to appear in the literature. The early based on these
algorithms limited to the possible model fluids because of
imposed the crude computers available at that time. hard sphere
properties were the first to be evaluated using this Both and
investigated. as these systems did not exist in
reality, the results of these studies a cornerstone of the and
testing of theories of the hard sphere and disk fluids. As computers increased
in and availability, complexity of the systems investigated increased
rapidly. At the time, variety of the algorithms developed to

various increased from Metropolis Monte
a canonical a
Markov of states with appropriate probability rules. and the first
dynamics (MD) which simulated the ensemble by
solving equations of motion for the particles that made up ensemble,
In early work and in nearly all subsequent studies, the of a
that treats only a few molecules in a box of equal to a
thousand cubic nanometers required that one eliminate the effects of the box
fluid properties and also find a to calculate the interactions
between a and the surrounding when molecule question is
near the box so is lacking neighbors on one or more sides. The techniques
of doing begin periodic boundary conditions which the box is
of itself and the in The walls of the box are
permeable to tile passage of in (and out) by any trace
of interaction between a and the box Because of the surrounding
passing out through a will be by an
molecule passing through the opposite thus conserving the of
molecules and finite-size effects. At the same the interactions
of a molecule its are evaluated by the (usually pairwise)
its real and its to some cutoff distance
chosen to precision minimizing the time required to perform
Both of these have been described in detail in many review articles
and books Their success relies the decay of spatial correla-
tions over a distance is significantly than the box dimension. For fluids
containing several this is usually not a problem (in three dimen-
sions. the number is often taken; in systems. 100 is
but of course the long-range spatial correlations present
solid phases can give rise to problems in the of the periodic boundary and
image
of these to the simulation of adsorbed phases began
to appear the 1960s. great niajority of are based on the idea that
adsorbent be treated as a rigid, inert field of force that an
fluid when adsorbate molecules are its vicinity. the
success of such depends on the quality of the model to represent
the adsorbate-adsorbent interactions, are
both molecular dynamics and algorithms have been

in of physical adsorption, this review is devoted primarily to the Monte
since these provide better to the potential
gives the adsorption isotherm), other thermodynamic structural proper-
ties. brief of sonie of the Monte
siniulation algorithms is given, In typical from are
presented. Conclusions are given in V.
It first be noted the introduction of a rigid solid adsorbent clearly
alters periodic boundary conditions for the fluid to be If the adsor-
bent is a planar free surface, one has periodic boundary conditions and
interaction in only the two dimensions parallel to the

uter Simulations
solid In order to confine the in the third dimension, in a

region near surface, it is conventional to introduce a hard that is parallel to
the solid surface and at a rather large distance from it-large that changes
in structure density) due to its presence be negligible. In practice, this
amounts to a separation of 5-10 molecular (of the adsorbed
between the adsorbing the hard boundary. course, this is
only encountered in sorption studies. particular, sorption in porous
solids is presently an active reseach area. If the pore is cylindrical, it is obvious that
the only periodic condition be the direction parallel to the pore axis, as
also be the case other pores with a closed cross section (triangular, elliptical, or
rectangular, for instance). frequently studied case of pores parallel planar
will be periodic in two dimensions and can be considered as an alternative to
the free surface-hard mentioned above, with the primary differences
being that two walls are usually assumed to be identical in their interactions
sorbed fluid that the properties of such fluids are frequently studied as
a function of the wall separation. Finally, the case of sorption the interstices of a
packed or in a network of connected cavities such as
those in zeolites are modeled three-dimensional periodic conditions.
Once the gas-solid energy and the gas-gas (adsorbate-adsorbate) interaction
energies are one pick a suitable algorithm large number
that are available. If transport, particularly diffusion, is of interest,
is usually the best starting (Ti-anspost can also be eval-
uated from Monte simulations but this approach has the disadvantage
the elapsed is taken to be to the number of Monte
trials While this seems to be reliable, the constant of proportionality is
hard to estimate.) original microcanonical dynamics algorithm has
now been by a number of approaches based on constrained dynamics
that yield other ensembles For example, addition of constraint
total kinetic energy of all constant can be imposed on the
Newtonian equations of motion (which means that the total energy is no
longer a conserved quantity), The result is a mechanical system with constant
since temperature is proportional to the total kinetic energy of a
A variety of Monte are available and the choice is
determined by the quantities that one to In parti-
cular, an adsorption isotherm be viewed as a of the chemical
potential of the fluid as a function of the number of adsorbed molecules. This
interpretation is based on the fact that adsorbed phases are usually studied
they are in with the bulk that is generally the ideal adsorbate gas.
This that the chemical potentials and of the two phases are equal.
Since is equal to + R T for an ideal gas should be replaced by the
fugacity at high pressures where the bulk phase is nonideal), one has a straightfor-
ward method of relating the isotherm pressure to potential of
adsorbed phase. The constant can be calculated from molecular properties by
standard methods In this way, grand canonical Monte (GCMC) algo-
rithms in which the constant quantities are can be to the
adsorption which is now obtained as a function of where the
brackets denote the ensemble average of a fluctuating quantity. A second

322
dynamic property of interest is the energy of the adsorbed phase o r , more specifi-
cally, the molar energy of adsorption from the gas. If this gas is ideal, the
energy change will be entirely to the potential energy acquired by the adsorbed
molecules when they are transferred from the bulk gas to a location that is a t or
near the adsorbate surface. This energy is to the interactions of the adsorbate
molecules with each other and with the solid, and these two contributions can easily
be evaluated separately in the simulation (and summed to give the experimental
energy of adsorption). The molar energy of adsorption is related to the enthalpy
change by an additive factor However, it should be noted here that
these energy and enthalpy changes are not the partial molar quantities usually
reported in experimental studies. F o r example, the heat is the partial
molar enthalpy change upon adsorption. If and the molar and partial
molar energy, respectively, then is equal to + Thus, to evaluate
the partial molar quantities from the molar quantities, one must either simulate the
energy to a sufficiently high precision to allow one to differentiate it with
respect to A' or take the alternative route of evaluating the partial molar energy
from the appropriate fluctuation in the GCMC ensemble. Theory gives
where is the total potential energy of adsorbed molecules. Neither of these

calculations is easy to perform accurately.
Equation (1) is a n example of the evaluation of derivatives of thermodynamic
quantities from fluctuation expressions valid for particular ensembles (Ref. 12,
Chap. 3). Heat capacities can be obtained from fluctuations in the canonical
(most convenient) and grand canonical ensembles; isothermal compressibilities
(which are related to the slope of the adsorption isotherm) can be obtained from
number fluctuations.
Two types of thermodynamic invariance of the local properties of a n
geneous fluid should be mentioned at this point. These arise from the mechanical
and equilibrium of such a fluid. The pressure tensor is a controlling factor
in the mechanical equilibrium. F o r fluids, this tensor is diagonal with
elements and If the coordinate is taken to be perpendicular to the
adsorbing surface, one can average over the and coordinates to obtain the
in-plane pressure tensor element. Mechanical equilibrium requires that there be n o
gradients in these elements in the direction of the force of the pressure element.
Thus, for any must be zero for all in the fluid. A similar relation holds
for the in-plane pressure tensor element. Clearly, this implies that a local pressure
tensor must exist and that simulations of the local pressure tensor elements must
obey these requirements. Such calculations rest upon expressions for the local
pressure tensor elements in a small volume element located a t r in the fluid.
It has been shown that the of the pressure tensor are given by a sum of
three terms: the contribution due to molecules moving toward the solid (the ideal
gas term); the contribution due to the forces between pairs of adsor-
bate molecules, denoted here by 1 and 2; and the contribution due to the gas-solid

force o n a denoted here by The explicit expression for the qth diagonal
element of the tensor can be written as
where can denote or and and are the lower and upper limits of the
position variables that define the adsorbed layer. Ordinarily, these are taken to be
zero and for where 1 is the thickness of the adsorbed layer, and and for
and the in-plane size variables. The brackets indicate an average over the
configurations generated in the simulation. Note that the in leads
to an integral equation for the number density in the fluid (Ref. 7. 3.131).
Chemical equilibrium in an fluid requires that the local chemical
potential be independent of F o r a fluid in a canonical ensemble (either
molecular o r Monte one considers a particular configuration of
molecules. An additional molecule is inserted a t this molecule interacts with
the molecules and the solid initially there but does not affect the simulated
molecular configuration in any way. If the energy of this interaction is denoted
by N), where the added particle is at and the A' particles are at positions
denoted by it can be shown that
where is the density of the fluid a t point , The average is taken over many
computer-generated configurations of the adsorbed molecules in ensemble.
Insertion points can be chosen a t any point in the fluid. chemical potential
of the fluid must be independent of location, o r there could be a spontaneous
transfer of molecules from the with higher chemical potentials to those
with lower, which would mean that such systems would not be at the lowest global
chemical potential and thus would not be in thermodynamic equilibrium. This
method of simulating has been used successfully for bulk model fluids,
including nonspherical molecules where the coordinate includes as
well as translational variables. the literature, one can see definitions of quantities
such as the electrochemical potential for a fluid in a n external electric field or the
intrinsic pressure in which the contribution of the gas-solid force is omitted from
the definition. These quantities will depend on position, but one still has the
requirement that their values plus the contributions due to the external potential
must be constant.
The expressions for the pressure tensor a t a surface can be integrated to give the
solid-fluid surface tension, which we here denote as (also called the spreading
pressure):
where is the volume of adsorbed fluid. This integral can be evaluated analytically
to give

The first ensemble average on the right-hand side of (5) is taken over all
particles in the fluid, and the second is taken over all pairs of particles. Here, the
gas-solid energy depends on the position of a particle in the fluid. and the
gas energy depends on the separation distance between particles 1 and 2. It has
been suggested that the term involving the transverse derivatives of does not
contribute to after averaging but the universality of this argument is not
completely clear at present.
Note that (4) and (5) are valid only for a fluid adsorbed on a free surface
where = In the general expression for that is as valid for pore fluids as
for fluids on free surfaces, must be replaced for In the notation of Magda
et (4) and (5) are actually for a quantity denoted by that becomes
equal to only in the limit of large pore size. For a fluid in a slit pore of width
= and the disjoining pressure Note that for a slit
pore is called the solvation force because it is the force on unit area of pore wall due
to the presence of fluid between the walls.
ALGORITHMS
The canonical Monte method has been altered to give averages in any
of interest. The basic idea is to generate a chain of Markov states (mole-
cular configurations) for the fluid that will eventually produce configurations that
characterize the desired statistical ensemble. T o do this, one specify the tran-
sition matrix a that takes the system from state to state by making a trial
change in state and then testing to see if the change leads the toward one
with the desired properties for the ensemble that is to be simulated. This means that
of the trial changes will be rejected, and one of the goals of the algorithm is to
maximize their acceptance probability. It can be shown that this probability
is where is a quantity that depends on the parti-
cular ensemble being simulated. To accept a trial change, a random number
between 0 and 1 is generated. If the random is less than
the trial change is accepted (note the trial is always accepted if is negative),
and if it is greater. the change is rejected. Even for acceptance ratios on the order of
0.1 or more, a chain of changes can contain upward of trials, depending on the
accuracy desired.
To generate a canonical ensemble, all the trials consist in changes in a position
variable for molecules in the computer box, and 5 is the change in the energy of
the molecule that is moved due to its interactions with the other molecules in the
fluid and with the solid. The scale of these moves is adjusted to be small enough
that 8 is not always large and positive so that molecules do not overlap too
much in the new state) but large enough that the phase space of the system is
explored in a reasonable number of trials.
Ensembles other than canonical ones can be generated by adding to the types of
moves that define the Markov chain. In the case of particular interest here, the

grand canonical Monte simulation is based o n supplementing with
other moves that yield and for the creations and destructions that
change the number of molecules in the system. Specifically, is equal to
+ O
- (p - p ) and is equal to - + - - pO),
where N is here the number of molecules present in the initial state. Thus, there are
three different types of moves in the G C M C simulation. Ordinarily, these are
attempted with equal frequency because this gives the most rapid convergence
of the chain.
A number of alternative adsorption simulation methods have been proposed
and tested, including for other ensembles such as the isobaric. isother-
mal and alternatives to the G C M C . One such method is the Gibbs ensemble
F o r example, adsorption isotherms have been evaluated using the potential
distribution theory which is based on Eq. (3) for the local chemical poten-
tial. Simulations of the Gibbs ensemble (which in adsorption consists of
two systems, one a dilute gas and one the actual dense adsorbed fluid, that are in
equilibrium due to exchange of molecules between the two) have proved useful in
studies of adsorbed fluids Variants on this method include those in Refs. 27
and 28.
The gas-solid interaction laws explored in the numerous simulation studies of

physisorption vary considerably in their complexity and level of realism. One starts
from the simplest case, which is that of the hard wall, which can be planar, either as
a free surface or as the boundaries of a slit pore. Of course. other geometries such as
the straight-walled cylindrical pore can be studied. These systems are, of course, not
very realistic, but they are very valuable in helping one understand the effect of
confinement o n the properties of a fluid without the complications of a n attractive
interaction at the wall. In fact, the hard wall idea can be extended by the addition of
an attractive square well next to the hard wall that gives a relatively simple repre-
sentation of the adsorption process as it might be observed in real systems. Still,
both the hard wall and the hard wall plus square well models yield only the basic
principles of adsorption, and one must go to more realistic representations if com-
parisons with experimental data are the goal,
Modeling of gas-solid interactions for real solids and gases is still done mostly
at the semiempirical level. Much of this work has been reviewed in the 1990s
29,301 and is not discussed in detail here. It is based o n an idea that is widely
used in the of bulk molecular fluids. In it, the shape of a
molecule is accounted for by assuming that its interactions with surrounding
molecules can be expressed as the sums of spherically symmetric func-
tions for a collection of sites that are most often located a t the positions of the
atoms that make u p the molecule in question. Variations on this theme have been
explored-in particular, by expressing the interactions of the methyl o r methylene
groups in hydrocarbons in terms of single sites rather than C plus two or three H
sites. However, recent work o n the structures of the dense arrays of hydrocarbon
chains in many self-assembled monolayers indicates that the all-atom potentials
give results that are significantly closer to experiment than the simpler

approximation by chains of methyl and methylene groups (The differ-
ences appear mainly at high densities, where the chains are in intimate contact.)
In any case, the success of these site-site models for the modeling of bulk phases
leads one to the idea the gas-solid potential energy might be represented in a
similar fashion---namely, that the total interaction of a site in an adsorbate
molecule with the solid can be given by a sum of spherically symmetric interac-
tions with the sites associated with the atoms in a rigid adsorbent. Most often, the
potential for a pair of sites separated by a distance is taken to be a
Lennard-Jones inverse 12-6 function:
where and are the well depth and size parameters for the pair of sites
under consideration. It is the choice of values for these parameters that is the
most serious problem in devising a potential to represent a specific system such
as a gas-phase interacting with carbon black or silica (two of the most
popular adsorbents in the experimental literature), whose bulk phase properties
have often been simulated using this kind of model. The problem is that one has
cross interactions for the adsorption energy functions with potential parameters
for the sites in the solid that are unknown. Even if one were to guess these
parameters (for C-C or 0-0 interaction, for instance), approximate combining
rules such as the Lorentz-Berthelot rule are needed. More often than not. the
validity of the results obtained is tested by comparison of experimental and
calculated low-coverage limits of the isotherm slopes and heats of adsorption,
and, if necessary, the parameters are adjusted to give agreement. [Low-coverage
limiting data are used because theory shows that in this limit the calculations
involve only isolated molecules on the surface (see Ref. 15, At
first glance, the summations over the semi-infinite numbers of sites in the solid
would seem to be a sufficiently formidable task to prevent one from successfully
carrying out simulations based on such models. However, there are ameliorating
factors: In the first place, the Lennard-Jones functions are sufficiently
ranged to allow one to truncate the sums after the nearest 100 (or so) sites
have been included; for higher precision, the long-range parts of these sums can
be replaced by integrals. Also, the atomic arrangements at planar surfaces of
crystalline materials that are made up of a single Miller plane are periodic and
thus will give site-site gas-solid interactions that have the same periodicity as
the exposed crystal face. This can be exploited to give useful expressions for the
interaction of a gas molecule site over a surface unit cell 2.2; The
two-dimensional vectors that define the size and shape of this unit cell can be
defined as and with reciprocal lattice vectors and such that
. = 1, . = 1 and 0, . = 0. If the vector connecting the
gas-phase atomic site and the center of the unit cell in the layer of solid
atoms is denoted by where denotes the orientation of the in-plane
vector one can show that

where denotes the site-surface interaction with the plane of solid atoms, a, is
the unit cell area, and
In this expression, is the angle between the vectors and the K,, are modified
functions of the second kind, and where and are
integers. Here we have given a simplified version of the theory for the case where
there is only one atom per surface unit cell and the solid atoms are all of the same
kind. Extensions to deal with these problems are straightforward but give rather
complicated expressions. T o obtain the total energy of interaction, these energies
should be summed over all values of a , but the sums converge very quickly and
convenient approximations to their values can be derived. A convenient approx-
imation to the value of the term with g 0 in (7) can be written
where is the distance between layers.

The terms in the expression for with not equal to zero represent the periodic
part of the gas-solid energy for a single crystal adsorbing surface, and the inverse
10 4 term in Eq. (9) gives the surface-averaged "smooth" interaction, which is
clearly independent of the position relative to the surface lattice of atoms.
Although the form of (8) tends to obscure its physical significance, it is easy
to show that a surface lattice whose unit cells are large compared to the size of the
gas atom site will have a large periodic component, especially when the site is close
to the surface.
Note that these results have been obtained for the inverse power law
Jones potential. If electrostatic interactions are present, expressions of this
that give the electrostatic energy in the form of a Fourier series are known
Perhaps the easiest way to see this is to represent the dipole, quadrupole, etc. of a
gas molecule by discrete charges attached to various sites in the molecule. the
charges having magnitudes chosen to correctly give the known
The electrostatic energy is then written as a sum of interactions over
charged species in the solid and the charge sites in the gas molecule. F o r
Coulomb interaction depending upon ( d i s t a n c e ) ' , the index of K is and,
since = one can derive closed-form expressions for the
electrostatic part of the interaction for polar molecules adsorbed o n charged
ionic) solids.
F o r other types of surfaces, one must either adapt the methods that lead to the
energy expressions shown here or, if all else fails, d o a brute force summation over
sites. Evidently, models of the interaction in porous solids require that one rethink

328 teele
the ideas presented here, since few if any porous materials exhibit internal surfaces
u p of single crystal planes.
Besides its accessibility for experimental work involving large-area adsorbents,
one of the reasons for the great popularity of materials whose adsorbing surface is
made u p of the graphite basal plane is the great simplicity of their surface. This
popularity extends not just to the free surface but also to porous materials such as
parallel-walled slits and even to the buckytubes that are presently under intense
scrutiny. the first the atomic spacing in the graphite basal plane
is so compared to adsorbate site sizes that the periodic component of the
solid energy can be completely neglected. If one keeps only the integrated site-site
energies that lead to the inverse 10-4 power part of the gas--solid energy for the
infinite plane, interesting modifications are available; in particular, it is possible to
derive a n analytic expression for the interaction of a gas a t o m with a truncated
graphite basal plane that is useful in a number of problems F o r example,
stepped graphite surface can be modeled pores that are more realistic than the
parallel-walled case can be treated, including those rectangular o r triangular
cross sections Although the grooved graphite surface that uses truncated
planes to make the grooves in the outermost surface layers might be considered
be of minor interest, one could take two such surfaces facing each other to make
parallel-walled pores with alternating narrow and wide regions with constric-
tions).
Finally, the planar graphite surface can be rolled into a cylinder to model a
The integrations that led to a 10-4 function for the infinite plane are
difficult in case, but a n analytic expression for the site-wall interaction in
a smooth cylindrical tube can indeed be derived for use in the simulations of
adsorption in this material
Well-known and widely used adsorbents that are amorphous include many
preparations of high-area oxides such as silica and titania. Models of the
adsorbing surfaces of these materials have been proposed that are based on
the ideas that first, the cations can be omitted for the site-site gas-solid poten-
tials, at least for nonpolar adsorbates, and second, that the arrangements of the
oxides is essentially that of a random close-packed array of hard spheres. An
algorithm is available that will generate such arrays with periodic boundary
conditions in two directions. If the hard spheres are then transformed into
sites with interaction that represent the oxide-gas site potentials,
the nonperiodic surface can be taken to be the adsorbing plane. In this case, the
brute force approach of summing over these oxides can be used to evaluate the
gas-solid interaction for such amorphous materials. The results of this modeling
have been analyzed in a number of papers (References to earlier work
are given in Ref. 40.) In addition, variants o n this procedure have been
explored: One can roughen the surface by randomly deleting atoms from the
layer o r o r pores can be created by deleting the oxides from a
of appropriate size and shape in the computer-generated random solid
This allows one to model pores of arbitrary shape and even pore net-
works of varying connectivity (if a computer of sufficient power is available to
handle such large systems).

The adsorption systems simulated to date range from elementary examples to
that are rather complex. We begin with cases that d o not exist in nature but that
exhibit features that may be encountered in other. physically realistic systems.
These solid-fluid systems involve hard interactions, either gas-solid o r gas-gas o r
both.
By "hard interactions", one means interaction energies that jump from zero to
infinity at a specific separation distance or where gg denotes gas-gas and
gas-solid. Obviously, such systems have n o interaction energies that could give
energies of adsorption that differ zero, regardless of the geometry of the
confining wall. Nevertheless, the presence of the wall o r walls produces profound
changes in the properties of a fluid of hard spheres in the vicinity of the walls.
Adsorption occurs with a n adsorbed amount denoted by the excess number of
(relative to the bulk) per unit area of surface when the pressure of the bulk
fluid is This is always greater than zero, which means that the hard sphere fluid
always wets the wall in these A t first this appear to be a surprising
result. because there are n o attractive interactions between the wall and the fluid,
but of course these interactions are absent in the adsorbed fluid as well. The number
of atoms adsorbed per unit area is given by the general expression
where denotes the density of atoms per unit volume and adsorption on a smooth
planar surface is assumed. are needed to evaluate the number
density in the fluid at a distance from the wall. Figure 1 shows the results of such a
simulation for a high density hard sphere fluid in a large hard-wall box with
walls apart. In this case, the density of the fluid far from the walls (and
thus the bulk fluid density) is = 0.90. W e see in the bottom panel that is
essentially equal to the bulk density for 5, which means that this system
behaves as if it were adsorption o n a free surface.
Specializing to planar walls for the one has the exact relation that
= where is the local density of the adsorbate in contact with the wall
when the pressure of the hard sphere fluid is F o r hard sphere mixtures
is the sum of the individual densities for each of the components in the fluid. Thus,
the pressure of the fluid can be obtained from estimates of the intercept of the curve
of versus for example, Fig. indicates that is between 8 and 9 for
this system. This result, taken together with the calculation of a t a given from
allows one to construct the isotherm Figure 2 shows the adsorption
of a hard sphere fluid on a hard wall as a function of the bulk-phase density
The simulation points compare well with results of two theoretical calculations
based on the scaled particle theory.
One also has a n exact expression for the hard wall-hard sphere fluid surface
tension which is

FIG. 1 Lower panel: Local density in reduced units for the hard sphere-hard wall
system discussed in the text. Upper panel: Local pressure tensor component
perpendicular (constant line) and parallel (oscillating curve) to the surface. These
properties are plotted as a function of the distance from the surface in reduced
units. The bulk density for this is 0.90 in reduced units, which is close to the
freezing value. (From Ref. 44.)
where is the separation distance between a pair of hard spheres, is the pair
correlation function for a pair of fluid particles, both of which are in contact with
the wall, is the pair correlation function for atoms 1 and 2 in the fluid).
A second general route to the surface tension (or spreading pressure) has been given
in I as the integral of difference between the components of the pressure
tensor parallel and normal to the surface. The local values of both of these pressure
tensor components for a bulk-phase fluid density that is close to freezing are shown
in the top panel of Fig. 1. The perpendicular pressure tensor element is independent
of as required for mechanical equilibrium. The transverse component (which is
not uniquely defined even though its integral with respect to z is oscillates and
approaches at large as it should. A11 integral over these functions gives the
gas-solid surface tension [see Eq. The alternative route to this surface tension
via (1 requires a simulation of the pair correlation function for spheres in
contact with the hard wall. Figure 3 gives a between this function and
that for the bulk fluid, and Fig. 4 shows surface tensions as a function of the bulk
fluid density.
Finally, a number of pair correlation functions with = but with varying z
are shown in Fig. 5 and compared with the bulk pair correlation functions for

.2 Adsorption (molecules adsorbed per unit area) for the hard sphere-hard wall
plotted in reduced units as a function of the bulk density = The
points show the simulations, and the lines show theoretical calculations based on the
scaled particle theory. (From Ref. 44.)
. 3 The pair correlation function for hard spheres at a hard wall plotted as a
function of the reduced separation distance between the pair of spheres for a reduced
density of 0.90. The points show the simulations; the curve is for the bulk hard sphere
fluid at the same reduced density. Ref.

FIG. 4 The gas-solid tension in reduced for the hard
hard wall system as function of tlie reduced density = The points
are the two curves are calculated different scaled particle theory
expressions, Ref.
same density. (The density profile for this is shown in Fig. 6.) In general,
simulations indicate that sphere pair correlation functions near a hard
the two equal are to those for the bulk fluid. Figure 7
shows the dependence of the pressure of the hard sphere in a slit pore o n the
spacing for a fixed value of the density oscillatory
behavior of the pressure is to the layering the the
The minima occur easily a n integral
of layers. Maxima are produced integral of layers cannot be
readily Note the periodicity function is slightly larger
Other hard simulation studies include pair correlations

fluids in cylindrical pores planar spherical pores isotherms in
slit pores and slit and cylindrical pores as well as hard on hard
and slit pores (Reference 58 gives references to other such
studies.)
T o this point, discussed only the adsorbed hard sphere fluid, it is
known spheres undergo freezing both two and three dimensions.
One expects to find analogous behavior in adsorption especially the
fluid is confined to a monolayer o r several layers stratified). can occur
in very narrow hard-wall slit o r it can be produced by the presence of a
holding potential a t the (which we discuss below). In fact, there has
been great interest in of simple systems,

Pair plotted as a function of reduced separation
distance for the system of Fig. 6 . For these curves. each sphere is at the same
distance from the so The number above each curve indicates
of distance. The points are and the solid curves are pair
for the bulk sphere at the same seduced density
as that of the
stimulated by the (KTHNY) theory,

predicts that not be a first-order
but occur as continuous and
the second, the hexatic phase one short-range trans-
lational order but long-range bond order. (The bonds are vectors
neighboring atoms in the of these 2 D
have been carried out the effort to or the theoretical
predictions. These include interaction potentials,
12-6 repulsive inverse screened and hard disks
[References to previous studies are given in Ref. 59, references to
experiments and simulations are provided Ref, the of
all this is still not entirely clear, it is clear that if 2 D melting is first-order, it is
very weakly first-order.

as local density in Fig. 1 ,except that the reduced density is
and the is 1 Ref.
, to the plotted reduced units as function of

separation in reduced units The are values, and the curves are
based on different integral (From

Recently, two dynamics hexatic
phases observed for particular systems. one phase observed for
a particular concentration in two-component of
in the other the hexatic appeared for a n inverse 12th-power
at a of the pressure. It is unclear these special conditions
produce the hexatic so of these appear to be
necessary.
any case, one should expect freezing in hard sphere-hard systems that
are under discussion. and indeed this is case slit pores. Schoen et al.
observed stratification very thin films o r three layers) the
atoms their layer to pack more efficiently for certain values of
the spacing. In a detailed Schmidt and determined the
phases present thin sphere films as a function of the density
for separations starting at strictly 2 D system of zero separation
excess of varying u p to As can be seen in F i g a rich
phase obtained, starting from the packing expected for
2D system, followed by a "buckled" analogous to that deduced by et
for excess separations slightly greater than zero. As
tinues to increase. a two-layered square 10) surface] is observed, fol-
by a two-layered rhombic phase finally a two-layered triangular phase.
The layers of atoms the and two-layered phases are all dovetailed to
The phase diagram for hard spheres enclosed in a hard-walled slit excess
ranging a value one layer of (or solid) = 0) to
layers shown as a of the sphere Phase boundaries and
coexistence regions are indicated. nature of the ordered phases observed is discussed
in Ref. 72.)

obtain densities. The forbidden region in Fig. 8 covers the range
densities higher than those allowed for the various close-packed hard sphere
structures.
proceed to more realistic of adsorption systems. As a preliminary

step, comparative of various gases on various model surfaces should be
mentioned. These hard spheres at a soft repulsive wall and
soft repulsive spheres, Lennard-Jones atoms between hard, soft repul-
sive, and soft attractive For coverages greater than one monolayer. these
the density is relatively to the detailed
of the interactions is the repulsive cores of the adsorbed atoms that
are determining factor. This point is Fig. 9 .
studies of phase diagram for
have been reported In particular, liquid-vapor coexistence curve has
the critical temperature and density
3
1.316, = 0.304) (Fig. 10). These studies are relevant to
layer adsorption of on smooth strong holding potentials,
since the on surfaces form quasi-two-dimensional systems.
Theoretical analysis (Ref. 15, Chap. 4 ) reveals the effect of this gas-solid
can be taken into by replacing p by with Henry's
constant by
where A is area of the wall. Tlie remainder of the relies on the 2D

of state of the model gas, In the case of spheres on adsorbing
simulated 2D equation of state is fitted by a expression
given by scaled particle theory, When this result is used in the 2D approximation to
the adsorbed one finds
and This expression resembles the

equation, but the definitions of coverage are different in the two cases
and, of course, the factor, which will strongly affect the equation at
coverage, is missing from the Langmuir equation.
two-dimensional also applies to Lennard-Jones gases
adsorbed on surfaces such as those presented by the basal plane of gra-
phite. Thus, classic experiments of Thoiny and coex-
istence and critical points for several simple gases adsorbed on graphite that appear
to be in good theory.
A number of of the adsorption of classical rare
gases on graphite appeared. In cases, potential of Eq. (9) was used,
parameters to represent the graphite-gas system under investigation.

Local densities for (a) a Lennard-Jones fluid at = 0.76 on an
atomically structured wall; (b) a hard sphere fluid with 0.75 on a smooth
hard wall; a hard sphere fluid on a wall with a soft repulsive potential; (d) a hard
sphere fluid on a wall with gas-solid potential given by Eq. (9). In all cases, 1.
(From Ref. 74.)
Argon krypton and methane (in attempts to reproduce the

ordered monolayers commensurate with the graphite lattice that have been
observed experimentally) and mixtures of argon and atomic models of methane
and a larger atom were studied As a n example, simulated multilayer isotherms
are shown in Fig. 11 for argon o n graphite a t two temperatures. Analysis of the
configurations gave unambiguous values for the coverage in each of the adsorbed
layers (which were shown to be well stratified) and, with less precision, for the
isosteric heat of adsorption. The waviness in the low temperature simulated iso-
therm is a remnant of layer formation that can be seen more clearly in the
experimental data shown in the figure.

FIG. 10 The coexistence curve for a two-dimensional Lennard-Jones fluid, showing
the estimated critical point as the filled circle The points are simulation results, and
the solid lines are fits to the 2D scaling law and the rectilinear diameter law; the dashed
line is an earlier equation of state. (From Ref,
The gradual increase to very thick layers as approaches unity is normal

behavior for adsorption o n a free surface; however, when the adsorption occurs in a
pore the isotherm terminates a t a coverage corresponding to a liquid-filled
pore. Vertical jumps in coverage are seen (for adsorption in a single pore with a
simple geometry) that are associated with capillary condensation and evaporation,
and hysteresis is usually present in which the capillary condensation and evapora-
tion occur a t different pressures. Examples of this behavior are shown in Fig. 12 for
a simulation of argon in a slit pore with walls made u p of graphite basal planes
The isotherms in these systems are very similar to those with a single
adsorbing wall, and this generalization holds good down to wall spacings that are
small enough that the gas-solid potential fields overlap significantly. The two
panels in Fig. 12 show capillary condensation at pressures that are much smaller
than the bulk vapor pressures at the slit widths taken for the simulations, but only
the large slit width isotherm shows hysteresis. In both cases. the jumps in coverage
start at values slightly larger than the and end at nearly full pores. These
two values can be viewed as those for coexisting capillary vapor and liquid phases.
As the temperature is changed in a pore of fixed width and potential function, a
capillary coexistence curve is traced out that has a critical temperature that is less
than the bulk-phase value-how much less depends o n the slit width and the
solid wall potential. (If the holding potential is strong, it may be convenient to treat
the monolayer adsorption as separate from the pore fluid phase when it comes to
calculating the pore fluid densities.) T h e detailed nature of the hysteresis loop is still
not very well understood even after a good deal of theoretical and

. 11Grand canonical Monte isotherms are shown for argon on the graphite
basal plane modeled by the potential of Eq. (9). Surface coverage is given in terms of
the number of layers of gas on the surface, and denotes the bulk vapor pressure of the
Lennard-Jones argon at the two temperatures of the simulation. points are
for 0.668 and 1.002; the solid curve is an experimental
isotherm for argon on graphite at 78 (From Ref. 78.)
effort. The width of the loop clearly depends on temperature, pore geometry, and
the gas-solid interaction potential. This phenomenon is a manifestation of meta-
stable behavior. Densities a t the tops and bottoms of the hysteresis loops may be
considered as corresponding to points o n the temperature-dependent lines
for the pore fluid, and the equilibrium densities can be obtained by calculating the

grand potential (the free energy in the grand canonical ensemble) by thermody-
namic integration along the isotherms starting from a low density point (which is
known to be at equilibrium) Then most stable thermodynamic states are
those with the lowest grand potentials. A detailed theory has been given that fleshes
out this picture Coexistence diagrams, which depend on pore size. geometry,
and interaction potentials as well as temperature, are now beginning to appear
(Liquid-vapor, liquid-liquid, and layering transitions in pores are reviewed by
et in Ref. 87.) One such diagram is shown in Fig. 13 for a
Jones fluid argon in a slit pore with = 5.9 and a repulsive ninth power
gas-solid interaction function The capillary critical temperature is
0.99 for this pore width, compared to the three-dimensional Lennard-Jones value of
1.25. This particular system is because of the absence of attraction
between the pore fluid and the walls. There can be problems with this approach
when the pore fluid is strongly attracted t o the walls, as is shown in Fig. 14. Here
the low density fluid is not homogeneous but is rather two layers of dense
fluid that d o not fill the entire pore volume. is not what is expected for the
coexisting "gaseous" pore fluid even though it has been taken the
of an isotherm point that is very close to the coexistence line.
Another pair of isotherms are shown in Fig. 15. Here a Lennard-Jones
model of methane has been used in a simulation of sorption in a graphitic slit
pore. The two sets of curves in the figure are for a single pore width but for two
Coexisting liquid- and gas-phase densities of argon in a slit of width

= 5.9 are shown a s a function of temperature. For bulk argon,
151 = 0.008 . The capillary critical point is shown by the circle.
(From Ref. 89.)

FIG. 14 Side views of the configurations of the adsorbed atoms in the system of Fig.
12a. Both panels are for pore fluid at = 0,516. which is in the hysteresis region of
the of Fig. 12: thus, panel (a) shows a configuration for the low density "gas"
in the pore, and panel shows a configuration for a pore that is nearly full of liquidlike
fluid. Note that the "gas" state is actually made up of layers on each wall.
(From Ref. 84.)
. 15 Simulated isotherms for a spherical Lennard-Jones methane model in a

graphite slit pore of width 5.0. (a, b) 0.5; (c, d) = 0.7. The
isotherms are given as the total number of pore per unit pore volume [in
reduced units obtained by multiplying by in (a) and (c) and by the number of
per unit area (in reduced units) in a given layer, in (b) and
layer 2. The isotherm data are plotted as a function of the configurational
chemical potential in reduced units [equal to a constant plus (From
Ref. 90.)

temperatures, both of which are low enough to show freezing. This becomes visible
when one takes the overall isotherm data and splits them into separate first and
second layer isotherms, as shown in parts and (d). With one exception, the small
jumps in coverage are due to freezing, while the large jumps are due to monolayer
and second layer condensation. The exception is the second small jump in the first
layer isotherm a t = 0.5. This is probably a consequence of the large change
in density of the second layer. which then provides a potential field for the first layer
atoms that allows to become slightly more dense.
Other simulations of the sorption of Lennard-Jones fluids in parallel-walled
graphitic slits include methane argon and nitrogen and as
well as methane- ethane mixtures and argon and Lennard-Jones mixtures in
slits walls of rare gas atoms Simulations of Lennard-Jones fluids in
straight-walled cylindrical pores include argon and argon- krypton mixtures in a
pore (weakly interacting compared to graphite) All these
simulations have been carried out at temperatures near the bulk boiling points of
the adsorbates where the density of the bulk vapor in equilibrium with the adsor-
bate is quite small, and freezing of the adsorbate can occur the density of an
adsorbed layer becomes sufficiently high (and the temperature sufficiently low).
in the
If one measures o r simulates adsorption at temperatures near the critical region of
the bulk adsorbate, new features appear in the isotherms, for adsorption both on
the free surface and in pores At slightly temperatures
where bulk-phase condensation is absent, multilayer formation can be observed, as
is shown by the local densities in Fig. 16. This simulation was for a simple model of
ethylene a graphitic surface (chosen so that comparison with experiment
was facilitated) at a temperature roughly 3 % higher than the bulk critical value.
N o t surprisingly. the density for the are not sharp as they are at
lower temperatures, and these layers build u p gradually compared to the low
temperature behavior. In addition, the simulated isotherm in Fig. 17 shows the
characteristic cusp in coverage versus pressure that is observed The
position and height of this cusp in the excess amount adsorbed depend o n the
choice of dividing plane that defines the adsorption excess; its existence is believed
to be due to a slow increase in absolute amount adsorbed compared to the more
rapid increase in the density of the bulk fluid increasing pressure, thus yielding
adsorption excesses that decrease with increasing pressure. The analogous simula-
tion for a graphitic slit pore shows the same phenomena. but with quantitative
features that depend o n slit width. In a more recent simulation the critical
behavior of fluid adsorbed in a graphitic slit pore was carefully studied. Although
not in contradiction to the earlier simulations. this high precision study revealed
that the pore fluid density for a rather large wall separation = 20) did not
approach the bulk-phase density in the core region of the pore far from the
walls). The explanation suggested for this is based on the observation that the
compressibility for the confined fluid was not the same as that for the bulk for a
temperature very near probably due to the fact that the capillary critical
temperature is slightly different from that for the bulk even for the large slit

16 The local density (in reduced units) an adsorbed gas slightly above its
bulk critical temperature is shown here for bulk densities equal to (a) 0.0015; (b)
0.0405; (c) and (d) 0.1935. Multilayer formation is clearly indicated at but the
lowest density. (From Ref. 104.)
17 Simulated excess adsorption in atoms per unit area (in reduced units) as a
function of pressure = for an adsorbed gas slightly above its bulk critical
temperature. The points show the adsorption for two slightly different definitions of the
excess adsorption (From Ref.

width taken in this work. In other words, the core fluid is effectively different from
the bulk, with the slight difference significantly magnified by proximity to the
critical point. This work did produce an interpretation of the experimental obser-
vations of near critical behavior in a system analogous to that of the simulation
At low temperature and high density, an adsorbed fluid with attractive interactions
such as those for the Lennard-Jones potential will freeze. We have already dis-
cussed the ordered phases that have been simulated for hard sphere fluids in slit
pores with small wall spacings. The 2 D Lennard-Jones fluid also forms an ordered
solidlike phase at high density, so there is every reason to expect that such phases
will be formed from adsorbed fluid. Indeed, this is the case. Experimental and
simulation studies of this behavior have been reviewed elsewhere and
need not be repeated here. It is worthy of note that the periodic boundary condi-
tions used in simulations for fluids can give problems when a solidlike phase with
long-range order in the directions parallel to the surface is Unless the
computer box happens to match perfectly with the lattice of the adsorbed
phase, there will be difficulties in matching up the images with the real sample. A
possible way of avoiding this problem is to simulate in the isobaric ensemble (for
the pressure components parallel to the surface), since this causes the area of the
computer box to fluctuate in order to hold the parallel pressure constant.
Nevertheless, questions remain concerning the correct expression for this pressure
in the presence of an adsorption potential that varies with position parallel to the
surface. A more reliable but more expensive method is to simulate a large isolated
patch of material on the periodic surface Simulations of the effects of periodic
variations in the gas-solid potential energy on the freezing transition d o reveal
some interesting features. Many of these simulations have been done with mole-
cular dynamics because of the power of this algorithm in revealing structures even
at low temperatures where Monte simulation becomes difficult. However, we
have chosen not to cover molecular simulations in the present review on
the grounds that this will be done in the near future
Starting with an early paper in which it was shown that an argon
layer on a graphitic surface at was ordered while the system in a
considerably weaker inverse 9-3 potential was not (see also Ref. a study
of the argon-graphite at 77.5 K with and without the periodic terms in
the potential revealed the periodicity had only a minor effect on the incom-
mensurate monolayer. This is definitely not the case for krypton-graphite,
where a commensurate layer does in the presence of the periodic potential
It is clear that the difference arises from the different sizes
of the two adatoms. Krypton with = 3.60 A , = 4.04 is the distance
for the minimum in the gg pair potential) matches up rather well with the
mensurate lattice spacing of 4.26 A , whereas for Ar is 3.82 A. In fact, the solid
K r monolayer undergoes an transition driven by
thermal expansion a t a temperature that depends on the magnitude of the periodic
part of the potential. Simulations of atoms on surfaces made up of exposed

planes of fee crystal have also been reported for both the free surface and
a narrow slit pore Methane is similar in size to 3.81 A for methane)
and the phase transitions for this molecule o n graphite have been simulated
with results that are quite similar to experiment and to those obtained for the
graphite system. Ordering in 2 D systems with periodic potentials due to vicinal
crystal planes has also been studied These molecular dynamics simulations of
ordered will be discussed elsewhere
Another kind of information obtainable adsorption simulations is the average

energy of interaction of the adsorbed atoms with each other and with the solid
The sum of these two energies is essentially equal to the thermodynamic molar
energy of adsorption, but the coverage dependence of the two separate terms gives
a revealing view of the energetics of formation of a submonolayer (and more)
As an example, these energies are shown in Fig. 18 for krypton adsorbed a t 120 K
in a n atomically rough-walled cylindrical pore The decay in shown there
is due to the successive adsorption o n the areas of the surface with the strongest
interactions, followed by coverage of the areas with weaker and weaker interac-
tions. The linear increase in with coverage nearly cancels the decay
when the two are added to give the total. Experiment gives the isosteric heat,
which is equal to R T the partial molar energy. F o r the nearly linear varia-
tions in energy shown in the figure, the molar energies are related to the partial
Average energies for krypton adsorbed at 120 K in a rough-walled cylindrical

pore. The dependence of the and the wall (gs) energies on the
of atoms adsorbed. are shown separately. together with their sum. The K r
monolayer capacity under these conditions is about 130 atoms. (From Ref. 125.)

molar energies by fact that they are equal in the limit of zero coverage, but the
slopes of the partial molar quantities are double those of the molar quantities.
Significant conclusions can be drawn from these simulated curves. First,
submonolayer total energies of adsorption are nearly constant because of the com-
pensation of the changing and The frequent assumption that the signature
of heterogeneity is a decaying negative energy (or enthalpy) of adsorption is seen to
be incorrect in this case. Second, the linear change in (which is often found in
simulations) implies that a mean field (or van der Waals) theory may be adequate to
account for the interactions between neighboring atoms in the The
extension of these simulations into the multilayer regime is capable of producing
physical insights into the nature of the average gg interactions in thick films as well
as into the continuing decay in the part. Thus, this straightforward evaluation of
the separate parts of the total energy helps elucidate the nature of film formation at
a solid surface.
As the holding potential for a gas o n a solid surface o r in a pore becomes weaker
and weaker, a point is reached where the competition between adsorp-
tion and the condensation of the adsorbate gas into a bulk liquid (or solid) is no
longer won by the adsorption process. I n effect, this describes the changeover from
wetting to In this picture, drying is a special case of adsorption
and can occur when the potential energy in the bulk phase is sufficiently negative
to that in the to stabilize the bulk relative to the adsorbed
film. Since these two energies change differently with temperature and density. there
is the strong possibility that marginal systems will undergo transi-
tions along some line in the temperature-density plane (evidently, other thermo-
dynamic variables can also be used to characterize the transition). The nature of
this transition (first-order o r continuous) is a question of considerable interest. A
nonwetting film comprises a thin layer (usually. a fraction of a in
equilibrium with the bulk fluid, and the transition to the wetting film corresponds
to the formation of a layer of fluid o n the surface whose thickness
infinity as approaches Other transitions are possible such as thin
film (but not infinitely thick) at a less than which is called
A study of argon o n a weakly interacting surface (an inverse 9-3 potential
with parameters characteristic of solid gives the local density curves for
= 1.1 that are shown in Fig. 19. The for this system, given in the
inset of the figure, indicates only small adsorption u p to a value of the bulk density
of about 0.048 (in reduced units) in equilibrium with the adsorbed film. When the
bulk-phase density exceeds this, a pronounced change in the nature of the local
density of the adsorbed phase occurs that is reflected in the jump in the amount
adsorbed. Thick layers appear, and the density of the monolayer is much greater
than that found for the thin films. Such behavior corresponds to a
transition. In contrast to this, Fig. 20 shows that the isotherm and the local den-
sities for this system a t = 0.95 grow continuously until saturation is
reached. However, the slope of the isotherm at saturation is finite, so that the
coverage is also finite a t saturation, indicating partial wetting.

Local densities of argon at 1.1 on a solid surface. The inset
shows the adsorption isotherm, and the curves show the local densities for bulk-phase
densities of (starting from the lowest) 0.0133. 0.0287. 0.0357. 0.0457, 0.0487, 0.0502.
0.0537, 0.0656 (in reduced units = The vertical dashed line the
inset figure is the saturated fluid density. A distance scale change is indicated for several
of the density curves. (From Ref.
A number of such simulations have been reported, including the prewetting of

A r on solid wetting and drying o n a weakly interacting surface
u p of (molecular dynamics was employed
to obtain density profiles and surface tensions); wetting and drying in a square-well
fluid--square-well wall system that was simulated a t various temperatures,
densities, and well depths; and the wetting behavior of solutions In addition
to deriving a number of useful exact expressions, the wetting properties of the
square-well systems have been obtained for several well depths. It was found that
the transition from a partially wetted to a dry surface is sharp; a combination of
theory and simulation data indicates that it is in fact first-order. Much of this work
has relied o n evaluations of the solid-adsorbed fluid surface tension This quan-
tity varies with film thickness and allows one to calculate the solid--fluid contact

Same as Fig. 19 but for partial wetting at = 0.95. The local density
curves are for bulk densities of 0.0126, 0.0189. 0.0212, and 0.0226. (From Ref. 126.)
angle 0 from Young's equation: where and denote the

solid interfaced with the vapor and with the coexisting adsorbed fluid, respectively,
and denotes the bulk liquid and coexisting vapor phases. Since this contact angle
has a value of zero for the completely wetted surface and 180" for the nonwetting
case (wetted by vapor only), with partially wetted surfaces having values between
these limits, evaluations of the contact angle can provide an alternative route to the
determination of wetting behavior.
I n this review, we have tried to show three aspects of the computer simulations of
condensed matter:
The study of extremely simple systems, as exemplified by hard spheres. In this
case. the computational power needed is relatively small, and the results are of
primary interest as tests of theories. They are a t best in qualitative agreement
with experiment.

2. Studies based on the introduction of interaction potentials that are sufficiently
realistic to produce results comparable with experimental data. In the adsorp-
tion case, these have been rare gases adsorbed on the graphite basal plane. The
agreement with experiment, while of theoretical interest, primarily serves to
validate the model potentials used.
3 . Simulation outputs that give new physical insights not available otherwise. For
example, the adsorbate local densities near the surface give information at the
atomic level concerning the temperature and density of the distributions of the
atoms near the surface that help reveal the mechanism of film growth. Splitting
the adsorption energy into its gas-solid and gas-gas components is helpful, but
not as helpful as the histograms that give the distributions of these energies.
The number of simulation studies of fluids physically adsorbed on solid surfaces

and in model porous materials is too large for them to be reviewed here in their
entirety. As stated above, most molecular dynamics studies have been omitted here
on the grounds that these will be reviewed in the near future Unfortunately,
this excludes much of the work on frozen ordered) layers in and
multilayer films. We have also excluded simulations on heterogeneous surfaces.
These are normally modeled by introducing imperfections that are either geometric
(such as cracks. steps, or surfaces rather than perfectly planar crystal-
line surfaces) or chemical (where impurity atoms are added to an initially chemi-
cally pure adsorbing surface). In fact, such studies are in their initial stages, and a
great deal more work is needed before these systems are well understood. Many of
the of adsorption in porous materials, particularly in model zeolites,
have also been omitted, as have the studies of solutions in pores and of nonsphe-
rical molecules adsorbed on surfaces and in pores In this last
case, orientational ordering in monolayers and its changes with temperature and
density have recently been reviewed elsewhere Simulations of nonsphe-
rical molecules in pores are scarce.
Consequently, the systems discussed here have been mostly limited to
ponent atomic adsorbates on completely smooth or mildly corrugated surfaces such
as the graphite basal plane. The large amount of experimental work carried out
with the graphite basal plane as an adsorbent certainly warrants the theoretical and
simulation studies reported for this system, and it seems fair to say that this parti-
cular adsorption problem is now essentially solved. Indeed, when one looks back to
the 1960s when the pioneering experiments of the adsorption on atoms on the
graphite basal plane were first reported by and and et
al. it is clear that enormous progress has been achieved in the understanding of
fluid and solid adsorbed phases on simple surfaces and in simple pores. Much of
this progress was at least assisted by computer simulation, which uniquely
suited for the investigation of complex systems such as physical adsorption. There
certainly remain many more problems in this area that will benefit the insights
afforded by thoughtful simulation studies.

N. Metropolis, A. Rosenbluth, A . Teller, and Teller,
Chem. Phys. (1953).
Woods, in Physics of Simple Liquids ( H . V . S. Rowlinson,
and Ruslibrooke. North-Holland, Amsterdam, 1968, Chap. 5.
A. Phys. Rev. A (1964).
P. Alien and Tildesley. Computer Simulation o f Simple Liquids,
Clarendon Press, Oxford, UK, 1987.
Binder, Monte Methods in Statistical Physics. 2nd (Topics
Phys.. Vol. 7), Springer-Verlag, Berlin, 1986,
K. Binder, Applications of the Monte Method in Statistical Physics (Topics
Phys., Vol. Springer-Verlag, Berlin, 1986.
Nicholson and Parsonage. Computer Simulation and the Statistical
Mechanics of Adsorption, Academic. New York, 1982.
Evans and P. Morris, Phys. Rep. (1984).
S. Nose, J. Chem. Phys. 81: 51 1 (1984); Phys. (1986).
J. Evans and Chem. Phys. (1985).
Hoover, Phys. Rev. A (1986).
Hill, Statistical Mechanics, New York, 1958, Chap. 3.
and (1996).
J. and Molecular Theory of Oxford
Press, New York, 4.2, 1987.
A. Steele, The Interaction of Gases with Solid Surfaces, New York,
1974. Chap. 3.
Nijmeijer and J. Bruin, Phys. A. (1990).
M. P. and C. Bruin, Chem. Phys. (1996).
Magda. and Davis, Chem, Phys. (1985).
Finn and P. A. Monson, Chem. Phys. (1986).
A. Phys. (1987).
A. Z. Pangiatopoulos, and J. Tildesley, Phys.
(1988).
in Computer Simulations in Chemical Physics Allen and
Tildesley, Elsevier, Amsterdam, 1993.
and W. A . Steele, 51600(1989); Sim. (1990).
V. and A . Steele, J. Chem. Phys. (1988).
Finn and A. Monson, (1988).
Jiang and Phys. (1995).
Svensson and Woodward, Phys. (1994).
Diestler, and Phys. (1993).
A. Steele. Chem. Rev. (1993).
Vidali, Kim, and Cole, Surf. Rep. (1991).
and Rev. Phys. (1992).
and Jorgenson, J. Phys. (1996).
A. Steele. Surf. (1973).
Lennard-Jones and Dent. Trans. Faraday (1928).
and A. Steele, (1988).

M. Bojan and A. Steele, (1993); Phys., (1998).
Bojan, Cole. and A. Steele, Adsorption (1996).
Bojan and A. Steele. Carbon, in press.
Feke, and A. Mann, Phys.
(1988).
V. A. and A. Steele, (1992).
V. A. Bakaev and A. Steele. Langmuir (1992).
Bojan and A. Steele, in 5th on Fundamentals of
Adsorption. 1995 Kluwer, Amsterdam, p. 17.
Bojan and A. Steele, in Characterization of Porous Solids (T. Mays
and Elsevier, Amsterdam, 1998. p. 49.
Henderson and F. van Phys, (1984).
I. Fisher. Statistical Theory Liquids, Chicago Press, Chicago, 1967,
Chap. 6.
Rickayzen and A. (1984).
M. and Rickayzen. Phys. (1989).
Tan, Marconi, and Chem. Phys. (1989).
I. Snook and Henderson, Chem. Phys. (1978).
A. Jech. and R Smith, Phys. (1993).
M. and Henderson, Phys. Chem. (1984).
A. Antonchenko, V. Makovsky. and V. M. Phys.
(1988).
and van der Eerden, (1987).
Labik and W. Phys. (1996).
Alejandre. and Mol. Phys. (1996).
and V. Phys. (1996).
A. Yethiraj and Hall, Chem. Phys. (1989).
A. Yethiraj and Hall. Phys. (1991).
Kaplan, and Phys. Rev. Lett. (1995).
Abraham, Phys. Rev. (1981).
J. A. Barker, Henderson, and Abraham. (1981).
Allen, Frenkel, and Chem. Phys.
(1983).
Evans. Phys. Lett. (1982).
Schnitker, and Weeks, Phys. (1993).
and Schnitter, Phys.
Weber, Marx, and Binder, Phys. Rev. (1995).
A. and V. Chester. Phys. Rev. B (1992).
Alder and Wainwright, Phys. Rev. (1962).
Li, Johnson. and A , Goddard Phys. Rev. E (1996).
Anderson, and W. Swope, Phys. Rev. (1994).
M. Diestler, and Cushman. Chem. Phys. (1994).
Schmidt and Phys. Rev. E (1997).
F. Abraham and J. Chem. Phys. (1977).
I. Snook and van Megen, Chem. Phys. (1979).
(a) Smit, Chem. Phys. (1992); (b) Smit and Frenkel,
Phys. (1991).

A. and X. Phys. (1970); (1970).
A. X. and Rend. (1969).
A. Rowley, Nicholson, and G . Parsonage, Phys. (1976);
(1970).
Whitehouse. D. Nicholson, and Parsonage, Phys. (1983).
Nicholson, Parsonage, and A. Phys. (1981).
Jiang, and A. Phys. (1993).
Finn and P. A. Phys. (1991).
S. Gregg and Sing, Adsorption, Surface Area and Porosity, 2nd
Academic, New York, 1982.
P. Walton and Quirke, Sim. (1989).
Peterson and Gubbins, Phys, (1987).
Evans, Marconi, and Tarazona, Chem. Phys. (1986).
M . Sliwinska-Bratkowiak, and Suh, (1996).
A. Phys. (1987).
J. and Woodward, Phys. (1997).
Jiang, L. Rhykerd, and Gubbins, Mol. Phys. (1993).
Piotrovskaya and Brodskaya, Langmuir (1995).
Aukette, Quirke, S. and Carbon
(1992).
Nicholson, in Characterization of Porous Solids 11 Rodriguez-Reinoso, J.
S. Sing, and K. Unger, Amsterdam, 1991,
11.
Kaneko, and K. Gubbins, Langmuir (1997).
Nicholson, and Quirke, Phys. (1993).
Schoen, Rhykerd, and Diestler, Phys.
66:1 17 1 (1989).
M. Schoen, D. Diestler, and J. Phys. (1987).
E. Curry, J. Cushman, Schoen, and Diestler, Phys.
(1994).
Curry and Cushman, Phys. (1995).
Peterson, Walton, and Gubbins, Proceedings the 2nd
International Conference on Fundamentals of Adsorption (A.
Engineering Foundation, New York, 1987, p. 463.
Heffelfinger, van and Gubbins, Mol. Phys. (1987).
P. Ball and Evans, Phys. (1988).
Tan. van and Gubbins. Phys. (1987).
van Megen and I. Snook. Phys. (1982); (1982).
Vermesse and Chem. Phys. (1994).
van Megen and I. K. Snook, Mol. Phys. (1985).
Specovius and G . Findenegg, Phys. Chem. (1978).
Schoen and M. Thommes, Phys. Rev. (1995).
and G . M. Findenegg, Langmuir (1994).
Findenegg. and Lewandowski, Ber. Phys.
Chem. (1994).
M. G. Findenegg, and Schoen, Langmuir (1994).

Cole, and A. Steele. in Surface Properties of Layered
Materials Benedek. ed.), Amsterdam, 1991, p. 219.
A. Steele, (1996).
J. Landau. and Knak Phys. Rev. E 50: 2007
(1994).
A.Steele. in Molecular Dynamics (P. and Seminario,
Amsterdam, in press.
Lane and T. Spurling. (1976).
E. Lane and T . Aust. (1978); (1981).
Nicholson and G. Parsonage. Faraday Trans. 2
(1986).
F . F. Abraham. Adv. Phys. (1986).
Abraham, Chem. (1978).
J. Knight and P. A. Chem. Phys. (1986).
A. Patrykiejew, Zeintarski, and Binder. Chem. Phys.
and Mol. Phys. (1993).

A. Patrykiejew. in Adsorption on and Modified Inorganic A.
Dabrowski and A. eds. New York, 1996, 599.
Bojan and A. Steele, Adv. Colloid Interface (1998) Vol.
76-77 p. 153.
Sokolowski and Phys. Rev. A (1990).
Finn and A. Phys. Rev. A (1989).
Fan and P. A. Chem. Phys. (1993).
P. and Henderson. Phys. (1991).
Sikkenk, 0. Indeku, van Leeuwen, and 0.
Rev. Lett. (1987).
Sikkenk, 0. van Leeuwen, 0. and A .
Bakker. Stat. Phys. (1988).
P. Bruin, A . Bakker, and J. van Leeuwen. Phys.
(1991).
van and R. Henderson, Chem. Faraday Trans. 2
(1986).
Y. Fan, Finn, and P. A. Fluid Phase (1992).
F. and Nicholson, Chem. Res.
(1995).
Marx and Weichert, Adv. Phys, (1996).

Department of
Theoretical Chemistry, Marie-Curie-Sklodowska University, Lublin, Poland
I. Introduction 355
I I . Theory 357
A. Principles of the absolute rate theory and of adsorption
equilibria 357
B. The condensation approximation approach to adsorption
equilibria and kinetics 359
C. Generalization of A R T for energetically heterogeneous solid
surfaces 364
D. Theories relating the rate of adsorption/desorption processes
to the chemical potential of adsorbed molecules 374
References 388
Most frequently equation is written in its differential form,
describing the dependence of the amount adsorbed on time Here, a and are
constants characteristic of the adsorption under investigation. Assuming
that at = 0, = 0, one arrives at the following integral form of the
equation:

which is also frequently used. However, the meaning of the equation
should be sought in its differential form, which was first used by and
in 1922 to describe the kinetics of oxidation of metals Later Roginsky
and Zeldovich were the first to report on the applicability of (1) to
describe chemisorption kinetics. As a result, (1) is also sometimes called the
Roginsky-Zeldovich equation. However, it was Elovich and coworkers who
later demonstrated the wide applicability of (1) and proposed a rationalization
for that empirical equation, and in the end the name of Elovich became attached to
the
That equation and its integral soon gained enormous popularity among
scientists investigating the kinetics of gas adsorption on solid surfaces. Correlations
of experimental data usually started with the application of that equation, and the
frequently noted deviations from "Elovich behavior" were the form of analysis in
dozens of published papers. Even reviews followed that general trend, discussing
the Elovich equations and "deviations" it.
Thus. in the introduction to a 1960 review, Low explained that "the Elovich
equation has been adopted for the present use in order to express, compare, and
contrast with precision the otherwise chaotic data of chemisorption kinetics."
That exhaustive review reflected a generally poor understanding of the phenom-
ena behind that equation. Ten years later an excellent monograph published by
Aharoni and [9] was oriented toward a theoretical study of the origin
of the Elovich equation. These authors pursued an idea proposed by Elovich in
1936, elaborated further in the 1940s by Temkin and later discussed by
in his 1949 monograph The proposal was that the Elovich equation
describes the kinetics of adsorption on solids where the adsorption equilibrium is
described by a isotherm. Two excellent reviews published by
and have appeared in the 1990s that shed more light on the
problem.
According to the latter authors the rationalization of the Elovich equation is
simple. It is based on the application of the absolute rate theory (ART) adsorp-
tion on an energetically heterogeneous surface with a rectangular distribution of
activation energies for adsorption. An implicit assumption is then made that the
adsorption process is at quasi-equilibrium, that the process proceeds in a
fashion, and that the activation energy for adsorption increases linearly with sur-
face coverage.
Recently, and Aharoni presented a new development of the
equation. based the statistical rate theory of interfacial transport (SRTIT) pub-
lished by Ward and coworkers The new development is more rigorous in a
thermodynamic sense and shows that the Elovich equation neglects the kinetics of
simultaneous desorption.
A neglect of desorption kinetics may be the source of the deviations from the
behavior" observed in real adsorption systems. The new theoretical
approach based on SRTIT also allows us to develop in a easy way a more general
form of the Elovich equation, taking into account desorption kinetics. The form
derived can now be tested against previously deviant adsorption kinetics
data.

We start with a brief review of the approaches used to describe adsorption and
desorption kinetics. First we consider the oldest, the still popular A R T approach.
According to A R T , in order for a to take part in a reaction it must first
form a n activated complex. This can be accomplished by an input of energy-the
"activation energy." It is assumed that molecules in the activated state are in
equilibrium with the reactants and that the rate of reaction is the rate a t which
molecules pass through the activated state. In the case of adsorption, the reactant is
a gas-phase molecule, and the activation energy for adsorption is the energy
difference between that of the gas-phase molecule and the energy of molecules in
the "activated" o r "transition" state. The most exhaustive and clearest develop-
ment of the A R T expression for the rate of adsorption & is probably the one given
by Clark
where is the rate of collision of gas molecules of mass with the

surface, is the pressure, and T are the constant and absolute tem-
perature, respectively, and is the fraction of surface available for adsorbing
molecules. The product is the rate at which molecules with
sufficient energy strike vacant adsorption sites, and a, the condensation coefficient,
is the fraction of molecules meeting this condition that actually adsorb.
F o r desorption, it is assumed that the desorbing molecule pass through an
activated state before desorbing. The activation energy for desorption, is defined
as the difference in energy between the adsorbed molecules and those in the acti-
vated state for desorption, and it is assumed that the adsorbed molecules are in
equilibrium with the molecules in the activated state. The absolute rate theory
offers the following expression for the desorption rate,
where is a constant and G(9) is another unspecified function of the surface cover-
age 0. The change of the surface coverage with time is thus given by
The absolute rate theory does not lead to explicit expressions for the functions
and G(0) and does not suggest that might depend o n the surface
coverage
First a coverage dependence of the preexponential factors was introduced o n a n
empirical basis, leading to the following expression for

where is the number of adsorption sites involved in an elementary adsorption--
desorption process. In the case of = 1 , when one molecule occupies one
adsorption site. at when 0, ( 4 ) yields the iso-
therm equation,
where K = and = and where the superscript refers to

brium.
The strategy of experimental studies in adsorption has long hindered progress in
understanding the nature of processes. This is because
experimental studies have proceeded along three quite separate paths with only a
minor interchange of ideas.
One group working on adsorption kinetics soon discovered the inapplicability of
A R T in describing the kinetics of adsorption in real adsorption systems. Instead,
the empirical Elovich equation was used, aided sometimes by other empirical
expressions. The Elovich equation was justified in using A R T assuming that the
activation energy for adsorption changes linearly with the surface coverage. At
the same it was assumed that the constant does not depend on the surface
coverage. The latter assumption is to be emphasized for reasons that will soon be
explained.
The second group investigating desorption kinetics also discovered the inapplic-
ability of the ART expression for correlating experimental data. Experiments on
isothermal desorption kinetics were rarely reported, and almost all experimental
studies were oriented toward the kinetics of thermodesorption. This was stimulated
by the development of the temperature-programmed desorption ( T P D ) technique,
which is easy to use The inapplicability of the A R T expression to describe
T P D experiment was solved by the assumption that the activation energy for
desorption also varies with surface coverage Also, the assumption had to
be made that the desorption constant varies with surface coverage. The neces-
sary procedures of deducing the functions and were reviewed by
Zhdanov in his intriguing review, which collected a large variety of results
from the literature. showed that the and functions obtained
by applying different assumptions were different as a rule rather than as an excep-
tion: varied with 0 by several orders of magnitude. This, as we will see, raises
serious questions concerning the applicability of the A R T approach. T o show this,
we first have to consider studies of adsorption equilibria.
The third group of scientists reported on the inapplicability of the Langmuir
equation for describing adsorption equilibria in real adsorption systems, starting as
early as the 1920s. Instead, empirical isotherm equations such as the
or Langmuir-Freundlich equations were widely used
As a matter of fact. it was Langmuir himself who first emphasized that there
may exist various sites solid surfaces that are characterized by various values of
the adsorption energy This idea was developed further by who
argued that in most cases the spectrum of adsorption energies will be so dense that
one can use a continuous function to describe the differenial distribution of the

number of sites with corresponding values of adsorption energy. T h a t function is
usually used in its form normalized to unity,
where is the physical domain of the adsorption energies found in a given

solid system. A t the same time it was assumed that the variation of the
constant with surface coverage should be so small that it can safely be
neglected. The experimentally monitored surface coverage was represented
by the integral
where T) is a n isotherm equation describing the adsorption on sites char-

acterized by adsorption energy &. Most commonly the Langmuir isotherm was used
to represent the "local" adsorption isotherm In the interest of sim-
plicity we first consider this isotherm equation, neglecting lateral interactions
between adsorbed molecules. Application of the Langmuir equation, along with
a Gaussian-like adsorption energy distribution in (9) leads to the result
that T) takes the form of the well-known Langrnuir-Freundlich empirical
isotherm equation. That isotherm has been successfully applied to describe experi-
mental isotherms in a variety of gas-solid systems In systems where strong
surface heterogeneity is assumed to exist. the rectangular function
for 5
elsewhere
is commonly applied to represent the dispersion of adsorption energies. Its applica-
tion in Eq. (9) yields a n expression for similar to the well-known
There is one approach in the theories of adsorption that is extremely useful in

considering the effects of solid surface energetic heterogeneity. It has been used
mainly to study adsorption equilibria, but it can also be applied to the study of
adsorption kinetics, as we will soon show. F o r that reason we repeat briefly the
principles of that approach, called the (condensation approximation)
approach.
We rewrite (7) in the form
where

Figure 1 shows the temperature dependence of the function
where = and = are appropriate values.

Figure 1 shows that the kernel becomes a step function in the hypothetical
limit 0,
This means that at moderate temperatures, adsorption on a hetero-

geneous solid surface proceeds in a fashion.
I n the limit 0 the kernel in Eq. ( 9 ) becomes the step function so
takes the simple form
where is the integral of
and the integration constant is chosen in such a way that = 0. T h e assump-

tion that the limits are (0, is commonly made in adsorption literature for the
purpose of mathematical convenience. It does not affect the calculations until
extremely low 0) o r extremely high (6, 1) coverages are considered
.1Temperature dependence of the function in (13). The

temperatures are 0.5 =4 - and = 8 -

Calculating in this way has been known for a long time in theories of adsorp-
tion equilibria and is called condensation approximation.
A t finite temperatures, is not exactly a step function, but it is a frequent
practice to replace the true kernel 9 under the integral of (9) by a step function
located at = defined by the condition
Thus, is the value of a t which the isotherm derivative reaches its max-
imum. Figure 2 shows the derivatives of the isotherms drawn in Fig. 1 .
Surprisingly, the above-described practice leads to pretty exact values of 9, for
adsorption systems (adsorption regimes) reported in the literature. The rea-
son for that is that appears to be the leading of a certain expansion for
a t finite temperatures F o r a better understanding of the forthcoming consid-
erations and the approximation made, we repeat briefly the principles of that
expansion.
The integral of (9) is evaluated by parts:
It can be shown that the first term on the right-hand side of (18) is negligible
compared to the second except for very small and very high values of 9,
Therefore the second term is evaluated by expanding into its Taylor series
around the point = at which reaches its sharp maximum,
The derivatives of the function in (13) for the three dimensionless

temperatures = 0.5 = 4 (- - - ), and = 8 (- -

Thus
The first term on the right-hand side of (20) is just the result obtained by
applying the condensation approximation, assuming the hypothetical limit
0. Because of the symmetry of the function the second term on the
right disappears, and the first nonvanishing correction to is the second
correction term. F o r the model of adsorption, to a good approximation
is not the only function that may appear in the theoretical considerations of
adsorption o n heterogeneous solid surfaces. However, one assumes that
This is equivalent to accepting the accuracy level corresponding to the condensa-

tion approximation. As in the case of calculating will always be a leading
term, thus usually approximating pretty well the exact value of the integral. It can
also be that at all finite temperatures yields a fairly exact value of
integral (8) if the variance of is reasonably larger than the variance of
The function calculated from the relation
as, cis,'
is usually called the condensation approximation for the actual adsorption energy
distribution and is defined as
In the limit 0, the derivative becomes the Dirac delta function

and becomes the exact function

As we have mentioned, the Elovich equation for adsorption kinetics is asso-
ciated with the systems in which the equilibrium adsorption isotherm is
described by equation
where and are constants. Temkin's equation is obtained accepting the

equation (15) along with the rectangular adsorption energy distribution. (10).
The expression for T) then takes the explicit
F o r the rectangular adsorption energy distribution, becomes applicable

when - >> 1 . In this limit the exact result of the integration in (9)
becomes the equation, (26). T o show this, let consider the exact
result of integration in Eq. (9) when is the rectangular adsorption energy
distribution, (10). Then the expression for T) reads
F o r physical reasons. when 1 , is expected to take relatively small

values, because cannot be unreasonably high. Then, except for very high pres-
sures is small compared to unity, whereas is
expected to be much larger than unity. Under such conditions (27) reduces to
isotherm, Eq. (26).
W e now note the following important fact: The assumption that only is dif-
ferent for different adsorption sites is sufficient to describe the behavior of adsorp-
tion equilibria. We not assume that K must also vary site to site; it
must also depend o n the surface coverage
On the basis of the ART approach, K = and = On the other
hand, the Elovich equation for adsorption kinetics was obtained by assuming that
only changes with surface coverage whereas is constant. This means that
= should be the same for all surface coverages, contrary to what is
found in the analysis of thermodesorption experiments This inconsistency
can be shown by carrying out a n analysis of data obtained in studies of kinetics of
adsorption, of the kinetics of and of adsorption equilibria at
same time. Such comparative studies are, unfortunately, extremely rarely
reported in the adsorption literature
We point out another difficulty. In trying to find an interpretation for the
physical meaning of the activation energy of desorption. it is frequent practice
to identify it with the heat of adsorption corresponding to the same
surface coverage. a t equilibrium conditions.
F o r the model of adsorption.

This is so only if the kinetics of adsorption is "nonactivated," = 0, and the
first term on the right-hand side of (29) is negligible. Meanwhile the Elovich
equation that exists and is a linear function of surface coverage.
These perceived difficulties in explaining the observed kinetics of adsorption/
desorption in terms of ART have led to attempts to modify its fundamentals.
In 1957 Kisliuk launched his idea of "precursor states." He assumed that
before being adsorbed form a weakly bound precursor phase. Molecules
that do not strike an empty adsorption site but an already adsorbed molecule may
still have a chance to be adsorbed by entering the precursor state the adsorbed
molecule and then jumping to a neighboring empty site.
Twenty years later, in 1977 King made further efforts to improve that
approach by assuming that two states may exist: one over a filled adsorption site
and one over an empty site. Still another edition of that idea was proposed in 1978
by Gorte and Schmidt Other versions of the precursor state theory have been
proposed and their description can be found in a review by Lombardo and Bell
The concept of precursor states and the corresponding expression for the rate of
desorption did not gain the attention of researchers using TPD. Nor did it solve the
problem of the large variations in the preexponential factor with surface coverage.
Seebauer et reviewed the existing theoretical interpretations of
factors; none of was able to explain the large variations found in the
analysis of experimental data.
The procedures used to generalize the A R T expressions for the case of energetically
heterogeneous solid surfaces are discussed in the excellent review by
The rate expressions were written in the following For adsorption,
and for desorption,
[
exp --
Contrary to what is found in the analysis of TPD spectra based on ART, the
preexponential coefficients and were considered to be independent of surface
coverage and of adsorption energy Also, the assumption was made that the
activation energies for adsorption and desorption, and are linear functions
of the adsorption energy

On the basis of A R T approach, the relation must be obeyed.
However, let us accept the assumptions in (32) and investigate their consequences
for adsorption kinetics. Let us also assume that is the rectangular adsorption
energy distribution, Eq. (10). This means that when T) is the
isotherm, the experimentally observed ("total") isotherm T) will be
the Temkin isotherm, (26).
Now we are going to check whether we can use (30) to arrive at the
equation by assuming that in Eq. (30) is the rectangular adsorption
energy distribution as (10) leading to the Temkin isotherm equation for adsorp-
tion equilibria. For that purpose, in (30) has to be expressed in terms of
the overall adsorption coverage
Like other authors using (30) or we assume that adsorption is running
at quasi-equilibrium conditions, so the chemical potential of adsorbed molecules
is the same across the whole surface and is a function of 9,. This means that the
function appearing under the integral in Eq. (30) is
A convenient way to express using 9, is to express through the equilibrium

pressure = corresponding to the surface coverage In practice this
means that the isotherm under the integral in (30) is taken to be the
Langmuir isotherm, Eq. and is expressed by the value appearing
in (26).
We now rewrite (26) in the form
Consider now the rate of the experimentally observed adsorption when

is the rectangular adsorption energy distribution leading to the iso-
therm for adsorption equilibria. With including assumptions (32).
Simple analytical expressions for can be obtained for the cases
0. 1 . For the case = the solution was given in review
Here we consider the two cases = 0 and 7 = 1.
When 7,. = 0 , the adsorption is not "activated" at all. The integral (30) takes the
where is the value of the equilibrium pressure at which the surface coverage is
equal to This is the same pressure that appears in Eq. (34). So we replace
in (35) by its value calculated from

The Temkin equation, Eq. was obtained as the limiting form of the related
expression when 1. In the interest of consistency this limit has to be
accepted in Eq. (36). Then for 0, the exponent in the denominator is small
compared to unity, and for 1, the exponent in the numerator is greater
than unity. Thus, except for very small and very high coverages, in (36)
reduces to
which is the A R T expression in which 9 is replaced by 9,.

Now let us consider the other extreme case when the coupling parameter in
Eq. (32) is equal to unity, when adsorption is strongly "activated." Then
similar simple algebraic exercises lead us to
,
(it
Replacing
- 1
by the expression in
+
+ 1
we obtain
where a function of the surface coverage 6,:
+ T
= exp
Now when 0 , 0 or 0, 111 the - 1, reduces to
! -
Thus, in the limit of a strong surface heterogeneity leading to the Temkin equi-
librium isotherm, the of a strongly "activated" adsorption leads us to
the rate expression
which is essentially the A R T expression in which = and 9 is replaced by

Thus, the generalization of A R T based on Eq. (30) does not lead to the Elovich
equation as the kinetic equation corresponding to the Temkin isotherm if t)
is assumed to be the isotherm. (7).
However, the equation can be obtained in another way by using the
A R T and the condensation approximation. That method was first used in the
theoretical work of Roginski and reviewed later by and

Let us neglect the second term on the right-hand side of
assume that 1. and carry out the integration of the resulting first-order differ-
ential equation, to arrive a t the function
- 6) = constant
Because at = 0, 9 = 0, we have constant = 0. so that
Figures 3 and 4 show the function and its change temperature

When T 0 the function 0 tends to the step function
The value of is given by the condition
which for the function 9 expressed in Eq. (44) yields
In the CA limit T 0.
Now let us assume that is the rectangular function,
-10 -5 0 5 10 15
The behavior of the function defined in (44) for three values of the
parameters = = 0.1 - -), and =5 -).

4 The behavior of the function defined in (44) for three different
temperatures: = 50 K T = 200 K (- 300 K - -).
From (47)-(49) we obtain
Differentiation with respect to t yields
Equations (50) and (51) are the integral and differential form, respectively, of the
Elovich equation.
Note that when linear relationship (32) between and is assumed, Eqs. (50)
and (51) still preserve the form of the Elovich equation. This raises the question as
to why along with the assumption that = and as a rectangular
function, does not lead to the Elovich equation in the limit
- This inconsistency has not been discussed to date. It seems
that the source of this problem lies in the use of the Langmuir equation
T) with the ART expression for adsorption kinetics.
We note here that there are other theories of kinetics that
offer expressions for the rate that are different from the
ART expression but also lead t o the Langmuir isotherm when = 0. It is
rather strange that adsorption systems with different kinetics of the adsorption/
processes have the same form a t equilibrium.
Let us consider the situation when adsorption runs at quasi-equilibrium condi-
tions. This means that a t a surface coverage all the surface correlation functions

are the same as at equilibrium a t the same surface coverage. That is, the chemical
potential of adsorbed molecules is the same across the whole solid surface.
F o r the Langmuir of adsorption o n a homogeneous solid surface, the
surface chemical potential of the adsorbed molecules, is given by
where is the molecular partition function of adsorbed molecules. consisting of the

product
In Eq. ( 5 3 ) . is the partition function for the vibrations normal to the surface,
is the vibrational partition function for vibrations parallel to the surface, and
stands for rotations, vibrations due to hindered rotations, and electronic and
spin degrees of freedom. Equation (52) can now be rewritten in the form
where
When the surface is heterogeneous, will be related to the total surface cover-
age and (54) will give the "local" surface coverage of sites having adsorption
energy The features of the function will be the same as those of function
(1 1) shown in Fig. 1 . F o r instance, the derivative will have its value
at 9 =
Meanwhile the derivative of the function defined in (44) reaches its
maximum value. at = = 0.368.
Although a comparison of these two pairs of values does not show dramatic
differences, it points to a n inconsistency when combining A R T kinetic expressions
with the Langmuir model for adsorption equilibria. W e can always imagine the
adsorption process running so slowly that conditions can be
assumed. However, even stranger is the necessity of invoking the existence of an
activation energy for adsorption and its variation on various adsorption sites. In
processes involving chemisorption, a n activation energy for adsorption can be
expected. However, an activation energy of adsorption processes of physical
adsorption is less certain. W e will show that the apparent activation energy for
physical adsorption is the result of neglecting the readsorption term in the A R T
expression, (6). Since the is obtained from a fully
rigorous consideration of adsorption kinetics should take into account both the
adsorption and desorption
Integration of (6) now yields

where
and
We can now rewrite (56) in the form
where
W e next accept the assumption that the activation energy for adsorption does not
exist, = 0, and that is equal to the isosteric heat of adsorption within the
accuracy t o a constant. Then,
Figures 5 and 6 show the behavior of the function defined in (59) for
various values of the parameters = and =
Figure 7 shows the behavior of in Eq. (59) a t three different temperatures.
We can see in Fig. 7 that once again we could use the condensation approximation
.5 The effect of the parameter on the behavior of the function defined in

(59) for the fixed value of the parameter 10. = 10 (- =
(- and = 0.01 - -).

.6 The effect of the parameter on the behavior of the function defined in
(59) when the parameter is fixed and equal to 10. = 10
= - -). and = 0.1 (- -
to calculate and then subsequently calculate putting the

derivative calculated from (59) to zero. Let and denote the
following functions:
.7 The behavior of defined in (59) a t three different temperatures: 300 K

Then the first derivative reads
whereas the second derivative takes the form
We see that (17) cannot now be solved analytically to obtain the function
In view of the fact that in (59) does not go to unity as Eq. (14)
has to be modified by replacing unity o n the right-hand side by the maximum value
of 8, which is reached in the limit Now (14) takes the form
From Eqs. (59)-(61) one can easily deduce that is given by

= -
Thus the generalized equation for takes the

1
where 0 must be taken.

The time dependence of in Eq. (66) is in fact a function of the variable
= and of the parameter K,, = Kp.
Thus. on the basis of A R T one cannot arrive a t a simple analytical general-
ization of the Elovich equation that takes simultaneous desorption into account.
The generalized Elovich equation for can be evaluated by using numerical
methods. I t consists of two steps:
1. Solve (17) for a pair of values K,, and is given by Eq. (64).
Use the calculated value of in (66) to calculate =
The choice of a particular value of will not affect the "Elovich plot,"
plotted against In It will only shift it o n the time axis. Therefore, in presenting our
illustrative model calculations, we plot 8, against In Figure 8 shows three such
Elovich plots for various values of the parameter
One can see in Fig. 8 that the linear dependence of 9, on In typical Elovich
behavior is observed over a region of moderate surface coverages, where

The "Elovich plot" of 6 In calculated from (66) for three of
the parameter K,, 10 1 (- - -), and 0.1 (- - -). I t is assumed that = 0, = 20
and T = 300
sorption kinetics d o not play an important role. What is important here is that we
have arrived at the "Elovich behavior"' without invoking the concept of activation
energy of adsorption.
In this way we have eliminated one in developing the Elovich equa-
tion on the basis of ART. However, there are other inconsistencies in the general-
ization of the A R T approach to suit the case of heterogeneous surfaces, using
(30) o r (31). Namely, the assumptions underlying the ART approach d o not involve
a particular mechanistic model of adsorption, for instance, the one leading to the
isotherm. To be fully consistent, one has to use expression (44) for 0 in
(30). In using the Langmuir equation for 9, one uses, in fact, the function
defined in which is the limiting form of 0 defined in (59) when t
and is obtained in addition by integrating full A R T expression, (6). Thus, it
will also not be consistent to use for in (30) any one generalization of the
Langmuir equation, taking into account interactions between the adsorbed mole-
cules. The same inconsistency, of course, must be observed in all the papers on
temperature programmed desorption based on applying (31) when is
represented by the Langmuir isotherm or its generalizations (Bragg-Williams,
quasi-chemical, etc.).
Therefore we would the results of our foregoing analysis of the A R T
approach as follows:
1. The A R T approach is not related to a particular mechanistic model of adsorp-
tion systems.
2. The fact that at equilibrium the A R T approach yields the isotherm is
not sufficient proof that this isotherm should be used to represent in the
generalizations of A R T for the case of heterogeneous solid surfaces. Other

approaches to kinetics different expressions for
but they also lead to the isotherm at equilibrium.
On the basis of ART it can be proved that the equation cannot be
used to represent except for the limit
4. Consequently, it is not consistent to use, in the A R T approach for general-
izations of equation that take into account interactions between
adsorbed molecules.
5. It is not easy to generalize the full A R T expression. taking into account
sorption, for the case of energetically heterogeneous surfaces, because one does
not arrive at simple analytical solutions.
In the following sections we show that almost all these difficulties related to the
use of the A R T approach disappear when we use one of the new approaches
relating the rate of adsorption kinetics to the chemical potential of
adsorbed
the Rate of
Potential of Adsorbed Mole
At the beginning of the 1980s a new family of approaches to
kinetics appeared. A common fundamental feature of these approaches is that they
relate the rate of kinetics to the chemical potential of the
adsorbed molecules, Although all these approaches appeared at approximately
the same time. the one proposed by Nagai in 1985 could, for historical
reasons, be considered first. This arises because the idea behind that approach
can be traced back to de Boer in 1956. In this approach the rate of desorption
is represented by
where is the chemical potential of an adsorbed molecule and is a constant.

approach, as well as two other approaches that are discussed soon, is
especially useful for describing processes that take place
not far from
It has often been assumed that the chances of doing an experiment at
equilibrium conditions depend on how fast the surface diffusion is. However,
puter simulations reported by and coworkers show that the essential
condition is rather the speed of the adsorption-desorption events. There is, in fact,
a body of evidence that a quasi-equilibrium condition is achieved in a majority of
reported TPD and experiments.
Nevertheless, Nagai's approach raised objections Later, however, Nagai
presented proof that his expression, our Eq. is superior to the expres-
sion derived from the A R T approach. Using the hard hexagon adsorption model,
Nagai showed that the ART approach underestimates the role of entropy changes
as a factor affecting the kinetics of processes.

Since we have taken the Langmuir for consideration. the chemical poten-
tial of the adsorbed molecules is given expression (52) Therefore in
( 6 7 ) non takes the
Another member of the new family of approaches using the chemical potential
of adsorbed molecules, is the approach proposed by Kreuzer and Payne in 1988
They followed the assumption by in 1938, that
the rate of adsorption should be proportional to the difference between the actual
pressure and the equilibrium pressure corresponding to the
same surface coverage. On this basis they wrote the expressions for and in
the form
and
where is a "'sticking coefficient." The equilibrium is now expressed

in terms of the gas-phase chemical potential,
where is the partition function of gas molecules due to the internal degrees of
A t equilibrium, so takes the form
which is similar to expression. (67). The sticking coefficient is

intended to account for the efficiency of energy transfer between the solid and
the adsorbate. It cannot be obtained from thermodynamic arguments and
be calculated from microscopic theory o r be derived from experimental data for a
particular system. If we postulate that = 9). then the
expressions for and take the form of the classical A R T expressions, with zero
activation energy for adsorption. F o r the rate of adsorption we have
By replacing in (72) by expression we obtain

On the other hand. since the physical meaning of sticking coefficient is
obscure, we may also put it equal to unit. Then, instead of we have
where
If we assume that it is proper to accept the low-temperature limit for the vibra-
tional partition function then will be an expression in which in the exponent
is increased by a constant, just as it is in (68).
We wrote that both for the case when the sticking coefficient is to be
constant and when it is assumed to be equal to Kreuzer's approach leads
to the isotherm at equilibrium, when = Ward and coworkers
present a different idea that applies the statistical rate theory to the
kinetics of adsorption.
When a molecule coming from a gas phase adsorbs on a solid surface, it must
both cross a gas-phase boundary and form an adsorptive bond when it encounters
the solid. Similarly, when a molecule desorbs. it must break the adsorptive bond
and enter the gas phase. Although the collision of a molecule with a solid surface
can be treated by using either classical o r quantum mechanics. the prediction of the
rate of bond formation o r breaking requires a quantum-mechanical description
because this process involves the probability of an event occurring.
Based on quantum mechanics and Ward and
developed a new approach to kinetics called the statistical
rate theory of interfacial transport. The objective of this approach is to obtain
an explicit dependence of the rate expressions on coverage so that the other proper-
ties appearing in the rate expressions are molecular properties. The S R T approach
leads to the rate expressions
and
where is a constant.
The physical meaning of is discussed in the paper by Ward and
as well as in the more recent publications of Ward and coworkers
is the exchange rate between the gas phase and the solid surface once an isolated
system has reached equilibrium. All these approaches relating the kinetics of
processes to the chemical potential of adsorbed molecules
share a n important feature. Because the expression for the chemical potential is
determined from a mechanistic model of the adsorbed phase, one arrives at a

Heterogeneity 377
dependence of and on coverage without invoking the obscure concept of a

sticking coefficient.
The problem that remains is to find a of generalizing the expressions for
and for the case of an energetically heterogeneous solid surface. The advantage
of the new theoretical approaches to kinetics lies in the fact
that is a function of Thus the knowledge of the behavior of a n adsorption
system at be helpful in predicting the kinetics of the adsorption/
desorption processes in that system.
The equation was successfully applied to a variety of kinetic data
obtained using different techniques with constant o r changing pressure This
indicates that surface heterogeneity is a factor governing adsorption kinetics in
these systems.
N o w let us consider the kinetics of adsorption at lower surface coverages. where
the desorption term defined in Eq. (78) can be neglected, compared to the
adsorption term defined in (77). Then,
where is the standard chemical potential defined as
Let
exp
Then (79) can be rewritten as
The above equation describes the rate of adsorption on sites having adsorption
energy W e also believe that the variation of from site to site is a factor
governing the kinetics of adsorption on the energetically heterogeneous surfaces.
I n two previous publications. we showed how the statistical rate theory of inter-
facial transport can be generalized for the case of heterogeneous solid surfaces.
Below we repeat briefly the principles of that method before we develop its general-
ization to take into account desorption kinetics.
The experimentally monitored surface coverage and the mean rate of adsorp-
tion are the values of 0 and averaged over all kinds of adsorption sites
taken with a n appropriate statistical weight. In the presence of a dense spectrum of
adsorption energies, that statistical weight is well described by a continuous dis-
tribution function relating the number of adsorption sites to corresponding values
of adsorption energy,
When the adsorption takes place a t quasi-equilibrium conditions, the function
in Eq. (82) can be written in the form

where
because the chemical potential of the adsorbed molecules will be the same through-
out the surface phase and will be a function of the average surface coverage The
explicit of this function will, of course, depend on the form of the adsorption
energy distribution
Thus, in the case of a heterogeneous solid surface, the experimentally determined
values and are given by the integral
because is not a function of
When T 0, both 9 and tend to the step function
The functions and have been discussed in previous sections.

At a given temperature T. pressure and the "adsorption front" is on
the sites whose energy is equal to defined in (86). At that instant the overall
adsorption rate is, in fact. governed mainly by the local rate of adsorption on
the sites whose adsorption energy is equal to and will, of course, be proportional
to the probability of finding such sites on the solid surface. Thus, the "local" and
the "total" rates of adsorption will be linked as follows
Note that at = 9= so that from (82) we have
and (88) takes the form
where
= x
Equation (90) is fully correct only in the limit 0, which represents the
accuracy corresponding to the condensation approximation. This means that it

ace Heterogeneity
can be used a t finite temperatures only when the solid surface is strongly hetero-
the variance of is larger than the variance of the derivative
Equation (90) is an equation from which other equations can be developed to
describe the kinetics of adsorption on surfaces characterized by various adsorption
energy distributions.
The in turn is related to the surface coverage
F o r the particular case of the rectangular adsorption energy distribution,
Replacing in Eq. (90) by its value calculated from we obtain
Thus. (94) looks something the Elovich equation and was developed here
as the kinetic equation corresponding to a Temkin isotherm for adsorption equili-
bria. This agrees with common experimental observations. W e see that the Elovich
equation should apply a t conditions where the desorption rate can be neglected,
since we developed it by generalizing the adsorption rate term in (77) and
neglecting the desorption rate term in (78).
As a matter of fact. most of the experiments o n adsorption kinetics were done a t
lower coverages. In the literature we often find statements such as. "After the first
very fast kinetics, a slower kinetics is observed, which follows the Elovich equa-
tion." The first fast kinetics, a t very low surface coverages. is probably governed by
purely kinetic (collision) factors. the effectiveness of the transfer of the energy
excess to the solid phase as well as possible changes in the structure of the adsorbed
phase. The last effect is due to the saturation of chemical bonds. which may be most
essential for the first molecules adsorbed on a strongly "unsaturated," "empty"
solid surface. After a certain first portion of molecules are adsorbed. the rate of
adsorption falls to the limits it runs as a t quasi-equilibrium conditions.
In addition to the above-mentioned common deviations from the Elovich equa-
tion at very low surface coverages, the other common negative deviation was
observed a t higher coverages. T o account for that negative deviation, Higuchi et
al. proposed the following modification of (I):
Other possible reasons for the deviations at higher surface coverages have also been
put forward. F r o m our derivation of the Elovich equation it follows that the main
reason may be the neglect of the desorption kinetics inherent the form of the
equation. Moreover, there is the role of the lateral interactions between the
adsorbed molecules.

As it happens, it can be shown that in the case of strongly heterogeneous
surfaces characterized by a rectangular adsorption energy distribution,
the effects of the lateral interactions d o not change the form of the Elovich equa-
tion.
As far as adsorption equilibria are concerned, the effects of the lateral interac-
tions are included in the function given by the function in (86). The
effects of the lateral interactions on were considered in much detail in a mono-
graph by and Everett They depend on the way in which various
adsorption sites are distributed on a solid surface. This two-dimensional spatial
distribution of various adsorption sites on the surface is called "surface topogra-
phy." T o prove that lateral interactions d o not change the form of the Elovich
equation but only its parameters, we consider here the two extreme cases of surface
topography.
In the case of high (infinite) spatial correlations, adsorption sites characterized
by the same value of adsorption energy are grouped into "patches" that are so
large that interaction between molecules adsorbed on different patches can be
neglected. The whole adsorption system can then be considered to be a collection
of independent subsystems in thermal and contact. For such
surface topography, the function takes the form
where is the number of nearest neighbors and is the energy of interaction

between two molecules adsorbed on two neighboring sites. Equation (96) was
obtained by applying the simplest mean-field (Bragg-Williams) approximation
and including the effects of the lateral interactions between the adsorbed
The other extreme model of surface topography, "random topography,"
assumes that sites characterized by different values of are distributed on an
energetically heterogeneous surface completely at random. The function now
takes the form
Now let us see how inclusion of the lateral interactions affects the form of the
Temkin isotherm for adsorption equilibria and the Elovich equation for adsorption
kinetics.
Using and we arrive at the isotherm For patchwise
topography, we have
whereas for random topography we obtain
By comparing (98) and (99) with we can see that including

the effect of lateral interactions affects only the parameters in

therm, not its analytical form. Now let us consider equations for adsorption
kinetics.
Including lateral interactions introducing an additional on
right-hand side of (86). If we accept the mean-field approximation, this inter-
action term will proportional to for patchwise topography or to
for surfaces with random topography. So for patchwise topography
(82) takes the form
whereas for topography we have
To replace the true kernels by their corresponding step functions we

use the criterion for choosing the value of
The problem is, in fact, similar to that in adsorption equilibria For both
and random topography, criterion (102) leads to the condition
== =-
7
Thus, from Eqs. (100) and (103). it follows that for patchwise topography,
whereas for random topography we have
So, from and (105) we arrive at the following generalized

forms of the equation. For topography.
whereas for random topography we obtain

Thus, including the lateral interactions between adsorbed molecules does not
change the analytical form of the Elovich equation but does change the parameters
of the equation. Deviations from the Elovich equation at higher surface coverages
cannot, therefore, be attributed to lateral interactions between adsorbed molecules.
Apart from peculiarities of some adsorption systems, the most probable source of
these deviations may therefore be the neglect of the simultaneous desorption rate,
the term in (4).
For and defined in (77) and along with the Langmuirian
expression for the chemical potential of adsorbed molecules, Eq. (4)
takes the following particular
= exp
where
At equilibrium. Eq. (108) yields the equation, in which the constant

K= So the meaning of the constants and is not the same as in
ART.
First we check whether (109) predicts the existence of a sharp adsorption/
desorption "front." For that purpose we solve (102) when 6 is given by Eq.
(108).
The first derivative is
where = and = Let denote the function in 10).

Then the second derivative, takes the form
where is the function

After some lengthy algebra. we arrive at the following form of the function
s):
From (1 1 it follows that condition (102) is fulfilled when = =

From (113) we can deduce that the above condition is fulfilled when
= = We now check whether there be some other solutions. F o r
that purpose we examine the solutions of the equation
We introduce a new variable, = and multiply (114) by so that
Now we solve the above quadratic equation with respect to This has only one
solution:
which is a nonphysical solution, because from its definition x can

never be negative. Thus, the only physical solution is
and (108) is consistent with the Langmuirian model for adsorption equilibria.
Equation (88) now takes the form

which takes both adsorption and desorption kinetics into account. The function
is obtained from (108) by taking into account the fact that when
= = Then
takes the form
where
= x and = x (121)
The first term on the right-hand side of (120) represents the adsorption rate
and is identical with the expression on the right-hand side of (90). The second
term on right-hand side of (120) represents the rate of the simultaneous
desorption. At quasi-equilibrium conditions, = and can be expressed
as a function of surface coverage. Equation (120) is a generalized form of
which allows us to calculate the rate of adsorption (desorption) if the adsorption
energy distribution is known. The latter function can be determined from the
observed equilibrium adsorption isotherms. For the rectangular adsorption energy
distribution replacing in (120) by defined in (93) leads us to
the following generalized form of the equation, taking into account both
adsorption and desorption occurring simultaneously:
The generalized equation is a three-parameter equation.
where
and
Parameter can be determined in equilibrium measurements as the tangent of

the adsorbed amount plotted against the logarithm of adsorbate pressure. The two
other parameters. a and are treated as best-fit parameters while fitting experi-
mental data with expression (123). Figure 9 shows the fit of Eq. (123) to the

-3 -2 -1 0 1 2
log I
.9The result of computer fitting of the experimental data for the kinetics of
oxygen adsorption on tungsten reported Lopez-Sancho and de
Segovia, by our generalized Elovich equation. (132). assuming that
a = 0.394, a = 3.85 = 4.374 The pressure and
temperature kept at the experiment were equal 5 ton" and 77 respectively.
experimental data of the kinetics of oxygen adsorption on tungsten

reported by Lopez-Sancho and de Segovia
Now let consider the integral form of the generalized equation,
which we write in the form
W e now introduce the variable
Then Eq. (126) takes the form
and we get
where
2
Assuming that = 0) = 0, we get the equation

+ arctanh z ] , 1
exp [
exp [ + -
The features of the generalized Elovich equation, are shown in Figs.
Now let us remark that both the differential equation, and its integral
form, (1 correspond to the level of condensation approximation. Therefore,
their application to the equation
obtained for equilibrium from (108) and the use of the rectangular adsorption
energy distribution (10) should yield the Temkin isotherm. One can easily prove
that this is true, by putting = 0 into (122). The total equilibrium iso-
therm T) then takes the form of the Temkin isotherm
to be with (26).
10 The effect of the parameter a on the behavior of the generalized Elovich

,
equation. Eq. a = 0.2 (- 0.5 -). and a = 0.8 (- - -). The
calculations were done by assuming that a b = torr' and
= 3.862

0
log t
. 11 The effect of the parameter a on the Elovich function defined in

7
(132). a = 5 x 3 a = 2 10
torr' (- - -). The calculations were done by assuming that = 0.5, b =
torr' and = 2.332 x torr.
The condition 1 in (131) and (132) arises from the application of the
condensation approximation and should not appear when, instead of (1
some exact solutions are applied. It can be easily checked that this condition is
equivalent to the condition that T) defined in (134) must be smaller than
unity.
-4 -2 0 2 4
log t
. 12 The effect of the parameter on the Elovich function, (132). =

1.5 -), and = 2 torr'
-). The calculations were done by assuming that a = 0.5, a = torr' and
= 5.21 torr.

and Z. Chem.
and Zeldovich, Acta Phys. Chem. USSR (1934).
Phys. USSR (1934).
Zeidovich, Thesis. Leningrad, 1936.
Elovich and Kharakorin, Prob. Catal, (1937).
Kharakorin and Elovich, Acta Physicochim. USSR (1936).
Elovich and Zabrova, (1939).
Low. Chem. Rev. (1960).
Aharoni and Adv. Catal. (1970).
I. Temkin. Fiz. (1941).
I. Temkin, Zh. (1957).
Roginsky. Adsorption and Catalysis on Heterogeneous Surfaces, Akad. Nauk
USSR. Moscow, 1949.
Theory of Physical Chemistry Processes at a Gas-Solid Interface,
Izd. Moscow, 1990; English C R C Press, Boca FL, 1991.
in Adsorption and on Inorganic (A.
Dabrowski and Tertych, New York, 1996.
Rudzinski and (1997).
Rudzinski, in Equilibria a n d Dynamics of Gas Adsorption or Heterogeneous
Solid Surfaces Rudzinski. A. Steele. and G. Zgrablich. Elsevier,
New York, 1996.
A. Ward and Surf. (1986).
A. Elliott and A, Ward, Phys. (1997).
A. Clark, The Theory of Adsorption and Catalysis. Academic. New York, 1970.
I. Cvetanovic and Y. Adv. Catal. (1967).
I. Cvetanovic and Catal. Rev. (1972).
I. I. Falconer and I. A. Catal. Rev. (1983).
I. in of Heterogeneous Catalysis
Marcel New York. 1984. Chap. 2.
and Do, Catal. Today 8-309
Y. and Cvetanovic, Phys. Chem. (1963).
A. Czanderna, and Colloid Interface
1970).
Cater. Vacuum (1962).
A. Grant and Cater. Vacuum (1965).
Dawson and Peng, Surf. (1972).
and Yoneda, Bull. Chem. (1978).
A. King, Surf. (1975).
Tokoro. and Yoneda, Catalysis (1979).
H. and P. Catal. Rev. (1978).
and in at the Gas-Solid and Liquid-Solid
Interface and Sing, Amsterdam, 1982,
p. 383.
V. P. Surf. Rep. (1991).

Jaroniec and R . Physical Adsorption on Heterogeneous Solids,
New York, 1988.
W. Rudzinski and D. Everett. Adsorption of Gases on Heterogeneous Surfaces,
Academic, York, 1992.
Adv. Colloid Interface (1995).
and Rudzinski, in Equilibria and Dynamics o f Adsorption on
Heterogeneous Solid Surfaces Rudzinski, A. Steele. and Zgrablich,
Elsevier, New York, 1997, p.
I. Langmuir. Am. Chem. (1918).
in Equilibria and Dynamics of Adsorption on Heterogeneous Solid
Surfaces (W. Rudzinski, A. Steele, and Zgrablich. Elsevier. New
York, 1997.
Kreuzer and Payne, in Dynamics of Gas-Surface Interactions
Rettner. and Royal Society of Chemistry, Cambridge,
U K , 1991, Chap. 6.
H. and S. Payne, in Equilibria and Dynamics of Gas Adsorption on
Heterogeneous Solid Surfaces Rudzinski, A. Steele and Zgrablich,
New York. 1997.
Phys. Solids (1957).
A. King, Surf. (1977).
Gorte and Schmidt, Surf. (1978).
Weinberg, in Springer Series in Surface Sciences, Vol. 8 ( M . and
Kreuzer. Springer, Berlin. 1987. 94.
S. Lombardo and A. Bell, Surf. Rep. (1991).
A. Kong, and Schmidt, Surf. Sci. (1988).
K . Nagai, Phys. Rev. Lett. (1985).
Nagai, Surf. (1986).
J. de Boer, Adv. Catal. (1956).
Talbot, X. and Wang, (1994).
V. Zhdanov, Surf. L469 (1986).
Nagai and A. Hirashima. Surf. 171: L464 (1986).
A. Surf. (1988).
Nagai, Surf. Sci. (1988).
Nagai, Surf. (1991).
Kreuzer and Payne. Surf. (1988): (1988).
Payne and H. Kreuzer, Surf. 205:153 (1988); (1989).
Kobukowa, Rev. Phys. Chem. Jpn. (1938).
A. Ward and Phys. (1982).
A. Elliott and A. Ward, in Equilibria and Dynamics Gas on
Heterogeneous Solid Surfaces Rudzinski, A. Steele and Zgrablich,
Elsevier, New York, 1996.
I. I. Ree, and J. (1955).
J. Lopez-Sancho and L. Segovia, Surf. Sci. (1972).

Department of Chemical Engineering. University of
Queensland, St. Lucia, Australia
I. Introduction 392
I I . Adsorbed Solution Theories 393
A. The ideal adsorbed solution theory 393
B. The nonideal behavior of the adsorbed phase 396
HI. Equilibria Models Based on Potential Theory 398
A. Direct application of D R and D A equations 400
B. Potential theory approach 403
C. Other methods 405
IV. Other Models 407
A. The extended Sips mode! 407
B. Isosteric heat as a function of loading model 408
C. The heterogeneous extended Langmuir model 409
D . The multisite Langmuir model 41 1
E. The multispace adsorption model 41 1
F. Martinez's model 412
G. Keller's model 413
V. Microporous Structure and Adsorption Equilibria on Activated
Carbon 414
A. The structure of activated carbon 414
B. Potential energy and slit pore configuration 415
C. Characterizing the PSD of activated carbon 417
D . Prediction of adsorption equilibria 420
VI. Microporous Structure and Adsorption Equilibria-the Micropore
Size Distribution Model 42 1
A. MPSD model for a pure component system 422
B. MPSD model for a multicomponent system 424

C. Comparison with experimental data-pure component equilibria
D . Comparison with experimental data- binary component
equilibria
E. Discussion of the MPSD model
VII. Statistical Mechanical Methods and Multicomponent Equilibria
A. Pore selectivity isotherm
B. Test of the IAST at the microscopic level
VIII. Conclusion
References
Knowledge of multicomponent adsorption equilibria is critical to the design of

practical gas separation processes based on adsorption mechanisms For
pressure or thermal swing adsorption technology, information such as adsorption
capacity, selectivity, driving force for mass transfer, and heat of adsorption
are necessary for design and operation. This information can be conveniently
obtained from an isotherm equation that properly represents the adsorption equi-
libria of the gas mixture in the However, finding a proper adsorption
isotherm equation for a gas mixture system is not a simple task if this equation
is meant to give good predictive results over a wide range of operating conditions.
Two main factors contribute to our slow progress in this area: (1) The adsorbents of
practical use such as activated carbon, carbon molecular sieve, zeolite, alumina,
and silica gels are heterogeneous solids; and (2) multicomponent gas mixture
adsorption data are scarce. and those that d o exist are mostly incomplete
Adsorption measurement for multicomponent systems is a function of the com-
position, temperature, pressure, and properties of adsorbate and adsorbent. As the
number of components increases, the number of needed to define the
adsorption equilibrium increases rapidly and eventually becomes infeasible.
Adsorption equilibrium models are therefore needed to correlate and predict the
multicomponent adsorption equilibria. These models should be able to predict the
equilibria using the information available on pure component equilibria, as
the latter are relatively easy to measure and furthermore there is an abundance of
pure component isotherm data available in the literature. As a result, predictive
models for gas mixture adsorption are necessary in the design and modeling of
adsorption processes.
In the past 30 years, great efforts have been expended to develop techniques for
predicting the adsorption equilibria based on pure component
data. However. until now only limited success has been achieved. Several publica-
tions provide good reviews of the work in this area Generally speaking,
these models can be classified into four groups: (1) Vacancy solution theory, (2)
statistical models, (3) ideal adsorbed solution theory (IAST), (3) theory,
and (4) various empirical or semiempirical models,
The vacancy solution theory or vacancy solution model (VSM) was proposed by
and Danner and later modified by Cochran The isosteric
heat of adsorption of this model was obtained by and Kabel In this theory,

the vacancy is a n imaginary entity defined as the vacuum space that acts as the
solvent for the system. Thus, the VSM treats pure component adsorption as a phase
equilibrium between two binary vacancy solutions of different compositions. The
detailed and its modifications were reviewed by Yang The VSM was
shown to be capable of predicting equilibria at high surface cover-
age, and even the adsorption azeotrope in some although it also
presents some deviation for a number of systems
The statistical models are based on thermodynamics arguments. But these mod-
els are normally specific to adsorbents with well-defined structures such as zeolites
o r carbon black There was also a statistical model reported for the
component adsorption equilibria of vapors o n activated carbon, but it was vali-
dated only for a particular system
This chapter reviews the progress in adsorption equilibrium models that have
received some general interest in the literature. Particular emphasis is put o n the
role of the micropore size distribution in single- and adsorption
equilibria.
Among the theories of predicting mixed-gas adsorption equilibria from pure

ponent adsorption isotherms, the ideal adsorbed solution theory (IAST) has
become the standard and often serves as a benchmark for the purpose of compar-
ison by other models. IAST is a thermodynamically rigorous theory based o n the
mixing of individual components at constant spreading pressure to form a n ideal
solution. It has the advantages that (1) n o mixture data are required and (2) the
theory is independent of the actual model of physical adsorption.
The basic equation of IAST is analogous to law for vapor-liquid
equilibrium:
where and are the fraction and spreading pressure. respectively. of

component i in the adsorbed phase. At adsorption equilibrium, the reduced spread-
ing pressures
for each component and for the mixture are the same,
The function is the pure component equilibrium capacity, and is the

pure component hypothetical pressure that yields the same spreading pressure as
that of the mixture. By assuming ideal mixing a t constant and the total amount
adsorbed, is found to be

with the constraint
Thus, for a system consisting of components wih specified partial pressure in the
gas phase, there are +
1 independent variables to be solved: P and
Equations (1)-(4) provide the +
1 necessary equations. Since the equations
are nonlinear and the integrals in (2) cannot be solved analytically for most
of the pure component isotherm equations, the + 1 equations have to be solved
numerically.
The IAST has the advantage that no restriction exists for the type of pure
component isotherm equation used for the prediction of multicomponent equi-
libria. The success of the IAST in predicting the multicomponent adsorption
equilibria depends on three factors: (1) the quality of pure component equilibria
data; (2) the correct choice of isotherm equation for the pure component. which
should fit the experimental data well over the entire pressure range; and (3) the
regression algorithm in deriving the parameters for pure component isotherm equa-
tions. Factors 2 and 3 were studied by Richter et Recently, and
Farooq investigated the performance of a number of popular isotherm models
in fitting pure component isotherms with various in a very compre-
hensive way. In the same article, they also examined the performance of each model
equation and compared it with the IAST in predicting the multicomponent adsorp-
tion equilibrium on activated carbon (experimental data on the ternary system
methane-ethane-propane were used).
1. The Procedure
To avoid the numerical integration procedure and iterative scheme in the IAST,
and Myers introduced an algorithm that allows for fast
of multicomponent adsorption equilibria. The pure component isotherm equation
was chosen as
where is the variance of the adsorption energy distribution and
Equations (6) can be analytically integrated for the spreading pressure The
equilibrium amount adsorbed of each species can be obtained from the solution
of a set of A' nonlinear algebraic equations This algorithm (called
was improved by Moon and and later refined by and Myers
It is about one to several orders of magnitude faster than the IAST, depending on
the nature of the problem,

Frey and Rodrigues proposed a method for the explicit calculation
component equilibria using IAST. They used a polynomial or Taylor series to
approximate the relationship between and However, their method has
limited application owing to the approximation used.
2. The Thermodynamics Requirement for the Isotherm Equation
Since the IAST has a rigorous thermodynamics basis, the pure component iso-
therms used are also expected to meet the thermodynamics requirement. Myers
and coworkers pointed out that the thermodynamic consistency for the
single-component isotherm equations means that they should not have a singular
value and be exact at the origin,
lim finite value

P
This is required by the integration procedure of ( 2 ) in calculating the spreading
pressure. For example, the Freundlich isotherm equation,
incorrectly gives a value of infinity in ( 7 ) and consequently may cause a large

error in the calculation of spreading pressure.
Toth recently suggested that additional criteria should be met for the
thermodynamics consistency of an isotherm equation. He defined the function
where is the fractional coverage of the surface. The spreading pressure of (2)
can then be rewritten as
where is the surface of adsorbent covered by 1 mol of adsorptive at 1.

According to Toth, the criteria for thermodynamics consistency of an isotherm
equation are
At 0 = 0,
lim =1
= finite value
Equation (1 1) corresponds to while (12) is the extra thermodynamics

requirement that basically stipulates that at full surface coverage the change in the
free energy of the adsorbed phase must be finite. According to the criteria of Toth,
however, most of the currently used isotherm equations are "thermodynamically
incorrect," since they cannot meet one or both conditions. Equation (12). however,
is controversial. Toth further pointed out that this inconsistency problem is

caused by the fact that most isotherm equations deal with the absolute amount
adsorbed. If the excess amount adsorbed is used instead, this inconsistency problem
can be avoided.
The thermodynamics consistency for binary adsorption equilibria isotherm
equations was studied by Myers and used similar
arguments and proposed the following differential criterion for
testing the thermodynamics consistency of adsorption isotherm equations of binary
systems:
where is the fluid phase concentration and is the adsorbed phase density
The IAST is based on the assumption that the adsorbed phase is ideal. This is valid
at low pressure, and the IAST is to give good prediction of multicomponent
adsorption equilibria at surface coverages up to about 50% At high pressure,
the model prediction tends to diverge from the experimental data (normally
predicting the experimental data), presumably due to the negative bias from
law. This is expected, because at high pressure the adsorbed phase
becomes more like a liquid, and liquid mixtures of dissimilar molecules are gen-
erally nonideal (nonideality). Sometimes problems also arise regarding the spread-
ing pressure when IAST is applied to binary consisting of strongly and
weakly adsorbing components; when the partial pressure of the 'strongly
adsorbing component approaches its saturation value, the spreading pressure of
the mixture can well exceed the saturation pressure of the weakly adsorbing com-
ponent.
The surface energetic heterogeneity further complicates this problem because
micropores of different sizes or sites with different adsorption energies will present
different adsorption affinities and selectivities toward different adsorbates.
Interactions among adsorbed molecules of the same type and among those of
different types also make this problem more complicated demon-
strated that either of the following two factors can cause erroneous predictions of
IAST even in cases where there are no lateral interactions among adsorbed mole-
cules: (1) a large difference in molecular size or (2) adsorbent heterogeneity. Under
certain circumstances, these two factors could result in an azeotrope that cannot be
accounted for by IAST, Dunne and Myers studied molecules of unequal size
adsorbing into a cavity by using a computer simulation technique,
and they found that the micropore size exclusion effect for larger gen-
erates nonideal behavior in the system.
1. Real Adsorbed Solution Theory
There are two approaches to dealing with these nonideal behaviors in the adsorp-
tion process. One accounts for the nonideality of the adsorbed phase with the
introduction of an activity coefficient; this is called the real adsorbed
solution theory, or RAST In RAST, Eq. (1) is replaced by

The RAST has extra parameters to be extracted from a binary system. so it is not,
in a strict sense. a predictive model. It has been suggested that can be correlated
as a function of at constant T by measuring a few data points for the binary gas
adsorption isotherm. This approach, however, has found only limited success,
2. Energy Distribution Approach

The other approach considers the surface hererogeneity, in which the heteroge-
neous surface is considered to be patchwise. T h a t is, adsorption sites having the
same energy are grouped together; thus the adsorption can be considered as
a collection of independent subsystems. The surface heterogeneity is then repre-
sented by a distribution function of adsorption energy. The overall adsorption
equilibria of a pure component is the integral of the local adsorption equilibrium
on each patch over the complete energy distribution range
For a with adsorbing onto a heteroge-
neous surface a t temperature pressure and coniposition = ...,
the overall amount of the component adsorbed, is
where
and ..., is the normalized distribution function in terms of

all adsorption energies, .., is the of component
adsorbed o n a surface characterized by a set of adsorption energies
and is the N-dimensional physical of ...
This integral equation represents a general adsorption isotherm of a
adsorption system involving heterogeneous surfaces
Equations (15) can be simplified with the assumption that the ordering of the
sites from low to high energy is the same, the perfect positive correlation. In
this case, reduces to
Equation (16) is the general integral equation from which many

equilibria models are derived also used the binary selectivity
instead of energy e in (16) t o study the role of adsorbent heterogeneity in
adsorption from a binary liquid mixture.
The local interaction energy in the adsorbed (or energy matching) between
different species in Eq. (16) is also a n interesting probledm. T o simplify the inte-
gration, Valenzuela et al. assumed that the cumulative energy distribution
function was the same for all components in the system,

and this is termed a cumulative energy matching Moon and
assumed regular matching (congruent distribution) or complementary matching
(opposite distribution) schemes, with binary data needed to obtain some relevant
parameters. It was shown in their study that with their matching schemes the IAST
can produce the nonideal behavior of some systems. The energy-matching
in the adsorbed phase is a fundamental issue in the multicomponent adsorption
equilibria on a heterogeneous surface, and we discuss it further in The
role of adsorption energy heterogeneity and the local isotherm in the prediction of
mixture adsorption equilibria was studied by H u and D o
Equation (16) has also been applied to IAST When it is assumed that the
gas mixture obeys the perfect gas law and that the adsorbed phase on each patch is
ideal, the resulting model is termed HIAST (heterogeneous IAST). Since the
HIAST requires the evaluation of the local equilibria on each energy patch with
IAST, much more computational effort is required, and yet the increase in perfor-
mance over the original IAST is only modest for most cases. Therefore, this theory
has not been widely used in practical applications.
The adsorbate-adsorbate interaction in the adsorbed chase was also considered
in some models However. the current trend in the literature is focused on the
surface heterogeneity resulting (1) the size difference between adsorbate mole-
cules or (2) the structural heterogeneity of the adsorbent. and
demonstrated that in a system where the lateral interactions are very significant, the
assumption of a locally ideal mixture can still be applicable. The other example that
strongly supports this point of view is the experimental results of binary equilibria
of the on BPL-activated carbon reported by Gusev et
Figure 1 presents experimental results of the selectivity of ethane to methane at
different pressures (dots) along with the prediction from IAST (solid line).
The selectivity is defined as
It can be seen that IAST predictions seriously deviate from the experimental data
for this system even though the adsorbed phase is known to be very close to ideal
under the experimental conditions. This deviation therefore points to the surface
heterogeneity. T o address this heterogeneity, new models continue to emerge in the
study of multicomponent adsorption equilibria.
The potential theory of adsorption was introduced by Polanyi in 1914.

and Stoeckli et improved the theory and it the theory of
volume filling of micropores (TVFM). This theory has been widely used in corre-
lating the effect of temperature on the adsorption isotherms of pure gases. The
modern formulationof TVFM is the (DA) equation, which is
expressed as

30
ethane (51 methane (48.8%)
IAST
Exp. data
Pressure (bar)
FIG. 1 Plot of selectivity of ethane to methane as a function of pressure.
Ref. 47.)
where
A RT
is the adsorption potential, V is the volume adsorbed at the relative pressure
is the saturation pressure at temperature is the affinity coefficient
of the adsorbate, is the micropore volume, and is the characterisic energy of
the adsorbent.
The plot of amount adsorbed versus A is defined as the characteristic curve. The
relationship between and pore size for adsorption in micropores was studied by
Chen and Yang The exponential term which lies in the range of
accounts for the structure of the adsorbent. Large values are found to be asso-
ciated with adsorbents that have a narrow micropore size distribution, while small n
values relate to adsorbents with a wide range of sizes and possibly with
and macropores as well.
The DA isotherm contains parameters related separately to the solid and the
adsorbates. The advantage of using the potential theory for predicting gaseous
mixture adsorption equilibria is that the pure component characteristic curves are
generally independent of temperature. Thus having obtained isotherm infor-
mation of a pure component at one temperature, the mixture equilibria at
other temperatures and pressures can be predicted. The DA equation and its

modifications have also been used extensively in the characterization of
adsorbents.
Some difficulties are also met in the application of the under certain
conditions, because
The liquid solution parameter molar volume is required to convert the volu-
metric adsorbed to molar adsorbed amount. When the adsorption
temperature is above the boiling temperature of the solution, empiri-
cal correlations have to be used to obtain the liquid molar volume
The adsorbate molar volume is normally assumed to be the same in both
single- and mixed-gas adsorption.
The calculation of the saturation pressure at supercritical temperature also
incurs some uncertainty.
The DA equation does not conform to Henry's law in the low pressure region.
The Henry's law for pore filling in adsorption
equilibrium was discussed by and Van
Various efforts have also been made to extend the D A equation for the study of
the adsorption equilibria of mixtures. The models for multicomponent
adsorption equilibria based on the D A equation can be roughly classified into three
groups: (1) the direct extension or application of the DA (or DR) equation; (2) the
potential theory approach, which is based on coalescing the potential curves of
different components; (3) other methods, which link the D A equation to other
adsorption properties, such as the adsorbed phase properties or adsorbate--pore
interaction potentials.
1. Bering Model
Early efforts to extend the D R equation were largely empirical. Bering et al.
extended the D R equation to binary system in a direct manner. For binary systems,
their model is
where and are volumes adsorbed for each species, and is the saturated
vapor pressure for the adsorbed mixture over which the saturated vapor has the
composition as that of the equilibrium vapor over the adsorbent. The para-
meter is the average affinity coefficient, taken as
= A-I +
T o calculate and the following Lewis relationship for a binary system
is used:

Adsorption Equilibria of Hydrocarbons 401
with the constraint

+ = 1
The Lewis relationship has its theoretical basis in micropore filling with the
assumption that the adsorbed phase is ideal
Thus the Bering model must be solved iteratively. When the temperature is
above the boiling point, it is difficult to justify the relationship between the partial
pressure of each species and the saturation pressure of the system, Hence this
model is less attractive than other methods.
2. Doong and Yang's Model
Doong and Yang proposed a simple way of predicting equi-
libria by using the concept of TVFM. Their model is based on the idea that in the
DA equation the total micropore volume of the adsorbent should be replaced by
the "maximum available volume" for each species. Other
are that (1) the adsorbate-adsorbate interaction is negligible compared to the
adsorbate-adsorbent interaction, (2) the parameters and for any species
are independent of the other species, and (3) the adsorbed phase is ideal. With
these assumptions, the DA equation for each species in a binary system can be
written as
Solving for and yields the equations
where
Equations (25) and (26) are sufficient for predicting the equilibria of binary mix-
tures. The simplicity of this method constitutes an advantage over other
based on TVFM. Besides, this model is not iterative and can be extended to
component systems in a straightforward manner.
For practical purposes, the volumetric amount adsorbed ( V ) in (25) should
be converted into molar amount adsorbed, with the equation

=
where is the liquid molar of the adsorbate. and Yang
recommended the following relationships:
where
molar volume of the saturated liquid at its normal boiling point
molar volume of the species at its critical temperature
This model is validated with the binary experimental data on several systems
(1) alkane and mixtures on activated carbon; (2) mixture on
zeolite; (3) binary adsorption on BPL activated carbon. The model was compared
with three other models (IAST, G-M, and Bering et model), and it was found
that under most this gives equivalent or even superior perfor-
mance.
3. Equation Used with IAST

The application of the D R equation in IAST was studied by Richter et and
was used recently by Lavanchy et to predict the binary adsorption equilibria
of tetrachloride vapor mixtures on activated carbon. This is
simply a application of the IAST theory with the D R equation being used to
describe the pure component isotherm. However, due to some special properties of
the D R equation, this combination generates some unique features and deserves
some elaboration here.
The DR equation is a special case of the DA equation, Eq. with = 1.
When the D R equation is substituted into the analytical form for the
spreading pressure of component i is
where = is the hypothetical pressure for pure component i that gives the
same spreading pressure as that of the mixture. The adsorbed phase concentration
for a binary system is obtained by simply solving (assuming an ideal adsorbed
phase)
The number of moles adsorbed at the hypothetical pressure can be calculated from
the equation as

It is seen 111 (29) that the DR equation requires two types of parameters: (1)
properties of the volume and characteristic energy (2)
properties of the adsorbate saturation pressure affinity coefficient and
liquid molar volume Principally, the first group of parameters are
independent and can be obtained from the adsorption of a reference adsorbate on
the same adsorbent, while the second group of parameters are
dent but can be simply estimated from the thermodynamic properties of each
species. Therefore, a set of data is required to describe binary adsorption
over a wide range of pressure and temperature. Thus the advantages of using the
DR equation deserve to be exploited.
The error in calculating the spreading pressure caused by the incorrect behavior
of the D A equation in the Henry's law region was estimated by Richter et as
This error is normally smaller than experimental error.
Theory
The basic assumption in the potential theory approach to predicting the adsorption
equilibria of a gas mixture is that the adsorption data for all pure components on
the same adsorbent will fall on the same line when plotted on some appropriate
coordinate (coalescing) This principle is reflected in the Lewis method
and in the Grant-Manes (G-M) method and its modification by and
Danner
1. The Grant-Manes Method

The G-M method for predicting the phase diagram of a binary system on activated
carbon is based on the equation
where and are the fugacities of components 1 and 2, and and are the
fugacities of each component at its saturation vapor pressure and at system tem-
perature T. Moon and showed that in most cases the bulk fugacity can be
replaced by pressure. Equation (33) can be used to derive the adsorbed phase
composition by using the concept of pseudoadsorbate pressure and its related
saturated liquid molar volume in an iterative manner.
According to Grant and Manes the affinity coefficient that coalesces the
potential curves of each species satisfies the relationship
where is the saturated liquid molar volume of pure component i at its normal
boiling point.

2. Method
However, and Danner and Moon et demonstrated that (34)
is too crude and too unreliable for various adsorbates on adsorbents of practical
interest. They pointed out that the potential curves of different pure gases on
activated carbon are usually close but do not coincide. T o overcome this,
and Danner suggested that (33) be replaced by the equation
where is the saturated liquid volume of the component i at the pseudoadsorbate

pressure. The parameter is the interaction parameter for the mixed
adsorbent system, which is related to the coalescing factors of each species:
For different adsorbate-adsorbent interactions, the following relationships were
proposed by and Danner:
For nonpolar adsorbents such as activated carbon:
For polar adsorbents such as silica gel and molecular sieve:
Upon generalization of the potential approach for adsorption equilibria of gas

mixtures, pointed out that the potential theory approach
corresponds to a special case of IAST for adsorption on a heterogeneous surface
and therefore it works better on strongly heterogeneous adsorbents such as acti-
vated carbon. Based on thermodynamic analysis, et also suggested
that an extra coalescing factor be introduced in (35) and in this way the pre-
dictive accuracy of the potential theory approach could be further improved.
and Myers used the same idea but proposed another way of coales-
cing the isotherms of different vapor adsorbates on a heterogeneous surface. Based
on a comprehensive analysis, they suggested the characteristic
equation
where is the adsorptive capacity at saturation, is the reduced pressure, and

is the free energy of immersion of the adsorbent in liquefied adsorbate, which can
be obtained from binary mixture data or by fitting the pure component isotherm
with Eq. (38). factor is the molar free energy of immersion of the
adsorbate vapor, which transforms the Polyani potential into a dimen-
sionless function that is the same for all adsorbates. Equation (38) is shown to work
well for adsorbates of different sizes and structures.

Dobruskin's Model
Dobruskin proposed a model for the adsorption equilibria of
vapor mixtures based on the concept of TVFM and an adsorbed phase in
which the adsorbate-adsorbent interactions predominate over the lateral interac-
tion between adsorbed molecules. The proportions of the component in the
adsorbed phase are determined by a statistical distribution based on Frenkel's
mechanism and kinetic gas theory In Dobruskin's study, the equili-
brium is viewed as a dynamic process in which the average molecular residence time
is the reciprocal of the rate constant for desorption, For adsorption of a binary
mixture in an elementary volume the ratio of the average times between two
components is
where and are the adsorption energy and vibration frequency, respectively. If
the adsorption energies of the components are independent of each other, the
following approximation is made:
in which the similarity of the characteristic curves at constant fractional loading is

taken into account.
At equilibrium of gas adsorbing onto a surface of area the of mole-
cules in the adsorbed phase is assumed to be equal to the product of the average
time and the number of molecules striking the surface. The average ratio of the
numbers of molecules to in this element is
By using kinetics theory and inserting = into

obtained the following equation for the selectivity of the binary system:
where b = The adsorbed phase equilibrium composition can be

obtained with further assumptions about the adsorbed phase, molar work. and
special points in the potential curve. This model is capable of computing multi-
component equilibria by using single gas isotherm parameters only, while the use of
liquid solution parameters is avoided.
2. Eiden and Schlunder's Model

Eiden and Schlunder proposed a model for predicting the adsorption equi-
libria of a binary vapor mixture based on pure component parameters. The model

combines the TVFM and IAST on the basis of a multiphase mixture, the
micropore space is divided into that are related to the adsorption
potential by means of the characteristic curves of pure components. Amount
adsorbed and composition are evaluated separately in each by assuming
that the adsorbate mixture forms ideal solutions.
and Schlunder used the modified D A equation,
V=
[ -
where is the saturation micropore volume and is the limiting potential.

When the are the same for all species, the micropore volume is divided into z
and each j corresponds to the adsorption potential
With related thermodynamics properties of the modified DA equation, the follow-

ing expression is derived for the molar fraction of component i in the subvolume /':
where is the relative pressure and is the ratio of the molar liquid
. With Eq. the phase compositions of the jth subvolume can be
calculated directly. The overall adsorbed quantity for a component is the sum of its
related quantities in each subvolume, provided the adsorption potential of that
component is larger than the limiting potential. It should be noted that this
model can describe the real adsorption behaviors when the exponents are
different.
In the case that are different owing to the steric effects, real adsoption
behaviors are to be expected. The multiphase real adsorbed solution (MRAS)
approach is used, considering the "real behavior" to comprise two factors: (1)
adsorbate-adsorbent interactions, which are taken into account by the multiphase
concept; (2) adsorbate-adsorbate interactions, which are considered by activity
coefficients and are taken as the same for the same species in each subvolume.
The MRAS model is
Their model was validated with the experimental data of

and systems on two activated carbons
Satisfactory agreement was found between predictions from (45) and
the data for all the systems, even for one that showed strongly pressure-dependent
azeotropic behavior.

In this section we review the models that do not fall within the scope of the previous
two sections. Some of these models may bear more or less empiricism in their model
development. However, their methodology and practical significance are of more
interest to practicing engineers. A large number of models of this type are to be
found in the literature, but here we cover only a few that exhibit some distinctive
features.
Sips proposed an empirical equation for the adsorption of a pure component

system that is a combination of the Langmuir and equations. This
three-parameter equation has been widely used for fitting the isotherm data of
numerous hydrocarbons on activated carbon Sips adsorption isotherm
for the pure component has the form
If we used the same analogy as that for the generalization of the Langmuir equa-
tion, we would obtain for the extended Sips model the equation
However, this direct extension corresponds only to the ideal situation where the
adsorption energies of the various components are completely
demonstrated that if the adsorption energies of each compo-
nent are highly correlated to each other by the relation
then the generalized Sips equation for a multicomponent system will take the form
This equation can also be derived by formally applying the IAST, using the Sips
equation, Eq. to describe pure component data. Hence, for adsorption of gas
mixtures onto activated carbon, the use of Eq. (50) is generally justified.
Because the exponential term a # 1 , the Sips equation does not have a correct
low pressure limit, which is contrary to experimental observations. T o overcome
this problem and extend the range of applicability of isotherms of this type, Staudt
et introduced the pressure- and temperature-dependent exponents to ensure
that Henry's law holds when pressure is sufficiently low. The new generalized form
of (48) is

where
a, =
is the new pressure-dependent characteristic exponent, which satisfies the two limits
a, = 1 and a, =
Staudt et suggested the following functional form of a,, which conforms to the
above limits:
where a, is a constant. The temperature dependence of the exponential terms is
where is the critical temperature and k is a constant. With these corrections, the
new model predictions were shown to be much for both single- and
multicomponent adsorption equilibria. This of Staudt et is also applic-
able to other isotherms. For for the Toth isotherm it is
where a is defined in the same way as in Eq.
D o and D o proposed an adsorption equilibria model on isosteric heat as

a function of loading (IHFL), In their model the variation of isosteric heat with
loading is regarded as the measure of the heterogeneity of the adsorbent. Since this
takes into account the surface heterogeneity and allows for the variation of
the isosteric heat during the adsorption process, it gives an excellent fit to the
equilibria of many species on the same adsorbent when fitted simultaneously.
The model was tested successfully for many systems on various adsorbents such
as activated carbon, zeolite, and The model equation for a pure
component is
- +
- RT
where is the surface coverage, t is the parameter specific to the
adsorbate pair, (3 is a parameter characterizing the solid property, E(0) is the
isosteric heat at zero loading, a is the extent of energy heterogeneity, is a
function describing the pattern of heterogeneity, and is the adsorbate-adsorbate
interaction energy. Equation (53) has also been extended to study multicomponent
adsorption equilibria. The general form of the extended equation is

where is the partial pressure in bulk phase. and
=
The cross adsorbate-adsorbate interaction energy is assumed to take the
k
The mole fraction of component in the adsorbed phase is calculated from
The form for the pattern of heterogeneity for the solid is assumed to be the same for
all components, and this pattern is assumed to take the form
To reduce the number of parameters in the the following assumptions

can be employed:
1. If the adsorbate-adsorbate interaction energy is negligible compared to that
of the adsorbate-adsorbent interaction, then is set to zero.
2 The saturation capacity is independent of temperature.
3. The saturation capacity is a function of the adsorbate.
Therefore. the number of parameters in the optimization can be reduced to
3 + = number of adsorbates) without seriously its perfor-
mance.
Kapoor al. proposed a heterogeneous extended Langmuir (HEL) model for

the description of equilibria on heterogeneous adsorbents. With
the integral equation approach of Eq. the general isotherm for a pure com-
ponent can be simplified as
where is the "local representing the amount adsorbed on a

homogeneous patch with an energy is the normalized distribution function;
and and are the minimum and energies, respectively. In one of
the cases where the local isotherm is described by the Langmuir equation
and follows the uniform energy distribution, Eq. (55) is analytically integrable,
and the final form is the famous LUD isotherm equation

where
In is the adsorption affinity at infinite temperature; and a are the

mean and square root of variance of energy, respectively; and is the heterogeneity
parameter related to the spread of the energy distribution.
When the system consists of a gas mixture containing components, the overall
adsorption equilibrium is described by the general integral equation, Eq. (16).
Kapoor et applied the extended Langmuir equation as a local isotherm and a
uniform energy distribution. The uniform distribution function is
elsewhere
The cumulative energy matching scheme is employed to address the energy
matching between different adsorbates in the adsorbed phase. With the uniform
energy distribution, it is
The HEL model then takes the form
Equation (59) must be solved numerically. The analytical solution is available only
under certain conditions (the variance of the energy distribution is the for all
components).
The HEL model is a noniterative model and has a reasonable prediction cap-
ability. With the pure component isotherm parameters derived the
equation, the model can be employed to predict the adsorption equilibria of gas
mixtures at any other temperature. Since the L U D equation is flexible in correlating
the pure component data, the HEL model prediction for a gas mixture is generally
good. Kapoor et demonstrated in their study that the HEL model gave
satisfactory results for a number of systems, which were comparable in cases
to the results from the IAST or even HIAST
However, arbitrary assumptions such as the energy distribution and
energy matching do exist in the HEL model. Meanwhile, the performance of the
HEL model depends on the L U D parameters derived pure component data.
When the quality of pure component data is not good or there are too few data
points, it is possible that the energy parameters (especially and derived
from the L U D equation are meaningless or inconsistent between different species.
In this case, the model prediction the H E L model must be treated with care.

The Langmuir (MSL) model is an extension of the Langmuir equation
to adsorption on multiple sites. Component i is to occupy a, sites (a, is not
necessarily an integer) in the adsorbed phase. The specific saturation capacity of
component is and for different components on the same adsorbent it satisfies
a, x = constant
The MSL equation can be derived from classical kinetics considerations or

from statistical mechanics. For a pure component, the MSL model equation is
and for a gas mixture, it is
MSL models present the correct Henry's law limit and obey the requirement of
thermodynamic consistency. This model is often used to account for the effect of
the size difference in the study of multicomponent adsorption equilibria
The MSL model can be extended to include effects such as lateral interactions in the
adsorbed phase as well as surface heterogeneity
et recently proposed a multispace adsorption model (MSAM) for

noncrystalline adsorbent. In this model, the pore is divided into two spaces:
Space I is close to the surface of the adsorbent where the adsorbate-adsorbent
interaction is strong, and space is further from the adsorbent, where the adsor-
bates interact with those in space I but have no direct interaction with the adsor-
bent. The amount adsorbed in space depends on the occupancy of space I. The
heterogeneity of the adsorbed phase is then due to the proximity of an adsorbed
molecule to the surface.
For species i in an N-component system, the amount adsorbed is then the
contribution both spaces,
+
The fraction of space I in the total pore volume is defined as
saturation capacity of i in space I
R=
saturation capacity of i on adsorbent
Hence the following equation holds:
The overall isotherm for species is

where
and are the constants for species i in spaces I and respec-

tively.
The adsorbed phases in spaces I and are assumed to be ideal, and hence the
IAST can be applied to each space separately The parameter R is the key
parameter for the MSAM; it reflects the structure of the adsorbent and is indepen-
dent of temperature and the adsorbing mixture. The value of R is determined by
using the MSAM to fit a limited amount of mixture data for an arbitrary binary
Because of this, the MSAM is more a corrective method than a predictive
It was demonstrated in the study of Gusev et al. that the MSAM gives a
good representation of both ideal and nonideal behavior of the adsorbed phase for
a number of systems that cannot be described by the IAST. For example. the
ideal adsorption behavior of the ethane-methane system on BPL carbon (Fig. 1 ) is
successfully simulated by the MSAM (dotted line in Fig. 1). It is
that the MSAM does introduce some arbitrary assumptions and unrealistic pictures
in describing the adsorbed phase; as the authors put it, "This model is intended for
engineering application rather than for the fundamental mechanisms of
ponent adsorption equilibria."
A similar method that treats the adsorbed phase as a combination of two phases
or patches was used by Li and Talu and G u o Li and Talu show
that with the combination of HIAST and the multicomponent van der Waals
model the highly unusual behavior of benzene-p-xylene vapor mix-
tures adsorbing onto silicalite can be more than qualitatively predicted. The
ideality of the system results in an S-shaped isobaric binary and a
crossover point of selectivity curves a t different pressures. Their results further
support the view that surface heterogeneity is the major contribution to the
'observed'"
. Martinez's
Martinez and proposes a general isotherm model for the adsorp-
tion of both pure gas and gas mixtures. This model takes into account a broad
range of factors such as adsorbate size loss of symmetry or chemical dissociation
clustering 111, and molecular interactions in the adsorbed phase The model is
based on the combination of the kinetics arguments of which account
for the molecular size effect, and statistical thermodynamics arguments which
account for other factors such as clustering and molecular interaction. This model
has two (1) the complete and (2) the reduced form in the Flory limit.
Only the latter form was investigated in detail by Martinez and
it

where 0, is the total coverage, M is the isotherm parameter, is the
of components in the system, and P is the bulk pressure. This model was
shown to reduce to simple isotherm equations such as those of
Sips. Nitta, and Fowler and Guggenheim in various limiting forms.
It can also be cast into a form similar to law as
=
where P is the pure component vapor pressure of species at the surface
coverage and temperature as the mixture. The parameters and are two distinct
parts of the activity coefficient, with the former reflecting the contribution due to
the size effect, segmentation, and clustering, while the latter addresses the molecular
interaction.
T o examine the model performance, Martinez and used the follow-
ing experimental data in their study: (1) the pure component and binary gas mix-
tures of hydrocarbon on activated carbon and on silica gel (2)
hydrocarbons and on zeolite (3) propane, and and their binary
mixtures on The adsorption nonideality of the binary systems
increases from mild (group 1) to high (group 3), and the model performance is
shown to be satisfactory for each group of data.
In spite of the merits of the model cited above, it suffers from the following
problems:
1. It needs binary data to derive the molecular interaction parameter thus it
is again a corrective rather than a predictive model (indeed, most existing
require binary data to represent the adsorption behavior of
the system).
It has five or more parameters to be optimized (depending on the
of the model), so there exist some uncertainties in the extracted para-
meters; for example, the optimized interaction energies may lack physical
meaning and consistency.
3 . In the simulation, the authors simplified the according to the degree of
nonideality of the experimental data; this may bear empiricism.
Keller and coworkers proposed a class of consistent

isotherm equations in which the size-scaling effects are encompassed by means of
the fractal dimension of the adsorbent. One interesting property of the Keller
model is that the influence of the adsorbent's geometry and roughness, expressed
by fractal dimension and the influence of the energetic heterogeneity, expressed
by the behavior of the adsorption energies with partial pressure,
clearly appear in the isotherm equation. The decoupling of these two effects may
enable us to achieve a clearer picture of the role of energeticgeometric hetero-
geneity in adsorption isotherms. The concept of fractal dimension was first

introduced by and Pefeifer in characterizing the surface structure in
adsorption processes and was recently investigated by Sahouli al. who
used the standard adsorption isotherm on various adsorbents. The Keller isotherm
IS
where is the fugacity and is related to the adsorption energy through the
equation
= = constant
The exponent a depends on the molecular size r and the fractal dimension of the
adsorbent:
where the subscript represents a reference species.

is the characteristic function. In the case of Langmuir monolayer coverage, it
IS
= +
and in the case of BET multilayer adsorption, it is
= - + Fx -
model was investigated in detail by Giona and coworkers and it

was shown that the model is able to simulate the highly nonlinear adsorption
behavior of the binary data provided by and
Activated carbon is produced from various carbonization and activation processes.

Its structure is known to be very complicated Microscopically, it is gen-
erally composed of amorphous carbons and graphitic crystals. In the adsorption
process, the contain void space (macropores and mesopores) for
talline transport, while the latter provide a microporous network in which most of
the adsorption capacity resides. The basic structure of the graphite crystal is
phite-like aromatic micro-crystals that accommodate slit-shaped micropores
between graphite layers (pore walls). The width of the is of molecular
dimension so that the interaction between adsorbate molecules and both sides of
the pore walls is enhanced, resulting in strong adsorption potential Under
a certain pore size, the adsorbate-pore interaction is highly sensitive to variation in
pore size; a small variation in pore size can result in a significant change in the

potential energy of interaction. When the system consists of adsorbates with dif-
ferent molecular properties, the difference in interaction energies will lead to
enhancement of one adsorbate relative to the others (selectivity). In physical
adsorption, where the dispersive interaction is predominant over the other interac-
tions, the micropore size distribution (MPSD) is the main source of surface hetero-
geneity, which dictates the overall adsorption equilibria of the system.
The slit shape of the micropore structure of activated carbon is supported by
several experimental observations: (1) sieving properties; (2) high resolu-
tion microscopy and (3) neutron diffraction and small-angle neutron scatter-
ing The most popular mathematical representation of the micropore
configuration in activated carbon is the slit pore contained in parallel graphite
layers of infinite extent. There are also other mathematical descriptions of the
pore configuration in the literature; for example, and Steele con-
sidered a pore with triangular section, and Segarra and Giant used circular
graphite platelets that are randomly oriented to represent the structure of activated
carbon.
Other sources of surface heterogeneity also exist in activated carbon. These
include functional groups and defects on the pore walls. the irregularity of the
shape of micropores, pore networking and blocking, and electrostatic interaction,
among others. These effects are too complicated to be quantified and treated indi-
vidually at present. A useful approach is to combine these effects in the '"equiva-
lent" MPSD, in which the interaction is taken as purely dispersive.
The role of surface heterogeneity in adsorption equilibria is well recognized
nowadays. The macroscopic equilibria properties such as the adsorbed,
selectivity. and heat of adsorption of each species reflect in general the aggregate
behavior of interaction between adsorbates and, more important, the interaction
between adsorbate and microporous network. Thus to correctly represent the
ticomponent adsorption equilibria on activated carbon, the surface heterogeneity
has to be properly addressed.
In this section, we briefly introduce a mathematical configuration of a
the potential energy of adsorbate-adsorbate interaction, and the
adsorption potential energy in micropores of different configurations. Then we
demonstrate two major applications based on the concept of adsorption potential
energy in slit-shaped micropores: (1 ) characterizing the pore size distribution (PSD)
of activated carbon and (2) predicting the adsorption equilibria.
In physical adsorption, the main interaction an adsorbate molecule and a

carbon atom on the graphite layer is the dispersive force. This potential energy is
adequately described by the Lennard-Jones 12-6 potential theory:
where z is the distance between the carbon atom and the adsorbate and
and are the cross LJ parameters, which are calculated using a geometric
mean and arithmetic mean (Lorentz-Berthelot rule), respectively,

The interaction potential energy between a n adsorbate molecule and all the carbon
atoms o n the pore wall is then the summation of all the interaction ener-
gies. This process of can be replaced by a n integration operation if the
shortest distance between the adsorbate molecule and the pore wall is greater than
the distance between carbon- carbon centers. Three configurations are commonly
used to describe the slit-shaped micropores in activated carbon: (1) The pore walls
are two parallel lattice layers of infinite extent; (2) the pore walls are two parallel
semi-infinite slabs; (3) the pore walls are a combination of many parallel lattice
layers separated by a distance A . The overall potential energy of molecules confined
in microporcs with the above pore configurations is 10-4, 9-3 and
10-4-3 potential energy respectively. T h e equations describing these potential
energies are described below.
1. The Potential
where
is the minimum interaction energy between the adsorbate molecule and a single
lattice layer, is the distance between the adsorbate and the central plane of the
pore, and is the number density of carbon centers per unit are of graphite surface
(38.2
2. The 9-3 Potential
where
is the interactive energy between an adsorbate molecule and a single slab

(pore wall) and is the number of carbon molecules per unit volume of the slab.

3. The 10-4-3 Potential
where
is the interaction energy between an adsorbate molecule and a single

lattice layer, and is the number density of carbon per unit volume.
The values of A and are taken as 0.335 and 114 respectively
These LJ potentials are temperature-independent, and among them the 10-4-3
potential is the most commonly used in the literature because its pore configuration
is the closest to the real pore structure of activated carbon. The 10-4 potential was
also found to be superior to the 9-3 potential in describing the microporous struc-
ture of activated carbon
Characterizing the microporous structure of activated carbon is an important and

yet very challenging issue. Until now there has been no method that was accepted
as a general tool even though many methods have been proposed in the literature.
Direct methods such as X-ray scattering, small-angle neutron scattering, and
nuclear magnetic resonance have the disadvantages of being of high cost and
doubtful interpretation. Classical adsorption methods such as M P molecular
probe, water adsorption, and TVFM are based on phenomenological
assumptions such as subcritical adsorbate homogeneity and incompressibility or
volume filling of pores and Gaussian distribution (TVFM). The inhomogeneity of
the adsorbed phase in smaller pores is not accounted for by these methods. More
specific reviews regarding the above methods for characterizing the PSD are avail-
able in Refs. 107 and 108.
The LJ potential theory in the last section provides us with a useful functional
link between adsorption potential energy and pore size, thus laying a foundation
for characterizing the PSD of activated carbon and for studying the molecular
assembly of simple gaseous molecules in graphitic micropores
Methods based on potential theory are available for describing the PSD of the
activated carbon. Three methods are briefly discussed here.
1. Horvath-Kawazoe Method
The Horvath-Kawazoe (H-K) equation was developed by and Kawazoe
to characterize the microporous structure of an adsorbent. They used the 10-4
potential to describe the potential energy of adsorbate molecules in a slit pore and

assumed that the average potential energy is equal to the free energy change upon
adsorption,
where
=
H = pore width
= number
density per unit area
= dispersion constants
From the amount adsorbed at Eq. (74) yields the pore width thus the pore
volume at this pore size can be obtained.
The implicit assumption behind the H-K equation is that the adsorbed phase is a
two-dimensional ideal gas, Henry's law region adsorption. Cheng and Yang
modified the H-K equation by including the nonlinearity of the isotherm. The
modified H-K equation is
where is the filling fraction. The modified equation generally results in

sharpening the PSD and shifting the peak position to smaller pore size.
2. Statistical Mechanical Methods
Statistical mechanics provides the of relating physical models for
adsorption in individual pores, expressed in terms of solid-fluid or fluid-fluid
interaction, to a single pore isotherm. Since it simulates the real adsorption process
in statistical mechanics gives a more accurate description of the of
adsorbents and has been used widely in characterizing the PSD of activated carbon.
Two methods are often seen in the current literature: (1) the approximate theory,
represented by the density function theory (DFT) and (2) the full Monte
simulation, represented by the grand canonical Monte (GCMC) or mole-
cular dynamics The single pore isotherm is related to the overall adsorp-
tion equilibria according to the equation

where is the number of moles adsorbed at pressure f is the differential
volume distribution of pore width, H) is the local molar density of adsorbate
at P (the single pore isotherm), and and are the lower and upper limits,
respectively, of the pore size range. In DFT, H ) is determined by the local
mean field density functional theory, whereas in GCMC, H ) is obtained by
taking the ensemble average. In both methods, the is assumed to be the sole
source of surface heterogeneity, the adsorbate-adsorbate interaction is described by
the 12-6 potential, and the adsorbate-adsorbent interaction is normally described
by the 10-4-3 potential. By inverting Eq. (76) respect to experimental isotherm
data, the PSD of the adsorbent can be derived.
(a) DFT Method The D F T is based on a mean field approximation of fluid-fluid
interaction. It provides a good description of a simple fluid (hard sphere) confined
in geometrically simple space and also has the advantage of being less
demanding than the full Monte simulation. et cal-
culated the single pore isotherm for nitrogen adsorption at 77K and applied the
results to obtain the micropore as well as the size distribution of the
activated carbon. Lastoskie improved the D F T method and proposed
the nonlocal density function theory (NLDFT). However, the D F T method also
has the problem of being inexact and cannot account for the steric effect of the
adsorbate molecule.
The consistency test of D F T was carried out by and more
recently by Quirke et Quirke et used the PSD obtained from nitrogen
adsorption data to predict the adsorption of methane at normal tempera-
ture. As a result, the methane isotherm is considerably overpredicted. This is
attributed to several reasons, two of which are that (1) at higher temperatures
other nondispersive molecular interactions are weakened and (2) the methane has
no dipole or which may enhance the adsorbate-adsorbent interaction
for adsorption.
(b) GCMC Method. The G C M C method can be conveniently applied to adsorp-
tion processes because its independent thermodynamic variables are temperature,
system volume, and chemical potential (which is related to the bulk pressure via an
equation of state). The adsorption isotherm for a given pore can be obtained by
evaluating the ensemble average of the number of adsorbate whose
chemical potential equals that of the bulk phase at a given temperature and pres-
sure The application of the G C M C method to characterize the of
adsorbent (especially activated carbon) has gained considerable popularity.
Gusev and used this method to characterize the PSD of BPL acti-
vated carbon with a methane isotherm at supercritical temperature (308 K). This
PSD is demonstrated to be capable of predicting the adsorption isotherm of
methane at two other temperatures (333 and with good accuracy. Later
the PSD obtained from methane adsorption was employed to predict the adsorption
of ethane on the same carbon Quantitative agreement was found between
simulation results and experimental data if the ethane molecule is considered as
the two-site LJ model. This is a strong indication of the consistency of the
PSD thus obtained.

Lopez-Ramon applied a more rigorous method to characterize the
PSD and pore connectivity (represented by the coordination number of the
pore network) of the activated carbon using the GCMC technique. Their study was
based on the idea that molecules of different sizes have access to different pore size
ranges. By combining and analyzing the PSD information obtained for each spe-
cies, the overall PSD for the adsorbent can be better represented. Three species
were used in their studies: and which are all nonpolar and approxi-
mately spherical but are different in molecular diameter. By analyzing the size
difference between the molecules, the obtained PSD is then represented by the
contribution of the PSD for each species. In their study, the tem-
perature was found to have some effect on the PSD derived for each species,
especially those for and
It was also found in their studies that to characterize the PSD of the activated
carbon at supercritical temperature, considerably higher pressure is needed to
obtain on micropores of larger sizes (near 2
. Prediction of Adsorption Equilibria

With the adsorbate-pore interaction potentials defined earlier, some researchers
attempted to predict the adsorption isotherm for some simple adsorbates based
on the known thermodynamic properties of the adsorbate and macroscopic infor-
mation of the adsorbent. So far only limited success has been achieved.
Fischer et proposed a model to predict the adsorption isotherm of
krypton in porous material at supercritical temperature. In their study, a model
pore of infinite length is formed by concentric cylindrical surfaces on which the
centers of solid atoms are located. The interaction between an adsorbate and an
individual center on the pore wall is described by the LJ 12-6 theory, and the overall
potential is the integral of this interaction over the entire pore surface. With ther-
relations, Fischer et obtained the functional dependence of the
saturation adsorption excess and the Henry's law constant on the pore structure.
The isotherm was then produced by the interpolation between Henry's law range
and saturation range. They tested their theory with the adsorption of krypton on
activated carbon. It was shown that, with information on the surface area of the
adsorbent and thermodynamic properties of the adso bate. their model gives more
than qualititative agreement with experimental data. If a few experimental data
such as the Henry's law constant at one temperature are available, the isotherms for
all temperatures and pressures can be predicted with good quality.
Forbert and Schlunder proposed a model that is able to predict the adsorp-
tion equilibria of arbitrary organic vapors on activated carbon at subcritical tem-
perature by using information of a reference isotherm or a characteristic curve for
the same adsorbent. The model is based on the idea that for organic vapor on
carbonaceous adsorbents, the molar binding energy is a function of the volume of
the adsorbed phase only while the characteristic curve can be regarded as an
integral volumetric distribution of the binding energy. The molar binding energy
between adsorbate and carbon surface is defined as the difference between heat of
adsorption and enthalpy of evaporation:

In the case of a temperature-independent characteristic curve. the following equa-
tion is valid:
= A
where is the volume adsorbed and A is the adsorption potential
A = - RT
So for organic vapors on carbonaceous adsorbents, is a function of the
of the adsorbed phase only. Forbert and then made the key assumption
that the interaction potential in the middle of the pore is crucial for the pore filling
and derived this interaction potential on the basis of the 12-6 potential theory as
where
=2 - 2.78 is the adsorbate diameter
= 0.35 the distance between layers of activated carbon
c,, and can be calculated from known thermodynamics parameters
H is the pore width
Following their assumption, is to by
Thus for a given adsorbate, the adsorption potential A is directly related to

pore size. However, due to its assumptions this model is valid only at subcritical
temperature for the adsorption of vapor adsorbate with spherical geometry.
Many researchers recognized the role of the micropore size distribution

(MPSD) in the study of adsorption equilibria on activated carbon. The MPSD
model is based on the idea that adsorption energy heterogeneity is induced
by structural heterogeneity, which can be characterized by the size distribution of
the slit-shaped micropores. The MPSD is an intrinsic property of activated
carbon, and in physical adsorption it dictates the adsorption equilibria through
the dispersive interaction between adsorbate and the network of
activated carbon.

422 ang and Do
If the MPSD of the activated carbon is represented by a distribution function

with being the fraction of having a half-width between r and
+ and the local isotherm is represented by the overall adsorption
of a species k at a pressure P and temperature can be expressed as the
integral of the local isotherm over the accessible micropore range:
where is the maximum micropore half-width and is the minimum pore

width below which that species cannot penetrate.
Equation (81) suggests that the surface heterogeneity is induced solely by the
MPSD. This approach is fundamentally sound than the approach of energy
distribution in since the distribution of energy sites, not follow
any particular type of distribution function. On the other hand, the microporous
structure of the activated carbon results from the random processes of carboniza-
tion and activation and therefore may possibly be characterized by some distribu-
tion functions. Hence, assuming a is more reasonable than arbitrarily
imposing a function In the MPSD approach, can be derived from
according to the relationship
For a given set of adsorption equilibrium data, (81) can be used (inverted) to
derive the MPSD if a local isotherm is specified. But such an inverse has been
known to be an ill-posed problem, and presents some difficulties in application
For example, for a given adsorbate and a given local adsorption isotherm,
one would find many different distribution functions that can yield the same overall
adsorption capacity (within experimental error). Furthermore, if the isotherm equa-
tion, is applied individually for each adsorbate, we might obtain very
different for the same activated carbon! This violates the basic premise
that the MPSD is an intrinsic property of the given carbon, independent of what
adsorbate is used. Realizing this inconsistency problem often seen in the literature,
we propose a different approach here, applying Eq. (81) to multiple temperature
data of many adsorbates simultaneously while constraining the MPSD as an intrin-
sic characteristic of the adsorbent. T o achieve this, the functional dependence of
interaction energy on size has to be defined first.
Adsorption Potential and Slit Pore Configuration

For the physical adsorption of a nonpolar adsorbate on activated carbon, the
adsorbate-adsorbent interaction is much stronger than the adsorbate-adsorbate
interaction. This interaction potential is represented by 10-4-3 potential,
and the adsorbate-adsorbent interaction energy can be taken as the negative of the
potential energy minimum inside the slit-shaped micropore. This concept was first
proposed by and Schwarz to characterize the MPSD of activated
carbon. and later it was refined and found other applications Although
this method does not take into account the adsorbate-adsorbate interaction, which

may be important a t high surface coverage. it catches the essential feature of the
adsorbate-pore There is also some justification for using it in the high
pressure range; for example, Lopez-Ramon that the
model isotherms obtained Monte simulation are fitted to the Hill-
de Boer isotherm, the variation of adsorption energy versus size closely
resembles the variation of the 10-4-3 potential pore size. Thus
solving Eqs. (73) for the potential the dependence of interac-
tion energy on pore can be obtained, and for a given MPSD the
energy distribution can be derived from Eq. (82).
The Model Equation

The local adsorption isotherm is assumed to take the form of the
equation
is the local adsorption affinity, which is assumed to take form
If the adsorbate--adsorbate interaction is taken into account, can take the
where 0 is the surface fractional of the adsorbate. is the

adsorbate interaction energy, and is the number of neighboring adsorbed mole-
cules. With the local adsorption isotherm is as
Guggenheim equation. For the adsorption of hydrocarbon gases or vapors on
activated carbon, is to be adequate to describe the local adsorption
affinity
In physical adsorption, the adsorption affinity at the infinite temperature level
of different species are correlated to
This equation is derived by assuming the collision of gaseous molecules toward the
pore mouth of and such a collision rate is known from the kinetic
theory of gases to be inversely proportional to the square root of the molecular
weight and to the square root of The affinity in (85) is
specific only to the solid The theoretical value of reported as
4.26 x
The maximum adsorption capacity of species, takes temperature-
dependent form
where is the adsorption capacity at reference temperature and is the

thermal expansion coefficient.

The MPSD to follow distribution function. A nonnegative
gamma distribution is used here:
where and are the parameters.

all the parameters defined. the overall adsorption isotherm of (81) can
be expressed explicitly of the MPSD for adsorbate as
where
The parameters and are the lower and upper limits, respectively, for the
inicropore size range. The lower limit is for a pore that has the same adsorption
potential energy as that of a single lattice layer or for a pore that has zero potential
energy. The of pores having sizes two values of is very
and hence value of will suffice. The upper is off at
for convenience, since the adsorption potential energy in pores with
widths larger than this is effectively zero,
The equations have distinct of parameters to be optimized: (1)
the characterizing carbon structure ( 2 ) the isotherm
By optimizing data
of many adsorbates on tlie same activated carbon, it is possible to derive
the results for size distribution as as relevant adsorption para-
meters for
The adsorption equilibrium a heterogeneous surface is

than that of a pure component system. As previous sec-
tions, two need to be addressed properly: (1) the size difference
between various adsorbate molecules and ( 2 ) the local interaction energy (or match-
ing energy) various adsorbate molecules, Issue 1 suggests there
sibly exists a certain pore range that adsorbates with larger size cannot
penetrate (size but that is accessible to adsorbates of smaller sizes,
issue 2 requires a relationship between interaction energy of each adsorbate the
pore size. These two are well accounted for by the MPSD approach.
The MPSD can be conveniently extended to describe the
equilibria of a gas mixture with components. The generalized MPSD
takes the extended equation as the local while using
the MPSD obtained from pure data:

This model has distinguishing features in addressing the two issues:
The equilibrium interaction between different adsorbates is confined
the same pore they are accommodated; adsorbates to a
pore d o not participate in the and are accounted for in
the local extended equation for that pore effect of
size is reflected in the lower
2. The matching energies between different species are related to
interaction strength with local micropores, scheme, which
is termed the adsorbate- pore interaction scheme, introduces the con-
dition for determining energies different species on a
heterogeneous surface.
the cumulative energy scheme lacks
a fundamental basis.
The size exclusion and the matching for the binary pair ethane- propane
is graphically in Fig, as an example It is seen that for pores of
width between 3.21 and 3.38 A, only ethane is allowed whereas propane is excluded.
Both ethane propane adsorb of larger than
What is interesting is that pores whose half-width is 3.38
3 4 5 7
Pore half-width
Plots of energy of interaction of ethane and propane versus pore

ethane actually has a higher than propane, and it is only pores
half-width is greater than 3.55 A that propane have a higher adsorption affinity
than ethane. This means that separation factor of the binary system can change
pore to pore. This example micropores of differ-
sizes the local interaction energy is a of not only the size of the
micropore but also the size of the adsorbate molecule.
In this section the MPSD model with pure component adsorption

data, that were measured for 11 species on activated carbons, Ajax carbon
and The properties of adsorbates and
tal temperatures for pure component system are listed in Tables and 2 for Ajax
activated carbon and Nuxit activated carbon, respectively.
In the model, the selection of the LJ diameter of the adsorbate is very
important, since the LJ diameter dictates the accessibility of that species the
micropore of the adsorbent. For inolecules having and
we propose to use the minor dimension, as molecules tend to
into micropores along their smallest configuration
1. Ajax Activated Carbon
The MPSD model the local equation and 10-4-3 potential is opti-
with respect to the isotherm data of eight species measured on activated
carbon (Table 1 ) at at least three different temperatures for each species. The
temperature for adsorption capacity, is chosen as The
fitting data for ethane and propane are
plotted Fig. The fit of other adsorbates (methane, benzene, toluene,
a
0
0 20 40 60 0 40 60 80
Pressure Pressure
. Plots of MPSD fitting and experimental data of ethane and propane on
activated carbon. (From

and Adsorption Temperatures for Ajax Activated Carbon
Property Methane Ethane Propane Butane Benzene Toluene
0
3.3 3.6 3.8 3.9 4.3 4.3 3.7 3.8
195.2 335.4 148.6 300.0 237.1 331.4 412.3 457.1
258, 273, 298, 323, 258, 273, 283, 303, 283, 303, 303, 303, 333, 303, 363,
303 373 283, 303 333 333 333,473 363 473
2 Adsorbates Adsorption Temperatures for Activated Carbon

Property Methane Acetylene Ethylene Ethane
3.3 3.8 3.3 3.9 3.9 4.5 4.3 4.3
195.2 148.6 231.8 224.7 300.0 248.9 237.1 331.4
293,313, 293,313, 293,313, 293,313, 293,313, 293,313, 293,313,
333, 363 333. 363 333, 363 333, 363 333, 363 333, 363 333, 363 333, 363
3 for Ajax Carbon

--
Property Methane Ethane Butane SO?
7 491 8 6 004 4 863 4 845 4 393 14 91
10 47 14 39 70 20 29 49 68 92
Ref 121

carbon dioxide, and sulfur dioxide) is also very good This is regarded as
excellent in the light of the fitting done on all isotherm data simultaneously (a total
of 27 curves). The optimized structure parameters are 88.69, and
= 1.04 x The optimized isotherm parameters are listed in
Table 3.
The derived for carbon are shown in Fig. 4, with the
average width This average value falls within the range of
micropore size determined by experimental methods for many samples of activated
carbon The affinity parameter for Ajax activated carbon, is approximately
one-fourth of the theoretical value of 4.26
Turning to the isotherm parameters optimally obtained, we found that
5 (the expansion coefficient) is very small, suggesting that the satura-
tion capacity is not affected by the change of temperature, at least within the
range studied.
2. (adsorbate-lattice layer interaction minimum) increases with the carbon
number for a given class of adsorbate family (ethane, propane, butane).
Aromatics have higher values of than those of paraffins, with toluene
being a more strongly adsorbing species than benzene.
3. of each species is very close to its theoretical value calculated by
4. The values of (adsorption capacity at 273 K) for and are a little
bit higher than those of the other species.
3 4 5 6 7
Pore half-width
.4 Plots of the derived micropore size distribution for activated carbon and
Nuxit activated carbon. (From Ref. 126.)

Figure 5 shows plots of the optimized interaction energy obtained from the
fitting versus the theoretical interaction energy calculated for the
10-4-3 potential. The theoretical parameters are calculated using the
properties given by Reid and It is noted that the optimal
and theoretical values for are very comparable to each other. On the other
hand, such good agreement is not observed if the 10-4 or 9-3 potential is
employed in the model equation (not shown in the figure). and this might suggest
that the 10-4-3 potential is appropriate to describe the dispersive interaction in
activated carbon.
Although the model fittings for and are good. we must be cautious
about the applicability of the MPSD model to these polar adsorbates, because the
dispersive force is not enough to describe the interaction between polar adsorbates
and carbon surface.
2. Nuxit Activated Carbon

Next, the equilibrium data on Nuxit activated carbon are used to test the model.
The data are collected for eight species at four temperatures (Table 2). To show the
predictability of the model, we choose the data at three temperatures (283, 303, and
333 K) for optimization and then use the results to predict the remaining data at
363 K. Like the case of Ajax activated carbon, the fit between the model and all the
Theoretical interaction energy
Plot of the optimized interaction energy versus the theoretical interaction

energy calculated from Eq.

data is excellent The optimized structural parameters are
17.93 = 81.11, and 3.84 The optimized
isotherm parameters are tabulated in Table 4.
As for Ajax carbon, the value of 5 is very small. The affinity parameter for
Nuxit carbon is a bit smaller than that of Ajax carbon, suggesting that the adsorp-
tion affinity at infinite temperature on Nuxit carbon is slightly lower than that on
Ajax carbon. The MPSD for Nuxit carbon is also plotted in Fig. 4, and we see that
the MPSDs of these two activated carbons are very comparable.
The optimized parameters are used in the model to predict the equilibria data of
these eight species at 363 K. The predictions are excellent clearly supporting
the temperature dependence of the model.
With the structural parameter and isotherm parameters obtained, we employ the
extended MPSD model to simulate the binary adsorption equilibria. As a criterion
to judge the model performance, the related simulation results from IAST are also
presented for the related systems,
1. Ajax Activated Carbon
The binary adsorption equilibria on Ajax activated carbon are measured on a
rig for the following systems (1) methane-ethane (2)
methane-propane (3) (273 and 303 K). For each binary
system, the total pressure for the bulk phase was kept at about 500 Figure
6 shows a typical plot of the amount adsorbed versus the mole fraction of methane
for the system methane-propane-Ajax activated carbon. The continuous lines are
predictions from the MPSD model, and the dotted lines are from the IAST. As the
figure shows, the predictions from the MPSD model are better than those from the
IAST. This trend is also observed for other binary systems studied.
2. Nuxit Activated Carbon
For Nuxit activated carbon, the binary equilibria of the following systems were
simulated: (1) methane--ethylene (2) ethylene-propylene (3)
ethane-propane (293 and 333 K); (4) (293 K); (5) ethane-methane
(293 K); (6) Like the case of Ajax activated carbon. the
model predictions from the MPSD model are again excellent, suggesting that the
MPSD model is appropriate to study multicomponent equilibria of hydrocarbons
on activated carbon.
of the
The MPSDs
A comparison of the MPSDs derived for the two activated carbons in Fig. 4 reveals
that Nuxit activated carbon has slightly larger average pore size and variance than
Ajax activated carbon does, although the difference is not significant. This obser-
vation reflect the general feature of the activated carbons. On the

Ajax carbon, 303K
methane (1)
propane (2)
Methane molar fraction in gas phase
FIG, 6 Plots of the amounts of methane and propane adsorbed versus the mole
fraction of for Ajax activated carbon. Total pressure is 500 torr, and the
temperature is 303
other hand, the MPSD derived from a KOH chemically activated nutshell-derived
activated carbon shows considerable difference in MPSD from those of commercial
carbons the physical and the reliability of the derived MPSD
deserve further investigation. The microporous structure of activated carbon is very
complex; the assumption of any form for will impose some constraint in the
inverse process of (81). Furthermore, the assumption of the slit-shaped model
of infinite extent is an oversimplified picture of a real micropore, in
which the other sources of surface heterogeneity, as we stated before, are not
accounted for. Due to the sensitivity of the LJ potential to the pore size, the
MPSD obtained will locate mainly in the region of one to two LJ diameters. So
the MPSD derived here represents the effect of micropores of all sizes on
the overall adsorption equilibria. Considering the complexity of the carbon struc-
and the two-parameter gamma distribution function, the MPSD obtained this
way should be treated qualitatively rather than quantitatively, in spite of its sig-
nificant role in describing both pure and binary adsorption equilibria on hetero-
geneous surfaces.
2. The Multicomponent MPSD Model

From the practical point of view, the extended MPSD model offers several
advantages:

1. The M P S D derived is based on the extensive equilibria information of many
pure component data, which is necessary to account for surface hetero-
geneity.
2. It gives a good fit for the equilibria of many species over a broad range of
temperature.
The model has the predictive capability for equilibria.
4. The model is noniterative and can be conveniently applied in the study of
adsorption kinetics o n a heterogeneous surface This is important
because the application of a practical kinetic separation process depends
critically o n a model that reliably describes equilibria,
In its theoretical aspect, the MPSD model also provides us with some insight
into the role of surface heterogeneity. It takes into account the size exclusion effect
and sets a physical criterion for the local matching energies between different
adsorbate molecules in the micropore. Although the local extended iso-
therm in the model is simplistic in its description of true equilibria, this problem can
be overcome by applying the concept of IAST in each and the overall
adsorption equilibrium is the contribution from the local equilibria over the com-
plete M P S D range. However. the model will become iterative and very
while its performance improves only marginally. The significance of this
sacrifice for purely theoretical purposes is questionable.
However, as we discussed before, the present model represents only a primitive,
picture of equilibria in activated carbon. All other
interaction mechanisms are also lumped into the potential minimum of the adsor-
bate-pore interaction, as the model does not account for interactions with the same
adsorbate molecule, nor does it reflect the importance of surface loading as the pore
filling progresses. These suggest that more fundamental work is needed for a full
understanding of the adsorption equilibria on heterogeneous
solids.
In recent years, the behavior of fluid molecules in small pores has been well
studied by computer simulation Computer simulation can provide us
with valuable information about the microscopic behavior of adsorbate mole-
cules confined in small pores in terms of intermolecular and surface forces,
thus enabling us to understand the fundamental behaviors of adsorbates in the
potential field of the pores. The current trend in the literature suggests
that for physical adsorption in activated carbon, the adsorbate-adsorbate inter-
action and adsorbate- pore interaction are well represented by the LJ potential
theory while the model micropore is a slit-shaped channel of infinite extent.
This forms the basis for the appliction of the statistical method in adsorption
processes.

Although the statistical mechanical method can provide us with accurate
descriptions of the of and the microscopic behavior of
adsorbates in micropores, it is still not feasible to use these methods to directly
compute the equilibria in practical processes because of the inten-
sive computational requirement. What is valuable to us is that it enables us to have
deep insight into the separation properties (especially selectivity) of certain systems
in pores of various sizes, thus possibly providing us with guidelines in making
or selecting the right adsorbent. An example is the study of the adsorbent structure
for natural gas storage the adsorption selectivity for a
system depends on several factors: (1) the adsorbate-adsorbate
interaction; (2) the adsorbate-adsorbent interaction; (3) the pore configurations,
including the length and width of the micropore; and (4) the shape and size of the
adsorbate molecules. All these factors fall within the capability of statistical
mechanical methods. At present, studies are carried out mainly for the binary
system consisting of species with simple geometry in graphitic model
pores,
Finn and Monson first tested the predictability of IAS theory for binary
systems using the isothermal isobaric Monte simulation on a single surface.
However, this does not represent real adsorption systems. Tan and
conducted detailed studies on the binary equilibria of the methane-ethane
system in slit-shaped micropores using the nonlocal density function theory
(NLDFT). The selectivity of ethane to methane was studied in terms of pore
width, temperature, pressure, and molar fractions.
The selectivity of ethane to methane is defined in terms of density or molar
fraction as pore phase, = bulk phase; 1 = methane, 2 =ethane)
At the of zero pressure, the selectivity is
which is determined by the ratio of Henry's law constants for each species and
hence independent of the bulk fluid D F T theory, this is expressed as
where = is the external field experienced by a fluid molecule at

for a given pore size So is entirely determined by the ratio of the integrated
Boltzmann factors of the external force of the two components, and this suggests
that the selectivity at zero pressure limit for a certain system depends only on the
temperature and pore size.

At various pore sizes, exhibits a for each temperature, the position
of which is approximately located at = 2. When temperature increases,
the position of this shifts slightly to pore size. In small pores,
decreases quickly, as ethane is geometrically more restricted than methane. When
H* 1.7, methane will be selectively adsorbed in comparison with ethane in such
small pores.
When pressure is high, the fluid-fluid force also plays an important role in
selectivity, and as a result the selectivity also depends on bulk-phase pressure
and molar fraction. Tan and Gubbins introduced a classification scheme for the
isotherm (selectivity versus pressure). They classified the shape of the
selectivity isotherm at different temperatures into four types: Type I is for cases
where the temperature is significantly higher than = capillary critical tem-
perature for the mixture), in which the selectivity isotherm shows a single max-
imum; type occurs where the temperature is just above in which the
selectivity isotherm has double maxima; types and IV occur when temperatures
are below and are characterized by discontinuities corresponding to condensa-
tion and layering transition. Later, another type of selectivity isotherm was dis-
covered by Cracknell and Nicholson and was appended to this classification
scheme as the type V selectivity isotherm. Type V selectivity presents a monotonic
decrease with an increase in pressure, a phenomenon frequently observed experi-
mentally.
Cracknell and Nicholson first studied the selectivity of the same methane-ethane
system using the G C M C technique in which the ethane molecule is treated as a
spherical fluid They also investigated the predictability of the IAST theory
for the binary mixture in pores of different sizes. Their study was carried out at the
supercritical temperature (T = 296.2 K), and the results were compared with those
of Tan and Gubbins. It was found that in most cases the D F T overestimates the
selectivities (of ethane), but there is qualitative agreement between the results from
the two methods. The following reasons are attributed for the discrepancies: (1) The
weighting function for density expansion does not have enough terms; (2) the mean
field assumption for the attractive force is inadequate.
Cracknell and Nicholson further investigated the same binary system using a
more realistic model for ethane, in which the ethane was treated as two sites
connected by a certain length of bond The simulation results showed that
the elongated shape of ethane strongly influences the pore size dependence of
selectivity and that the confinement in may cause a significant
decrease in the selectivity. Their studies were also extended to the ethane-propane
system in which the ethane and propane were considered as two- and
site LJ fluids, respectively. Again, using the simulated single-component data as
input, the IAST was found to be able to predict the selectivity with reasonable
accuracy.
et al. also studied the selectivity of the ethane-methane system with
the G C M C method in model graphite micropores of various sizes at supercritical
temperature and under a broad range of pressure. Both adsorbates were treated as

spherical LJ fluids. It was found that in the low pressure range the IAST describes
the binary system behavior well for all pore size ranges. This is in accordance with
our expectation, as IAST becomes exact in the Henry's law region. At the simula-
tion temperature (308 K), the predictability of IAST is seen to be the best at pres-
sures less than 10 bar, and it deteriorates sharply at higher pressure. This is partly
attributed to the fact that the IAST calculation procedures require extrapolation of
the simulated methane isotherm to pressures well above the mixture pressure and
even the normal pore-filling pressure in many cases. The overall predicted selectiv-
ity on model solids (with different is found to be reasonable and to be as
good as that in a single pore.
The issue of multicomponent adsorption still poses a fundamental problem to

researchers working in the adsorption area. Many mathematical models have
been proposed to address this and there have been reviews such as
those cited in this chapter. The choice of an appropriate model rests on the
balance of simplicity and capability of the model as well as the intended appli-
cation of the model. For example, for quick calculation of the
equilibria of similar adsorbates (such as low-order paraffin gases), simple
such as the and the extended Sips model are adequate. However, to
study the influence of the porous structure of activated carbon on the adsorption
equilibrium properties of nonpolar adsorbents, the MPSD model is complex
enough, while still retaining some simplicity, to explain the effects of factors
such as size distribution and adsorbate molecular properties on the
adsorption equilibria.
Two factors play important roles in the study of adsorption equilibria of mix-
tures on heterogeneous adsorbents like activated carbon and hence have received
attention in the development of various models. They are (1) the size difference
between the adsorbate molecules and (2) the adsorption energetic heterogeneity, or
the structural heterogeneity of the adsorbent if the adsorbate is nonpolar. For
conditions, these two factors are adequate to account for the nonideal beha-
vior of the adsorbed phase, and they are readily accounted for by the MPSD model.
This model needs to be further applied and tested in many systems to verify its
capability, and it needs to be modified (fine-tuned) to account for many factors
such as the irregular structure of the micropore as well as the inclusion of functional
groups to deal with polar adsorbates.
Statistical mechanical methods like G C M C or D F T have a good prospect in
characterizing the PSD of activated carbon and in multicomponent
adsorption equilibria at the microscopic level. However, the study of nonideal
adsorption behavior with these methods is not a trivial task. The nondispersive
interaction in the adsorption process also needs to be quantified.
Systematic and reliable experimental data of single adsorbates and binary and
ternary mixtures on samples of activated carbon with various pore structures and
surface chemistry are badly needed for the critical evaluation of models of
component adsorption equilibria.

This work is supported by the Australia Research Council
Ruthven, Principles o f Adsorption and Adsorption Processes. New

York, 1984.
Yang, Gas Separation by Process. Boston, 1987.
Eng. Res. 30:1032 (1991).
Ind. Eng. Chem. Dev. (1984).
Jaroniec and Madey. Physical Adsorption on Heterogeneous Solids,
Amsterdam. 1988.
and P. Danner, AIChE (1980).
and P. Danner, (1980).
Kabel, and Danner, AIChE (1985).
0. Talu and Kabel, AIChE (1987).
and Phys. (1988).
and P. Danner, J. (1982).
Ruthven and (1985).
M. P. Valentin, V. C. and Guiochon, Colloid
Interface 141:127 (1991).
A. L. Myers and AIChE (1965).
Richter, and A. L. Myers. Chem. (1989).
A. and Farooq. AIChE J. (1996).
A. and A , Myers, Ind. Chem. Process Des. (1985).
Moon and Res. (1987).
A. and A. Myers, (1988).
and A . Rodrigues, AIChE (1994).
A. Myers. Fundamentals p. 365. (A. Myers and G. Belfort, eds.)
Engineering Foundation. New York (1984).
and A. Myers, Adsorption Data Handbook,
Prentice-Hall, Cliffs. 1989.
Toth, Colloid Interface
Toth. Colloid Interface (1997).
Toth, Colloid Interface (1997).
A. Myers and (1972).
Frances, Siddiqui. Ahn, Chang. and
(1995).
A. Myers, Belfort, and M . Int. Eng. (1992).
AIChE 1 1 35 995).
and A. L. Myers. Chem. Eng. (1994).
Costa, J. and C. Marron, AIChE J. (1981).
0. and I. (1986).
Paludetto. G . G. S. Carra, and M. Eng.
(1987).

Adsorption and Catalysis on Heterogeneous Surfaces, Sci.
USSR, Moscow, 1948.
S. Ross and P. Oliver, On Physical Adsorption, Interscience. New York, 1964.
Hoory and W. Chem. Eng. Sci. (1967).
Jaroniec and Rudzinski, Phys. Rev. Lett. (1975).
Rudzinski and Everett, Adsorption of Gases on Heterogeneous Surfaces,
Academic, Orlando, 1992.
W. Rudzinski, Moon, and Rhee, Chein. Eng. Sci.
(1995).
Langmuir (1987).
P. A. Myers, Talu, and I. (1988).
Moon and Chem. (1988).
X. Hu and D. Do, (1995).
K. W. Riemsdijk. Wit, and J. Colloid
Interface 1 (1994).
Giona, and Chem.
(1995).
Li and 0. Talu, Chem. (1994).
V. A. R. and A.
(1996).
M. Carbon (1989).
Stoeckli, Kraehenbuenl, and Carbon
27: 125 (1989).
Chen and R. Yang, Colloid Interface (1996).
Jaroniec and Choma, in Equilibria and Dynamics o f Gas Adsorption on
Heterogeneous Surfaces Rudzinski, A. Steele, and Zgrablich,
eds.) New York, 1997.
Bhatia and H. K. Langmuir (1994).
Kraehenbuehl, H. Stoeckli, and Carbon (1990).
Jaroniec, Madey, and Carbon (1989).
Carbon (1988).
V. Dobruskin, Langmuir (1996).
and Le Van. (1988).
P. Bering, V. V. Serpinski, and I. Surinova, Dokl. Phys. Chem. (English
translation) (1963).
Lewis, R . Gilliland, B. and W. Cadogan. Ind. Chem.
(1950).
S. Doong and R. Yang, Chem. Res. (1988).
A. Lavanchy, and Stoeckli, Adsorp. Technol.
996).
Cook and Can. Chem. Eng. 42:146 (1 964).
Lewis, E. Gilliland, and P.
(1950).
Grant and Manes, Eng. Chem. 490 (1966).
and R. P. Eng. (1985).
H. Moon and Chem. (1988).

68. Moon and C. Sep. Technol. (1993).
69. and A. Myers, AIChE (1986).
70. I. Kinetics Theory of the Liquid State. Akad. Moscow, 1945.
71. C. Henry, Phil. Mag. (1922).
72. A. Physical Chemistry of Surfaces, New York. 1976.
73. U. and U. Schlunder, Eng. Process. (1990).
74. Liedy. U. Schlunder, and Turek, Process. (1987).
75. R. Sips, Chem. Phys. (1948) (1950).
76. B. Hsu, and Rudzinski, Can. Chem.
(1985).
77. R. Dreisbach. and U. Keller, 5th Int. Conf. Fundamentals of
Adsorption, Boston, 1995.
78. Do and H. Do, Chem, (1997).
79. A. Kapoor, A. Ritter, and Yang, (1990).
80. T. Nitta, Shigetomi, Kuro-oka. and J. Chem. Eng.
(1984).
81. Guo. 0 . and Hayhurst, (1989).
82. A. and A. Myers, in Fundamental of Adsorption p. 451,
Engineering Foundation, New York, 1987.
83. Martinez and Diran Chem. Eng. (1996).
84. M. Martinez, Chem. Eng. (1994).
85. K. Lewis, Gilliland, and Am. Chem.
72:1160 (1950).
86. Lewis, Gilliland, and Am, Chem.
72:1 1 57 (1950).
87. Szepsey and V. Chim. Hung. (1963).
88 L. Szepesy and V. Chim. Hung. (1963).
89. Keller, Bunsenges. Phys. Chem. (1988).
90. Keller, A. (1990).
Sailer, Dreisbach, and Keller. AIChE Meeting, St. Louis,
Missouri, 1993.
92. Avnir and P. Chim. (1983).
93. and (1997).
94. Giona and Giustiniani, Chem. Res. (1995).
95. Giustiniani, M. and Chem. Res.
(1995).
96. S. Gregg and Sing, Adsorption, Surface Area and Porosity, Academic,
London, 1982.
97. P. Carrott, A. Roberts. and K. Sing, Characterisation of Porous
Solids, Elsevier, Amsterdam, 1994.
98. Do, Chem. Eng. (1996).
99. Everett and Chem. Faraday Trans. (1976).
100. A. Steele, The Interaction of Gases with Solid Surfaces. Pergamon, New York.
1974.
101. H . Marsh. Crawford, and A. Wennerberg, Carbon
(1982).
102. Kaneko, and Nicholson, (1994).

and W. A. Steele, Surf. (1988).
and A. Steele, Ber. Bunsenges. (1990).
E. I , Segarra and Chem. Eng. (1994).
and de Boer, (1965).
P. Russell and Carbon (1994).
Kaneko. Membrane (1994).
Evans and Chem. Phys. (1986).
Kaneko, Colloids Surf. (1996).
Horvath and Kawazoe, Jpn. (1983).
Cheng and Yang, Chem. Eng. (1994).
Gubbins, M. and Nicholson, Res.
(1995).
A. P. and A. Quirke, Carbon (1989).
C. Gubbins, and Quirke, Phys. Chem, (1993).
Jiang, and K. Gubbins, Phys. (1994).
Quirke and Tennison, Carbon (1996).
Nicholson and Parsonage, Computer Simulation and Statistical
Mechanics Adsorption Academic, Orlando, 1982.
V. and A. Langmuir (1997).
V. Gusev and A. Langmuir (1997).
V. Lopez-Ramon. Jagiello, Bandosz, and A. Langmuir
(1997).
Fischer, and Findenegg. Chem. Eng.
(1983).
Forbert and Schlunder, Chem. 34-49 (1995).
Jagiello and A. Schwarz, J. Colloid Interface (1992).
Jagiello and A. Schwarz, (1993).
Wang and Do, (1997).
X. and Do. (1994).
Jagiello, Bandosz, and A. Schwarz, (1996).
P. J. Chein. Phys. (1965).
Rao. R. Jenkins, and A. Steele, Langmuir, (1985).
W. Breck. Zeolite Molecular Sieves: Structure, Chemistry and Use, New
York, 1974.
J. Fryer, and F . Stoeckli, Carbon (1989).
Reid. J. and Polling, The Properties o f Gases and
A. Wang, and Do. AIChE (1998).
Do and Wang, (1998).
P. and Computer Liquids, Clarendon,
Oxford. 1987.
Evans. in Fluids Henderson, p. 85. Marcel Dekker,
New York, 1992.
K. A. L. Myers, and Chem. (1992).
E. Finn and P. A. Monson, Phys. (1991).
Tan and Gubbins. Phys. (1989).
Tan and Gubbins, Phys. (1992).

142. F. Cracknell and Nicholson, Chem, Faraday Trans. (1994).
143. Cracknell and Nicholson, (1993).
144. Cracknell, D. Nicholson, and (1994).
145. V. A. Jensen, and A. in Fundamentals
of Adsorption Boston, 337 1996.

Department of Chemistry,
Kent State University, Kent, Ohio
I. Introduction
I I . Methods of Characterizing Porous Silicas on the
Basis of Nitrogen Adsorption Data
A. Introductory remarks
B. Evaluation of the specific surface area and pore volume
C. Calculation of the pore size and pore size distribution
D. Estimation of the adsorption energy distribution
III. Studies of Porous Structures and Surface Properties of
Organic-Modified Conventional and Ordered Mesoporous Silicas
A. Nitrogen adsorption properties of unmodified conventional
and ordered mesoporous silicas
B. Physically coated mesoporous silicas
C. Conventional mesoporous silicas with bonded organosilanes
D. Ordered mesoporous silicas with bonded organosilanes
IV. Conclusions and Future Research Perspectives
References
Porous silicas are widely used in areas of modern science and technology as
adsorbents, catalyst supports, additives, and chromatographic packings, to men-
tion just a few Many of the applications result from their high surface area,
tailored particle size and pore size, and high mechanical, thermal, and chemical
stability Another advantage of porous silicas is their high surface reactivity
toward coupling agents, which makes it possible to obtain a broad range of
materials with desired surface properties that are useful for the separation of

mixtures (high performance liquid chromatography, ion-exchange chromatogra-
phy. size exclusion chromatography, gas chromatography), catalysis, and biochem-
ical applications For instance, chemical bonding of was found
to be a very convenient procedure for attachment of various ligands of different
size, surface coverage, and functionality The of a proper silica
support and an organosilane used for the modification provides an opportunity to
obtain materials with high mechanical and stability; tailored particle size,
pore size, and surface area: and the desired functionality and surface properties.
The discovery of ordered mesoporous materials opened new
perspectives in the field of synthesis and application of mesoporous silicas. OMMs
were reported for the first time in 1990 by Kuroda and coworkers who
synthesized sieves using layered polysilicate precursors.
However, it was the discovery of scientists from Mobil Oil published in 1992
that attracted considerable attention and initiated a significant and still
growing number of studies in the direction of synthesis, application, and character-
ization of ordered materials
The Mobil scientists reported details of the synthesis and structure of a new
family of mesoporous silicas and called and put forward a
hypothesis concerning the synthesis mechanism The materials were
obtained from inorganic precursors in the presence of
surfactants. Depending on the synthesis conditions, three different types of pro-
ducts were obtained: hexagonally ordered MCM-41, bicontinuous cubic MCM-48,
and lamellar MCM-50. The MCM-41 structure exhibited a system of long straight,
approximately cylindrical channels arranged in a honeycomb structure with no
connections between the channels. MCM-48 featured a three-dimensional porous
structure of cubic symmetry, exhibiting curved interconnected channels with
constant diameter. MCM-50 was a lamellar phase silicate or
aluminosilicate layers separated by surfactant layers. As-synthesized MCM-41 and
MCM-48 materials had their pore voids filled with surfactant molecules, but after
the surfactant was removed by calcination at elevated temperatures (ty ically about
823 K), porous structures of remarkable pore volumes (above /g) and sur-
face areas (above were obtained. The structure of the lamellar MCM-50
collapsed during calcination. The Mobil scientists proposed that the synthesis of
these remarkable materials proceeds via a liquid crystal mechanism. It is
well known that surfactants in water form various kinds of micelles, which can
assemble into liquid crystalline phases. For example, cylindrical micelles can be
formed and aggregate in a honeycomb structure. It was speculated that silicate or
aluminosilicate precursors condense on such preformed liquid crystalline matrices
and, for instance, the MCM-41 phase results from the condensation of inorganic
precursors around a honeycomb assembly of micelles. Alternatively, cylindrical
micelles may first be covered by inorganic species and then assemble into the
honeycomb structure.
Subsequent studies showed that the liquid crystal mechanism is gen-
erally correct, but its details are usually somewhat different from the two ways
outlined above It was demonstrated that in most cases the synthesis of
novel mesoporous materials involves the formation of ion pairs between surfactant
ions and inorganic precursors, their assembly into the liquid crystalline phase, and

subsequent condensation of the inorganic species. The liquid crystal templating
mechanism was directly confirmed by a successful synthesis of several new phases
most of which have their counterparts in liquid crystalline
factant-water systems.
In accord with the reports of Mobil scientists novel materi-
als were found to exhibit surface areas of about most of which was the
surface area of the so-called primary mesopores, structures with
uniform pore diameters. The external surface area was found to be about an
order of magnitude smaller. most cases, there was no evidence of the presence
of micropores (pores of width below 2 The sizes of
typical MCM-41 materials range from about 2 to 4.5 and synthesis procedures
for good quality large-pore materials with pore diameters up to at least 6.5 have
recently been developed Usually, the primary volume of
ordered mesoporous materials is in the range of 0.5-1 The capillary con-
densation in primary mesopores of MCM-41 manifests itself as a pronounced and
often almost vertical step on the adsorption isotherms Only in the
case of small-pore samples with pore diameters close to 2 n m are there no notice-
able condensation steps, and the increase in the adsorbed amount at higher pres-
sures is very small, indicating that primary mesopores are filled with the adsorbate
at low relative pressures (below about 0.2)
The Mobil scientists reported that the surface properties of MCM-41 are close to
those of silica gels which is not surprising when one takes into consideration
their composition and the lack of crystallinity of their pore walls
Consequently, surfaces of MCM-41 and other ordered mesoporous silicas feature
groups (single o r and bridges. groups may be
hydrogen-bonded with one another or with physisorbed water molecules. Water
adsorption studies revealed that the surface of both siliceous and
MCM-41 samples is relatively hydrophobic in with that of
conventional silica gels. Several adsorption studies indicated strong surface hetero-
geneity in accord with well-known surface properties of silica gels
and silica-aluminas
The presence of silanol groups on the surface of MCM-41 and other ordered
mesoporous silicas provides an opportunity to modify the surface via chemical
bonding of organic ligands. Many surface modifications of ordered
materials by chemical bonding have been reported and some of the
resulting materials have been shown to be catalysts with interesting properties
or efficient adsorbents for the removal of heavy metal ions
Moreover, it was demonstrated that surface modification via
significantly improves the mechanical and hydrothermal stability of silica-based
ordered materials It is also interesting to note that mesoporous
molecular sieves with covalently bonded organic groups can be synthesized via
condensation of and in a one-step proce-
dure
The aim of this chapter is to provide the methodology for characterizing unmo-
dified and modified porous silicas by using nitrogen adsorption. Owing to the
significant and still growing interest in the synthesis, modification, and application
of ordered mesoporous silicas, problems of their characterization are emphasized

and considered detail. In a possibility of complex characterization of
both porous structure and surface properties of modified porous silicas on the basis
of nitrogen adsorption measurements in a wide pressure range is presented.
Methods for determining the surface area, pore and pore size distribution
are discussed. Moreover, calculations of adsorption energy distributions from low
pressure nitrogen adsorption data are outlined, because they provide new oppor-
tunities in characterization of surface properties of modified porous silicas. In
Ill, several recent studies of modified conventional and ordered mesoporous silicas
are presented to demonstrate the usefulness of the adsorption data analysis
described in Conclusions are discussed in IV. Due to immense amount
of literature about the characterization, modification, and application of porous
silicas, the list of references is by no means complete and is focused mostly on the
synthesis, characterization, and modification of ordered mesoporous materials,
which currently attract a good deal of attention.
. Introductory
Nitrogen adsorption at 77 is a standard and widely used method to determine the
specific surface area and pore size distribution of adsorbents
This technique is also applied to determine microporosity
However, it is rarely appreciated that analysis of low pressure nitrogen adsorption
data affords considerable insight into the surface properties of porous materials
Adsorption instrumentation capable of performing low pressure measure-
ments (down to relative pressures of about is currently commercially avail-
able. Therefore, it is convenient to carry out nitrogen adsorption runs over the
entire pressure range in order to obtain information about surface area, porosity,
and surface properties from a single adsorption isotherm. It is well known that
porous usually exhibit strong structural and surface heterogeneities, the
former caused by the presence of pores of different sizes and shapes, and the latter
resulting from surface irregularity, roughness, and the presence of impurities and
various functional groups. Adsorption processes on solid surfaces are influenced by
both structural and surface heterogeneities, and it is often difficult to recognize
which of these factors is responsible for adsorption properties observed experimen-
tally.
Our recent studies of nitrogen adsorption on mesoporous silicas
were aimed at exploring the possibility of separating pore struc-
ture and surface functionality effects in order to develop better characterization
procedures for these materials. The low pressure adsorption properties of porous
silicas were found to be markedly different from those of many other adsorbents,
such as graphitized and carbons and porous and zirconia or
even alumina However, low pressure nitrogen adsorption on numerous
conventional and ordered mesoporous silicas proceeds in essentially the
way, indicating that nitrogen probe molecules are quite insensitive to details of
silica surface and structure The similarity of the surface

properties of different silica samples allows for the easy detection of micropores,
even if they are present in relatively small amounts Moreover, nitrogen
adsorption was shown to be significantly affected by the introduction of
silanes in chemical bonding modification procedures, and the changes in adsorption
properties were clearly dependent on the structures of the bonded species
These studies demonstrated that nitrogen adsorption is a powerful
technique to analyze both the porous structure and surface functionality of
dified and modified mesoporous silicas. Before the results of our studies are
reviewed, it is important to discuss selected fundamental aspects of adsorption
data analysis and to present characterization techniques that were found to be
very useful in studies of porous silicas.
According to IUPAC recommendations pores are classified with respect
to their size as (width below mesopores (width between 2 and
50 nm), and macropores (width greater than 50 For mesoporous silica gels,
nitrogen adsorption isotherms of type IV (IUPAC classification are commonly
observed. These isotherms usually exhibit pronounced hysteresis loops at relative
pressures above about 0.4, arising from capillary condensation of nitrogen in
pores. However, if mesopores present in the material d o no exceed a diameter of
about 4 nm, there is no adsorption-desorption reversibility. If, in addition, the pore
size distribution is narrow. which is the case for ordered mesoporous silicas, sharp
reversible steps of capillary condensation are observed, and the resulting adsorp-
tion isotherms may be referred to as reversible type IV isotherms When a
material has mesopores of sizes below 4 n m with a broad pore size distribution
or the size approaches the micropore size limit an adsorption
isotherm resembles type I behavior characteristic of microporous adsorbents (rever-
sible adsorption isotherm, which is concave to the "relative adsorption" axis and
levels off at low relative pressures). In this case, the character of the
sample can be verified by using the comparative plot in the range of low
pressure nitrogen data.
The IUPAC recommendations classify hysteresis loops into four types
and two of them, H I and H2, are typically observed for many
silicas. Hysteresis loops of the type have two parallel, nearly vertical branches
and were often reported for materials composed of agglomerates (assemblages of
particles rigidly joined together) or compacts of approximately spherical molecules
arranged in a fairly uniform way Recently, type H I loops were also observed
for materials of cylindrical pore geometry and a high degree of pore size uniformity
Thus the presence of the type of hysteresis loop on the adsorption
for a porous solid appears to indicate high pore size uniformity and facile
pore connectivity.
A type H2 hysteresis loop has a triangular shape and a very steep desorption
branch. Such behavior was observed for many porous inorganic oxides and was
attributed to pore connectivity effects which were often defined as the presence
of pores with narrow mouths (inkbottle pores), but the latter identification may be
grossly oversimplified Indeed, triangular hysteresis loops were observed even
for highly ordered MCM-41 materials with pore sizes of about 4-5 nm
For such samples, desorption (capillary evaporation) of nitrogen from primary
takes place at relative pressures of 0.4-0.5, in the region where

hysteresis loops usually close. This results in very steep desorption branches of
nitrogen adsorption isotherms and triangular hysteresis loops. It can be concluded
that the H2 hysteresis loops may result not only from poor pore connectivity but
also from the transition from the irreversible to reversible adsorption-desorption
behavior in a relative pressure range characteristic of a given adsorbate (0.4-0.5 in
the case of nitrogen at 77
valuation of the
The Brunauer-Emmett-Teller (BET) method provides a standard procedure to
calculate the specific surface area of mesoporous materials However, if an
adsorbent has an appreciable number of mesopores with sizes close to the
pore size range (2-3 nm), adsorption data from the region of capillary condensation
in these mesopores should not be used in the BET calculations In such cases,
the accuracy of surface area determination can be improved by using the compara-
tive plot method as described below. The total pore volume can be estimated
from the amount adsorbed at a relative pressure close to the saturation vapor
pressure (usually about 0.99) by converting the adsorbed amount to the corre-
sponding volume of liquid adsorbate The conversion factor is equal to
0.0015468 for nitrogen at 77 K when the adsorbed amount and the pore volume
are expressed in cubic centimeters at STP per gram and cubic centimeters per
gram, respectively, and it is assumed that the density of condensed adsorbate is
equal to the density of liquid adsorbate at the temperature of measurement. This
procedure provides the volume of pores with widths below 100-200 depending
on the pressure value used in the calculations. In the case of adsorption isotherms,
which exhibit a steep increase in the adsorbed amount at pressures close to the
saturation vapor pressure, the estimation of the total pore volume may be quite
inaccurate.
In many cases, information about structural properties of high surface area
materials can be obtained by comparing the adsorption isotherm of a solid under
study with the adsorption isotherm of a reference macroporous material of similar
surface properties The comparison is made by plotting the amount
adsorbed for the solid under study as a function of the amount adsorbed for the
reference material at the same pressure. The adsorption for the reference is
usually expressed either as the statistical film thickness of the adsorbate on the
surface (in the t-plot method) or as the standard reduced adsorption (in the
plot method) The quantity a, is defined as the ratio of the adsorbed amount at
a given pressure to the adsorbed amount at a relative pressure of 0.4.
In the case when the solid under study and the reference material have the same
surface and structural properties and differ only in the amount of surface area, the
comparative plot is linear and starts from the origin of the graph. However, usually
the plots obtained on the basis of experimental data exhibit deviations from linear-
ity, which can be analyzed to obtain information about the structural and surface
properties of a given solid. If the solid under study has they are filled at
low relative pressures, and the amount adsorbed in them does not change signifi-
cantly with further pressure increase. The increase in the gas uptake for higher
pressures results mostly adsorption on the external surface area of the sample,

which proceeds via multilayer adsorption, as is the case for the reference material.
Therefore, the comparative plot curves for many solids rise steeply at
low pressures and then gradually level off and exhibit linearity at higher pressures.
The linear part of the comparative plot can be extrapolated to the adsorbed amount
axis, and the intercept provides the amount adsorbed in micropores, which can
be converted to the micropore volume by multiplying by a proper conversion
factor. slope of the linear part of the comparison plot can be used to calculate
the external surface area of the sample, as described elsewhere
Another type of comparative plot curve can be observed for mesoporous sam-
ples. the region of low pressures, the adsorption on both the mesoporous solid
and the reference adsorbent proceeds via the mechanism,
multilayer formation, so the initial part of the plot is linear, and its slope can be
used to calculate the total surface area of the mesoporous sample However,
in the region of higher pressures, capillary condensation takes place in mesopores,
and the increase in the adsorbed amount is relatively much larger for the
ous solid than for the macroporous reference, and consequently the comparative
plot is bent upward. Subsequently, the plot levels off when mesopores
are already filled with condensed adsorbate. For ordered mesoporous solids, those
higher pressure parts of the plots can be used to calculate the primary
volume and the external surface area (the surface area of other pores present in
these materials) in a way analogous to that described above for microporous
adsorbents. If the material under study contains both micropores and mesopores,
the deviations of the comparative plot linearity can be observed in both low
pressure and high pressure regions.
The considerations presented above are based on assumptions that surface prop-
erties of the reference adsorbent are close to those of the solid under study.
However, it is also possible to use the comparative plot analysis to study the surface
properties of adsorbents. It was recently reported that the shape of low
pressure part of the plot reflects the differences in interactions of the adsorbate with
the compared surfaces, provided microporosity is not present. This finding is
important in the analysis of modified silicas, whose surface properties
are dramatically different from those of unmodified silicas due to the removal of
surface silanols and the introduction of organic groups of certain functionalities.
When macroporous silica gels are used as reference adsorbents. comparative plots
for modified materials exhibit marked downward swings resulting from the
decrease in nitrogen-surface interactions as a result of the modification with
organic species. These results are discussed in detail in the next section.
Many methods to calculate the pore size are currently available (see Refs. 74, 75,
and 84 and references therein), so it is important to choose a proper tool for the
analysis of mesoporous silicas. Our studies showed that the
Halenda (BJH) method is a very useful characterization technique In
the BJH approach, the adsorption process in mesopores was considered to consist
of the formation of the adsorbed layer on pore walls and subsequent condensation
of the adsorbate in the interior of the pore. The desorption process was assumed to

proceed in the opposite way, by capillary evaporation the interior of the
pore followed by a gradual decrease in the thickness of the adsorbed film the
pore surface. In order to develop a useful method for the analysis of porosity, the
relationships between (1) the pore size and the capillary
pressure and (2) the statistical film thickness and pressure need to be known.
Barrett et al. that the formation of the adsorbed film is essentially
independent of the pore size and used the statistical film thickness curve for
a macroporous reference adsorbent. To describe the relation between the capillary
evaporation pressure and the pore radius (denoted here as the Kelvin equation
with the correction for the statistical film thickness was used:
where is the molar volume of the liquid adsorbate, 7 is its surface tension, R is
the universal gas constant, and T is the absolute temperature. In the case of nitro-
gen adsorption at = 8.88 and =
In the BJH method, calculations are performed starting from the
desorption isotherm data for the highest pressures attained during the experiment,
where it is that all pores are completely filled with the condensed adsor-
bate. When the pressure is subsequently lowered, a certain number of the largest
pores are by the capillary evaporation, but the walls of these pores are still
covered by the adsorbate film. The decrease in the amount adsorbed, which was
caused by the lowering of pressure, can be recalculated to obtain the volume of
these emptied pores. but in order to do that, assumptions need to be made about
pore geometry and density of condensed adsorbate. The latter can be approximated
by the density of liquid adsorbate at the saturation vapor pressure and temperature
of the As far as the geometry is concerned, Barrett
assumed that pores are cylindrical in order to make calculations more tractable.
Hence, the volume of the inner space in the pores that were emptied during the first
evaporation step is equal to the volume of liquid adsorbate corresponding to the
measured decrease in the adsorbed amount. T o arrive at the total volume of those
pores, one simply needs to account for the thickness of adsorbate on the pore walls.
The pressure for which the considered decrease in adsorption took place can now
be related to the pore radius through the Kelvin equation [Eq.
A reasoning similar to that described above for desorption from the largest
pores during the first pressure-lowering step can be applied to subsequent decreases
in the adsorbed amount. But in these cases, the decrease in adsorption is caused not
only by the capillary evaporation but also by thinning of the adsorbed layer in
pores in which the capillary evaporation has already taken place. The calculations
are repeated until a sufficiently low pressure value is reached or the cumulative pore
volume obtained using the procedure described above equals or exceeds the total
pore volume. The BJH procedure follows the sequence of events during desorption
measurements; it actually does not make any difference whether adsorption or
desorption data are used. However, different relations between the pressure and
the pore size may need to be used for capillary condensation and evaporation
It needs to be noted that the Kelvin equation [Eq. was shown to underestimate

the size of pores below 7.5 nm increasingly inaccurate as the pore
size decreases. Moreover, the assumptions of the BJH method are invalid for
micropores (width below since the formation of the adsorbed film on the
pore wall is affected by the proximity of the opposite pore wall and cannot be
adequately described by the statistical film thickness curve for the macroporous
reference adsorbent.
As was already mentioned, nitrogen adsorption-desorption at 77 K on the
majority of porous solids is reversible at relative pressures below about 0.40
and only above this pressure limit can the irreversibility, the pre-
sence of hysteresis loops, be observed. Therefore, lower closure points of hysteresis
loops are often observed at pressures close to this limiting value. The existence of
the pressure range for which crossover between irreversible and reversible adsorp-
tion behavior takes place has significant implications for characterization of porous
solids on the basis of adsorption data. The commonly used method to calcu-
late pore size distributions is based on the Kelvin equation which
was shown to predict adsorption-desorption irreversibility during adsorption in
pores of the commonly encountered shapes cylindrical, slitlike)
Therefore, in the cases of materials for which experimentally measured hysteresis
loops close at relative pressures of about 0.4-0.45, exhibiting steep desorption
branches, calculations of the pore size distribution from desorption data will indi-
cate narrow pore size distributions, which is likely to be grossly misleading
In our studies of modified conventional mesoporous silicas that had pore

sizes greater than the BJH method was used, and calculations were per-
formed on the basis of data. These materials exhibited narrow HI
hysteresis loops, which indicated facile pore connectivity and permitted the appli-
cation of desorption data, which be inappropriate for many other mesoporous
silicas, as discussed above. The statistical film thickness curve used assumed the
form of the equation: [nm]
A reliable assessment of pore size appears to be a major problem in the char-
acterization of ordered mesoporous materials, for which pore sizes are
perse and can be precisely controlled in the range from 2 to at least 6.5 by using
proper synthesis procedures. At the same time, traditional adsorption methods of
pore size estimation, those based on the Kelvin equation, are expected to be
inaccurate for such small mesopores Therefore, there is a need to
develop new accurate and reliable approaches to the pore size analysis for unmo-
dified and modified ordered mesoporous silicas and other nanostructured porous
materials. In some cases, one can take advantage of the simple geometry of ordered
porous structures (referred to as primary inesopores) of these materials. For
instance, in the case of long, approximately cylindrical primary mesopores
(MCM-41, HMS, etc.), one can calculate their diameter on the basis of their surface
area and pore volume using the = The primary
size of MCM-41 materials can also be calculated from X-ray diffraction
(XRD) data and the primary volume using a based on
geometrical considerations of the honeycomb structure characteristic of the
41 materials

where is the (100) interplanar spacing obtained from X-ray diffraction, is the
density of siliceous pore walls (which can be assumed to be equal to that of amor-
phous silica, 2.2 and is a constant dependent on pore
geometry assumed and equal to 1,213 for cylindrical pores. Calculations based on
(2) were shown to be consistent and are expected to give very accurate results.
especially for good quality MCM-41 samples Equation (2) can also be
applied for such materials as SBA-15 It should also be noted that the assump-
tion of a constant pore wall thickness for MCM-41 samples can be employed for
approximate evaluation of the pore size, which can simply be estimated by
tracting the assumed pore wall thickness from distance between pore centers
The latter can be obtained from XRD measurements. Our extensive studies of
MCM-41 materials [based on Eq. confirmed that in cases the pore wall
thickness is approximately constant (0.8 0.2 but such an assumption is not
always valid, especially for samples that exhibit small primary volumes
(below 0.5
The methods of pore size evaluation for ordered mesoporous silicas described
above are not generally applicable for all kinds of ordered mesoporous structures,
and some of them cannot be applied for modified samples. Therefore, computer
simulation studies of ordered mesoporous materials were undertaken to improve
the understanding of adsorption on these materials and improve the ability to
characterize their porous structures Adsorption on MCM-41 was modeled
using nonlocal density function theory (NLDFT) and computer simulations
to establish the relation between the pore size and the pressure of the capillary
condensation in primary mesopores. Based on N L D F T data a program was
developed to calculate pore size distributions for MCM-41 These advanced
computational approaches seem to be very promising, but there are still
questions concerning the interpretation of computational adsorption isotherm
data. Namely, there is no agreement in the literature as to how the
branches of the calculated isotherms and the point of equilibrium
transition between these branches correspond to the experimental
branches of adsorption isotherms. Moreover, the programs to calculate
pore size distributions for ordered mesoporous materials on the basis of computer
simulation o r N L D F T data are still not available to the majority of the scientific
community,
In response to the demand for an accurate method for evaluating pore size
distributions of ordered mesoporous silicas and other materials, empirical equa-
tions were recently developed to describe relations between the capillary condensa-
tion pressure and the pore size in cylindrical siliceous These formulas
were derived using good quality MCM-41 materials with pores in the range of
6.5 nm as model adsorbents, and their pore size was estimated using Eq. (2). The
following relation was found between the pore radius and the pressure of nitrogen
capillary condensation in the pores a t 77 K

+
where is the equilibrium vapor pressure, is the saturation vapor pressure, and
the statistical film thickness curve t is described by the equation
Equations (3) and (4) can conveniently be used in different methods of

size evaluation, of which the BJH approach appears to be particularly suitable
These relations were carefully tested on numerous MCM-41 samples and to
be in excellent agreement with pore sizes calculated using with differences
usually below 0.1-0.2 Pore size calculations based on Eqs. (3) and (4) can
conveniently be performed for both modified and unmodified ordered mesoporous
silicas and can be expected to provide highly accurate results. especially for siliceous
samples with cylindrical pores. Therefore, the BJH method of calculations based on
these equations was used in our recent studies of modified MCM-41 materials
It needs to be noted that Eq. (3) was derived for nitrogen capillary con-
densation (adsorption) data and should not be used in calculations based on
orption data. Moreover, it was shown that the desorption data are not appropriate
for the pore size evaluation for mesoporous materials with pore sizes in the range of
about 4-5.5 due to the transition between reversible and irreversible adsorption
behavior at relative pressures of about 0.4, as already described above. Because of
this behavior, the relation between the capillary evaporation (desorption) pressure
and the pore size would have a more complicated and less useful than Eq. (3)
derived for the capillary condensation process
It is well known that the adsorption of a gas on a solid surface depends the
interactions between the surface and gas molecules These interactions
can be described in terms of the adsorption energy distribution (AED) for a given
adsorbate-adsorbent The AED can be analyzed to obtain information
about the surface, about its structure and functional groups
In order to calculate one needs two fundamental pieces of information.
First, adsorption isotherms for homogeneous surfaces interacting with the adsor-
bate with specified adsorption energies need to be known Second, one needs
to account for the topography of adsorption sites random, o r inter-
mediate between these two extremes) The topography is expected to
influence the adsorption on a heterogeneous surface unless lateral interactions
(interactions between adsorbed molecules) are negligible. T o deal with these pro-
blems, one needs to make suitable approximations. Under certain assumptions
concerning the properties of the adsorbent and the adsorbed film (geometry of
the surface, number of layers of adsorbed molecules, lateral interactions, mobility
of adsorbed molecules, etc.), one can derive analytical expressions for adsorption
isotherms on homogeneous surfaces For instance, the Fowler-Guggenheiin
(FG) equation was derived for localized adsorption with lateral

actions on homogeneous surfaces. The equation appears to be particularly
interesting, since it has a simple analytical form and yet accounts for lateral inter-
actions in the adsorbed layer. During the last decade, it became possible to generate
adsorption for homogeneous surfaces with specified adsorption energies
by employing the nonlocal density function theory (NLDFT) or computer
simulations Therefore, accurate model adsorption data for homogeneous
faces are now readily available. On the other hand, effects of topography of adsorp-
tion sites are rather difficult to account for adequately in calculations of adsorption
energy distributions using the integral equation of adsorption discussed
below) and often have to be neglected to make the calculations feasible.
T o obtain the adsorption energy distribution from experimental adsorption
data, one can use a formalism of the integral equation for the overall adsorption
This equation simply states that the overall adsorption is a sum of the
contributions from adsorption on all of the adsorption sites present on the
surface:
where is the total adsorbed amount, is the degree of surface coverage on

sites of adsorption energy as a function of pressure (later referred to as the local
adsorption isotherm), and after integration is the monolayer capacity of
adsorption sites with energies from the interval to It is sometimes more
convenient to use an alternative form of (5) that assumes the form
where the overall adsorption and were divided by the monolayer capa-
city of the adsorbent in order to obtain the overall degree of surface coverage
= and the fraction of adsorption sites of energy
The monolayer capacity can be calculated using the standard BET
T o calculate the population of adsorption sites as a function of adsorption
energies normalized AED), the integral equation for the overall adsorption
(5) or needs to be inverted. The required overall adsorption isotherm is
directly measured experimentally, but the form of the local adsorption isotherm
must be specified. The F G adsorption isotherm was used in our calculations, and
the reasons for its selection are discussed elsewhere One of the useful
features of the F G equation is that it can easily be modified to account for differ-
ences in lateral interactions in the cases of different topographies of adsorption
sites. For patchwise topography, the standard expression for the adsorption
isotherm can be used:

where the lateral interaction energy [accounted for by the exponential in [Eq.
for a molecule located on a site with adsorption energy equal to is dependent
on the average number of nearest neighbor molecules (assumed to be proportional
to the degree of surface coverage on the surface patch of this adsorption energy.
In the case of random topography of adsorption sites, it is more realistic to assume
that the lateral interactions of a molecule located on a certain adsorption site are
dependent on the degree of the surface coverage on the whole surface rather than
on adsorption sites of the same energy. Therefore, the local adsorption iso-
therm for the random topography of adsorption sites the form
In Eqs. (7) and T is absolute temperature, is the Boltzmann constant, is the

number of nearest-neighbor molecules in the monolayer, and is the interaction
energy between two nearest-neighbor In our studies of mesoporous
silicas, was set to be equal to 380 and the random topography of adsorp-
tion sites was assumed. is the constant,
The preexponential factor can be calculated from partition functions for an

isolated molecule in the gas and surface phases
To invert (6) with respect to the normalized distribution of adsorption
sites, the INTEG program was used This advanced numerical procedure is
based on the regularization method, which is required to overcome the difficulty
caused by the inherent numerical instability of the inversion of integral equations
that have the general form of Eq. (6) It also should be noted that there is
a commercially available program to calculate adsorption energy distributions,
which is part of the D F T Plus package of software for the pore size and adsorp-
tion energy calculations developed by (Norcross, GA). It was
reported that the local adsorption isotherms were calculated using the
N L D F T formalism and that the inversion of the integral equation for overall
adsorption [Eq. was performed using the regularization method to provide
numerically stable results.
T o successfully synthesize modified silicas with the desired properties,

it is important to characterize the structure of the silica support used for modifica-
tions. For that purpose, nitrogen adsorption is particularly useful, because it pro-
vides an opportunity to obtain such crucial data as specific surface area,
size distribution, and micropore volume. In addition, low pressure

adsorption data can be used to study the surface properties of porous materials, as
was discussed in the previous section. However, in the case of silicas,
low pressure adsorption was rather insensitive to the details of their porous struc-
ture. Shown in Fig. 1 are adsorption isotherms for a macroporous silica gel
LiChrospher Si-1000 (BET specific surface area of 25 and an ordered
mesoporous MCM-41 silica with a pore size of 5.0 nm For
1000, nitrogen adsorption proceeds via multilayer formation up to pressures close
to the saturation vapor pressure, where capillary condensation may also contribute
to the observed steep increase in the amount adsorbed. In the case of the MCM-41
material, multilayer adsorption takes place on the surface of primary
and on the external surface at relative pressures below about 0.4; then capillary
condensation in primary mesopores is observed and the adsorption isotherm levels
off. Despite completely different porous structures and the resulting differences in
high pressure adsorption behavior, these two silicas exhibit very similar low pres-
sure adsorption properties, which can be seen from the strong similarity of their
adsorption energy distributions calculated from submonolayer adsorption data (see
Fig. 2). One can conclude that the difference between the and
porous structures of silicas d o not affect their surface properties with respect to
nitrogen probe molecules This observation has two important implications.
First, it opens an opportunity to obtain a detailed comparative plot analysis of not
only high pressure data, which are quite insensitive to surface properties of
bents but also low pressure data for mesoporous silica. Second, changes in low
pressure adsorption properties that may result from surface modification can be
attributed to different surface properties of the modified material rather than to
changes in the structures (the latter usually involve mainly a decrease
the pore diameter).
0.0 0.2 0.4 0.6 0.8 1.0

Relative Pressure
.1 Comparison of nitrogen adsorption isotherms for the silica gel
LiChrospher Si-1000 and an ordered MCM-41 material with pore size of
5 (Data partially taken from Ref. 59.)

Adsorption Energy
.2 Comparison of adsorption energy distributions for the silica gel
Si-1000 and an ordered mesoporous MCM-41 material with pore size of
5 (Data partially taken from Ref. 59.)
Nitrogen adsorption was used to study the applicability of a physical coating

procedure for surface modification of conventional and ordered mesoporous
silicas. A LiChrospher Si-100 silica gel with an average pore diameter of
(estimated using the conventional BJH on the basis of desorption
data) was coated with 4'-cyano-4-biphenyl [4-(4-pentenyloxy)] benzoate (CBPB)
samples are denoted as CBPB-AX, where A indicates the type of
ificaton procedure physical coating; chemical bonding; and BT, chemical
bonding and subsequent coating) and X provides the weight percent of CBPB in a
given material. For comparative purposes, a LiChrospher material with
bonded CBPB was also prepared and will be referred to as
Moreover, CBPB-B22 was coated with an additional amount of CBPB, and the
resulting material is denoted as The MCM-41 material chosen for the
physical coating procedure had a pore size of 3.8 nm and its properties are
described in detail elsewhere The MCM-41 samples coated with CBPB are
referred to as CBPB MCM-41, where X denotes the weight percent of CBPB.
As can be seen in Fig. 3, the introduction of physically coated species led to a
significant decrease in the adsorbed amount and consequently in the pore volume
and surface area but the overall shape of the adsorption isotherms remained
unchanged, indicating that the mesoporous structure of the silica support was not
altered. However, the average BJH pore size decreased after the
depending on the amount of the organic species introduced and the
procedure used. The largest decrease was observed in the case of chemical bonding
(from 12.9 for the unmodified silica gel to for CBPB-B22) and was much

LiChrospher Si-100
CBPB-B22
0.0 0.2 0.4 0.6 0.8 1.0

Relative Pressure
.
3 Nitrogen adsorption for the unmodified silica gel
LiChrospher Si-100 and physically coated (CBPB-T24) and chemically
bonded with CBPB. (Data Ref. 77.)
pronounced than in the case of physical coating. F o r instance, CBPB-T24

with the CBPB loading similar to that of CBPB-B22 had a larger pore diameter
1 2 vs. The reasons of such behavior became clear when the pore size
distributions for these samples were examined (Fig. 4). Namely, the physical
coating resulted not only in a shift of toward smaller pore sizes, but also in a
depletion of smaller pores. Therefore, the physical coating procedure can be
expected to provide much less uniform surface coverage than the chemical bonding
LiChrospher Si-100
CBPB-T24
CBPB-B22
10
Pore Size (run)
FIG. 4 BJH pore size distributions calculated using desorption isotherms for the
unmodified silica gel LiChrospher Si-100 and samples physically coated
and chemically bonded (CBPB-B22) with CBPB. (Data from Ref. 77.)

procedure. This conclusion was supported by examination of low pressure adsorp-
tion data, which is discussed below.
To show the influence of the modification procedure on the surface properties of
the materials studied, the comparative plot method can be used to analyze low
pressure adsorption data. As can be seen in Fig. 5 , initial parts of the comparative
plots for the silicas exhibit pronounced deviations from linear behavior,
which was observed for the silica support, confirming that the latter has essentially
the same surface properties as macroporous reference silica (LiChrospher
1000). The plots for modified materials are bent downward to an extent dependent
on the CBPB loading and the modification procedure. Increases in the amount of
organic modifier caused a systematic decrease in the low pressure adsorption.
Examination of the comparative plots for CBPB-T24 and CBPB-B22 shows that
the plot for the chemically modified material is bent downward to a slightly larger
extent, indicating somewhat more pronounced changes in surface properties. The
latter may be due to a more uniform surface coverage. in agreement with the pore
size data shown in Fig. 4 due to removal of high energy adsorption sites in
the course of chemical modification. The presented results indicate that in the case
of a silica with pore sizes much larger than the size of coating mole-
cules. the physical coating procedure used may lead to the accumulation of the
organic modifier in smaller pores. still it provides a quite uniform surface
coverage only slightly inferior to the chemical bonding procedure.
T o quantify the gas-surface interactions for the materials under study, adsorp-
tion energy distributions were calculated. It was shown that the highest
energy adsorption sites (above 12 gradually disappeared (see Fig. 6) as the
amount of coated CBPB was increased up to in the case of CBPB-T24
LiChrospher Si-10
CBPB-T24
0.0 0.2 0.4 0.6 1.0

Standard Adsorption
.5 Comparative plots for unmodified and modified LiChrospher Si-100 samples,

The adsorbed amount for the samples was divided by the adsorbed at a relative
pressure of 0.4 to allow for a better comparison. (Data partially taken from Ref. 77.)

CBPB-T24
CBPB-B22
Adsorption Energy
FIG. 6 Comparison of adsorption energy distributions for unmodified
100 silica gel and physically coated (CBPB-T24) as well as chemically bonded (CBPB-
samples with similar loadings of CBPB. (Data taken from Ref. 77.)
material, which corresponds to a surface coverage of 2.45 Only a minor

further decrease in the number of high energy sites was noted when the surface
coverage was increased to 6.58 This indicated that the surface
coverage of 2.45 allowed a layer of coated CBPB to form on of the
silica surface, which is also supported by the fact that the sample with chemically
bonded CBPB had a similar loading and consequently the similar surface
coverage (2.15 as that for the CBPB-T24 sample. The disappearance of
high energy adsorption sites was explained as an effect of shielding (in the case of
physical coating) or of and shielding (in the case of chemical modification)
of surface which decreased their population accessible to nitrogen mole-
cules It should be noted that even for the highest coating and/or chemical
bonding levels, there was an appreciable population of 10-12 adsorption
sites, which were not present on for silicas modified by
groups with long aliphatic chains as will be discussed later. Therefore, it was
concluded that these sites are polar or aromatic segments of the organic modifier,
such as ester, ether, or groups and benzene rings. However, it was impossible
to perform a further identification, since, for example. the presence or absence of
the group in the structure of organic modifiers did not result in any notice-
able changes of the adsorption properties of the modified surface
To further examine the usefulness of the physical coating procedure, a
of MCM-41 material with pore size of about was modified by introducing
several different amounts of CBPB It was shown that in this case the
physical coating procedure was not suitable for surface modification, probably
because of the relatively large size of the organic modifier (CBPB) in comparison
to the pore diameter of the MCM-41 material used, which did not allow for

successful deposition of the organic modifier inside the pores. After the coating
procedure, samples exhibited a significant lowering of the primary
volume, but at the same time the pore diameter and the surface properties did
not change significantly. As can be seen in Fig. 7 , the A E D for the CBPB
MCM-41 is much closer to the A E D for the unmodified MCM-41 than to the AED
for the CBPB-BT33 sample (the surface of the latter is expected to be completely or
almost completely covered by CBPB). It was concluded that coating of the
MCM-41 material under study resulted in a very nonuniform surface coverage of
CBPB and that the organic modifier filled o r blocked some of the pores, making
them inaccessible to nitrogen adsorbate. As will be discussed later, the chemical
bonding procedures have the potential to overcome the shortcomings of physical
coating procedures, providing a means of modifying the surface of primary
pores by bulky without pore blocking.
An interesting problem was to assess the fraction of the surface covered by the
physically coated organic modifier T o d o that, a patchwise model of a mod-
ified surface was employed. A part of the surface of the MCM-41 material that was
not covered by the organic species was assumed to exhibit the same adsorption
properties as the unmodified MCM-41. and the surface of the organic modifier
(CBPB) coated on the silica surface was assumed to exhibit the same adsorption
properties as pure CBPB. The latter is nonporous and is expected to have a very
small surface area, so a direct measurement of its nitrogen adsorption was not
attempted. Instead. adsorption data for the previously studied CBPB-BT33 silica
gel chemically bonded and subsequently coated with high loading of the organic
modifier were used. T o test our model, the fitting procedure was applied to the
MCM-41
36% CBPB MCM-41
CBPB-BT33
0 4 8 12 16 20
Adsorption Energy
7 of adsorption energy distributions for an unmodified MCM-41
material, MCM-41 sample coated with CBPB, and a silica gel sample that was
chemically bonded and subsequently physically coated with CBPB. (Data from Refs.
46 and 77.)

with physically coated CBPB that were described ear-
lier. Calculations were performed using adsorption isotherm data for relative pres-
sures below 0.1, and it was shown that the low pressure data for the physically
coated samples can be fitted by a proper combination of data for the unmodified
sample and the CBPB-BT33 (see Fig. 8). The results indicated that in the case of the
samples with CBPB surface coverages (estimated from the elemental analysis) of
0.67 0.96 and 2.45 (CBPB-
and respectively, of the surface was covered with CBPB.
T o reproduce these surface coverages on the basis of calculated percentages, an
assumption about the monolayer surface coverage of CBPB must be made. When
one assumes that the latter is equal to or smaller than about 1.83
coverages for CBPB-T8 and samples can readily be reproduced, and
only in the case of CBPB-T24 would much lower surface coverage be obtained on
the basis of adsorption data. This result can easily be explained, since it was already
shown that the coverage of CBPB on the surface of CBPB-T24 is expected to be
quite nonuniform and some CBPB molecules may lie on top of other CBPB mole-
cules or fill smaller pores present in the sample, leaving some uncovered silica
surface, even though potentially the amount of CBPB is enough to cover the entire
surface of the support. In principle, calculations of the surface coverage on the basis
of adsorption data make it possible to obtain the fractions of bare and covered
silica surface. This provides the way to estimate the minimum surface coverage of
CBPB molecules the number of molecules forming a that would
result in the observed low pressure adsorption properties of the material.
Consequently, one can compare the total surface coverage obtained from elemental
Si-100
CBPB-BT33
Relative Pressure
Results of the fitting of low pressure adsorption data for a physically coated
silica gel sample with a relatively low loading of CBPB using low pressure
data for the unmodified silica gel and the modified sample (CBPB-BT33) with large
amounts of CBPB chemically bonded and physically coated. Data for CBPB-BT33 are
assumed to represent adsorption on pure CBPB. (Data from Refs. 46 and 77.)

analysis or other methods with the surface coverage assessed from adsorption data
in order to estimate whether the coverage of physically coated molecules on the
surface is uniform.
Since the fitting procedure was found to provide very reasonable results for
modified silica gels, it was interesting to use it in studies of the MCM-41 materials
coated with CBPB It was shown that the low pressure data for the modified
samples can be well fitted by data for the unmodified MCM-41 and the selected
reference material covered with CBPB The results obtained by
fitting indicated that the surface coverages of CBPB molecules did not vary sig-
nificantly for 1 1 and loadings of CBPB and were estimated to be less
than o r equal to 0.32, 0.40 and 0.48 respectively. Total surface coverages
of CBPB estimated from the elemental analysis were 0.27, 0.86, and 1 . 1
respectively, which suggests a very nonuniform coverage of CBPB for the materials
with higher CBPB loadings. The relatively small increase in surface coverage with a
significant increase in the CBPB loading that was estimated from the adsorption
data paralleled the observed changes in the primary size of the materials,
which were also small and suggested that a limited amount of CBPB actually
formed a monolayer on the surface of the primary mesopores. Therefore, it appears
that the organic modifier either completely filled of the pores or blocked them
deposited on the external surface area of the material. It can be concluded
that low pressure nitrogen adsorption data can be useful for a qualitative estima-
tion of the coverage of organic species the surface of silicas. T o our
best knowledge, similar estimations of the coverage of surface species using adsorp-
tion of nonpolar gas molecules have not been reported before.
As was discussed above, studies of conventional and ordered mesoporous silicas

with coated organic species allowed us to examine the effects of surface coverage on
the adsorption properties of modified materials. Another interesting problem was
to investigate how the structure of organic surface groups influences low pressure
nitrogen adsorption. First, a study of several silica gels with chemically bonded
ligands undertaken Surface coverages of octyl groups
were in the range of 3.4-4.1 It was found that for all of the octyl phases,
low pressure nitrogen adsorption was greatly reduced in comparison to that on
unmodified silica gels. Consequently, adsorption energy distributions for the mod-
ified samples did not feature adsorption energy sites above about (see
Fig. 9), which are characteristic of the silica surface (Figs. 2, 6, 7 ) . The high energy
adsorption sites were hypothesized to be isolated groups, and it was con-
cluded that most of these sites reacted during the modification procedure. It was
also suggested that nitrogen adsorption energy for hydrogen-bonded with
one another or with water molecules, for siloxane bridges, as well as for octyl
ligands lies in the range from about 3 to but further identification
appeared to be difficult. As will be shown in there is evidence
that for somewhat lower surface coverages of bonded ligands
there are still high energy adsorption sites present on the silica surface. and,
depending on the size of the bonded ligands, they may be accessible or inaccessible

Partisphere
Silasorb
Adsorption Energy
9 Adsorption energy distributions for silica gel samples with bonded
octyldimethylsilyl ligands. (Data from Ref.
to nitrogen molecules. Consequently, the lack of high energy sites on the AEDs for
the octyl phases under study does not necessarily indicate the removal of these sites
from the surface, as they could be effectively shielded by the long-chain
groups.
In another study low pressure adsorption properties of polymeric
and bonded phases were examined. It was shown that the introduction of
relatively small polymeric 3-aminopropylsilyl groups led to a decrease in the num-
ber of high adsorption energy sites in the range of A slight further
decrease in the number of these sites was observed when the amine ligands were
converted to with phenyl, cyanophenyl, and octyl groups,
especially for the latter. The similarity of AEDs for and
bonded phases confirmed previous findings that it is impossible to
detect the presence of groups attached to aromatic rings of the ligands,
which indicates that these two groups exhibit essentially the interaction ener-
gies with respect to nitrogen molecules. However, it was possible to distinguish
between adsorption properties of and bonded phases,
which is also consistent with the previous studies discussed in and
It was shown earlier that polar parts of organic modifiers gave rise to adsorption
energy sites with energies up to about 12 However, the bonded
phases exhibited adsorption sites with energies up to about 15 One can
expect that the latter sites are groups present in the structure of the bonded
phase due to hydrolysis of side alkoxy groups and/or on the silica surface. This idea
was supported by the examination of the polymeric octyl phase which exhib-
ited a much smaller population of adsorption sites between 10 and 13 but
showed a pronounced peak at about 13 Since octyl chains interact with

nitrogen with adsorption energies below about it was concluded that the
AED peak at 13 can be attributed to the presence of groups acces-
sible to nitrogen probe Moreover, these results confirmed that polar
parts of organic modifiers give rise to adsorption energies in the range from
about to 12
Although it is possible to study pore size changes resulting from the modification of
conventional mesoporous silicas, as was shown in such changes are
much easier to analyze in the case of ordered mesoporous silicas due to their
narrow pore size distribution and relatively small sizes. Moveover,
there has recently been a significant and rapidly growing interest in studies of
organic-modified ordered materials It was therefore
interesting to investigate the surface modification of a good quality large-pore
MCM-41 material via chemical bonding of trimethylsilyl (TMS), butyldimethylsilyl
(BDMS), and octyldimethylsilyl (ODMS) monomeric ligands and polymeric 3-ami-
nopropyl (APS-P). hexanoyl-3-aminopropyl (HAPS-P), and octyl phases
The obtained surface coverages were in the range of 2.5-3 It was
shown that the increase in the size of bonded ligands led to a systematic shift of the
step of capillary condensation in primary toward low pressures (Fig.
10). The overall shape of the adsorption isotherm remained unchanged, showing
that the modification procedure did not change the structure of the ordered
porous support. Pore size distributions for unmodified MCM-41 and
selected modified materials (Fig. 11) indicated that the modification procedures
0.0 0.2 0.4 0.6 0.8 1.0

Relative Pressure
IG. 10 Comparison of nitrogen adsorption isotherms on a large-pore MCM-41 with

pore size of unmodified and chemically bonded trimethylsilyl
butyldimethylsilyl (BDMS), and octyldimethylsilyl (ODMS) ligands. Adsorbed
amounts for the materials were divided by the maximum adsorbed amount at a
relative pressure of about 0.995. (Data from Ref. 59.)

I 1.0 MCM-41
APS-P
0.8 . ODMS
3 4 5 6 7 8 9 1 0
. Pore Size
FIG. 11 Pore size distributions calculated from adsorption isotherms using the BJH
method with (3) and (4) for an unmodified large-pore MCM-41 and
bonded materials with polymeric 3-aminopropylsilyl (APS-P) and monomeric
(ODMS) ligands. (Data from Ref. 59.)
used markedly decreased the pore size of the material but did not lead to an
appreciable broadening of the PSD. It was shown that the increase in the size of
bonded ligands resulted in a gradual decrease in the pore size. For example, the
initial pore size of support decreased to 4.5, 4.3, and after
modification with BDMS, and ODMS ligands, respectively. Moreover,
there was good agreement between the pore size changes in the case of different
ligands of similar sizes. For instance, the pore diameters of the materials with
monomeric and polymeric octyl phases were 3.8 and respectively.
sizes of and aminopropyl phases of similar organic chain length but different
functionality were also very similar (4.3 and 4.4 respectively). It was concluded
that when reproducible surface coverages of bonded ligands can be achieved, it is
relatively easy to predict the pore size of the modified materials on the of the
pore diameter of the support and the size of the ligand. Such predictions can be
based on the data, which were presented and discussed in detail elsewhere
Another interesting problem was to monitor changes in the surface properties of

ordered mesoporous silicas caused by attachment of organic ligands. As can be seen
from the linearity of its a, plot, the unmodified MCM-41 (Fig. 12) exhibits surface
properties typical of mesoporous silicas. The modifications via bonding of
led to a decrease in the low pressure adsorption, which manifested itself in
downward bends of the comparative plots. As expected, these deviations from
linearity were especially pronounced for bonded phases. Figure 13 shows
adsorption distributions for selected modified materials. I11 comparison
with the AED for the bare MCM-41 material, the AED for the
phase exhibited a smaller number of higher energy sites the range

0.0 0.2 0.4 0.6 0.8
Standard Adsorption
. 12High resolution a, plots for an unmodified large-pore material and
samples with chemically bonded polymeric 3-aminopropylsilyl (APS-P) and monomeric
octyldimethylsilyl (ODMS) ligands. (Data from Ref. 59.)
10-16 in agreement with previous studies reviewed in The com-

parison of for monomeric (ODMS) and polymeric octyl bonded
phases indicated that in the case of the latter a small number of high energy
adsorption sites are still accessible to nitrogen adsorbate. It should also be noted
that the T M S MCM-41 had a significant number of high energy adsorption sites at
Adsorption Energy
FIG. 13 Adsorption energy distributions for an unmodified large-pore

material and samples modified with chemically bonded polymeric
(APS-P) and octylsilyl ligands as well as monomeric octyldimethylsilyl (ODMS)
groups. (Data from Ref. 59.)

11-16 In contrast, only a small population of these sites were observed for
BDMS and they were not present on the A E D for ODMS MCM-41 (see Fig. 13).
The surface coverage of the bonded groups was shown to slightly decrease with an
increase in the size of bonded ligands. Consequently, the larger the size of the
bonded groups, the fewer silanols had to react during the modification. This
leads to the conclusion that the number of residual surface silanols is probably
the greatest for the ODMS MCM-41 among the three materials considered. Since
silanols are expected to be high energy adsorption sites with respect to nitrogen, one
can conclude that the increase in the size of bonded ligands decreases the accessi-
bility of nitrogen to sites on the silica surface, and for a sufficiently large surface
coverage and size of bonded groups, the silica surface may become inaccessible to
nitrogen adsorbate.
The results presented here demonstrate the great usefulness of nitrogen adsorption
measurements for the characterization of unmodified and modified mesoporous
silicas. High pressure adsorption data allow for a detailed characterization of
mesoporous structures of materials, providing their surface areas and pore size
distributions. Low pressure measurements provide an opportunity to study the
surface properties of materials and to monitor changes in these properties that
result from surface modifications. For instance, it is possible to attribute the
appearance of adsorption sites of certain energies in the adsorption energy distri-
bution to the presence of some types of surface groups as silanols. aliphatic
chains, or or polar segments of organic ligands. Low pressure data can
also be used to estimate the surface coverage of organic species on the silica surface.
Therefore. nitrogen adsorption measurements carried out over a wide range of
pressures it possible to characterize both surface and structural properties
of silicas.
However, further studies are needed to improve our understanding of the rela-
tion between the adsorption properties of modified mesoporous silicas and their
structure and surface functionality. First, additional studies of modified materials
with different functional groups would be desirable to gain additional insight about
the types of surface groups that can be detected and distinguished from one another
by nitrogen probe Second, other nonpolar or weakly polar probe mole-
cules need to be examined in order to assess their applicability and sensitivity to
details of the surface, and it is very likely that some of will exhibit properties
very useful for surface characterization. Third, it is important to improve methods
of pore size analysis. New approaches based on studies of model adsorbents with
well-defined structures MCM-41) appear to be promising and successful, but
they require further refinement to improve their accuracy for materials with differ-
ent pore geometries and surface properties. Finally, the currently proposed analysis
of the effects of the pore structure and surface functionality on adsorption proper-
ties of materials is relatively easy due to the possibility of separating
these effects. Namely, high pressure adsorption data are influenced mostly by
properties of the porous structure, whereas low pressure data reflect mainly the

nature of the surface. A major challenge will be to develop methods of pore struc-
ture and surface analysis suitable for materials for which separation of the surface
and structural effects o n adsorption properties is difficult, for instance, for modified
microporous adsorbents.
The Chemistry of Silica, Wiley, New York, 1979.

Bergna (ed.). The Colloidal o f Silica. American Chemical Society,
Washington, DC, 1994.
J. Brinker and Scherer, Sol-Gel Science: The Physics and of
Sol-Gel Processing, Academic Press, San 1990.
Vansant, P. Van der Voort, and K . C. Vrancken. Characterization and
Modification of the Silica Surface, Amsterdam, 1995.
K. Unger Packings and Stationary Phases in Techniques,
Marcel Dekker, New York, 1990.
Scott, Silica Gel and Bonded Phases, Wiley, Chichester, 1993.
Wirth, R. and Science (1997).
J. Am. Chem. (1995).
S. T. and A. R. Sanger, Phys. Chem. (1996).
Guidotto, and Suter, Langmuir (1996).
A. Alden, Walter. Walter, Petersen,
and Neue, Anal. Chem. (1995).
A. Grabe, A. Michalske, and Smith, Phys. Chem.
Vrancken, P. Van der Voort, Possemiers, and F. Vansant, Colloid
Interface (1995).
Vrancken, De P. Van der Voort, P. J. Grobet, and Vansant,
Colloid Interface Sci. (1995).
P. Tripp and Hair, Langmuir (1995).
Hair and Tripp, Colloids Surf. A (1995).
D. and (1997).
Jaroniec. and B. Chromatogr. A (1996).
C. P. Jaroniec, K. Gilpin, and Jaroniec, Chromatogr. A (1998).
Gilpin, Jaroniec, and J. Liquid Chromatogr.
(1996).
V. Jaroniec, and R. Gilpin, Liquid Chromatogr.
(1996).
T. Czajkowska and Jaroniec, Chromatogr, A (1997).
Jaroniec, Chromatogr. A (1996).
Jaroniec, and Gilpin, Chromatogr. A (1994).
P. Van der Voort and F. Vansant, J. Liquid Chromatogr. (1996).
L. Ascah, A. Szafranski, and Liu,
Liquid Chromatogr. (1996)
B. Buszewski, Jaroniec, and Gilpin, Chromatogr. A (1994).
T. Shimizu, K. and C. Kato, Bull. Chem. Jpn.
(1990).

70 and
and Bull. Jpn.

(1990).
Beck, Vartuli, Roth, Leonowicz. Kresge. K.
Chu. Olson, Sheppard,
and L. Chem. (1992).
Kresge, Leonowicz, Roth, Vartuli, and Beck, Nature
(1992).
I. Margolese, Demuth. P. Feng, Gier. P.
A and D . Stucky, Chem, Mater.
(1994).
Huo, I. andStucky, Chem. Mater. (1996).
and Mater. (1996).
A. Sayari, Stud. Surf, Catal. (1996).
A. A. Tanaka, Garces, M.
and Coombs, Adv. Mater. (1995).
Park, and
Gladden, Chem. Faraday Trans. (1997).
A Corma. Navarro, and Chem. Mater.
(1997).
A. Sayari, P. Kruk. and Jaroniec, Chem. Mater. (1997).
Zhao, Feng, Melosh, G. B. Chmelka, and
Stuck), Science (1998).
L. and Mann, Chem. 1367 (1996).
Macquarrie, Chem. 1961 (1996).
Lim, and A. Stein. Am. Chem. (1997).
A. Koyano. T. Tanaka. and Nakata, Phys. B
(1997).
X. Zhao, A. Miller, and Phys.
Chem. (1997).
Jaroniec, Kruk, C. P. Jaroniec, and A. Sayari, Adsorption, in press.
X. Feng, E. Fryxell, Wang. A. and
Science (1997).
Mercier and Adv. Mater.
J. X. Feng, Fryxell, Wang, A. Kim. and Gong, Adv. Mater.
(1998).
Brunei, A. Cauvel, and Stud. Surf. Catal.
(1995).
A. and Brunei, (1997).
V. Subba De and P. A, Jacobs, Chem. Int. Ed.
(1997).
M . Kruk, M. and A. Sayari, Phys. B (1997).
Kruk, Jaroniec, Ryoo, and Kim. Mater.
(1997).
Jaroniec, and A. Sayari. (1997).
I. Ravikovitch, A. V. F. Schuth, and
Unger, Langmuir (1995).

P. I. and A. V. Phys.
Chem. B (1997).
M. Maddox, P. and Gubbins. (1997).
C. P. Jaroniec. M. Jaroniec, and A. Phys. Chem.
( 1 998).
0. Rathousky, Starek. and Zukal,
Chem. 724 (1993).
P. Tanev and Pinnavaia, Science (1995).
A. Bagshaw, Prouzet, and Pinnavaia, Science (1995).
and C. Shin, Phys. Chem. (1996).
Li, and Davis, Microporous (1993).
P. P. Hall, Treguer, and K. Sing, Chem. Faraday
Trans. (1995).
P. Sing, and White, Faraday Trans.
(1997).
P. Llewellyn, Rouquerol. Martin, and Surf. Sci.
(1996).
P. Hall. and Sing. Chem. Chem. 1257
(1993).
Kruk, Jaroniec, and A. Sayari, Microporous Mater.. in press.
Bereznitski, Jaroniec, and Kruk, J. Liquid Chromatogr. (1996).
Jaroniec, Kruk, and Liquid Chromatogr.
(1996).
Sing, Everett, R. A. W. Haul. L. R . A. J.
Rouquerol. and Pure Appl. Chem. (1985).
Rouquerol, Avnir, Fairbridge. H. Everett. H.
K. Sing, and Unger, Pure Appl.
60:1739 (1994).
Gregg and Sing. Adsorption, Surface Area Porosity. Academic
Press. London, 1982.
Jaroniec and R . Physical Adsorption on Solids, Elsevier,
Amsterdam. 1988.
P. Jaroniec, and Jaroniec. Phys. B (1997).
Jaroniec, and Y. Zhou. Langmuir (1997).
P. Jaroniec. Jaroniec. and Kruk, Chromatogr. A (1998).
M. Jaroniec, Kruk, and in Characterization Porous Solids
J. Mays, Rouquerol, Rodriguez-Reinoso, S. Sing, and
K. Unger, Royal Society of Chemistry. London, 1997, pp.
Zhang, and A. Colloid Interface (1993).
Brunauer. Emmett, and Teller. Am. (1938).
Jaroniec and Kaneko, Langmuir (1997).
Kaneko, Ruike, and Kuwabara, Carbon (1992).
Jaroniec, in Access to Materials Pinnavaia and
Thorpe. Plenum, New York, 1995, pp.
P. Barrett. and P. P. Am. Chem. (1951).
Gubbins, Maddox, and Nicholson, Chem. Res.
(1995).

Dabadie, A , Guizard, Cot, and P. Chem.
(1996).
in Fundamentals in Adsorption ( M . Kluwer, Boston,
1996,
Jaroniec and Brauer, Surf. Rep. (1986).
Rudzinski and D. Everett, Adsorption of Gases on Heterogeneous Solid
Surfaces, Academic Press, New York, 1992.
Jaroniec, in Adsorption on New and Modified Inorganic Adsorbents (A.
Dabrowski and A. Tertykh, Elsevier, Amsterdam, 1996, pp. 41
L. Riccardo, A. Steele. A. Ramirez and Zgrablich,
13:1064 (1993).
A. Steele, Rev. (1993).
Szoinbathely, P. Brauer. and Jaroniec, (1992).
Jaroniec, Gilpin, Brauer, and
Langmuir (1993).

KO Department of Chemistry, Chiba University, Chiba, Japan
I. Introduction
II. Solid Nanospace of Enhanced Molecular Field
III. Magnetic Interaction of Molecular Spins
IV. Magnetic Properties of Bulk and Two-Dimensional 0 2
V. Pore Structures and Magnetic Properties of Activated
Carbon Fibers
VI. Adsorption of in a Graphitic Nanospace
VII. Magnetic Susceptibility of Confined in a Graphitic Nanospace
VIII. Dependence of Magnetic Susceptibility of Confined on Pore
Width
IX. Effect of Coexistent on Magnetic Susceptibility of
Confined
X. Magnetic Distinction of Weakly Interacting Molecules in
Nanospace
References
The interaction energy of a simple molecule with a solid surface as it undergoes

physical adsorption is quite small compared with that of chemical bonding. The
physical adsorption mechanism changes according to the type of solid surface-flat
or macropore surface, mesopore, or micropore The molecule-surface inter-
action increases in the order of adsorption on the flat surface, mesopore, or
micropore. Adsorption on a flat or macropore surface leads to a type adsorp-
tion isotherm, whereas adsorption on a gives type IV. An enhanced
potential from opposite pore walls in the micropore is great enough to provide a

characteristic adsorption isotherm of Type I which has a significant uptake
within a low relative pressure region. Although lateral interaction between
adsorbed molecules should be important, it is neglected in the well-known
(BET) theory On the other hand, the so-called
capillary condensation theory about adsorption on is based on lateral
interaction. Capillary condensation theory presumes that the adsorbed layer is the
same as the bulk liquid, that the lateral interaction is not specific but is the as
in the bulk Although the interaction of a molecule with a solid surface is
markedly enhanced in micropores, an adsorbed layer in micropores has been
believed to be identical to the bulk liquid. That is, no enhanced lateral interaction
is presumed. Thus, the intermolecular interaction between adsorbed
molecules is less understood than molecule-surface interaction Although recent
computer simulation approaches have taken the lateral interaction into account
using the potential, the importance of the interaction
in physical adsorption. particularly on inicropores, is not sufficiently recognized
If physical adsorption is used for controlling interactions
the aid of various solid surfaces from flat surfaces to it should be
helpful to understand special molecular behavior.
In the molecular sciences, elucidation of the structure and property of a van der
Waals molecule is an important current research subject. The van der Waals
cule is unstable at room temperature, and it is quite difficult to get a concentrated
system of van der Waals molecules There is another weakly interactive
molecular complex compound, called a molecular clathrate. Usually, molecular
are unstable at ambient conditions. Supercritical gas molecules such as
and NO are said to tend to produce in In particular,
NO molecules are adsorbed in micropores of activated carbon fiber (ACF) at
ambient conditions in the form of that are typical van der Waals molecules.
Water form an organized structure in the carbon micropore The
formation of NO hydrate and hydrate in at a subatmospheric
pressure has also been suggested Hence micropores accelerate the forma-
tion of van der Waals or clathrate hydrates. In the case of
that have a deep potential well, many molecules tend to be adsorbed in
the deep potential well. Molecules confined in micropores should form the best
dense structure according to the geometry. Therefore, we can control the
intermolecular structure with inicropores even at supercritical conditions of the
bulk gas. The molecular field of micropores can stabilize van der Waals molecules
and molecular clathrates.
Even a weak interaction is much stronger than a spin-spin
A spin-spin interaction is so weak that it is sensitively affected by an inter-
molecular interaction. Therefore, measurement is a powerful method for
the study of the weakly bound structure between molecules. Furthermore, in gra-
phitic microporous carbons such as that are mainly composed of diamag-
netic the electronic state of is triplet, leading to paramagnetism in the
gas phase. Hence, the magnetic properties of confined in inicropores of A C F
should provide essentially important information on the intermolecular structure of
simple molecules particular, we can compare the intermolecular structure
of in a with the intermolecular structure of on a graphite surface

Ma
which has been actively studied The formation of a linear of in the

cylindrical pore was suggested by magnetic susceptibility measurements at tempera-
tures above 77K Understanding the assembly in a graphitic
inicropore should be helpful to the understanding of the specific role of in
biological activities and new separation technologies.
II.
interaction between a molecule and a surface atom as a function of
the distance between them can be expressed by the Lennard-Jones potential,
where and are the well depth and effective for the
graphitic carbon atom complex. These cross parameters are calculated according to
the Lorentz-Berthelot rules, = + Here, and
are the Lennard-Jones parameters for a surface atom and a molecule,
respectively. The interaction potential for a molecule and a single graphite
slab is given by the Steele 10-4-3 potential
where is the vertical distance of the molecule above the surface, A is the separa-
tion between graphite layers 0.335 nm), and is the number density of carbon
atoms in a graphite layer (= 1 As the of activated carbon can
be approximated by the slit spaces between the predominant basal planes of
graphitic units, the whole interaction potential of a molecule with the
micropore of an intergraphite surface distance can be given by the equation
Consequently, we can evaluate the potential profile of the molecule adsorbed in the
graphitic Here is not the effective pore width determined by the
adsorption experiment. The difference between and is a function of and
In the case of the slit pore system, is equal to 0.26 Figure 1

shows potential profiles of an molecule in a slit-shaped graphite pore as a
function of using the one-center approximation. Here, the molecular position
in Fig. 1 is expressed by the distance from the central plane between two surfaces.
The solid line shows the potential profile for the single graphite surface. The poten-
tial becomes deeper as decreases. Thus, micropores have stronger adsorption
fields than flat or graphitic surfaces. However, the potential
for 1 nm is still 1500 K, which is not enough to induce predominant adsorption
of at ambient conditions.

1Interaction potential profiles of with the graphitic slit pores of different
widths. Here z denotes the vertical distance from the of two graphitic surfaces.
The broken line indicates the profile for the single surface.
We can estimate the effective pressure from the potential profile; molecules
confined in a slit pore of 1 nm width are presumed to be exposed to the high
pressure of 100 Therefore, the graphitic micropore can offer a high pressure
field from the macroscopic view
When a homogeneous magnetic field is applied to the sample, the magnetization

M is induced.
Here is the magnetic susceptibility. When the magnetic field is weak enough, is
independent of molecular magnetism the cgs emu system has been used
instead of the SI The volume magnetic susceptibility is a dimensionless
quantity traditionally expressed in the dimension of the electromagnetic
unit (emu) is formally the cubic centimeter. The specific magnetic susceptibility
which is obtained by dividing the volumetric magnetic susceptibility by
the density, is often used. is the of the diamagnetic ( y o ) and paramagnetic
susceptibilities.
+
The value is negative and is positive. When dominates, the sample is said
to be diamagnetic; it is repelled by the magnetic field. When is the leading
contribution, the sample is paramagnetic: it is attracted by the applied field.

is an underlying property of matter. It is always present, even when
it is masked by paramagnetism. Diamagnetism is due to the interaction of the
magnetic field with the motion of electrons in their orbits. is independent of
the temperature and the strength of the applied field. The diamagnetic susceptibility
of a molecular compound can be approximated by the additive form of the con-
stitutive corrections (Pascal's law). The additive method may be sufficient to esti-
mate the diamagnetic contribution for paramagnetic compounds of low molecular
weight. The observed susceptibilities are corrected for the diamagnetic contribution
for a discussion of paramagnetic susceptibility
Paramagnetic susceptibility characterizes the in which an applied magnetic
field interacts with the angular momenta associated with the thermally populated
states of a molecule. The magnetization is related to the energy variation under an
external magnetic field.
A molecule has a discontinuous energy - 1 , 2. 3 . . . in the presence of a

magnetic field. For each energy level we can define a microscopic magnetization
The macroscopic magnetization M is then obtained by summation of the micro-

scopic magnetizations using the Boltzmann distribution.
where T and are the temperature and Boltzmann constant, respectively. This
combines macroscopic and microscopic quantities. However, the application of the
general equation (9) is difficult without the knowledge on = Van
introduced the approximation
where is the energy level in zero magnetic field. and are called the
first- and second-order Zeeman coefficients respectively. When
exp
T
When all energies

= exp
[- are linear in these approximations yield the equation
Here and the operator

where I, is the orbital momentum of electron i and is the spin momentum of the
same electron. is the factor of the free electron (= and is the
= = 0.9174 x erg gauss).
We assume the simplest situation in molecular magnetism of the molecule of the
ground state irreducible representation). When the field is applied, the
energies of the 1 components are given by the equation
If the first excited states are to be too high in energy to couple with the
ground state. g is in principle isotropic and equal to As is linear in the
above Van formula with = 0 can be used.
which leads to
Here is number. Thus the magnetic susceptibility varies as

the constant depending on the spin multiplicity of the ground state. This is the
Curie law, which describes the representative paramagnetism of mutually isolated
spins. In the cgs system, 0.12505.
In real solid-state or molecular magnetic species, spins are not perfectly free.
Hence the Curie law is modified to account for weak intermolecular interactions
using the assumption that the molecular field approximation of a perturbation is
added to the term. This perturbation takes the form where
is the mean value of the component of the spin operator. is the interaction
parameter between two nearest-neighbor magnetic species, and is the number of
nearest neighbors around a given magnetic When is positive or nega-
tive, the intermolecular interaction is said to be ferromagnetic or
respectively. The neighboring spins tend to align in a parallel fashion in the case of
the ferromagnetic interaction, while they tend to align in an antiparallel fashion in
the antiferromagnetic interaction. The total spin Hamiltonian is
H-
where the magnetic field is assumed to be along the z direction and the tensor is
to be isotropic. The eigenvalues My) are given by
-
is given by the distribution law and is expressed as
is negative because the components with negative are more popu-

lated than those with positive may be expressed as

Magnetic
Hence 7 can be derived using =
is rewritten as
-
This is the Curie-Weiss law. is the Curie constant, and is the Weiss
constant defined by = +
A plot of for a system obeying the Curie-Weiss law gives a straight
line with a slope o f c ' . The intercept with the T axis yields both the sign and the
value of 6. In the framework of this model, a positive indicates ferromagnetic
interactions and negative 6 indicates antiferromagnetic interactions. Hence the
vs. T plot can determine the sort of magnetic interaction. In the case of ferro-
magnetic interactions the Curie-Weiss is limited to the temperature
range T Accordingly, if we use a paramagnetic probe such as an
the low temperature magnetic susceptibility measurement can provide an
essential information on the intermolecular states of confined using this
analysis.
The ground state of an molecule is the triplet and therefore gas shows
the representative paramagnetism. The effective spin concentration and con-
stant of bulk gas are 1.5 and 0 K . gas liquefies at 90.2 and
the liquid solidifies at The solid oxygen has three
K T and T The
magnetic properties of solid oxygen have been actively studied since the days of
Kammerlingh Onnes. Solid oxygen has two sharp magnetic transitions cor-
responding to a-(3 and crystallographic changes. Active studies [35 showed
that the a phase is antiferromagnetic, with two-dimensional ordering. It is
suggested that the (3 phase has a "frustrated" nature
It was indicated that the and liquid phases have a common feature
Although is a simple elemental solid. there are still many obscure pro-
blems with respect to the properties and structures of
Oxygen gas of high purity (99.99%) was sealed in a highly pure quartz tube. Its
magnetic susceptibility was measured with a SQUID magnetometer system over a
temperature range of under a magnetic field of 1 Figure 2 shows the
temperature dependence of the magnetic susceptibility of condensed The
value in Fig. 2 was determined just after the equilibration of the measuring tem-
perature. The relations on cooling and heating between 2 and almost
coincide with each other. There are two anomalies at 23.8 and 43.8 which are
ascribed to the well-known and transitions

2
Temperature dependence of the magnetic susceptibility of bulk condensed
with and without coexistent + molar ratio
0.22).
However, we observed a long time dependence of in the temperature range of

44-55 K over several hours upon cooling from liquid Figure 3 shows the y-T
relation at equilibrium. Here the value at 4 4 K after 2.2 is used in Fig. 3,
which should be smaller than the perfect equilibrium value. The equilibrium
values are smaller than nonequilibrium ones by one o r two orders of magnitude (see
Fig. 2). An obscure jump near 24 K and a very steep jump at 44 K are shown. The
latter jump should become more pronounced if we measure it after a much longer
time. The shape of relation near 4 4 K is close to a type Consequently,
the well-known relation in Fig. 2 is obtained at conditions.
It is presumed that both the a and (3 phases should form a single phase at perfect
equilibrium. The obscure transition at 2 4 K should stem from residual
of the phase due to a finite cooling rate from liquid Marked time
dependence is one of the characteristics of random magnetism, and therefore the
higher temperature solid phase 24 K) should show random magnetism. As the
value of the lower temperature phase ( 24 K ) decreases with the decrease of tem-
perature, it should be antiferromagnetic.
The effect of coexistent on magnetic susceptibility was examined under
nonequilibrium conditions. The effect of coexistent on the relation is
also shown in Fig. 2. The coexistent gives rise to a remarkable increase of in
the a phase as temperature decreases, but it does not change that of the phase. In
contrast, of the phase becomes smaller due to the addition. The appear-
ance of the paramagnetic behavior in the a phase is ascribed to the decoupling of

Ma
Temp
FIG. 3 Temperature dependence of the magnetic susceptibility of bulk at an

equilibrium state.
spins by structural destruction. The insensitive behavior of the phase may come
from the disordered nature of the structure Jezowski et reported that
doping of or Ar in had a drastic effect on the thermal conductivity, but the
phase was insensitive to the doping, which has a similarity to the effect of coex-
istent
The magnetic interactions of adsorbed on the graphite surface have been
extensively studied The mono- or of molecules adsorbed on the
two-dimensional graphite surface has phases such as 5, and that are different
from the bulk a , and phases; the phase change depends on both temperature
and coverage. The and phases appear in the multilayer of adsorbed on the
flat graphite surface. Also the to phase transition leads to a magnetic anomaly.
A of the two-dimensional magnetic structure of on the graphite basal
plane is shown in Fig. 4. This diagram was obtained at nonequilibrium conditions.
In the following discussion, the magnetic properties of confined in nanospace
obtained at nonequilibrium conditons are used for comparison with published
results.
Activated carbon fibers are mainly composed of nanographitic units that

have sheet structures (stack height stack width = 3.5 nm). The
unit size depends on the degree of activation, and therefore a low-burnoff A C F has
a greater nanographitic structure. There are many slit-shaped spaces between the
nanographites, giving slit-shaped micropores with pore volume in the range of

liquid
4 Phase diagram of on graphite surface
0.3-1 There are several kinds of ACFs depending on the precursors.

Cellulose-based A C F contains a large number of oxygen groups, while
lonitrile-based A C F has polar bonds. Pitch-based A C F (PIT) is the purest
carbon, and PIT samples of different pore widths are available. Therefore, PIT is
the best A C F for elucidation of molecular states in carbon micropores. Three kinds
of pitch-based ACFs were used. These are denoted P-0.75, P-0.96, and P-1.45, using
the micropore width in this chapter. the values were 0.8 for for
P-0.96, and for P-1.45.
Figure 5 shows adsorption isotherms of on ACFs at 77 K The adsorp-
tion isotherms of on P-0.75 and P-0.96 are typical type I, indicating the presence
of uniform The adsorption isotherm of P-1.45 has a steep uptake at the
low range and a gradual adsorption until = 0.4. The initial adsorption
uptake and the gradual adsorption at the medium range are ascribed to the
monolayer adsorption on the micropore walls and to the second- and/or third-layer
adsorption of on the monolayer-covered micropore walls, respectively. This
adsorption behavior suggests that the pore width of P-1.45 corresponds to a thick-
ness of than four adsorbed layers. As the monolayer is completed below the
relative pressure corresponding to adsorption on the flat surface, the routing BET
evaluation of the surface area leads to an overestimation. The adsorption in micro-
pores is composed of both an enhanced monolayer adsorption on the micropore
walls and adsorption in the residual space between the monolayer-covered
pore walls (a kind of condensation). Hence the enhanced adsorption should be
subtracted. A high resolution plot of the isotherm is quite effective for evaluation
of the specific surface area. The data of high resolution adsorption isotherms of

.5 Nitrogen adsorption isotherms of pitch-based ACFs at (0)P-0.75;
P-0.96; P-1.45.
nonporous carbon black were used for the standard carbon. The subtracting pore
effect (SPE) method for the high resolution a,, plot can remove the enhanced
adsorption, giving the correct area The high resolution a,, plot
for P-1.45 has two swings linearity in the regions of 0-0.3 and
The line between the two swings passes through the origin, and its slope provides
the total surface area, that is, the sum of the and external surface
areas. The slope of the line of = 1 gives the external surface area. and its inter-
cept leads to the micropore volume. Hence the high resolution plot provides the
total surface area, the microporous surface area, and the micropore volume. The
micropore geometry of ACF can be approximated by the slit, and thus the average
slit width (micropore width was determined from the microporous surface area
and the micropore volume. The a, plots of P-0.75 and P-0.96 had the only swing
below 0.3. Their surface area and volume were determined in a
similar way. Table 1 lists the total surface area, micropore volume, and average
micropore width for the three ACFs.
Micropore Structural Parameters of Pitch-Based ACFs

Surface area Pore volume Pore width
ACF 1

Porous carbon is composed of nanographites. The electronic struc-
ture of porous carbon is estimated from that of graphite. Carbon atoms in a
graphite structure are combined with hybrid orbital) and it orbital)
bonds. it electronic wave functions extend over the basal plane, so TI electrons
can move freely in the basal plane. The conjugated it-electron system of the
graphite sheet structure gives rise to the representative diamagnetism
which is associated with the extension of the conjugated TI-electron system.
Thus the magnetic susceptibility provides information on the electronic struc-
ture of nanographites in porous carbons. It was shown that the diamagnetic
susceptibility 7 of porous carbon at room temperature is proportional to the
nanographitic unit size, as give by the equation. A unit nanographitic unit
structure of about 60 carbon atoms for activated carbon was suggested
Thus, most ACFs show representative diamagnetism, and the diamagnetism is
associated with (24).
= 0.0090 (number of carbon atoms) + 0.47
At the same time, the presence of an electronic defective structure in the

conjugated structure and a dangling bond at the edge induces paramagnetism.
Also, Volpilhac and Hoarau studied the relationship between graphitic unit
size and diamagnetism theoretically; they suggested that magnetic susceptibility
decreases with increases in graphitic size until the stack width equals and
becomes constant thereafter. Hence, the magnetism of the porous carbon indi-
cates sensitively the electronic states through the nanographitic unit and their
stacking structures. In particular, the low temperature magnetic susceptibility
measurement is useful. ACFs having very high surface area show paramagnetism
near room temperature regardless of the absence of magnetic impurities
they have electronic defect structures and dangling bonds due tot the extremely
minute graphitic unit structure. The mechanism of the paramagnetism of ACFs
having very high surface area is itself a n important research subject.
Figure 6 shows temperature dependence of the susceptibility of
over a wide temperature range. Basically P-0.75 and P-0.96 are diamagnetic
above 10 K. the other hand, P-1.45 has positive magnetic susceptibility. being
temperature-independent above P-1.45 is highly activated. and it should
have many dangling bonds at the edges of nanographites, giving rise to para-
magnetism as suggested by et The magnetic susceptibility
increases steeply below This indicates clearly the presence of spins, which
should be associated with the nanographitic structure. A preliminary electron
spin resonance (ESR) study showed that there are two kinds of electron spins
having TI and character in ACFs, and only the spin of the electrons is
affected by the adsorbed at low temperature The paramagnetic behavior
below 2 0 K should be associated with the it-electron spin. As the magnetic
susceptibility of these PIT samples is much lower than that of the reliable
magnetic susceptibility of confined of PIT can be determined,
as shown later.

Magnetic
FIG. 6 Temperature dependence of the magnetic susceptibility of pitch-based ACFs

at low temperature, P-0.75; (A) P-0.96; P-1.45.
The adsorption isotherms of o n A C F s a t 77 K are close t o those of as shown

in Fig. 7 Although the adsorption isotherm was measured a t which is
below the boiling temperature n o condensation occurred during the
measurement. T h e adsorption isotherm was described by the
(DR) equation
0.0 0.2 0.4 0.6 0.8 1
.7 Oxygen adsorption isotherms of pitch-based ACFs at (

Here W and are the amount of adsorption at and the micropore volume,
respectively. and are the affinity coefficient and characteristic adsorption
energy, respectively. A is the adsorption potential. The DR plot was almost linear.
suggesting that molecules are filled homogeneously. The liquid density of 1.15
at 87 K was used for calculation of the volume of the adsorbed layer at 77
The micropore volume from the plot for adsorption was similar to that
adsorption. The heat of adsorption at = was
evaluated with the equation
Here is the enthalpy of evaporation at the boiling temperature. The

pore volume and values for and are shown in Table 2. Here, 6.82 and
5.59 were used as A H , , values for at 9 0 K and at respectively.
The two pore volumes are similar; the Gurvitch rule holds for and adsorp-
tion. The for is slightly greater than that for irrespective of differences
in molecular size and quadrupole moments. Probably adsorbed in is
not ordinary liquid, but molecules should form an organized structure.
SUSCEPTIBILITY OF
The magnetic susceptibility of adsorbed on A C F after sealing was measured

with a SQUID magnetometer system over a temperature range of 1.7-100 and at
a magnetic field of 1 T. The magnetic susceptibility of confined in carbon
nanospaces of PIT depends on the temperature, the coverage, and the pore width
The coverage is defined as
where is moles of oxygen dosed, 6.36 is the reference density

for the registered x commensurate structure, is the specific surface
area and is the weight of (g). The commensurate structure is not
observed for adsorbed layers in the A C F micropores but is used to define the
standard unit for comparison with results in an system. Fractional
filling which is the volume fraction filled with adsorbed is calculated
as = using the pore of and the liquid density of oxygen
( I.14
2 Micropore Volume and for and from DR Plot
Micropore Micropore volume

ACF
P-0.96 0.49 12.4 0.45 11.52
1.45 0.79 9.22 0.88 8.68

Ma
The temperature dependence of the magnetic susceptibility of

adsorbed in micropores of P-0.96 is shown as a function of the coverage in Fig. 8.
No critical change in the curve for 0.2 is observed in the temperature
range of increases steeply with decreasing temperature, which is char-
acteristic of paramagnetism. The y,-T relation is well described by the
equation. The Curie constant of 2.8 emu K and the constant (0)
of are obtained from the curve for 9, = 0.046 (the average
distance d 1.9 These Curie and constants are quite close to those for
bulk gaseous 3.1 emu K and 0 = K ) . Although the sub-
monolayer on the graphite surface gives rise to a magnetic anomaly due to
the melting of the 5 phase at 2 6 K for 1 , the low coverage layer in
pores of P-0.96 at < 0.5 provides no evidence of such a magnetic transition.
Hence. the adsorbed layer of 0.5 is completely different from that on the
flat graphite surface. Furthermore, the spin system of in the behaves
like a free spin regardless of the fact that is confined in the micropore by the
strong surface-molecule interaction potential.
Temperature dependence of magnetic susceptibility of adsorbed in

micropores of P-0.96 at overages of 0.046, 0.11, 0.24, (V) 0.46,
0.75, 1.1, and ( ) 1.3. (a) curves; (b) curves.

The temperature independence of 7 with the increase of is observed, especially
for the temperature range of 20-50 K at 0.46. However, this change is not due
to the effect of by the formation of 5, or a phase but to the
effect of the random magentism developed in the micropore as described later.
Clear jumps in the y-T curve of = 1.3 = 0.35 nm and = 0.82) are observed
at 24 and which correspond to the a-(3 and transition temperatures of the
bulk oxygen, respectively, whereas either phase corresponding to the bulk oxygen
does not appear below = 3.5 in the system This result shows
that even layers in the micropore can form a dense structure identical
to that of the bulk solid, since the pore width of P-0.96 is almost three times the
molecular diameter of This phenomenon may be ascribed to the strong
pore field, which offers a high pressure effect
The Curie and constants calculated in the several
temperature ranges are plotted in Fig. 9 to illustrate how the intermolecular inter-
action changes with an increase in since the curve is not straight at
0.2. The and values were evaluated from the linear approx-
imation for the curves in the temperature range of to where and
represent the low and high temperature limits, respectively, in the temperature
range. The and values change at = 1.2. The changes in
these constants reflect the phase transition of the bulk solid to Since the
values change steeply at 8, 1.2, these constants also correlate
to the phase transition. and however, show little change over the
whole range of and indicate the values relatively close to those for gaseous
This may be because in the very low temperature region, contributions from
free spins to the susceptibility are predominant but those the
phases or spin clusters are very small, provided that the mixed states of the
free spins and interacting spins coexist. We can regard and as the
parameters related to the amount of the free spins.
The curve for 6, 0.75 = 0.46 and = 0.47) in the low temperature
range is shown in Fig. l0a. The curve has a at 2.1 K after zero-field
cooling whereas it has no anomaly and can be described by the Curie-Weiss
equation for the field cooling (FC) process. The time course after a temperature
drop from 30 to 5 K is also shown in Fig. The moment slowly
increases with after the temperature drop. This magnetic behavior should be
attributed to random magentism such as spin glass or mictomagentism This is
because molecules form clusters consisting of a few or several molecules and the
cluster behaves as the system of an effective spin. Since such spins are frozen in the
low temperature range, we can observe the relaxation phenomenon for spin orien-
tation in the magnetic field. results are obtained for 0.24 1.1. The
spin-spin interaction of the spin clusters gives increases in and and
decreases in and before the bulk phase transition. as are shown at
0.2 1.0 in Fig. 9. The random magnetic behavior is not observed for
9,. 0.2, because a very small cluster or an isolated molecule behaves like a free
spin. On the other hand, the busy packing of the molecules inhibits the free motion
of the spin and facilitates the formation of the bulk solid phase, so that the
adsorbed of 1.3 shows no magnetism.

Magnetic
The Curie and Weiss constants in the several temperature

ranges for the system at = 0.75. The (a) and (b) values were
evaluated from the linear approximation for the curves (Fig. 8b) in the
temperature range of to where and represent the low and high
temperatures limit, respectively, in the temperature range.
The spin association of confined in micropores depends on the pore width.

Results similar to those of ACF-0.96 are obtained for adsorbed in the
pores of P-0.75, although the dependence of the curve on differs from that
of P-0.96. The curve for = 0.056 (d gives the constants of
C x emu K and -1.8 K. N o 2D magnetic lattice formation is
observed at 1.1 (d = 0.38 and = but the bulk phases appear at
1.1. The changes in and are not so great relative to those for
P-0.96. The magnetic relaxation process of the spin cluster in the of

10 Magnetic of the system at = 0.75. (a) curve.
The measurement was carried out after cooling to 1.7 K with the magnetic fi
zero-field cooling (ZFC)]. followed by application of the field to 1
after ZFC; field cooling (b) Time course after temperature drop from 30 to 5 K
at H = 1 The was dropped at the point indicated by an arrow.
P-0.75 is also observed a t 0.3 1.2, but the relaxation is much faster than that
of P-0.96. Therefore, P-0.75 offers a restricted molecular environment less suitable
for cluster formation than P-0.96.
The curve of adsorbed in the micropore of P-1.45 a t = 0.021
(d nm) satisfies the Curie-Weiss law 2.6 emu and
2.3 K) as is shown in Fig. 1 l a . The curves for 0.42, however, give different
features from those of the other two samples described above. T h e curves for
= 0.42, 0.60. and 0.78 have broad peaks a t 28, 36, and respectively.
Moreover, the bulk phase appears a t 1.1 (d 0.38 n m and = 0.63). This
is a smaller coverage than with either 0.96 o r P-0.75. The changes in and

. 11 (a) curves at coverages of 0.021, (A) 0.063,
0.42. 0.60, 0.78, and 1.1. (b)
are different from those for P-0.96 or Accordingly, and

especially, give discontinuous changes in the region 0.5 0.7. The
sample of = 1.1 gave the slowest spin-relaxation phenomenon in the
1.45 system regardless of the bulk phase formation at the coverage. A
of upon heating after ZGC and spin-relaxation after the temperature drop
are observed. These results indicate the coexistence of the bulk phase and clusters in
micropores of P-1.45, which is not observed in the of P-0.75 or P-0.96.
The detailed curves for = 0.42, 0.60, and 0.78 are shown in Fig. 12. These

FIG. 12 curves of system at coverages of of (a) 0.42, (b) 0.60, and (c)
0.78.
are somewhat similar to a curve for the 0 phase in the system obtained
by Kobler and Marx These authors propose that liquid and phases coexist in
the region of 1 0, 1.8 and 32 T 40 (32 K is the melting point of the 8 phase;
40 that of the 0 phase) for the system. The phase is presumed to
have a liquid crystalline structure in which molecular axes are ordered like those of

a smectic crystal, but the molecular centers are disordered in 2D plane. The
clusters formed in the micropores of P-1.45 should be in a less ordered state than
that of the phase. The values of and at = 0.42 are comparable to
those for P-0.96 at = 0.75, but the results of spin relaxation show that the cluster
formation proceeds to a lesser extent than in the latter case. This is because the
spins in the state of the 0-like phase also contribute to the two constants.
The magnetic behavior of the clusters formed in the of
depends on the pore width and tends to approach that of the system
with an increase in the pore width. The and values are related to the
amount of the free spins as described above and are shown in Fig. 13. These results
indicate that the amount of free spins decreases with increasing at a constant
pore width and with increasing pore width at a constant although the
values for ACF-0.96 and ACF-1.45 deviate noticeably after the phase forma-
tion owing to the influence of the antiferromagnetism of the a phase. Each
0.02
lad 0.015
Dependence of and on for three systems. (a)

curve, (b) vs.

to zero gives the Curie and Weiss constants of the free spins confined
of
in the = 2.6 x emu K and = for P-0.75;
= 3.0 x emu K and = for P-0.96; and = 2.9 x
emu K and = for All values obtained are very close to
those of gaseous The results for each system also indicate the
spin cluster is formed at 0.015. The phase diagrams of the systems
deduced the total data in this work are shown in Fig. 14. The phase diagram is
simpler than that of the system because few phase transitions occur
before the bulk phases appear in the two-dimensional system. It is very interesting
that none of the liquid or solid phases are observed for the in the
gas cluster
30
,
gas + cluster
a gas
rams for confined in of ACF. (a) P-0.75; (b) P-0.96;

in (c) denote the value at the peak and at the trough, respectively,
in Fig. 12.

Ma
low temperature range. It is plausible that molecules confined in the slit pore
of 1 n m width cannot form a two-dimensional lattice similar to that on the flat
graphite surface. The relationship between the disappearance of the 2 D lattice
formation and the slit width should shed light o n the unusual state of
confined in a narrow space.
Water molecules can form a network structure that accepts a guest molecule to
produce a clathrate compound under application of high pressure o r low tempera-
ture Recent in situ X-ray diffraction studies showed that molecules
confined in the graphitic of A C F form an ice-like structure and
produce a clathrate compound different from the bulk one even near ambient
conditions Such a clathrate structure can isolate an spin system.
Thus, knowledge of the magnetic susceptibility of the mixture confined
in the of A C F a t different compositions and fractional fillings over the
temperature range of is necessary for understanding assembly.
The results for P-0.96 are shown here. The diamagnetic contribution of ice
(-0.699 x emu g) was subtracted from the observed value The coexistent
affects seriously the magnetic structure of even the bulk condensed as
mentioned earlier. Earlier studies showed that the y-T curve of confined
for 0.16 0.25) has no critical change in the temperature range of 2-100
and confined shows within a low range; an almost
free spin state was confirmed by the Curie-Weiss parameters close to those of
gas. Temperature independence of was observed in temperature range of
20-50 K for 0.16 which is associated with magnetism due
to spin cluster formation having an explicit relaxation. The temperature depen-
dence of of confined in the of A C F is shown in Fig. 15 as a function
of in the presence and absence of Here. the content is expressed in
terms of the molar ration R,,. Although the coexistent brings about a
paramagnetic nature as mentioned above, it does not sensitively affect the magnetic
behavior of confined for because of the free spin state even without
the coexistent O n the other hand, the addition of excess noticeably
changes the relation, which indicates the perfect isolation of by coexistent
Figures 16 and 16b show the Curie-Weiss plots for R,, and 28.7,
respectively. The coexistent bends the plot upward markedly, indicating that
the coexistent accelerates the spin cluster formation. inducing random mag-
netism. In the case of the Curie-Weiss plot bends down at for both
cases, with and without T h e spin concentration is decreased slightly by the
presence of while the constant does not change, as shown in Table 3.
The decrease of the spin concentration by the presence of supports the for-
mation of spin clusters. Association of molecules should lead to a n island
formation of molecules, producing the spin cluster. The Curie-Weiss plot in the
absence of is linear. whereas that with a n excess of is convex. The
constants from the Curie-Weiss plot are also shown in Table 3. The spin

FIG. 15 Temperature dependence of the magnetic susceptibility of confined in the
graphitic nanospace at different fractional fillings by and amounts of coexistent
= 0.029, 0; ( + ) = 0.23, = 0; 4) 0.45, = 0;
4) = 0.18. = 0.12; 4) 0.30, = 0.21; 0.20. = 0.35;
) 4) = 0.045, = 8.7; 4) = 0.0057, = 76; = 0.0023. = 172.
0 10 20 30 40 50 5 10 15 20 25
Temp K Temp K
. 16 The plot of confined in graphitic nanospace. (a) A

content of (V) = 0.20, = 0.35; ( + ) 4) =
excess amount of = 0.029, = 0; = 0.045, R,, = 8.7; (@)
0.0057, = 76; 4) 0.0023, 172.

3 Curie-Weiss Constants in the Temperature Range of 2-10 K
Fractional filling radio, Spin concentration Weiss constant
concentration of diluted by is smaller than that of the bulk gas, while

the Weiss constant is close to zero. The decrease in the spin concentration suggests
a strong interaction between molecules such as formation and/or
of with molecules in the graphitic nanospace. The diamagnetic
of surrounded by molecules should reduce the value. Accordingly,
single and form a clathrate structure in the nanospace to give rise
to almost perfect paramagnetism. Thus, the doping of molecules in an
assembly can control the magnetic interaction of in the graphitic
nanospace.
Chemisorbed species solid surfaces have been widely studied. and established
methods are available for determination of the different chemisorbed species. F o r
example, five representative species of oxygen- 0;.
and 0 - h a v eb een evidenced with the aid of I R and ESR for oxygen
o n the surface of transition metal oxides However, interactions in physical
adsorption, even micropore filling. are too weak, and we have no established
method for assigning the different physisorbed states. The interaction potential
profiles of a n molecule with slit graphite are shown in Fig. 1. Although the
potential minimum difference of three shown in Fig. 1 is only 150 K, which is
very small, it is large enough for magnetic interaction. There is a possibility that a
slight difference in a weakly interacting system can be measured by the magnetic
method. The magnetic method is quite sensitive, and we identify different oxygen
states in micropores with the aid of magnetic susceptibility measurements at low
temperature, as described above. The magnetic behaviors of micropores can
provide a n adsorbed mechanism. The average intermolecular distance d is calcu-
lated from the fractional filling using the assumption of isotropic and homogeneous

distribution of in the pore. Generally speaking, of 1 n m is necessary
for the spin dipole-spin dipole interaction. It has been recently elucidated that a
new spin relaxation of bulk gaseous occurs through a quadrupole-mediated
mechanism a t the contact distance of Lennard-Jones molecules
Consequently. the critical distances are and about 1 n m in the magnetic
interactions of Oxygen molecules d o not magnetically interact with each other,
and they are isolated a t a distance of more than 1 n m below = 0.2 0.3). In
the middle fractional filling range of 0.2 some isolated
molecules are associated with each other to form a cluster that causes
magnetism. Clusters are aggregated to produce a n organized assembly near com-
plete filling. The condensed phase in micropore exhibits same magnetic
transitions as the bulk phase. However, the jump at the magnetic phase
transition is not sharp, and therefore the condensed molecular layer should contain
defects due to geometrical restrictions.
The grand canonical Monte simulation of adsorption by a graphite slit
at 100 K showed that molecules are orientationally ordered on the micropore
walls If molecules have an orientationally ordered structure in the micro-
pore as suggested by the simulation work, such a n ordered structure coincides with
the cluster and the organized formation. Thus magnetic measurements are
quite effective for identifying weakly interacting molecular states in
I a m grateful for D r . Dr. Ishii, and M r . J. for collaborative

works. This work was funded by a Grant-in-Aid for Scientific Research on Priority
Areas N o . 288, "Carbon Alloys," from the Japanese government.
A. Physical Chemistry New York, 1990.

J. Gregg and W. Sing, Adsorption, Surface Area, and Porosity, Academic,
London. 1984.
Sing, D. H . Everett, A. Haul, A. Pierotti, and
Pure
Everett and C. J. Faraday Trans. I (1976).
Emmett, and Teller, (1938).
Gregg and K . Sing, Adsorption, Surface Area, and Porosity. Academic,
London, 1984, p. 113.
P. Broekhoff and Linsen, in Physical and Chemical Aspects of
Adsorbents Catalysts Linsen. ed.), Academic. London. 1970, Chap.
H. Everett. in The Solid-Gas Interface A. Flood, Marcel New
York, 1967. Chap.
and D. Everett, Adsorption of Gases on Heterogeneous Surfaces,
Academic. London. 1992.
A. and Quirk. Carbon
A. V. Bakaev and W. A. (1992).

F. Cracknell, D. Nicholson, and Gubbins. Chem. Trans.
(1995).
Cracknell, Nicholson, Gubbins, and Chem. Res.
(1995).
Smith, A. Wakehani. and The Forces between
Molecules. Oxford Oxford, UK. 1986.
0. Cluster Chemistry (in Japanese), Tokyo, 1992.
K . Kaneko. (1987).
Kaneko, and Ozeki. Chem. Phys. (1987).
Sounia, Ozeki, and Kaneko. Phys. Chem. (1991).
Kaneko, and Suzuki, Chem. Phys, (1992).
Moreh and Levant, Phys. (1990).
Iiyama. Otowa. and Kaneko, Phys. Chem.
(1995).
Minagawa. Suzuki. and Kaneko, Chem. Phys. Lett.
(1995).
Kanda. Y. Maeda, and Kaneko, J. Phys. Chem.
(1998).
Suzuki. Kasuh, and Kaneko, Phys. Lett. (1993).
Kanoh and Kaneko, Phys. Chem. (1995).
Kanoh and Kaneko, Chem. Phys. Lett. (1995).
A. Steele, Chem. Rev. (1993).
Takaishi. Chem. Faraday Trans. (1997).
A. Steele, Surf. (1973).
Kaneko, Cracknell. and Nicholson, Langniuir (1994).
Van The Theory of Electric and Magnetic Susceptibilities, Oxford
Univ. Press, Oxford, UK, 1932.
A. and Witte, (Translated into Japanese by Sorai).
Tokyo, 1980, Chap. 4.
Molecular Magnetism, VCH. Weinheim, 1993.
Statistical Theory of Magnetism (in Japanese), Tokyo, 1975.
Phys. Rev. (1981).
Kanda, and A. (1954).
P. Stephens and Majkrzak. Phys. Rev. B (1986).
and Date, Phys. (1985).
Dunstetter, V. P. Plakhti, and Mater.
(1988).
Barrett, Meyer, and Phys. (1967).
P. and Phys. C (1983).
A. J. da and Falicov. Chem. Phys. Lett. (1994).
and Leroi, Chem, Phys. (1969).
Kaneko, and C. Phys. Rev. Lett. to be submitted.
A. A. P. Stachowiak, and V. J.
Phys. Matter (1995).
Gregory. Phys. Rev. Lett. (1978).
Lewis, and Phys. Rev. Lett. (1983).
48. and Marx, Phys. Rev. B

50
and Magnet. Magnet Mater. (1990).

K . Kaneko and Ishii, Colloid Surf, (1992).
Setoyama, Suzuki, and Kaneko, Carbon, (1998).
M. Jaroniec and K . Kaneko, (1997).
J. Phys. Rev. (1956).
Kaneko, Ishii, and Suzuki, Chem.
Phys. Lett. (1991).
and Rev, (1978).
A. Suzuki, Enoki, Ishii, Kaneko, and
Solid State (1995).
C. and Kaneko, Chem. Phys, Lett. (1995).
Ishii and Kaneko, Prog. Organic Coat. 1 (1997).
Ishii. Suzuki, and Kaneko, Porous Solids (1997).
A. Zamma, Sugihara. and Kaneko, Phys. Chem, to be submitted.
Kanoh, A. Setoyarna, and Kaneko,
(1997).
Carbon (1987).
Kaneko, Ishii, Kanoh, and Suzuki. Adv.
A. Mydosh, Spin Glasses: A n Experimental Introduction, Taylor Francis,
London. 1993.
Sparks and Texter, Chem. (1992).
Hydrate of Natural Gases, Marcel New York.
1990.
T. T. Suzuki, and Kaneko, Lett.,
(1997).
Suzuki. Ishii, and K. Kaneko, Chem. Phys. Lett. (1998).
A. and A. Gedanken, Chem. Trans.
(1996).
Transition Metal Oxides: Surface Chemistry and Catalysis,
Amsterdam, 1989, Chap. 7.
and Kaneko, Phys. (1996).
Sokotowski, Mol. Phys. (1992).

0 Corporate Science and
Technology Center, Air Products and Chemicals. Allentown. Pennsylvania
I. Introduction 50 1
II. Heats of Adsorption by Gibbsian Thermodynamic Model 502
III. Heat Removal from and Supply to an Isothermal Adsorption System 506
IV. GSE Model for Mass and Energy Balances in Adiabatic
Adsorption Columns 507
V. Indirect Estimation of Isosteric Heat of Adsorption 508
A. Pure gas isosteric heat 508
B. Multicomponent gas isosteric heat 510
C. Shortcut method 510
VI. Direct Estimation of Isosteric Heat of Adsorption 513
VII. Characteristics of Pure Gas Isosteric Heat of Adsorption 516
A. Homogeneous adsorbent 518
B. Heterogeneous adsorbent 518
C. Limiting value of isosteric heat of adsorption 520
D . Special cases of adsorbent heterogeneity 520
E. Simple models for isosteric heat of adsorption 52 1
VIII. Characteristics of Multicomponent Gas Isosteric Heats of Adsorption 521
A. Limiting values of isosteric heat of adsorption 524
B. Simple models of multicomponent gas isosteric heats of adsorption 524
C. Example of model heterogeneity for a binary system 526
References 527
Adsorption and desorption are, respectively, exothermic and endothermic

processes. Consequently, all practical adsorptive gas separation processes such as
pressure swing adsorption (PSA) and thermal swing adsorption (TSA) are

isothermal. These processes are generally carried out using nearly adiabatic adsor-
bers, and the local temperatures of the gas and the solid adsorbent b
tinuously along the length of the adsorber during various steps of the and TSA
processes The local gas adsorption equilibria and kinetics within the adsorber
depend on the local gas pressure (P), the gas and adsorbent temperature and
the gas-phase composition (mole fraction of component Thus, it is critical
that accurate knowledge of the heats of adsorption and desorption of the adsor-
bates as a function of and be available so that correct temperature changes
within the adsorber can be estimated during the process steps. A theoretical sensi-
tivity analysis of the performance of a simple PSA process (Skarstrom cycle) for
separation of a binary gas mixture using a detailed mathematical process model
showed that small errors in the estimation of heats of adsorption of the adsorbates
and their variation with adsorbate loadings could significantly alter the process
performance Furthermore, the study clearly demonstrated the pitfalls of ignor-
ing adsorber
Traditionally. a variety of heats of adsorption and desorption for pure and
gas-solid systems have been defined by using thermodynamic
models Experimental techniques have also been developed to measure
these heats These generally use the actual amounts adsorbed as the
primary variables for representing the extents of adsorption of the adsorbates.
Unfortunately, the Gibbsian surface excesses (GSE), and not the actual amounts
adsorbed, are the only true experimental variables for measuring the extent of
adsorption In view of this fact, a detailed thermodynamic model for
component gas adsorption equilibria using GSE as base variables has already been
developed
This chapter (1) elaborates on the thermodynamic properties of multicomponent
heats of adsorption and desorption using the GSE framework, (2) demonstrates the
practical utility of these heats by developing mathematical models to describe mass
and energy balances in adsorption columns using GSE, (3) describes direct and
indirect experimental methods for estimating the heats, (4) provides experimental
and theoretical examples of the heat of adsorption of pure gas and binary gas
mixtures on microporous adsorbents (energetically homogeneous and hetero-
geneous), and ( 5 ) comments on the complex nature of heats of
adsorption. Only solids are considered in this chapter because of their
practical values.
Figure 1 shows the GSE model for equilibrium adsorption from a bulk
gas of i components N) characterized by and
The adsorption system contains a unit amount of an inert adsorbent whose total
helium void volume is It is assumed that all pores of the adsorbent are
accessible to the nonadsorbing helium gas. The dotted line in Fig. 1 represents the
interface separating the Gibbsian adsorbed phase and the bulk gas phase. It
is arbitrarily located inside the actual bulk gas phase. The adsorbed phase has a

1 gram of adsorbent
.1
Schematic of Gibbsian surface excess (GSE) model for gas
adsorption.
volume of and the actual amount of component adsorbed is

The density of the adsorbed phase is = The bulk gas-
phase molar density is p The pressure (P)and temperature (T)of the
bulk and adsorbed phases are the same, is the total amount of component
g) in the entire system.
The magnitudes of variables and o n the location of the Gibbs
interface. and therefore the actual amount adsorbed cannot be unequivocally mea-
sured. O n the other hand, the G S E for component of the gas mixture
is independent of the location of the Gibbs interface and can be estimated by
measuring P. and vO is defined as
All conventional experimental techniques such as volumetric, gravimetric,

loop recycle, total and chromatographic techniques the GSE
of a pure gas as functions of P and T o r those of the components of a gas mixture
as functions of and It can be easily demonstrated that when
(1) for pure gas adsorption and (2) for multicomponent gas
adsorption, where is the adsorbed phase mole fraction of compo-
nent Otherwise, there may be a substantial difference between and
An excess internal energy( a n excess enthalpy a n excess
entropy a n excess Gibbs free energy and an excess
volume can also be defined for the thermodynamic system of Fig. 1
7
-
where and can be the internal energy the molar enthalpy the
molar entropy the molar Gibbs free energy (g). and the molar volume of the
adsorbed and bulk gas phases, respectively, is the corresponding excess function.
It should be noted that is equal to zero by this definition, and thus
(= + is equal to

Integral and differential thermodynamic relationships between the different
excess quantities defined by (2) and the set of experimental variables
or ( T , can be derived analogously to those for conventional
phase thermodynamic properties However, an additional intensive property
called the surface potential is necessary to completely define the
adsorbed phase. The surface potential can be calculated by using the
relationship
For pure gas adsorption, Eq. (3) can be used to calculate T) by measuring
the pure gas surface excess as a function of gas-phase pressure at constant
T (adsorption isotherm for pure gas). For a multicomponent gas mixture, Eq. (3)
can be used to calculate by measuring the surface excess of component
as a function of total gas-phase pressure (P) at constant T and The
variable of (3) in this case is the total surface excess for all components
R is the gas constant.
The total internal energy and the total enthalpy of the entire
adsorption of Fig. 1 at and are given by
where K)] is the heat capacity of the adsorbent and is a reference

temperature. Equations (4) and (5) can be combined for the case where the gas
phase (pure or mixture) of Fig. 1 is ideal (h = + = RT: = t o get
Both and can be expressed as functions of and or and Thus, one

can show by using the chain rule of calculus that
One can now define the following thermodynamic variables

It can be shown by using the thermodynamic relationships between different
extensive and intensive variables of the of Fig. 1 that
where is the molar heat capacity of pure gas It is assumed in the

derivations of Eqs. (11)-(14) that the gas phase of Fig. 1 is ideal. The variables
and defined by Eq. are two different forms of differential heats
They represent the differential changes in the total internal energy or the total
enthalpy of the closed system of Fig. 1 due to a differential change in the surface
excess of component i while the surface excesses of other components and the
system temperature are held constant. They are related by Eqs. (11) and (12).
They can be indirectly estimated from experimental adsorption equilibrium data
using (11). and respectively, are called the differential and isosteric heat of
adsorption of component i from a gas mixture. For pure gas adsorption. and
are simply obtained by setting 1 in Eq. (11). Traditionally, and q, are
defined by Eq. substituting The present derivations show that
these variables can be analogously defined by using GSE as the base variables.
The isosteric heat of adsorption of component can be a function of and
or of T and
The variables and defined by Eq. are two different heat capacities
of the adsorption system of Fig. 1. They represent the changes in the
total internal energy or total enthalpy of the closed system of Fig. 1 due to a
differential change in the system temperature at constant Equation (14) relates
these two different heat capacities.
Equation (13) can be used to calculate from experimental adsorption equili-
brium data. For pure gas adsorption, or can be obtained by simply setting
1 in (13). The summation terms of (13) are then replaced by a single
The total surface excess of all components of Fig. 1 can be

expressed as functions of T and gas-phase partial pressures of component
Thus, by using the chain rule of calculus, one has

Equation (15) can be differentiated with respect to T at constant combined
with (11) to get
It can be shown from Eq. that for a closed system (constant with an ideal
gas phase, one has
Equation (3) can be differentiated with respect to T at constant and and

combined with Eqs. (16). and (17) to get
For the special case where is not a function of and Eq. (18) simplifies to
Equation (19) was previously derived for adsorption of a pure gas
A certain amount of heat must be supplied to or removed from the closed

adsorption system of Fig. 1 to maintain system isothermality when the GSE of
component is changed by a differential amount By applying the conven-
tional energy balance for the closed system I], one gets
The negative sign for the quantity in Eq. (20) represents the case where heat is
removed from the system in order to maintain system isothermality. Equations
and (20) can be combined to obtain
Equation (21) was previously derived by many authors using the actual amount
adsorbed instead of the GSE as the base variable. The equation demonstrates
the practical value of the differential heats defined by Eq. (9).

It is demonstrated in this section that the GSE framework of adsorption system can
be easily used to describe instantaneous unsteady-state mass and energy balances in
an adiabatic adsorption Figure 2 is a schematic diagram of gas flow
through a differential section (AZ) of an adiabatic adsorption column located at
a distance Z from one end of the column at any given The column is packed with
the adsorbent particles. The gas flow is in the direction of coordinate Z. It is
that (1) the gas flow through the adsorber is plug flow in nature, (2)
there is instantaneous thermal equilibrium between the gas and the adsorbent,
(3) there is no axial dispersion of mass or heat, including axial conduction of
heat between the adsorbent particles, and (4) the column is adiabatic. The variable
Q represents the instantaneous molar flow rate of gas per unit cross-sectional
area of the adsorber It can be a function of Z and and are,
respectively, the gas pressure, gas and adsorbent temperature, and gas-phase mole
fraction of component at Z and The instantaneous mass balance for component
i across the differential section at time yields
where is the total amount of adsorbate i per unit volume of the

adsorber at Z and
where is the instantaneous of component at Z and is the

helium void fraction of the adsorber, and is the gas-phase
density at and is the bulk density of the adsorbent.
The instantaneous energy balance across the differential section A Z (open adia-
batic system, no shaft work) at time t yields
Adsorbent Particles
Schematic of instantaneous mass balance across a differential section in an

adsorbent column at time

where is the molar heat capacity of the gas phase at and Equations
(12). and (24) can be combined to get
Equation (25) represents the instantaneous column energy balance using the
framework. The key thermodynamic variables in the equation include q, and
defined by Eqs. (9) and respectively. The last term on the right-hand side of
Eq. (25) vanishes for isobaric process steps (constant P).
The quantity in (25) represents the instantaneous rate of change of
surface excess of component i at Z and t. This replaces the local rate of adsorption
term in the conventional energy balance equation derived by using the actual
amount adsorbed as the base variable. A simple model to describe the rate of
change of GSE is the analog of the conventional linear driving force model
where is the equilibrium GSE of component at the local values of and

is the mass transfer coefficient for the ith component of the gas mixture.
Equations (22)-(26) show that the unsteady-state mass and energy balances
within the adiabatic adsorber can be written using the surface excess of each com-
ponent of the gas mixture as the primary variables to define the extent of adsorp-
tion. The isosteric heat of adsorption of component i and the heat capacity of
the adsorption system defined using the framework, become the appro-
priate thermodynamic properties to describe the energy balance.
For pure gas adsorption of component (1 1) and (15) can be modified

m
= = = P) and differentiated with respect to at constant to
obtain
Equation (27) shows that the isosteric heat of adsorption of pure gas i can be
estimated at a given value of P and T or T and by (1) measuring a set of pure gas
adsorption isotherms a function of P at constant T ) a t different tempera-
tures, (2) generating a set of adsorption isobars as a function of a t constant
P) from the isotherms or measuring them independently, and (3) estimating the
isotherm and isobar slopes at the desired values of P and The use of Eq. (27) for
calculating for pure gas is, however, very sensitive to errors in measuring the
isotherm and isobar slopes. On the other hand, it is very useful to develop analytical
expressions for as functions of T and the adsorption isotherms can be
described by analytical models

A more popular method of estimating a, for adsorption of pure gas = 1) is to
integrate (1 1) by assuming that q, is not a function of T . Thus,
where c is a constant or a function of According to a plot

of P against at constant values of yields a straight line with a slope equal to
Thus q, can be estimated as functions of measuring pure gas adsorp-
tion isotherms at different temperatures. The linearity of Eq. (28) is experimentally
validated over a large range of T for many pure gas adsorption iso-
therms on microporous adsorbents, indicating that for a given is indeed a very
weak function of T . Figure 3 shows a family of linear plots according to (28)
for adsorption of pure on BPL carbon at different values of . These plots
were generated from pure gas adsorption isotherm data at and 301.4 K
One practical problem in using (28) to calculate as a function of

is that the procedure requires very high pressure isotherm data at the higher
temperatures
. 3 Plots of P versus at constant adsorption of pure on BPL

activated carbon.

The common methods of measuring the extent of adsorption of
gas systems are to measure (1) as functions of a t constant and (2) as
functions of T at constant P and and (3) as functions of at constant and
T Adsorption thermodynamics can be used to generate expressions for multi-
component gas isosteric heats of adsorption using the slopes of the three function-
described above. These expressions are analogous to (27) derived for
pure gas adsorption. For a binary system = 1, 2) one gets
where is the gas-phase mole fraction of component 1, and
It can be seen that the experimental protocol to generate data for use in (29)
and (30) can be extremely time-consuming and impractical. We are not aware of
any such data in the published literature. The degree of complexity increases enor-
mously when the gas phase has three of more components.
It is also unfortunate that Eq. (1 1) cannot be easily integrated for
nent adsorption as in the case of pure gas adsorption because no multi-
component adsorption experiment can be designed to generate as functions of T
at constant = 1. 2, . . . , N). Consequently, there is no easy way to estimate
multicomponent (i 2) isosteric heats of adsorption by measuring as functions
of P. and
A shortcut for estimating the isosteric heat of adsorption of pure and

multicomponent gas has been proposed It consists of packing a
container with a unit amount of the adsorbent of interest and then saturatin the
adsorbent (initial state) with a pure gas at and or with a gas mixture at P ,
and The container is placed in a thermostated bath at and is equipped with a
gas-mixing device (optional) and instruments for measuring the gas-phase pressure
and composition. The amount of pure gas i or the amounts of the components
(i = .., of a gas mixture introduced into the container to reach the initial
state of saturation are measured. The temperature (T) of the thermostated bath is
then periodically changed, and the changes in the and of the gas phase within
the sealed container are measured after every new equilibrium state is achieved at
different values of The total amount of each adsorbate inside the closed con-
tainer ($, remains constant during the process.

It is argued that is approximately equal to according to pro-
vided that >> This can be met by having a small specific helium void
volume for the container and choosing appropriate values of
and Consequently, it is assumed in the shortcut method that
The apparent heat of adsorption of component i (pure or

component gas) defined by (32) is estimated by the shortcut method. According
to Eq. a plot of the pressure of pure gas 1) or the partial
pressure of component against at a constant value of should yield
a straight line with a slope equal to if is independent of T. The linearity of
such plots has been experimentally verified for many systems using the experimen-
tal data obtained by the shortcut method Thus and hence can be
estimated as functions of by the shortcut method.
Unfortunately, the approximation assumed by (32) may not be valid
even if the approximation is reasonable and a, are equal only if is
equal to zero, which is not practically possible. The minimum value of for a
container tightly packed with the adsorbent particles is given by
where and are, respectively, the chemical and bulk densities of the adsorbent.
Most practical microporous adsorbents (zeolites, activated carbons, aluminas,
silica gels, etc.) exhibit a value of and values ranging between
0.4 and Thus, the typical minimum values of range between 0.5 and
2.0
The exact relationship between and for closed system
(constant can be derived by expressing in functional form in terms of
and to estimate the change in for differential changes in those variables and
combining the result with the differential form of (1).
It can be shown that for adsorption of pure gas the difference between and
is given by
Equation (33) shows that (1) the difference between and q, increases as P or
increases (isotherm slope decreases) and (2) a, because R T ) is generally
positive.
For a binary system one gets

where is the gas-phase mole fraction of component 1 and variables and are
given by (31).
Table 1 shows the differences between and and and for binary
adsorption of (component 1) + (component 2) mixtures on BPL carbon
at 303.1 K They were estimated at three different total gas pressure levels and
at different gas-phase compositions. The table also gives the fractional adsorbate
loadings (6,) of the components is the saturation capacity
(surface excess) for both components. The pure and binary gas adsorption iso-
therms for this system can be described by the model The model
parameters are given in Table 1 . A value of was used as for these
calculations. It can be seem from Table 1 that the differences between and are
relatively small because (component is more selectively adsorbed than
on the carbon. The differences between and are, however, much larger.
Furthermore, the differences between and are much larger than the corre-
sponding differences between and The differences between and get even
bigger as the system pressure and the mole fraction of component in the
gas phase are increased. These examples demonstrate the weakness of the shortcut
method.
Equations on the other hand, show that a, can be calculated using the
values obtained by the shortcut method provided that are estimated
from some other data source. For pure gas adsorption, the denominator of Eq. (33)
can be easily obtained by independently measuring the adsorption isotherm for the
pure gas at the base temperature of the shortcut method. For a binary gas
adsorption system, can be estimated by independently measuring the
total surface excess as functions of at constant and and as
functions of at constant P and
1 Examples of Errors in Estimating by Shortcut +

on carbon at 303 K]
Conditions Adsorbate loadings Estimated error
parameters: in =4.35 2.682 atm); atm);

5.29 = 4.17

at constant T and
Equations (36)-(40) are derived using adsorption thermodynamics Even

though the above-described experimental protocols for a binary system are tedious,
they provide the additional information to calculate from using (33)-(35).
We are not aware of any work where the above-described method is tested. This
thermodynamic approach has not yet been extended to systems containing more
than two adsorbates.
It must be very clear from the above discussions that the available indirect methods
for estimation of may not be very useful for 2) adsorption
systems. A calorimetric method is necessary where the heat evolved (or consumed)
due to specific changes in the adsorption system is directly measured. Recently, a
heat flux microcalorimeter was developed to measure the heat of
adsorption Figure 4a is a schematic diagram of the calorimeter. It consists
of a thermostated cell whose walls are surrounded by thermopiles for directly
measuring the heat flow signals through the cell walls. Figure 4b shows the typical
profile of thermopile response (in microvolts) versus time during an experiment.
The area under the curve of Fig. 4b gives the total energy removed from the
calorimeter during an experiment. It is necessary to calibrate the response curve
using a known heat generation system within the cell.
The cell is connected with instruments for measuring the gas-phase pressure and
composition. It is also connected to a thermostated gas injection system (dosage
side) through a valve. A unit amount of the adsorbent is placed in the cell (sample
side), and it is initially saturated with pure gas i at and or a gas at
and The experiment consists of very slowly introducing a very small quan-
tity of pure gas i (AN,, into the sample side from the dosage side through
the valve. The temperature of the gas entering the sample side is monitored
as a function of time in conjunction with the thermopile response profile. The
dosage side pressure and temperature are also as functions of time.
The system temperature is at both the start and end of the experiment.
The amount of pure component introduced during the experiment is given by

(a) Schematic of type of calorimeter and
response curve.
where is the rate of introduction of pure gas into the side. It can
be calculated by knowing the volume of the dosage side and the rate of
change of gas density in the dosage side during the experiment. t*
is the time interval of the experiment. The amount of energy (enthalpy,
introduced into the sample side of the calorimeter during the experiment
is given by

where is average molar enthalpy of pure gas i introduced into the
sample side. is equal to zero if during the experiment.
The mass balance for the sample side of the calorimeter during the experiment
can be described by
and are, respectively, the changes in the surface excess and the
phase partial pressure of component in the sample side. is the helium void
volume of the sample side.
An energy balance for the side of the calorimeter (open
no shaft yields [1
where is the total change in internal energy of the sample side of the
calorimeter during the experiment. is the total of heat removed
from the sample side of the through the during the experi-
ment. Equations (1 and (42)-(45) be combined to get
where ... is the change in the gas-phase partial pressure atm)

of component j in the sample side of the calorimeter when moles of pure gas
is introduced into the sample side.
For a multicomponent gas system = . .., a total of N experiments
must be carried out around the base saturation conditions of (corre-
sponding surface excess of component by separately introducing differ-
ential quantities of each pure gas i into the sample side. This
generate N independent equations 2, . , . , analogous to (46)
relate N different unknowns Consequently, the isosoteric heat of adsorption
of component can be estimated as functions of and
For adsorption of pure gas (46) simplifies to

The direct calorimetric method for measurement of for multicomponent gas
mixtures (i 2) may still be tedious, but it is simpler, more accurate, and
less time-consuming than the indirect methods.
Figures 5 and 6, respectively, isosteric heats of adsorption of pure

and as functions of adsorbate loadings (surface excess) on three different
microporous adsorbents crystals, zeolite crystals, and granular
activated carbon). 7 shows similar plots for adsorption of three differ-
ent pure gases and on zeolite crystals and pellets. The data
for adsorption of and on carbon and those for adsorption of
on zeolite pellets calculated with (11) using the pure gas adsorption
isotherms at different temperatures The isotherms on the zeolite pellets
were measured in our All other data were measured by using a
type of calorimeter
Silicate is a crystalline, hydrophobic, nonpolar, microporous adsorbent. The
zeolite is a crystalline, hydrophilic, highly polar microporous adsorbent.
The activated carbon is an amorphous, moderately polar, micro-mesoporous
Isosteric heats of adsorption of pure as functions of adsorbate surface

excesses on zeolite, and BPL activated carbon.

FIG. Isosteric heats of adsorption of pure as functions of adsorbate surface
on zeolite, silicate, BPL activated carbon.
Isosteric heats of adsorption of pure and as functions of their

surface excesses on zeolite.

adsorbent. pores the zeolite are extremely homoge-
neous and well defined, while the activated carbon has a network of various
and The adsorbates range from nonpolar to polar
to polar and they have sizes, molar volumes, and
abilities. Table 2 lists the key properties of the adsorbates.
An adsorbent is said to be energetically homogeneous for adsorption of a pure gas

if all the sites of adsorption for that gas have the same energy of adsorption.
Consequently, the heat of adsorption of a pure gas i on that adsorbent
remains constant for all values of the surface excess The data of Figs. 5 and 6
show that silicate is energetically homogeneous for adsorption and
within the range of the data.
Even for an energetically homogeneous adsorbent, can vary with if (1) the
lateral interactions in the adsorbed phase are significant, (2)
the adsorbate orientation on the adsorbent surface changes with coverage; (3) there
is adsorption or condensation of the adsorbate the and (4)
there is a change in the physical structure of the adsorbed phase. The lateral inter-
actions cause to increase with increasing n,: particularly the surface cover-
age is large. The adsorption of (Fig. 6) and (Fig. zeolite are
examples of significant lateral interactions in the adsorbed The interactions
between the adsorbed can be very different from those in correspond-
ing bulk gas or liquid phase due to the presence of the adsorbent molecules,
create interactions in the adsorbed phase.
An adsorbent is said to be energetically heterogeneous for adsorption of a pure gas

if consists of a distribution of sites of different adsorption energies. Even though
all sites are populated by the adsorbate molecules at all gas pressures, the higher
energy sites are predominantly occupied by adsorbate molecules a t the
gas Lower energy sites are progressively occupied by the adsorbate
as the gas pressure increases. Consequently, for pure gas adsorption
decreases as increases (P increases).
Molecular Properties of Various Adsorbates

Liquid molar Kinetic
Dipole moment moment
Adsorbate (cm3, (A) (x (x . (x ,
37.7 3.8 26.0 0.0

53.6 44.7 1.56
72.7 5.8 65.4
33.3 3.3 26.5 0.0 4.30
normal point.

The adsorption of on zeolite and BPL carbon (Fig. 5 ) and the adsorp-
tion of on BPL carbon (Fig, 6) indicate that these systems exhibit various
degrees of adsorbent heterogeneity. The highly polar zeolite is more hetero-
geneous for adsorption of polar molecules than the less polar carbon.
The data of Fig. 7 show that the same adsorbent can exhibit vastly different
solid interactions depending on the nature of the adsorbate. The adsorption of
on zeolite is nearly homogeneous. The adsorption of exhibits significant
lateral interactions. The adsorption of is significantly heterogeneous. These
behaviors can be qualitatively explained by the properties of the adsorbates. The
nonpolar adsorbates and are not affected by the polar heterogeneity of
the zeolite. However, the larger and more polarizable molecules introduce
significant lateral interactions within the zeolite cavities. The polar molecules,
on the other hand, experience the polar heterogeneity of the zeolite.
The energetic heterogeneity of the adsorbent is physico-chemical in nature. For
microporous amorphous adsorbents (activated carbons, aluminas, silica gels), the
heterogeneity is caused by (1) the presence of a network of micro- and mesopores of
different sizes and shapes, (2) the presence of different polar and nonpolar surface
groups on the pore walls, and (3) the presence of strongly adsorbed trace impurities.
For microporous crystalline adsorbents (zeolites), where the physical structure of
the crystal lattice is well defined and uniform, the heterogeneity is caused by (1)
structural defects, (2) the presence of different types of cations in the structure, (3)
the location and accessibility of the cations, (4) hydration of the cations, and (5)
variation in the ratio in the zeolite framework. It is not possible to a
priori characterize the adsorbent heterogeneity at the microscopic scale (pore size
distribution or surface chemistry) by using today's technology. Measurement of
overall isosteric heats of adsorption of pure gases remains the best way to char-
acterize the macroscopic heterogeneity of adsorption of gases on
ous solids.
Despite the above-mentioned problem, numerous works have been published to
micromodel adsorbent heterogeneity by assuming that the adsorbent comprises a
patchwise distribution of homogeneous sites of different isosteric heats of adsorp-
tion According to this concept, the overall surface excess isotherm of a pure
gas on a heterogeneous adsorbent is given by
at constant T
where is the local surface excess isotherm of pure gas on a homogeneous patch
of sites characterized by the isosteric heat of adsorption is the normalized
probability density function for the distribution of isosteric heats of adsorption:
The lower and upper limits of in the energy distribution function are generally
assumed to be zero and infinity for mathematical convenience.
Equations (48) and (49) have been used (1) to calculate the pure gas isosteric
heat probability distribution function for a heterogeneous adsorbent from

the experimental overall isotherm assuming a local homogeneous
adsorption isotherm model as the equation, and (2) to develop
analytical overall adsorption isotherm equations by assuming a functionality for
in conjunction with a local homogeneous adsorption isotherm model
of the key works in these areas are described in Ref. 23. Unfortunately, as
mentioned earlier, the estimated or assumed functions cannot be experi-
mentally verified.
of
Statistical thermodynamics of adsorption dictate that the pure gas adsorption iso-
therm of component be linear at the limit of 0:
where is called the Henry's law constant. It is a function of T only.

It follows from (1 1) that in the Henry's law limit (linear isotherm range) is
given by
0
q -
where is the isosteric heat of adsorption of pure gas in the Henry's law region.
is a constant. Equation (52) is derived by integrating Eq. assuming that
is independent of T (generally valid). It follows from (51) that is a constant
in the Henry's law region.
An adsorbent may not be homogeneous even if remains constant as a function of

For example, an adsorbent can be heterogeneous by having a distribution of
adsorption patches with different Henry's law constants but with the same isosteric
heats of adsorption for the sites. Thus the constant of Eq.. (52) may vary from
patch to patch while remains constant. The overall isosteric heat of adsorption of
a pure gas in that case will not vary with even though the local adsorption
isotherms on different patches will be different. This was experimentally demon-
strated by synthetically creating a heterogeneous adsorbent for adsorption of pure
by (1) physically mixing crystals for pure and zeolites and (2) binary
ion-exchanging and ions into an X zeolite structure Both pure and
zeolites were homogeneous adsorbents for pure Both materials had prac-
tically the same isosteric heats of adsorption for but the values were sub-
stantially different for these materials. The end result was that the synthetically
produced mixed zeolite sample was a chemically heterogeneous adsorbent for
which the isosteric heat of adsorption of did not vary with surface coverage.
It is sometimes argued that the observation of a constant isosteric heat of
adsorption of a pure gas over a range of surface coverage is caused by mutual
cancellation of two effects: (1) an increase of a, with increase of due to lateral

interactions and (2) a decrease of with increase of to adsorbent hetero-
geneity Although this is possible, such cancellation is fortuitous. The energy of
adsorbate-solid interactions on practical adsorbents is generally much greater than
that for lateral interactions in the adsorbed phase.
for
Table 3 lists several simple, analytical, adsorption isotherm models and the corre-
sponding isosteric heats for pure gas adsorption. These models have been fairly
successful in describing pure gas adsorption on microporous adsorbents (type I
isotherms
The homogeneous Langmuir (MSL) model assumes that a,
adsorption sites are occupied by each molecule of adsorbate The saturation sur-
face excess for the pure adsorbate i is given by which is not a function of
T. is the fractional coverage for pure component is the
gas-solid interaction parameter for component b, is a function of T only (Table
3), and is a constant. The isosteric heat of adsorption of pure component i
for this model is independent of coverage. The MSL model reduces to the original
homogeneous Langmuir model [4] when = 1 .
The MSL model has been extended to incorporate lateral interactions in the
adsorbed phase In this case, the adsorbent is energetically homogeneous, but
increases linearly with increasing due to lateral interaction. is the pure gas
lateral interaction energy parameter for component This model reduces to the
original homogeneous Fowler-Guggenheim model [7] when a, 1. Furthermore, it
reduces to the Langmuir equation when a, = 1 and = 0.
The Toth model is derived for pure gas adsorption on a heterogeneous adsor-
bent without lateral interactions The parameters and b, are identical to
those for the Langmuir model. The parameter which is a function of T only,
is the heterogeneity parameter. The adsorbent is heterogeneous if 0. The
adsorbent heterogeneity increases as the value of the parameter k decreases. The
model reduces to the homogeneous Langmuir model when = 1. The Toth model
can be derived using (48) and (49) by assuming that the adsorbent is patchwise
homogeneous, consisting of a collection of Langmuir sites with a skewed
like distribution of site isosteric heats of adsorption The for the Toth model
decreases as increase. An inconsistency of the Toth model is, however, that
as 1.
ill.
There are very few published experimental data on multicomponent gas isosteric
heats of adsorption Even those are limited to the study of binary gas
adsorption systems. Figures 8 and 9, respectively, show the isosteric heats of
adsorption of and from their binary mixture on zeolite crystals
measured as functions of loading (surface excess) at approximately constant

8 Isosteric heats of adsorption of from binary with as
functions of surface excesses on zeolite.
0 1 2 3 4 5
Isosteric heats of adsorption from binary mixtures with as functions

of surface excesses on zeolite.

loadings at a temperature of 302 The data were measured using a
type of calorimeter
It can be seen from Fig. 8 that decreases significantly as increases
within the range of the data. It also shows that is larger at any given value of
when the coadsorption of is increased. This result is
intuitive. Figure 8 also plots as functions of for pure gas adsorption. The
from the binary mixture is than that for pure at any given value of
loading. Figure 9 shows that slowly increases as is increased at both
levels of reported in the figure. Again, these results are counterintuitive.
Clearly, the factors determining the isosteric heats of adsorption of gas
are complex. Many more binary experimental data have to be gathered and ana-
lyzed before any acceptable theory to explain the observed behavior emerges. There
is also an urgent need for generating a comprehensive and systematic database for
multicomponent gas 2) isosteric heats of adsorption for aiding practical
adsorptive process designs.
Values of isosteric Heat of

Statistical thermodynamics of adsorption dictates that the adsorption isotherm of
component from a mixture be a linear function of the partial pressure of that
component = in the Henry's law region
= P 0
= 0 ; constant P
The Henry's law constant in Eq. (53) is equal to the Henry's law constant of
pure gas [Eq. and it is a function of T only. The variable in Eq. (54) is a
different Henry's law constant, and it is a function of both P and It follows from
(1 and (54) that
Equation (55) shows that the isosteric heat of adsorption of component of a

mixture is equal to the pure gas isosteric heat of adsorption of component
in the mixture's Henry's law region at the limit of 0. On the other hand, the
isosteric heat of adsorption of component i of a mixture is equal to at the limit of
0 (constant
The three pure gas adsorption models described by Table 3 have been extended to
cover adsorption equilibria. Table 4 gives the multicomponent

gas adsorption isotherm equations and the corresponding expressions for
component isosteric heats of adsorption.
The saturation capacity (surface excess) of component in the
model is given by in,, and the fractional coverage of component is given
by The total fractional coverage of all adsorbates is 9(= The
multicomponent MSL equation is thermodynamically consistent provided that
the constraint = constant is obeyed. The isosteric heat of adsorption of com-
ponent is equal to that for the pure gas.
The multicomponent MSL model with lateral interaction exhibits a linear
increase in the isosteric heat of adsorption of component i with increase of
of all components = 1, 2, . . . , N). is the adsorbate i-adsorbate j
lateral interaction energy parameter in the adsorbed phase.
The multicomponent Toth model is thermodynamically consistent only when
the variables and k for all components are equal The mixed gas adsorp-
tion model is based on the criteria that the local site selectivity of adsorption for
component pairs = is constant for all adsorption sites, but the
solid interaction parameters of component vary from site to site. The functions
and of Tables 3 and 4, respectively, are identical except that the inde-
pendent variables are 9, and 9.
It should be noted here that all models of Tables 3 and 4 were originally devel-
oped by using the actual amount adsorbed as the base variable. We adapted
them to describe the actually measured surface excess variable as functions of
and because they describe the experimental surface excess isotherm data
very well.
The Toth model describes the adsorption isotherms of pure (component 1)

and (component 2) as well as those for their binary mixtures on char-
coal fairly well The multicomponent model was used to calculate the
binary gas isosteric heats of adsorption for a mixture as functions of
gas-phase mole fraction at a total as pressure of and at a
temperature of 293 K Figure 10 show the results, decreases as the
mole fraction of that component in the gas phase increases. increases as the
mole fraction of that component in the gas phase increases. This is a typical char-
acteristic of the Toth model.
Figure 10 also shows the isosteric heats of adsorption of the pure gas (dotted
lines) at the adsorbate loading (surface excess) as that of component i from
the binary mixture at and It can be seen that the binary and pure gas
isosteric heats are comparable for only a t the higher values of There is a
large difference between them at lower values of On the other hand, the differ-
ence between the binary and pure gas isosteric heats are striking for at all
values of The pure gas isosteric heats for component 2 are much larger than
those from the mixture at the same of The reason for these differences is
that the isosteric heat of adsorption of the components of the mixture, according to
the Toth model, depend on the total adsorbate loading 9. For this 0 is closer
to at higher values of due to the high selectivity of adsorption of

FIG. 10 heats of adsorption of (1) and (2) from their binary
mixtures at constant pressure and temperature on charcoal, calculated with
the (Parameters: 1 = 7.802 x
7.602 x = k at 293.1 K
= 0.34; =
compared to Thus the and pure gas heats are comparable only for
component 1 at higher values of . These results show that it may not be correct to
use pure gas isosteric heats of adsorption as functions of its coverage to represent
the isosteric heat of adsorption of the components from a gas mixture at the same
coverage for process design purposes.
1. in Fundamental of Adsorption, Proceedings of Engineering Foundation

Conference held at Sonthofen, Germany (A. B. Mersmann et
Engineering Foundation, p. 8
2. G. Hartzog and Adsorption (1995).
3. T. Hill, (1949).
4. Young and A. D. Physical Adsorption of Gases, Butterworths,
Washington, DC, 1962.
5. Everett. Int. CNRS (1971).
6. Cardona-Martinez and J. A. Adv. (1992).

S. Ross and P. Physical Adsorption, Interscience, New York, 1964.
Gibbs, The Collected Works o f Gibbs, and Green, New
York, 1928.
Chem. Faraday Trans. (1985).
S. Adsorption (1996).
E. M. R. Samuels, and Chemical Engineering
Thermodynamics, Prentice-Hall, Englewood Cliffs, 1972,
Ind. Eng. Chem. Res. (1992).
Langmuir (1991).
Bulow and P. of Adsorption, Proceedings of Engineering
Foundation Conference held at Santa Barbara, CA (A. I. ed.), Engineering
Foundation, 1987, p. 119.
P. A. Hughes, and V. Rees, Gas Separ.
and R. Sep. (1986).
E. Handy, Sharma, B. and J. A. Technol.
(1993).
A. Mariwala. M. Rao, Gorte and A. Myers,
Langmuir, (1996).
A. Dunne, S. R. Gorte and A. Myers,
(1996).
A. Gorte. and A. Myers, Langmuir.
(1997).
Reich, Dissertation, Georgia Institute of Technology, 1974,
and A. Myers. Surf. (1988).
Golden and J. Colloid Interface (1991).
Nitta, M. and Katayama, Chem. Eng.
(1984).
(1995).
Martinez and Chem, (1996).
Jaroniec and J. Colloid Sci. (1976).
H. Kowalczyk, Thermal. Anal. (1995).
Szepesy and Chim. Hung. (1963).

Department of Materials Science and Engineering, The
Pennsylvania State University, University Park, Pennsylvania
I. Introduction
I I . Surface Chemistry of Activated Carbons
III. Adsorption of Inorganic Solutes
A. Adsorption of cations
B. Adsorption of anions
IV. Adsorption of Organic Solutes
A. Role of electrostatic interactions
B. Role of dispersion interactions
C. Toward a comprehensive adsorption model
V, Summary and Conclusions
References
It is well known that the uptakes of gaseous, vapor, or liquid adsorbates by differ-
ent carbon adsorbents can vary over a remarkably wide range. Many of these
variations are due to differences in the physical surface properties, to surface
area effects, as well as to different degrees of molecular sieving because of their
widely varying pore size distributions. However, when such factors are allowed for,
large uptake differences per unit of total (or accessible) surface area still remain
These are especially important in liquid-phase adsorption; they have been well
documented, but they are not well understood. They may also be important in
vapor- and gas-phase applications. This undesirable state of affairs is surprising,
having in mind the fact that of the four technologically relevant adsorbents acti-
vated carbon is by far the most important one, with estimated worldwide sales of
about being an order of magnitude higher than those of its competitors

The results of a very recent study by Attia summarized in Fig. 1 , are a good
illustration. Carbons of varying surface area were produced by carbonization and
activation of date pits in steam, air, or carbon dioxide (at different temperatures
and to different extents of burnoff). The correlations of water vapor and pyridine
uptakes with the total surface area are seen to be very poor. The author concludes
that "water vapour adsorption is related to the chemistry of the surface rather than
to the extent of the surface area" but does not identify, let alone quantify, this
chemistry. Regarding adsorption, the author notes that "activation of
carbons at low temperatures created acidic sites while treatment at high tempera-
tures led to the generation of basic sites on the surface," but she does not identify
these sites.
.1 Correlation (or lack thereof) between the surface area of an activated carbon
and the uptake of (a) water vapor at 298 K and (b) pyridine at 423 K. (
area (from adsorption at 77 K); Dubinin-Radushkevich (DR) surface area (from
adsorption at 298

In this review I first present what is arguably the state of the art in the under-
standing of the surface chemistry of activated carbon This knowledge is a
necessary condition for rationalizing the varying uptakes of both inorganic
and organic and pollutants from the aqueous phase. I illus-
trate this general conclusion by presenting the results obtained in my research
group and by analyzing those of others. Finally, in I rescue a11 appro-
priate quantitative model from undeserved oblivion, and I conclude that a t long
last real progress is being made toward achieving predictive capabilities and tailor-
ing the chemistry of carbon adsorbents for the removal of specific pollutants from
aqueous and other waste fluid streams.
It is not my intention to present an exhaustive review here. A comprehensive and

critical review is provided elsewhere Boehm has also presented a brief but
authoritative review Instead I summarize here the issues that are now known
to be essential for predicting adsorption uptakes, especially in aqueous solutions.
The main features of carbon surface chemistry are presented first, and the conse-
quent acid-base behavior of carbons is briefly discussed to illustrate their unique
amphoteric character. (These same features should be important in the adsorption
of gases and vapors, but a discussion of the relevant evidence, beyond that pre-
sented in conjunction with Fig. 1, is not within the scope of the present work.) In
it is shown that these phenomena govern the adsorption of inorganic
compounds. In IV I argue that they are sometimes dominant in the adsorption
of organic compounds, but more often they are only a part of the "whole story"!
The essential features of the surface chemistry of carbons, relevant to their
behavior in aqueous solution, are summarized in Figs. 2 and 3. Purists may
argue that the two figures are representations that d o not do justice
to the acknowledged complexity and heterogeneity of carbon surfaces
Obviously, the three-dimensional nature of the micropores, the details of their
topology, the proximity of the pore walls, and the distribution of energies of the
various surface functional groups should all be somehow taken into account if
reliable adsorption potentials are to be obtained. However, a judicious selection
of activated carbons for many practical applications water treatment, removal
of volatile organic compounds) requires the resolution of much more basic issues.
Perhaps at the top of the list of such issues are the separability and relative impor-
tance of electrostatic and dispersion adsorbate-adsorbent interactions. I argue here
that Figs. 2 and 3 are necessary and sufficient for addressing this issue.
Figure 2 illustrates the macroscopic chemical features of the activated carbon
surface, both internal and external. In response to pH changes in aqueous solution,
the surface develops coexisting electric charges of opposite sign, whose prevalence
depends on solution chemistry. Therefore, attractive or repuslive electrostatic inter-
actions between the adsorbate and the adsorbent must be taken into consideration.
The development of negative charge is well understood, well documented, and
noncontroversial: It is the result of the dissociation of acidic surface functionalities
such as carboxyl, and phenolic groups. Two controversial points in this

Carbon
surface
FIG. 2 Principal features of the activated carbon surface that give rise to surface
charge development in aqueous solution.
figure deserve special attention: (1) the origin of positive charge and (2) the
relative contributions of functional groups vs. graphene layers to the development
of surface charge. In Fig. 3 one such graphene layer is depicted, albeit in two
dimensions only. It consists of saturated basal plane sites, (2) unsaturated edge
sites, (3) edge sites saturated with heteroatom-containing functional groups, and (4)
electrons delocalized over the entire graphene layer. The small size and poor
alignment of the crystallites make them highly porous and reactive. The high
osity, especially in the micropore range (with equivalent diameters of
exposes large surfaces to gaseous and liquid environments. The crystallite faces
are composed of fused benzene rings (graphene layers). The functional groups at
the crystallite edges are ubiquitous because of carbon's great affinity for oxygen;
they possess a wide range of thermal and chemical stability. Most of them are acidic
carboxyl and groups), but some may bc basic carbonyl plus ether,
or groups). Some edges are terminated by thermally stable groups,
while others are much more reactive because are terminated either with free
radicals unpaired electrons) or with somewhat stabilized electron pairs.
In contrast to more ordered carbons heat-treated carbon blacks), where the
overwhelming majority of the surface consists of basal planes, in activated carbons
the ratio of edge to basal plane carbon atoms can be as high as 10-20%.
Three possible sources of the positive charge are highlighted in Fig. 2: basic
oxygen-containing functional groups pyrones or chromenes), protonated
amino groups (for nitrogen-rich carbons), and graphene layers acting as Lewis
bases and forming electron donor-acceptor (EDA) complexes with molecules.
When nitrogen-containing functional groups are not abundant, which is the case in

. 3 Schematic representation of a graphene layer in a carbon crystallite including
the various oxygen-containing functional groups (carboxyl, phenolic, carbonyl,
pyrone) at the edges of graphene layers. Many such layers are stacked more or less
parallel to each other to form carbon crystallites connected by a network of slit-shaped
pores. The structure is representative of an activated carbon with a crystallite width of
15A and an elemental analysis (by weight) of 87.5% 11.3% 0, 1.2%
Unpaired sigma electron; an "in-plane sigma pair," where is a localized
electron.
most practical situations (unless nitrogen is added deliberately, by treatment

with ammonia), the relative contributions of basic oxygen and graphene layers to
the positive charge are the key issue. Such information is not readily available in the
literature, even though it has important practical implications (Just based on sta-
tistical arguments): In the absence of shielding effects, electrostatic repulsion of
adsorbates from the graphene layers (where most of the surface area resides) is
thought to be much more detrimental for adsorbent effectiveness than repulsion
from the basic oxygen functional groups.
Electrophoretic studies on activated carbon and carbon black particles, and
their comparison with the results for amphoteric oxides, contain clues about the
contributions of functional groups and graphene layers to the development of
surface charge. Figure 4 illustrates a typical set of results. Electrophoretic mobilities
of particles are measured as a function of pH and are seen to be an order of
magnitude smaller than those of hydrogen ions under comparable conditions.
They are commonly converted to either zeta potentials or surface charges using
standard electrical double layer (EDL) equations. In theory, each graph represents
"a titration of surface charged groups and as such is strictly comparable with a

Furnace black Alumina
Electrophoretic mobility as a function of pH for (a) a-alumina and a

furnace carbon black and (b) an activated carbon
potentiometric titration" The so-called H carbons have a high isoelectric

point 7), while the so-called carbons have a low isoelectric
point 7). A common observation in the surface charge characteriza-
tion of many carbon materials is that the absolute value of the positive charge is
often much smaller than that of the negative charge This has been obvious
for more than half a century but its implications were not discussed until
recent years; for example. from the data of Steenberg, as reproduced by
and the uptake of on an carbon (sugar-derived char activated
at was 700 while the uptake of on an H-type carbon (same
char activated at was In contrast, for carbon blacks the
situation can be reversed or the electrophoretc mobility vs. pH diagrams

tend to be much more "symmetric" More recently, for example, Dobrowolski
et reported the charges for a "basic" carbon (outgassed in argon at
and an "acidic" carbon (oxidized in of (at pH = 2.4) and
(at pH = respectively. To put these numbers in perspective, an
activated carbon (- that is heavily oxided with may con-
tain as much as 10 oxygen; if all this oxygen is in the form of carboxyl groups,
this is equivalent to 3 mmol of C O O per gram, or 1.5 V).
The comparison with amphoteric oxides is also instructive. In an early
review, Snoeyink and Weber compared the surface functional groups on car-
bons and silicas but did not point out the resulting differences in the "symmetries"
of their electrokinetic behavior. For amphoteric oxides containing ROH surface
groups, the symmetry (see Fig. 4a) is a consequence of the equilibrium
+ + + OH'
and at pH = =
Taken together, these results strongly suggest that (1) there is an important
contribution of the graphene layers to the positive surface charge of carbon
bents and (2) the widely held belief (especially reflected in the literature reviewed in
that surface charge develops as a consequence of both deprotonation and
protonation of the same oxygen functional groups should be discarded.
Protonation of aromatic carboxyl groups in aqueous solution is highly unlikely.
[Thus, for example, only acids whose strength exceeds - 7, HI
= can protonate aromatic carboxyl groups, while is, of
course, a much stronger acid than and is the
dominant species in water.] Instead, protonation of the basal plane is consid-
ered to be much more important for the applications of interest here.
The use of titrations is commonly considered to be an alternative to electro-
phoresis for the estimation of surface charge. Actually, in the case of solid carbons,
and especially in the case of high surface area activated carbons, this turns out to be
a complementary method and failure to make the distinction can be mislead-
ing. Electrophoretic mobility is considered to be a response to primarily external
surface charge on the particles, while titrations probe both the external and internal
surfaces of particles. For convenience I shall refer to the titration-derived para-
meter (pHpzc) as the point of zero charge, while the electrokinetically derived
parameter will be referred to as the isoelectric point. I acknowledge that
a more fundamental distinction between these two parameters requires a dis-
cussion of specific adsorption (in an ideal situation, the two should be identical in
the absence of specific adsorption), but such a discussion is typically not included in
adsorption studies, especially of organic compounds. In most cases of practical
interest, because of preferential (diffusion-controlled)
air-induced oxidation carboxyl group generation) on the external surfaces of
carbon particles. Hence, the smaller the difference between the two values, the more
homogeneous the distribution of surface charges. This is illustrated in Table 1. A
dramatic example of this difference between and has been shown
(albeit implicitly) by Dixon et for the case of a commercial activated carbon,
10.2 vs. 3.5.

Electrochemical Properties of As-Received and Chemically Modified
Activated Carbons
Carbon sample"
Nc
ox
16h
"A = treatment in air at 250-C for 3 N treatment in at for 3 N A same as N but

subsequently oxidized in air at 250Â° for 7 min; = treatment in at 950Â° for 3 h;
room temperature exposure of sample H to ambient air for days; = treatment of
sample H with concentrated = c) or dilute (X = d) for
Source: Ref.
A compilation of and values for several commercial activated

carbons shows that the values for adsorbents commonly used in water treat-
ment can vary by several pH units. This reinforces the point that was clearly
illustrated by Lau et Because of the large affinity of carbon for oxygen,
the storage and thermal history of the adsorbent often have a strong influence on its
chemical surface properties.
Figure 3 is an attempt to synthesize and highlight the key molecular features of
the edges and basal planes of graphene layers. Only some of the most prevalent
surface functional groups are depicted Special attention should be given to the
delocalized TI-electron system that acts as a Lewis base in aqueous solution
+ + OH-
In addition to giving rise to a positive surface charge, the graphene layers parti-
cipate in TI-TI interactions with aromatic adsorbates. These are known to
be sensitive to substituent effects on the aromatic ring The substituent
effects on the carbon adsorbents are examined in detail in and
Suffice it to note here that there are at least three ways in which the
density of a graphene layer can decrease (1) deliberate of
the edges with oxygen functional groups, by controlled oxidative treatments; (2)
accidental (and more or less inevitable) decoration of the edges with oxygen
functional groups upon adsorbent exposure to ambient air; and (3) additional
localization of electrons at the edges by forming "in-plane pairs" with the

unpaired electrons (remnants of high temperature treatment of the material). The
degree to which these processes take place can be fine-tuned by carbon pretreat-
ment in inert or reactive gases, as illustrated in Fig. 5. The dramatic decrease in
both the uptake (at room temperature) and the heat of adsorption upon
0 50 100 150 200 250 300

Coverage carbon)
0 50 100 150 200

Pressure
.5 Effects of treatments in at etc.) and at

etc.) on the surface chemistry of a commercial activated carbon C extra).
(a) Differential heats of adsorption of at (b) adsorption isotherms of at 25
and (aw acid-washed).

treatment at (Fig. 5a) is not a consequence of the creation of stable
groups at the edges of the graphene layers; rather, it is achieved by selective
removal (hydrogasification) of the most carbon atoms. The resulting
structure is less upon subsequent ambient air exposure even though
it contains many unsaturated sites. Figure confirms this interpretation;
adsorption of on these less carbon sites does occur at the higher
temperatures.
Interaction of systems with water, as well as with cations, has been a topic of
considerable fundamental and practical interest. Even though solid
carbons are generally considered to have a hydrophobic character (which is
certainly true in comparison with adsorbents such as most zeolites and clays),
they can be very effective in removing both organic and inorganic acids and
bases, especially from aqueous solutions A critical review of the
water interactions has long been overdue, but space constraints do not allow me
to offer it here. (The reader is referred to a recent review by Vartapetyan and
Voloshchuk in Chem. Rev. 64, 985, 1995). Suffice it to say that progressive
incorporation of acidic oxygen functional groups can transform a very hydro-
phobic carbon into a very hydrophilic one while treatment at an opti-
mum temperature in can result in both hydrophobic (basic) and stable carbon
surfaces The latter point is illustrated in Fig. 6: A relatively high
is maintained for this carbon over a period of several months, even though
some external surface oxidation does occur.
0 20 40 60 80 100 120
Time of atmospheric exposure (days)
Effect of atmospheric exposure on the isoelectric point (determined by

electrophoresis) and point of zero charge (determined by mass titration) for an
activated carbon treated in at

The adsorption of inorganic solutes is a topic of great practical interest because of
water treatment and metal recovery applications. In particular, the uses of activated
carbons in removal and gold recovery have been of interest for
decades. The fundamental aspects of the process are also of interest in the prepara-
tion of carbon-supported catalysts A critical and comprehensive review is
offered elsewhere Here I briefly discuss the removal of selected metals. In
particular, I explore the "symmetry", or lack thereof, in the electrostatic attraction
between a positively charged surface and adsorbate anions at pH and
between a negatively charged surface and adsorbate cations at pH If
such symmetry existed, then the uptake of, say, species would be equally
maximized when cations are contacted with a negatively charged car-
bon as when are contacted with a positively charged surface. Early
(albeit indirect) evidence against the existence of such symmetry was provided by
et (see also Radovic and Rodriguez-Reinsoso in a
study of catalytic activity of carbon-supported palladium: The catalyst was much
more active, and presumably adsorption of Pd occurred over a much greater sur-
face (thus minimizing catalyst sintering), when anions were contacted with
a positively charged surface (at pH 0.1).
The key issue from both a practical and a fundamental viewpoint is whether or
not the entire adsorbent surface is available for adsorption. (Multilayer adsorption
or volume filling of micropores is not considered here, nor is the possibility of
molecular sieving.) Its resolution depends on the answers to two questions:
1. Is the surface electrostatically accessible, has electrostatic repulsion been
eliminated?
2. Is the affinity of the uncharged surface for the (usually charged) adsorbate
species sufficiently strong to attract them from solution through van der
dispersion) interactions?
These factors are considered in more detail elsewhere Here I analyze only one
of the clues that contains the answers: the fraction of the carbon surface covered by
selected and anionic adsorbates.
Table 2, from the study of Ferro-Garcia et provides the relevant informa-

tion, which turns out to be typical for adsorbate species (see, Refs.
68-70). The isotherms were obtained at pH 5 because a large increase in uptake
was observed as the pH was raised from 3 to 5. The authors reasoned that the
"lies between these two values," despite the fact that the pH of the slurry
was higher (see Table 2). Therefore, residual electrostatic repulsion of the
and cations by the still positively charged surface (pH
may have counterbalanced the ion exchange that has increasingly taken
place on some of the dissociated carboxyl groups at pH 5 . Indeed, in the earlier
work of and Smith using a chemically similar carbon =
removal efficiency continued to increase well beyond pH 5, and especially so

T 2 Physical and Chemical Surface Properties of Selected Activated Carbons
and Their Metal Adsorption Capacities Expressed as a Fraction of Available Surface
Area
Percent surface occupiedc
Carbona pH
"The activated carbons A, P were prepared from almond shells, olive stones, and peach stones,
respectively. Their aqueous suspensions all had pH 7.0.
surface area, total pore volume, and micropore volume (in pores 7.5 were determined
by adsorption and mercury porosimetry.
'Calculated using Langmuir adsorption capacities obtained at 293K and pH 5.
Source: Ref. 67.
when pH was above (even though some of this increase was attributed by
the authors to hydroxide precipitation). The authors calculated for "a
hydrated radius of approximately 25 A which is larger than the reasonable size of
hydrated cadmium ions."
These results indicate that a much smaller fraction of the adsorbent surface is
typically covered by inorganic adsorbates than by organic (especially aromatic)
adsorbates (see IV). A logical inference is that, in contrast to the mechanism
of adsorption of organic solutes (see IV), the graphene layers are not involved
in the adsorption of inorganic cations from dilute aqueous solution. Instead,
cation-exchange processes or interactions with functional groups at the graphene
layer edges are primarily responsible for the adsorption of metals. The study by
et is interesting in this context. These authors found that "saturation
adsorption [of an activated charcoal cloth] increases slightly as the p H increases
from 3 to 5.5," which is consistent with a decrease in electrostatic repulsion or a
higher degree of ion exchange. They assumed that at the saturation adsorption
capacity at pH 5.5, mercury exists as a species (even though Baes and
Mesmer suggested that neutral is predominant) and thus calculated
a surface coverage of 3.5% for the unhydrated cation; however, when a single
sphere of hydration was assumed to surround it, an unusually high coverage of
90% was obtained. One possible reason for such a large uptake is that this
carbon (whose surface chemistry was not described) had a very high surface func-
tional group concentration; perhaps a more likely explanation is that the metal
concentration used was much higher (by several orders of magnitude) than those
typically encountered in wastewater streams.
Clearly, further fundamental studies are needed to confirm these conclusions.
For example, it will be interesting to find out whether much higher surface cov-
erages can be accomplished on a carbon whose maximum number of
exchangeable) adsorption sites, - is not only created but
also made electrostatically accessible by adjusting the solution chemistry (and in the

Chemistry of 1
process distinguishing clearly between adsorption and precipitation). Under these

conditions, for example, the theoretical (ion-exchange) capacity for a
cation is as much as 50% of the total surface area of a typical activated carbon
Alternatively, it will be important to verify whether the highest uptakes can be

achieved under conditions of electrical neutrality of both the adsorbate species and
the adsorbent surface, for example, at some pH. This has often been
observed for organic adsorbates (see IV), because departure from these con-
ditions has led often to electrostatic repulsion than to electrostatic attraction.
Not as many such reports are available in the literature on the adsorption of
inorganic solutes, or if they are, their discussion does not follow this line of reason-
ing. I offer just one example to illustrate this point. and Cheng
studied the adsorption of Cd-Ni on a commercial activated carbon
= and reported a set of intriguing results: When the metals were
plexed with ethylenediaminetetraacetic acid (EDTA), a maximum in uptake was
observed at intermediate pH; in contrast, when they were complexed with a
that yields a positively charged species over the entire pH range a mono-
tonic increase in uptake was observed.
The adsorption of anions is a much more controversial topic than the adsorption of
species. It is quite intriguing that in many cases the uptake does not appear
to be much higher, as one might expect if the positively charged graphene layers
(see Figs. 2 and 3) were electrostatically accessible and "active" in adsorption.
Indeed, in a review of anion adsorption by Hingston the use of carbonaceous
adsorbents is not even mentioned.
For example, and Hancock cite and discuss a study of
and in which it was "calculated that the surface coverage amounted to less
than one gold atom per 5 n m 2 of available carbon surface." They thus "excluded
basal planes, carboxylic acid groups and basic oxides as so-called 'points of attach-
ment' and proposed that the adsorption sites were probably quinone-type groups"
or "special type of In sharp contrast, Klauber notes that "it is
difficult to convert the pmol loadings into a sensible figure for aurocyanide
coverage" but proceeds to calculate (based on X-ray photoelectron spectroscopic
results) that the "extrapolated loading of 2000 pmol approximately equates to a
0.65 monolayer equivalent"; he thus concludes that "the graphitic planes of the
activated carbons [are] the location onto which the anions adsorb."
Table 3 summarizes the results of Solar et which clearly illustrate the
dramatic improvement in molybdenum adsorption once the electrostatic repulsion
is eliminated. Thus, for example, the surface coverage of the as-received carbon
black increased by more than an order of magnitude at pH despite the
uncertainties regarding the surface area occupied by the adsorbed species and the
complex speciation diagram of these high values could well be inter-
preted to represent a large fraction of the adsorbent surface.

3 Molybdenum Loadings on As-Received and Chemically Modified
Carbon Blacks
Mo adsorbedh
M (Monarch 700, Cabot = 6.2; 8.5 0.06

7.0; 202 2.6 1.39
1.3; 2.6; 246 8.5 0.02
1.8 0.85
M-2500-HN03 3.0; 5.6; 8.5 0.02
13 1 1.9 0.52
= treatment in boiling 6 N nitric acid for 1 2500 = hat treatment at in
a solution of initial pH 2.0; determined by atomic absorption spectroscopy.
Source: Ref. 18.
The need to purify both municipal and industrial wastewaters places increasingly
stringent demands on the performance of activated carbon adsorbents. Reliance on
the physical surface properties high surface area, adequate pore size distribu-
tion) is thought to be particularly insufficient for the cost-effective removal of
aromatic pollutants and the subsequent reactivation of the spent adsorbent. A
thorough discussion of the opportunities that judicious design of surface chemistry
offers in these applications is presented elsewhere together with a comprehen-
sive and critical review of the relevant literature. Here I summarize one of the key
issues: the relative importance of electrostatic versus dispersive interactions in the
adsorption of aromatics,
It is well known that methylene blue and other dyes are used to estimate the
adsorption capacity of carbons and even to characterize their porous structure
This implies that nonspecific dispersion adsorbate-adsorbent interactions are
dominant in these systems and that the entire surface is accessible to the adsorbate.
However, as Fig. 7 illustrates, the adsorption capacity depends very much on the
chemistry of the carbon surface and the pH; unless the uptake by this carbon is
determined at a relatively high pH, the measured surface area can be underesti-
mated by a factor of 2 from 500 to This means that electrostatic
interactions can be very important as well.
Based on extensive experimental evidence regarding the importance of and
surface chemistry there is no question that the arguments discussed in the
previous section in the context of adsorption of inorganic solutes are applicable
also-and indeed are required--for understanding the adsorption of organic
solutes, many of which are weak electrolytes. What does need careful consideration
here are the answers to the following two questions:
1. Which specific features of carbon surface chemistry need to be known, in both
qualitative and quantitative terms, in order to be able to optimize an activated
carbon for a specific pollution control task?

Equilibrium concentration,
FIG. 7 Adsorption isotherms for methylene blue on BPL carbon Carbon

Corp.) under different pH conditions: pH 3.1; pH 6.29; pH 9.0; (A) pH
11.9.
2. What are the relative roles of electrostatic and dispersive interactions in the
various adsorbate-adsorbent systems of interest?
In the preface to their landmark monograph, and Mark wrote,

"Carbon researchers have, for the most part, considered the surface chemistry of
activated carbon to be in such a state of disarray that they want to avoid lengthy
discussions of surface phenomena." More than two decades later, with vast
improvements in our knowledge of the role of surface chemistry in carbon gasifi-
cation and in the use of carbons as catalyst supports an argument is
presented here that this is no longer true.
There are probably no more contentious issues in the use of activated carbons
than the role of surface oxygen functional groups in the aqueous phase adsorption
of weak aromatic electrolytes. In a 1980 discussion in print, Suffet (see Ref. 83)
made the following and posed the key question: "There is some
confusion in the literature regarding the major sites of adsorption [on granular
activated carbon]. Is an oxidized surface adsorbing material on the oxidized func-
tional groups, or is it adsorbing on the nonpolar carbon surface?"
Even though the components of the total interaction potential between such
complex adsorbents as solid carbons and a wide range of adsorbates can be
grouped in many different ways it is convenient to consider only the
London dispersion (induced dipole) forces and the electrostatic (double layer)
forces

Elsewhere my coworkers and I discuss in detail the benefits of establishing con-
necting points, and thus achieving cross-fertilization, between the disciplines of
surface chemistry, electrochemistry and liquid-phase adsorption Here I
briefly review the effects of pH on surface chemistry (or vice versa) and on the
uptake of weak organic electrolytes. (It should be noted, however, that there con-
tinue to appear published reports in which the pH of the adsorption experiment is
not even mentioned, let alone controlled.)
In the pioneering work of et a decrease in uptake
with increasing pH was noted, but it was attributed to increasing solubility rather
than electrostatic repulsion. Vaccaro noted that the pH effects "originate from
a variety of many of which remain inadequately understood" but
chose not to address the relevant issues. Zogorski et al. noted a decrease in
the rate of adsorption of substituted phenols as pH increased. Peel and Benedek
noted that the "influence of pH is complex and potentially could have . . . a
significant effect on the results of studies conducted in the vicinity of the
Fuerstenau and Pradip reported a small but consistent decrease and a small but
consistent increase in the uptake of o-cresol = 10.2) for two coals with
values of 3.6 and 4.6, respectively; they dismissed these trends by stating that "no
effect of pH on adsorption was observed" and that this was "not surprising since
cresol is a neutral molecule and presumably adsorption is occurring through hydro-
phobic interactions of the benzene ring with the coal surface." discussed
the decrease in the pH values of the aqueous suspensions of oxidized activated
carbons and its effect on the uptake of ammonia but did not consider this effect
in the adsorption of aromatic species such as aniline. et reported
that as the pH was increased from 1.9 to 7.7 the uptake of phenol also increased
slightly. Grant and King reported a large suppression of reversible phenol
uptake as the pH increased from 1.8 to 12.1, but they attributed this primarily to
a solubility effect ("phenolate anion has more affinity for the aqueous phase than
does phenol"); electrostatic repulsion was not associated with the charge on the
adsorbent surface but rather to "anions in the surface layer." Vidic al.
reported similar results for 2-chlorophenol uptakes under anoxic conditions but
invoked the unconvincing explanation offered by ad Manes "in
terms of physical adsorption of the undissociated acid," which does not recognize
the importance of carbon surface chemistry, Nakhla also studied the
effect of pH on activated carbon adsorption of phenolics in both oxic and anoxic
systems. Surprisingly, they stated that "little is known about the influence of para-
meters such as pH . . . on the retention capacities'' of activated carbons (under
conditions when surface-promoted reactions such as oxidative coupling
are possible). Their findings confirmed those of Grant and King and Vidic et
and the explanations offered are similar, eluding any reference to the role of
carbon surface chemistry, especially under anoxic conditions. Analysis of pH
effects on the kinetics of adsorption of phenols by the same research group
also ignores the role of surface chemistry. Similar findings and conclusions are
reported by Cooney and Xi with some notable exceptions. They do state
that "ionized forms of species adsorb less effectively to activated carbon than do

their undissociated forms" but do not give an explanation for this fact; and they
discuss the role of electron-donating substituent groups in accelerating oxidative
coupling, an important issue that I address in and Ha et al.
studied the adsorption equilibrium of the protein bovine serum albumin on
surface-modified carbon fibers. They report that "the saturation amount adsorbed
is significantly affected by surface characteristics as well as Maximum uptakes
were typically observed at intermediate pH values "near the isoelectric point of
Curiously, however, these authors that carbon fibers usually possess a
net negative charge over the pH range of their experiments (pH 3-9). They d o
conclude that the '"surface charge attributed [to] the surface functional groups"
or the "dissociability of the surface functional groups" is one of the "controlling
factors for BSA adsorption," but they leave the reader somewhat confused about
how exactly such control is exerted. Longchamp et studied the effect of pH
adsorption behavior of an enzyme) on the surface of a
paste. They also concluded that "maximal efficiency in the adsorption of was
attained at pH 7.7," which is close to the isoelectric point of the "major fraction of
and the "local isoelectric potential distribution of the supporting surface is
also of concern"; but they did not elaborate on this latter point. et al.
state that the "pH of the system could be critical in
determining the amount adsorbed" and that "the affinity for adsorption increases
(hence larger adsorbed) as the acidity of the adsorbate increases," but they
do not provide the explanation for these effects nor d o they direct the reader to the
appropriate references where such explanations are offered.
In these and many other studies, the pH effect was noted, but its critical link
to both surface chemistry and adsorption behavior had not been made. As early as
1937, however, in a paper whose dubious distinction is that it has zero citations in
35 years (according to the Science Citation Index, Frampton and
Gortner studied charcoal surface chemistry using electrokinetic techniques.
They did not report the variations in the point of zero charge as a function of
charcoal pretreatment at at etc.), but they did
measure the electrophoretic mobilities in water and found them all to be negative
[on the order of 3 4 and insensitive to treatment conditions. They
concluded therefore that (1) "electrokinetic data d o not indicate a difference
between 'basic' and 'acidic' charcoal" and (2) "apparently the adsorption of
weak electrolytes from aqueous solution on charcoal is molecular rather than
ionic." In the first edition of Active Carbon Hassles cites this work to affirm
that the "electrokinetic behavior of carbon does not appear to be altered by the
adsorption of weak organic acids or bases"; in the subsequent edition how-
ever, the same statement is made without any references. Also intriguing is the
omission from the second edition of a graph showing the influence of pH on the
electrophoretic mobility and amphoteric character of activated carbons. In fact, the
relevant changes between the two editions appear (in hindsight) to have hindered
our understanding of electrostatic interactions (see pp. 89-92 in the first edition and
pp. 231-233 and 360 in the second), because for a long time (see below), when
surface charge was invoked, if at all, carbonaceous adsorbents were assumed to
possess only a negative charge.

The important work of Muller and coworkers regarding pH effects on
the adsorption of weak organic electrolytes has been ignored by the environmental
electrochemistry community but in the adsorption community it resonated
clearly, albeit slowly (see below). Their model for single-solute systems is discussed
in more detail in et al. note that "certain combinations of
solute solute concentration, and surface charge" are predicted to result in "an
adsorption maximum with hydrogen ion concentration," this being a frequently
reported experimental finding. In particular, such an "adsorption maximum is
found in a pH range where the solute exists mostly as a counterion and where
the effect of on ionization of the solute is stronger than that on the decrease of
the magnitude of the surface charge" Clearly, the emphasis here is on elec-
trostatic adsorbate-adsorbent interactions. The authors do acknowledge that "aro-
matic solutes adsorb onto activated carbon by interaction of the aromatic ring with
the carbon surface" and that "changes in the electron density of the aromatic ring
affect the affinity to adsorb" although their reference for the latter statement
should be (see Ref. 37) the work of Coughlin and coworkers and not the
cited study of et The adjustable parameters in their model were not
subjected, however, to a scrutiny that would reveal the importance of this
electrostatic contribution to the overall adsorption potential (see The
dominant effect of electrostatic interactions was illustrated by showing (1) a dra-
matic decrease in the uptake of benzoic acid = 4.2) as the pH was raised from
4.5 to 11.0 for a carbon whose was 6.7; (2) a gradual increase in the uptake
of aniline = 4.6) by the same carbon as the pH was raised from 2.0 to 10.8;
and (3) a gradual decrease in the uptake ofp-nitrophenol = 7.3) as the pH was
raised from 2.05 to 10.6 for a carbon whose was 5.8.
In the mainstream literature on the adsorption of organics, this appreciation of
the importance of electrostatic interactions and its further development was a very
slow process. It was hampered by uncertainties regarding the key aspects of the
chemistry of the carbon surface and by efforts to find a direct relationship between
specific features of this chemistry and carbon's adsorptive properties. In the pio-
neering work of Snoeyink and Weber it was concluded that "there is strong
indication of the need for continuing research on the definition of the nature of
surface functional groups, and on the particular conditions of preparation which
will produce an active carbon best suited for a specific application." Snoeyink et al.
were somewhat more specific a few years later, after establishing that the
uptakes of both phenol and are suppressed as pH increases; but
they still concluded that the "precise reasons for the capacity differences [of a
coal-based versus a coconut-shell carbon] with respect to the two solutes are not
known." Thirty years later, a better understanding of surface chemistry has
been achieved, but, as argued below, it was not necessary to wait for this develop-
ment in order to be able to optimize the adsorption of organics on activated
carbons. A brief historical development of the key ideas is presented below; it is
discussed in more detail elsewhere
Chapter 7 of the landmark review by and Mark emphasized the pH
dependence of the uptake of acetic acid, n-butylamine, substituted benzoic acids,
anionic benzenesulfonates, and methylene blue. In particular, the suppression of
methylene blue uptake at pH below the isoelectric point of the carbon was

explained by stating that formation of positive surface sites in that region
certainly would not enhance adsorption." The important work of
Getzen and Ward was also cited, in which the concept of the isoelectric
point (or point of zero charge) was not mentioned (presumably because this work
was based on the historically popular concept of hydrolytic adsorption), but which
clearly emphasized the importance of electrostatic adsorbate-adsorbent interac-
tions: "There is a distribution of residual forces on the surface of the charcoal
(which be varied by changes in pH) such that they may exert an attraction
for both electropositive and electronegative ions as well as for the uncharged
molecules.. . . They are, in the case of ions, the forces and, in the case
of uncharged molecules, van der and similar forces."
Even though Getzen and Ward state that the "existence of such forces is
recognized today and their description is generally understood," the subsequent
three decades of published work in this area have shown too many examples to the
contrary. A series of papers by Urano and coworkers - in none of which
is the work of et mentioned is a good example. In an early
paper on the adsorption of strong acids and bases the authors provided a
very useful determination of the of a commercial activated carbon.
Subsequently, uptakes of a large number of both aliphatic and aromatic adsorbates
by five commercial carbons were reported, but among the properties of the
bents there was no mention of their or of the p H of the solutions
Instead, and surprisingly, emphasis was placed on the adsorbent's physical surface
properties, and the authors concluded that they could "predict the adsorption
capacities of organic compounds in aqueous solutions for any activated carbon
whose pore size distribution is known." Even more surprising is the introduction
of new arguments regarding the mechanisms of adsorption in the later papers
even though many of the same adsorbates are reconsidered. In particular,
the authors emphasize the combination on the carbon sur-
face" (see They do acknowledge the influence of "several electrophilic
groups . . . on the surface of activated carbon" and that "the influence of the
pH on the adsorptions of organic compounds in water has not been made clear"
but then attribute the decrease in uptake of weak anionic electrolytes with
ionization in basic solutions only to the fact that "the apparent solubilities of the
ionized forms increase much more than those of the non-ionized forms" They
emphasize the behavior of phenolic compounds and conclude, rather
unconvincingly, that "the deviation of phenol [from predictions based on
Freundlich isotherm parameters for the nonionized and ionized species] may be
due to the combination of the hydroxide ion with the electrophilic groups on the
surface of activated carbon."
By this time, a brief review from an authoritative industrial source had
identified the reason for the fact that "quite a number of phenomena [including
adsorption of polar substances and electrochemical properties] cannot be explained
at all with the usual carbon characterisations" and had postulated that "the pre-
sence and the configuration of oxygen-containing functional groups are responsible
for the differences hitherto not explained." The author briefly discussed the zeta
potential but concluded, rather that oxidized carbons adsorb a
polar compound such as methanol in preference to toluene and in proportion to

the amount of surface oxygen present. In a more comprehensive treatise acti-
vated carbons, et did not do justice to this issue either; their 17-page
section on the removal of organic solutes from water in the active carbon applica-
tions chapter cites only 13 papers (from the pre-1983 period), none of which illus-
trate the important role of pH in the adsorption of weak organic electrolytes. The
authors do point out "the importance of the surface chemistry of activated carbons
for [the] removal [of organics]," but how exactly and which of the surface
chemistry important was not settled. Even in the more fundamental chapter
on the characterization of active carbons, electrostatic interactions are mentioned
only in passing, as they relate to heat of immersion measurements. In this respect,
monograph is more enlightening; it contains a detailed discussion
of the zeta potential and surface charge, but not in the context of liquid-phase
adsorption phenomena.
The relevant studies by Mazet and coworkers published within a span
of six years, are both illustrative and instructive. In the first of these papers, elec-
trostatic interactions between a powdered activated carbon (PAC) and calcium
cations and zeta potential measurements are discussed, but the emphasis is on
the beneficial effect of in neutralizing the negative charge on the carbon
surface. The last paper focuses much more on the chemical nature of the carbon
surface and its impact on the adsorption of a wide range of organic compounds.
Both positive and negative zeta potentials are reported, with values ranging
from below 3 to above 6 Particularly instructive are the results for phenol and
benzoic acid (see Table 4), both showing monotonically decreasing uptake with
increasing pH in agreement with the reversal of zeta potential from positive to
negative in the same pH interval. The authors attribute these results to the effects
of solubility, adsorbate dissociation (ionization), surface charge, and electrostatic
repulsion and conclude "that the electric charge of PAC is an important parameter
for the adsorption of organic solutes." Table 4 also illustrates the neglected point
Effects of pH on the Uptake of Phenol and Acid for As-Received

and Chemically Modified Activated Carbons
Adsorption capacity
Zeta potential
Adsorbenta Phenol Benzoic acid
Heat-treated
Filtrasorb 400
As-received
Filtrasorb 400
Oxidized
Filtrasorb 400
H e a t treatment in at for 4 h; oxidation in boiling for 1 h
Source: Ref. 127.

that, on a given adsorbent, the uptake of phenol is comparable to that of benzoic
acid, despite the fact that the latter is much less soluble (see the authors
mention the "existence of a polarity due to it electrons (in phenol for example),"
but they interpret this also as an electrostatic rather than a dispersion effect.
The recent studies of King and coworkers have addressed the effects of
pH and surface properties both on the uptake of lactic and succinic acids and on
the adsorption and regeneration of phenol. In many cases, uptake decreased as pH
increased, and all oxidized carbons "exhibited a significant reduction in uptake at
the lower solute concentrations" Table 5 summarizes the relevant results. The
observed uptakes were interpreted as a complex net effect of several poten-
tially conflicting trends: (1) a greater extent of oxidative coupling at high pH; (2) a
lesser extent of oxidative coupling on oxidized surfaces; (3) a suppressed uptake at
high p H "because of an increase in the negative charge on the adsorbent surface
and an increased degree of phenol ionization"; (4) a decrease in available surface
area upon surface oxidation due to preferential adsorption of water; and (5) a
decrease in the adsorbate-adsorbent interaction energy for oxidized surfaces, pre-
sumably as a consequence of suppressed overlap, as originally argued by
Effect of pH and Carbon Surface Chemistry on the

5 Adsorption
a
Parameters for Phenol
Adsorbent
F400
WVB
CG6
= adsorption capacity ("uptake at a solution phase concentration equal to unity"); =

"intensity of adsorption" (low values reflect high uptake at low concentration).
and H6 = heat-treated at and respectively, for 24h in =
washed then heat-treated at (in followed by oxidation in for at
temperature Z (in
Source: Ref. 129.

Coughlin and coworkers (see Ref. 37) and not, as the authors state,
proposed by Mahajan al. Thus, for example, even though the data
shown in Table 5 are not unambiguous, the authors invoke the dominant effect
of oxidative coupling for the observation that the "uptake at pH 9 was either higher
RO, CG6) or remained essentially the same (WVB) as that at pH
They did characterize the surface chemistry of their carbons and did cite the work
of in this context, but they did not determine the surface charge
variations with pH.
The recent studies by Newcombe and coworkers concerning the com-
plex fundamental issue of adsorption of natural organic (NOM) are impor-
tant contributions to our understanding of electrostatic interactions. This is also an
important practical issue in water treatment because NOM adsorption leads to
carbon fouling. They performed measurements of both the point and
the point of zero charge and investigated their relationship with the adsorption and
regeneration behavior of several commercial activated carbons. Table 6 summarizes
the surface charge characteristics of virgin and spent commercial activated carbons.
The significance of the differences between and was discussed earlier.
Of interest here is the decrease in with increasing adsorption of humic and
fulvic acids, as well as the increasing negative charge of the adsorbent. This result
led Morris and Newcornbe to conclude that "the surface charge properties of
the G A C are largely determined by the adsorbed material, rather than the carbon
surface itself." When applying these concepts to the chemical regeneration of G A C
from an operating water treatment plant, Newcombe and Drikas note that in
prior studies "no attempt was made to investigate the surface chemistry of the
interactions taking place at the carbon-water interface. As a result the findings
have often been poorly understood, incorrect conclusions have been drawn and
the possibility of chemical regeneration has not been fully explored." T o remedy
this situation, Newcornbe and coworkers analyzed the surface chemistry
of virgin, spent, and regenerated activated carbon and speculated that the effec-
tiveness of the selected regeneration protocol was due to the creation of envir-
onment of lower pH within the pore structure, enhancing adsorption of organics"
by virtue of "a large decrease in the electrostatic repulsion between the surface and
Electrochemical Properties of a Granular Activated Carbona Before and

After Adsorption of Natural Organic Matter
Surface charge
Virgin GAC 7.3 3.0 -0.0006

Spent GAC (3.5 mg 6.4 2.7 -0.0019 -0.0046
Spent GAC 4.5 2.4 -0.0053 -0.0084
300 Carbon Corp.).
dissolved organic carbon (measure of concentration of NOM)
Source: Ref. 133.

the adsorbing species." In a more incisive study of the same issue, Newcombe
set out to determine "how surface concentration and affect the degree of
ionization, the conformation, and the change in the free energy of dissociation of
the adsorbed material, and how the effects may be explained in terms of the proper-
ties of both the humic material and the carbon surface." Most recently, the same
research group tackled the key issue of the relative importance of electrostatic and
nonelectrostatic effects in the adsorption of NOM by studying the effective-
ness of electrostatic screening (using an added salt) in the case of two chemically
different commercial activated carbons. Those results are discussed in more detail
elsewhere
Nonspecific dispersion interactions between the adsorbent and the adsorbate are
the dominant driving force for the adsorption of gases and vapors. Therefore, in the
absence of molecular sieving effects, the entire carbon surface is "active'" in the
removal of gaseous or vapor-phase pollutants. It was seen in that adsorp-
tion of inorganic solutes takes place predominantly on specific sites on the carbon
surface, by ion exchange, and that very high concentrations of solute or sur-
face functional groups are necessary for achieving high removal efficiencies. Here I
show how the interplay of electrostatic and dispersion interactions governs the
uptake of aromatic pollutants on activated carbons.
My interest in this topic was sparked by intriguing initial results for the adsorp-
tion of benzoate, and fumarate anions on chemically modified (oxidized vs.
nitrided) activated carbons For example, both oxidation (which reduces the
of the carbon) and nitriding (which raises the were found to sup-
press the uptake of benzoic acid; this is contrary to the expectations based on
electrostatic arguments alone. On the basis of these and other findings, my co-
workers and I proposed the following mechanisms of adsorption
1. Adsorption of benzoate (aromatic) anions occurs primarily on the basal plane
of carbon, and the electron-withdrawing effects of nitrogen and carboxyl
functional groups suppress the interaction of the basal planes with the
adsorbate's aromatic rings.
2. Adsorption of aliphatic anions occurs also predominantly on basal plane,
and the same electron-withdrawing effects enhance its interaction with the
carboxyl anions.
More recently, we studied in more detail the relative importance of dispersion and
electrostatic adsorbate-adsorbent interactions as we examined the uptake of
methylene blue and nitrophenol as well as that of benzoic, oxalic, and fumaric
acids In agreement with our earlier findings, we concluded that while electro-
static interactions are important, dispersion interactions appear to be dominant in
the adsorption of aromatic solutes. On the other hand, electrostatic repulsion
appears to be much more important for the adsorption of aliphatic anions.
Further validation of these conclusions was achieved by contrasting the
behavior of three chemically different carbons in adsorbing two vastly different
aromatic solutes; nitrobenzene is a very weak Lewis acid that possesses the

Iron-withdrawing group. whereas aniline is a predominantly species
at pH 4.6 that also possesses the electron-donating group. The chemical
differences among the carbons were produced by incorporating 0- or N-containing
functional groups on the carbon surface. While the effects of the former on the
adsorption of phenol and other aromatics are well documented (though not neces-
sarily well understood) the effects of the latter, as well as the interplay of (1)
surface groups, (2) solution pH, and (3) effective charge on the carbon surface, had
not received adequate attention in the literature.
Table 7 summarizes some of the physical and chemical characteristics of the
samples used. Only samples exhibiting minor changes in physical surface properties
(which can be tailored to meet widely varying regarding surface area and
pore size distribution) and major changes in chemical surface properties were used
in subsequent adsorption tests.
Figure 8 shows the behavior of as-received and modified carbons in adsorbing
aniline under widely varying solution chemistry conditions. A significant effect of
solution chemistry (pH) on the uptake is observed. The effect of surface chemistry
depends in turn on pH. At p H 2 (pH when the surface charge is pre-
dominantly positive and cations are dominant in solution, carbon oxida-
tion is seen to enhance aniline adsorption whereas nitriding is seen to suppress it
(with respect to the as-received and carbons). This is exactly what one
would expect if electrostatic adsorbate-adsorbent interactions were more important
than the dispersive interactions. In contrast to oxidation, which does not lead to the
formation of positively charged sites on the carbon surface and shifts the to
a value that is close to solution pH, nitriding obviously enhances the
adsorbent repulsion. Note that this behavior of the cation-forming aniline is in
contrast to that of the anion-forming benzoic acid as the adsorbate, for which
the dispersive interactions were found to be dominant even in highly basic aqueous
solutions (pH- 10) (This again confirms the lack of "symmetry" in the
7 Characteristics of Carbons Used in Figs. 8 and 9

b
C H N
Sample
GCW Americas),
"BET surface area.
'Micropore volume obtained from the equation applied to the
adsorption isotherm.
treatment in inert atmosphere at 1223
'Treatment in at 363 K for 6
with at 873 K for 3 h.
Source: Ref. 138.

Chemistry of Activate
Reduced concentration
0.0
0.000 0.001 0.002 0.003 0.004 0.005
Aniline adsorption isotherms at pH 2 (a) and pH 1 1 (b) for as-received and

odified carbons. As-received; nitrided:
adsorbate- adsorbent electrostatic attraction.) A t the other extreme ( p H 1 1 ,

pH when the surface charge is predominantly negative and aniline mole-
cules are dominant in solution, both oxidation and nitriding are detrimental. This
suggests that dispersion forces control the adsorption process. In agreement with
the behavior of benzoate anions TI-TI interactions are not favored by the
decrease in TI-electron density o n the graphene layers caused by localiza-
tion and/or withdrawal due to the presence of oxygen-containing and some nitro-
gen-containing functional groups. A t p H - aniline uptake is relatively

high on all adsorbents. This is consistent with the minimized electrostatic repulsion
on the (overall) neutral surface. Closer inspection of the data also reveals
the following trends as a function of pH: a monotonic increase for as-received,
devolatilized, and nitrided carbons and a maximum (at pH-pHpzc) for oxi-
dized carbon. These results are discussed later in conjunction with those shown
in Figs. 11 and 12.
Figure 9 shows the behavior of the same carbons in adsorbing nitrobenzene. In
contrast to the findings for aniline, solution chemistry (pH) is seen to have little
Nitrobenzene adsorption isotherms at pH 2 (a) and pH 11 for as-received

GCW carbons. As-received; nitrided;

effect on the equilibrium uptake. This is consistent with the fact that is
a very weak Lewis acid and molecular species are dominant in the entire pH range.
The effect of surface chemistry is obviously much more important. The as-received
and carbons exhibit the highest uptake, while adsorption is suppressed
for both nitrided and oxidized carbons. This behavior is consistent with that of
aniline and benzoic acid at high pH (pH Dispersion forces are dominant
once again; reduced adsorption is the consequence of withdrawal and/or localiza-
tion of electrons decrease in density on the graphene layers)
brought about by carbon oxidation and nitriding.
The importance of dispersive interactions is also apparent when the results
shown in Fig. 8 are compared with those shown in Fig. 9. For example, aniline
is much more soluble in water than nitrobenzene (35 vs. at yet its
uptake does not reflect this. A similar result was reported recently by and
Okazaki In agreement with the foregoing arguments, this result can be
attributed to the beneficial effect of the electron-donating group. Enhanced
adsorption of aniline is due to the resulting increase in the negative charge density
on the graphene layers of the adsorbent.
There are several possible theoretical approaches to the formulation of a suitable

model for the adsorption processes discussed above. The most fundamental one is
based on the perturbation molecular orbital (PMO) theory of chemical reactivity
in which the wave functions of the products are approximated by using
the wave functions of the reactants. A key issue in the use of Klopman's
theory is the relative importance of the two terms in the expression for the total
energy change of the system, which is taken to be a good index of reactivity,
The question is whether the extent of adsorption is determined by charge control

or electrostatic interactions) or by orbital control
attractive interactions between the filled orbitals of the adsorbate and the
empty orbitals of the adsorbent).
In a recent interesting study, and coworkers implicitly took
the approach that orbital control is dominant. They analyzed the adsorption and
desorption characteristics of a series of aromatic compounds at 308 K and uncon-
trolled pH. (They assume that "since the concentration of aromatic compounds is
very low, pH seems to be almost equal to the value of distilled water" They
do not mention, however, whether the presence of carbon modifies the pH of the
suspension, as it often does.) They used the semiempirical method to
determine the HOMO energy levels for these adsorbates and the LUMO levels
for several adsorbents (including chemically modified activated carbons). The
range of calculated values for adsorbates ranging from p-nitrobenzoic
acid to aniline on a commercial activated carbon was Low values
coincided with those systems for which irreversible adsorption was
found; those adsorbates that possess electron-donating groups aniline),

with the highest HOMO levels (or lowest ionization potentials), could not be
orbed easily (using distilled water at 308 K). In contrast, those aromatic species that
possess electron-withdrawing groups nitrobenzene) were more weakly (rever-
sibly) adsorbed. These findings are in remarkable agreement with the conclusion
that when dispersion forces are dominant, the electron-withdrawing
effects of nitrogen and carboxyl functional groups suppress the interaction of the
basal planes with the adsorbate's aromatic rings. Based on these findings, et
proposed a two-state model to explain the appearance of irreversible
adsorption of electron-donating compounds, according to which the barrier for
going from the precursor (reversible) adsorption state to the irreversible state is
higher for the electron-withdrawing aromatic compounds than for the
donating compounds. They do not provide a justification for this assumption;
instead, they invoke the Hammond postulate according to which "the structure
of the transition state will resemble the product more closely than the reactant for
endothermic processes whereas the opposite is true for exothermic processes"
How exactly this is related to the exothermic adsorption process is not clarified. A
more straightforward alternative theory is offered below.
Functionalization of either the adsorbate or the adsorbent that increases the n-
electron density leads to either enhanced or stronger adsorption when the adsorp-
tion process is governed by (dispersion) interactions. The converse is also
supported by available experimental evidence: Functionalization (of the carbon
adsorbent or the aromatic adsorbate) that decreases the density leads
to suppressed or weaker adsorption.
As shown in and however, some adsorption systems involving
aromatic adsorbates are very much influenced by electrostatic interactions.
Additional examples are cited and discussed elsewhere Clearly, a model is
needed that takes both electrostatic and dispersion interactions into account.
Such a model has been presented by and coworkers My co-
workers and I used this model to illustrate the possibly dramatic
effects of modifications of carbon surface chemistry on equilibrium uptake of
p-nitrophenol. We also extended it further to evaluate the relative importance of
electrostatic and dispersive interactions
The wide range of surface chemical properties of activated carbon adsorbents is
illustrated schematically in Fig. 10. (The intriguing lack of "symmetry" between the
low pH and high pH regions, observed for the vast majority of activated carbons
has been mentioned earlier and will be addressed in more detail elsewhere.) The
"acidic" (Cl) and carbons, whose values are 3 and 10, respec-
tively, can be easily designed to develop an electric charge on the order of 0.03
m 2 , which corresponds typically to a surface potential of At the other
extreme, high temperature treatment and stabilization in of an "acidic" carbon
or low-temperature treatment in the presence of hydrogen atoms pro-
duces a "basic" carbon whose remains high and constant over extended
periods of exposure to room temperature air, as illustrated in Figs. 5 and 6.
The key equation describing competitive Langmuirian adsorption on such
amphoteric carbons in terms of fractional surface coverage of molecules (M)
and ions or of a partially dissociated organic solute on a homogeneous
surface is

10 Surface potential vs. pH for three (hypothetical) activated carbons:
carbon 3.0): C2, typical as-received (amphoteric) carbon
- 6.5); "basic" carbon = = 10.0).
In the above isotherm equation, K is the equilibrium constant for the molecules
(due to London dispersion interactions), while K' is the equilibrium constant for
the ions and has the form
exp exp
In the above expression, which consists of two driving forces for adsorption, is
the valence of the charged species, F is the Faraday constant, is the interfacial
electrical potential due to the development of surface charge, and is the potential
arising from the dispersive adsorbent-adsorbate interactions. Implicit in this
expression is the realization that the co-ionic solute species an
cation interacting with the surfaces a t pH can adsorb only if the (attrac-
tive) adsorption potential is greater than the repelling electrostatic potential
Explicit exclusion of co-ionic species from the surface is included in the
model for a heterogeneous adsorbent surface, which conforms to the Freundlich
isotherm:
with

Here again it is assumed that the electrostatic potential is on the same order of
magnitude as or lower than dispersion potential,
The remarkable potential flexibility of activated carbon adsorbents is readily
apparent from inspection of Figs. 10-12. As is intuitively obvious from simple
electrostatic arguments, an "acidic" carbon is most appropriate for adsorbing
the weakly basic aniline = especially at low pH, when a large fraction
of the adsorbent surface is unavailable for adsorption of co-ionic species. These
figures show the results of a parametric sensitivity study of the effects of modified
dispersion potentials, as a consequence of the removal (Fig. 11) or incorporation
(Fig. 12) of electron-withdrawing groups as well as the incorporation of
donating groups (Fig. 11) at the edges of graphene layers of an activated carbon.
Conversion of an "acidic" carbon (Cl) to a "basic" carbon (C3-E) increases the
point of zero charge from 3 to 10, and this is detrimental for the adsorption of
anilinium cations at low pH; but it also enhances the density in the
graphene layers and thus increases the dispersive potential, say, by a factor of 2
(C3-ED x 2 ) to 10 (C3-E-D x 10). It is seen in Fig. 11 that the electrostatic effect is
dominant in suppressing aniline uptake at low pH, but the dispersive effect both
dampens the electrostatic effect at low pH and enhances the uptake at high pH.
. 11 Combined effects of electrostatic and dispersive interactions on changes in

aniline uptake upon treatment in or of an "acidic" carbon and
its conversion to a "basic" carbon. C l , original "acidic" carbon; C3-E. carbon
(only electrostatic interaction adjusted); "basic" carbon (dispersive
attraction potential enhanced by a factor of 2); "basic" carbon
(dispersive attraction potential enhanced by a factor of 5); C3-E-Dx "basic"
carbon (dispersive attraction potential enhanced by a factor of 10).

. 12 Combined effects of electrostatic and dispersive interactions on changes in
aniline uptake upon oxidation. C2, original (as-received) carbon; oxidized carbon
(only electrostatic interaction adjusted); oxidized carbon (dispersive
attraction potential reduced by a factor of 2); oxidized carbon (dispersive
attraction potential reduced by a factor of 5); oxidized carbon (dispersive
attraction potential reduced by a factor of 10).
Conversion of an amphoteric carbon (C2) to an "acidic" carbon (Cl-E)

decreases the from 6.5 to 3.0, and this is beneficial for the adsorption of
cations at low pH; but it also reduces the density in the
graphene layers and thus decreases the dispersive potential, say, by a factor of 2
to a factor of 10 The net result is shown in Fig. 12:
enhanced adsorption at low pH, made less pronounced by the detrimental disper-
sive effect, and suppressed adsorption at high pH.
In contrast to the optimization of physical surface properties of activated carbons

where we know what is needed for a particular adsorption or water treatment
task (an adequate surface area and a specific pore size distribution) and we also
know how to achieve it very few guidelines were available for the optimization
of the chemical surface properties. Much progress has been made in our under-
standing of the surface chemistry of carbons and in developing techniques for
modifying it for catalyst support applications. So, in using carbons for water
treatment we have long known how to modify their surface chemistry, but only
recently have we learned what kind of surface chemistry we need for a specific

task. Not unexpectedly, we have also realized that compromises need to be made
to the sometimes conflicting demands imposed by electrostatic or dispersion
forces.
In the adsorption of inorganic solutes, the main fundamental challenge remains
how to "activate" the entire surface to achieve maximum removal efficiencies. In
the adsorption of organic solutes, the influence of carbon surface chemistry is
decidedly more complex. Both electrostatic and dispersive interactions can influ-
ence or control the equilibrium uptake of a weak aromatic electrolyte.
Typical modifications of carbon surface chemistry affect the extent of both
electrostatic and dispersive interactions. For example, carbon oxidation not only
lowers its point of zero charge; it also reduces the dispersive adsorption potential by
decreasing the density in the graphene layers. And conversely, the
removal of electron-withdrawing oxygen functional groups not only increases the
point of zero charge, it also increases the dispersive adsorption potential by increas-
ing the it-electron density.
A relatively simple theoretical model that accounts for both electrostatic and
dispersive interactions successfully describes the qualitative trends observed in this
study. In the particular case of aniline (a weak electrolyte), both theory and experi-
ments indicate that maximum adsorption uptake is attained on oxidized carbon
surfaces at a solution pH near the adsorbate's Accordingly, aniline adsorp-
tion is deemed to involve two parallel mechanisms: (1) electrostatic interactions
between anilinium cations and negatively charged carbon surface groups and (2)
dispersive interactions between aniline molecules and graphene layers. On the other
hand, for nitrobenzene, maximum adsorption uptake is found on
heat-treated carbon surfaces, particularly at a solution pH Therefore,
adsorption results primarily from dispersive interactions between nitrobenzene
molecules and graphene layers. Dispersive interactions in general are promoted
by conducting the experiments at solution pH values near the adsorbate's
at which the repulsive interactions between charged surface groups and
uncharged molecules are effectively minimized. The theoretical model invoked to
account for all the above trends can in principle be used as a much needed predictor
of the adsorbent and solution characteristics that will act in concert to maximize the
adsorption uptake of aromatic compounds by activated carbons.
It is expected that similar considerations of both electrostatic and dispersive
forces will be necessary for developing more reliable interaction potentials in the
case of the adsorption of gases and vapors on chemically heterogeneous activated
carbon surfaces.
The collaboration of my colleagues, Joshua Ume, Isabel Angel

Menendez, and Jonathan Phillips, as well as stimulating discussions with Carlos
and Jose Rivera-Utrilla (University of Granada, Spain) are grate-
fully acknowledged. Financial support was provided by the U.S. Department of
Energy and the Carbon Research Center at State
University.

van in Fundamentals of Adsorption (A. I . New York,
1986, p. 589.
Humphrey and G . Separation Process Technology,
A , A. Attia, Technol., (1997).
A. Leon y Leon and L. R. Radovic, in Chemistry and Physics o f Carbon, Vol.
24 ( P . A. Thrower, ed.), Marcel Dekker, New York. 1994. 213.
P. Boehm, Carbon (1994).
Jagiello, Bandosz and A. Schwarz, Colloid Interface
Bandosz, Jagiello, and A. Schwarz, (1993).

Jagiello, Bandosz, and A. Schwarz, Carbon (1994).
A. in Surface and Colloid Science (R. Good and
Plenum, New York, 1979. p. 121.
V. A. and Rev. Pure (1957).
P. Huang and F. Ostovic, J. Eng. (1978).
P. and H. Smith, in Chemistry in Water Reuse Cooper,
Ann Arbor Science, Ann Arbor, MI, 1981, p. 355.
Jevtitch and Bhattacharyya, Chem (1983).
P. Huang and Fu, Wat. Contr. Fed. (1984).
Jaroniec, and M. Carbon (1986).
C. P. Huang and Vane, Wat. Fed. (1989).
Abotsi and A. W. Scaroni, Carbon (1990).
Solar, A. Leon y Leon, and Radovic, Carbon
(1990).
V. Lopez-Ramon, Rivera-Utrilla, and R.
Alvarez, Carbon (1993).
A. Menendez, A. Leon y Leon. and Radovic,
Carbon 33: 1655 (1995).
Babic, and V. Carbon
1998, in press.
A. Garcia, A. Cuesta, A. A. Martinez-Alonso, and M.
Tascon, Colloid Interface Sci. (1997).
V . Frampton and A. Gortner. (1937).
Kratohvil and E. Colloids Surf. (1982).
Fuerstenau, Pure Appl. (1970).
P. Pure Chem. (1978).
Schultz, Grimm, and Burckhardt, Solid State (1993).
V . L. Snoeyink and J. Weber, in Progress in Surface and Membrane
Science, Vol. 5 Rosenberg, and D. A. Cadenhead,
Academic, New York, 1972, p. 63.
0. James and A. Parks, in Surface and Colloid Science, Vol. 12 ( E .
Plenum, New York, 1982, p. 119.
A. A. Schwarz, Huang, and S. Catal. (1990).

and A. Faraday
Trans. (1990).
P. Jayaweera, Hettiarachchi, and Ocken, Colloids A (1994).
Lyklema, Fundamentals of Interface and Colloid Science, Solid-Liquid
Academic, London, 1995.
A , Leon y Leon, Solar, V. and R. Radovic, Carbon
(1992).
Cho, and C. Ser. 75 (1978).
Radovic, and Carbon 27
(1998). in preparation.
A. Furlong, and Colloids Surf.
(1986).
and D. Schleifer, Carbon (1984).
Papirer, and Carbon (1987).
A. Hunter and J. M. Sanders, J. Am. (1990).
C . A. Hunter, and J. Thornton, Mol. (1991).
C. A. Hunter, M. and van Koten,
Trans (1995).
Jaffe, A. and A. Am.
118:11217 (1996).
Prog. (1976).
R. Taft and Prog. (1987).

Abe, and Okazaki,
(1995).
A. Phillips, and Radovic. (1996).
A, Menendez, Radovic, Xia, and Phillips,
(1996).
J. A. Menendez, Phillips, and Radovic, (1997).
and Street, J. (1989).
G . Green, Dasgupta, A. Goddard and
A. Blake. Science (1992).
A. Dougherty. Science (1996).
A. A. Menendez, and Suarez, J.
B in press,
and T. Res. (1997).
V. Snoeyink and Jenkins, Chemistry, New York, 1980.
Walker, and Janov, Colloid Interface Sci. (1968).
Kaneko, and Kaneko,
(1995).
E. A. Vega, and Gubbins,
(1996).
R. Farmer, and Kovacic, Am. Fuel
Prepr. (1996).
A. Greinke and I. Bretz, U.S. Patent 5,582,811 to UCAR Carbon
Technology 10, 1996).

Humenick, and Schnoor. J. Environ. (1974).
J. Davidson,
Radovic and Rodriguez-Reinoso, in Chemistry and Physics of Carbon,
Vol. 25 (P. A, Thrower, ed.), Marcel Dekker, New York, 1997, 243.
J. Solar, and R . Radovic. International Carbon
Conference ('Paris), 1990, 672.
A. Ferro-Garcia, J. Rivera-Utrilla, Rodriguez-Gordillo, and
Toledo, Carbon (1988).
Petrov, Budinova, and I. Khavesov, Carbon (1992).
Budinova, Gergova, Petrov, and V.
60:177 (1994).
Polovina, A. Surbek. M . Lausevic. and Serb.
(1995).
A. Sangster, and Wilkinson, Carbon
(1987).
and E. The Hydrolysis of Cations, New York,
1976.
Bhattacharyya and Y. Cheng, Environ. Prog. (1987).
J. in Adsorption of at Solid-Liquid Interfaces A.
Anderson and A. Ann Arbor Science. Ann Arbor, MI, 1981, p.
51.
and Gold Bull. (1981).
(1991).
Graham, (1955).
D. Duff, Ross, and (1988).
(1991).
and F. Cannon. J. Am. Water Works
S. and H. Mark, Activated Carbon: Surface Chemistry and
Adsorption Solution, Marcel Dekker. New York, 1971.
and P. Ehrburger Fundamental Issues in Control Carbon
Reactivity, Kluwer, Dordrecht. The Netherlands, 1991,
Roundtable Discussion, in Activated Carbon Adsorption Organics from the
Aqueous Phase, Vol. 1 ( I . H . Suffet and Ann Arbor
Science, Ann Arbor, MI, 1980, p.
Israelachvili, Intermolecular and Surface Forces, Academic. London, 1992.
Evans and The Colloidal Domain: Where Physics,
Biology and Meet, New York, 1994.
S. and K . D. Papadopoulos, (1995).
R. Deshiikan and Papadopoulos, Sci. (1995).
Mark, Weber, and
Crittenden, Colloid Interface (1969).
Vaccaro, Environ. (1971).
Zogorski, Faust, and J. Colloid Interface
(1976).
G . Peel and A. Benedek, Environ. Sci. (1980).
Fuerstenau and Pradip, Colloids Surf. (1982).

B. in Treatment of Water by Granular Activated Carbon J.
and I. Suffet, American Chemical Society, Washington, DC, 1983, 77.
and Fedkiw, (1985).
Grant and King, Eng. Res. (1990).
R. Vidic, Suidan, and Brenner, Environ. Technol.
(1993).
Rosene and Manes, (1977).
Nakhla, and Farooq, Water Environ. Res. (1994).
Abuzaid and F. Nakhla, Hazard. Mater. (1996).
D. 0. Cooney and Xi, (1994).
Ha, Hinago, A. Sakoda, and in Fundamentals of Adsorption
ed.), Tokyo, 1993, p.
S. Longchamp. Randriamahazaka, and Colloid Interface
994).
Dargaville, N. Guerzoni, Looney, and Solomon, Colloid
Interface 182: (1996).
Hassles. Active Carbon, Chemical Publishing Co.. Brooklyn. NY, 1951.
Hassles, Activated Carbon, Chemical Publishing Co., New York, 1963.
Muller, Radke, and Prausnitz, J. Chem. (1980).
Muller, Radke, and Prausnitz, Colloid Interface
(1985).
Muller, Radke, and Prausnitz, Colloid Interface
(1985).
W. Coughlin and S. Ezra, Environ. Technol. (1968).
Coughlin, F. Ezra. and Tan. (1968).
Coughlin and Tan, in Water-1968, New York, 1968, 207.
V. Snoeyink, and Weber, Environ. Technol. (1967).
V. Snoeyink, Weber, and H. Mark,
(1969).
Getzen and M. Ward, J. Colloid Interface (1969).
Ward and Getzen, Environ. Sci. 4-64 (1970).
Urano, Sonai, and Y. Kobayashi,
1976:1773.
Urano, Koichi, and Nakazawa, Colloid Interface (1981).
Urano, Koichi, and Yamamoto, Colloid Interface (1982).
Urano and Kano, Bull. (1984).
Urano, and Tabata, Bull. Chem. (1984).
R. C. and Stoeckli, Active Carbon, Marcel Dekker, New
York, 1988.
Carbon: Electrochemical and Properties,
Interscience, New York, 1988.
Mazet, A. and Lafrance, Water Res.
P. Lafrance and Mazet, Am. Water Works (1989).
Allali-Hassani, 0. Dusart, and Mazet, Water Res. (1990).
0. Dusart, and Mazet, (1991).
Mazet, and Water Res. (1994).
L. A. Tung, and King, Eng. Res. (1994).

Kilduff and King, (1997).
0. and Walker, Sep. Sci.
(1980).
and Drikas, Water Res. (1993).
Newcombe, Hayes, and Drikas, Colloids (1993).
Morris and G. Newcombe, Colloid Interface
Newcombe, Colloid (1994).
and Drikas, Carbon (1997).
J. I. Ume, A. Scaroni, and R. Radovic, Biennial Carbon Conference,
Buffalo, NY, 1993, p. 468.
L. R. Radovic, I. Ume, and A. W. Scaroni, in Fundamentals of Adsorption
Le Van, Kluwer Academic, MA, 1996, p. 749.
Radovic, I. Silva, J. I. Ume, A. Menendez, A. Leon y Leon, and A.
W. Scaroni, Carbon (1997).
and Okazaki, Colloid Interface (1996).
Am. (1968).
B. Jensen, The Lewis Acid-Base Concepts, Wiley-Interscience. New York,
1980.
H. Atsushi, and Okazaki, Colloid (1996).
Jean and Volatron, An Introduction to Molecular Orbitals, Oxford
Press, New York, 1993.
Atsushi, and Okazaki, in Fundamentals of Adsorption
Le Van, Academic, MA. 1996, p. 961.
I. Silva, I. A. Scaroni, and R . Radovic, Prep. Fuel
(1996).
A. Yopps and D. Fuerstenau, Colloid Sci. (1964).

Institute for Surface
Chemistry, Stockholm, Sweden
I. Introduction
I I . Characterization of Porous Structure
A. Nitrogen adsorption
B. Mercury porosimetry
C. Limitations of the cylindrical pore model
D . Determination of surface charge density and surface potential
III. Thermodynamic Background of Charge Regulation
IV. Surface ionization
A. Protolytic properties of partly oxidized silicon nitride: Heuristic
ideas
B. Distribution of surface potential and surface charge
over pore sizes
C. Application of charge regulation model for determination of
equilibrium constants of surface reactions
References
When two different phases containing some carriers of electricity, such as electrons
or ions, are brought into contact with each other, charge separation occurs, leading
to the formation of an electrical double layer This is a rather general contact
phenomenon that occurs in a variety of systems, not necessarily involving a solid.
Known examples are emulsions containing an ionogenic surface-active component,
or solution systems. Indeed,
the origin of the charge and the mechanism of charge transfer may be different,

but the driving force is always the same: a difference in the energy levels of the
charge carriers in the different phases.
This chapter is concerned with the mechanisms of formation of the electrical
double layer at a dielectric solution interface. When such a contact
occurs, the solid's surface acquires a certain charge due to the dissociation of sur-
face ionizable groups and adsorption of ions from solution. Since the whole system
is to be the solution has to bear an equal charge of opposite
sign. This charge is effectively confined to a thin layer near the dividing surface,
termed the diffuse double layer. Its thickness is characterized by the reciprocal
length,
where is the number density of uni-univalent background electrolyte ions, and

other symbols have their conventional meaning.
For aqueous solutions, the diffuse double layer will have a decay length varying
from a few up to a few thousand angstroms, depending on the concentration of the
background electrolyte. However, taking into account that the the-
ory fails for electrolyte concentrations higher than about 0.01 M and that it is
difficult to maintain a constant ionic strength below about 0.0001 for
based colloidal systems-mainly because of some solubility of solids and
of soluble species-a much narrower range, about 30 up to about
300 A, appears to be of practical importance.
When the surface charge is formed, its sign and magnitude are governed by
thermodynamic factors such as the affinity of adsorbed ions for the surface and
their activity in the bulk solution and conservation laws, including the charge and
mass balance. making the whole process self-regulated. This is, basically, where the
term "charge regulation" comes from. Many facets of this phenomenon have
already been studied in detail and summarized in several reviews Other
aspects have remained untouched until recently. In particular, this concerns charge
regulation at the surface of porous solids.
When talking about a "porous solid", it is usually meant that the pore system has a
fixed geometry that is not affected by the surrounding environment. Although more
alleged than proven and not thermodynamic at all, this assumption of absolute
rigidity makes it possible to bring together results of porosimetric measurements
normally performed with a dry solid with potentiometric and electrokinetic data
that bring information about the surface charge and potential of the solid in con-
tact with an aqueous electrolytic solution.
Compared to nonporous solids, charge regulation at the surface of porous solids
reveals some new features if the mean pore size is (1) large enough to justify
consideration of the surface charge, which is discrete by its very nature, as con-
tinuously smeared over the surface, but (2) small enough to ensure sufficient

lapping of diffuse double layers that extend from opposite walls of a pore. These
are met whenever the mean pore size happens to be much greater than
the mean distance between surface ions, being at the same time comparable in
magnitude with the screening radius, dilute aqueous electrolytes,
it is the mesoporous region of pore sizes (20-200A) that is of interest with respect
to the charge regulation problem. Consequently, the following discussion is, for the
most part, limited to systems of this type.
Since an analysis of the charge regulation problem is intimately related to certain
parameters used to characterize the porous structure of solids, it is useful to start
with an elementary description, outlining the most common experimental routines
used for characterization of porous solids and elucidating the involved model
assumptions and limitations. For more details on the subject, the reader is
referred to special monographs, Refs.
Gas adsorption is by far the most common technique used to gain information
about the porous structure from the isotherms. As adsor-
bent, nitrogen is the most commonly used. The treatment of results here is based on
a cylindrical pore model.
The adsorption isotherm is measured at 77 gradually increasing the nitrogen
pressure, from zero up to the saturation pressure, Indeed, the boundary
points are excluded because of instrumental limitations. When the relative pressure
exceeds a certain limit, capillary condensation commences. resulting in filling
pores with liquid adsorbate. As continues to increase, larger and larger pores
are filled, until all porous space is flooded at 1. Then, by reversing the course
of events, the desorption isotherm can be obtained.
Assuming that the surface is perfectly "wetted" by adsorbate, the contact
angle is zero, the maximum pore radius, filled at the point of the isotherm,
corresponding to the relative pressure can be found from the Kelvin equa-
tion,
where is the surface tension and 34.7cm 3/mol is the molar

volume of liquid nitrogen, and other symbols have their standard meaning.
Take any two neighbouring points, the ith and (i + on the adsorption
isotherm. By the moment the first point is reached, all pores having radius less
than or equal to are filled. Accordingly, at the second point, all pores with radius
less than or equal to are filled. Hence, the volume of pores with radius between
and is given by
where stands for the mole amount of nitrogen (per unit mass of solid) adsorbed
at relative pressure

Next, from Eq. (2) it follows that
1
-
Plotting vs. yields the so-called differential pore size distribution, prob-
ably the most widely used characteristic of a porous solid.
In a more rigorous treatment, a correction should be made to account for the
finite thickness of a multilayer molecular film preadsorbed on the pore walls prior
to the capillary condensation. Most modern instruments automatically execute all
these routines.
The same data can be used for an evaluation of the specific surface area. On
noting that the internal surface area of cylindrical pores with radius in the range
to + is
As, =
and substituting expressions (2) and (3) for and respectively, the well-known
Kiselev formula for the specific surface area, follows at once:
which more commonly is expressed in an integral form, useful if the experimental

points are interpolated by a continuous function.
Mercury porosimetry makes use of the fact that the pressure, required to intrude
liquid mercury into a cylindrical pore with nonwettable walls is inversely pro-
portional to the radius, of the pore. This is a direct consequence of the
Young-Laplace equation,
where 0 is the contact angle of mercury with solids, taken to be

and 7 is the surface tension of mercury. Here, the interpretation of
results is also based on the cylindrical pore model.
By measuring the volume,, of mercury intruded into pores under pressure
the differential pore size distribution, vs. can be arrived at by noting
that
cos
and expressing from Eq. (7). Further processing of data follows essentially in
the same ways as in the case of nitrogen capillary condensation described in the
previous section.

It is worth noting that in mercury porosimetric measurements, the finest pores
are accessed the last at the highest pressure allowed by instrument design, whereas
in nitrogen adsorption measurements they are accessed first, as soon as the capillary
condensation commences. As a result, the accuracy of nitrogen adsorption data
progressively deteriorates with increasing pore size, basically because of the
destructive effect of small pressure fluctuations while approaching the saturation
limit, whereas the accuracy of mercury porosimetric data has the opposite ten-
dency. For this reason, mercury porosimetry is deemed to be a useful complemen-
tary tool for scanning the upper part of the mesoporous region of pore sizes.
Like any idealized model, the cylindrical pore model has many limitations, knowl-
edge of which is essential for proper interpretation of experimental data. In parti-
cular, the following should be kept in mind:
1. Except when cylinders are plugged a t one end, the model fails to explain the
existence of hysteresis between the adsorption and desorption branches of
isotherm. The pore size distributions calculated from these two branches
may differ considerably. A typical example is given in Fig. 1. The desorption
branch always gives a more narrow distribution shifted to the fine-pore
region, which is, in general, in better agreement with that calculated from
mercury intrusion data. At the same time, unless different pores are isolated,
the desorption isotherm may be influenced by some network effects, such as
pore blocking, which do not affect the adsorption branch.
2. The actual pore size may be underestimated at the lower end of the
ous region because of some decrease in surface tension of capillary liquid with
increasing curvature of its
3 In gas adsorption measurements, there is a critical lower value of relative
pressure, 0.4 for nitrogen, below which capillary condensate can-
not exist as a separate phase, no matter whether the pore system extends to
finer pores or not; this is a consequence of the existence of so-called tensile
strength limit As a result of this, a very narrow false maximum may
appear on the differential pore size distribution at its fine-pore end, typically
between 17 and 20 A in the case of nitrogen.
Last, it should be realized that although it is not impossible to build up a

crystalline structure with a close to ideal geometry of the porous space-a compu-
ter-generated example can be seen in Fig. 2 - it is rather unlikely to find something
like that in existing porous materials. In most cases, including precipitated silica
gels and related metal oxide systems, a porous solid is represented by an agglom-
erate of primary particles, of micrometer size or less, rigidly joined together-by
fusion, by sintering, by application of pressure, or otherwise. A molecular model of
such an agglomerate produced by irreversible random coagulation of a few primary
spherical particles, each one having the structure of high cristobalite, is shown in
Fig. 3. The voids between the particles form a pore system.

ADSORPTION
1
0 DESORPTION
0.8
0.6
0.4
0.2
0.0
. (a) Adsorption and desorption isotherms of nitrogen at the surface of

mesoporous gel with the specific surface area and specific pore volume
Experimental points are continuously interpolated by cubic splines. (b)
Differential pore size distributions obtained from the adsorption (- - -) and desorption
branches.

Hypothetical silicon oxide structure (molecular formula shaped
in the form of an empty cylinder with internal radius 16A and external radius 23 A.
Constituent Si and 0 are situated in the lattice of and the
remaining atomic valences are filled with hydrogen atoms to simulate a
surface.
Neither the surface charge density nor the surface potential at the
interface can be measured directly. They are to be retrieved from pH potentiometric
o r electrokinetic data on the basis of certain theoretical concepts and models.
pH Titration of Reactive Surface Groups

The most straightforward and widely applied routine for quantification of reactive
surface groups possessing either basic o r acidic properties and thus capable of
binding o r elimination of protons o r hydroxyl ions is probably pH potentiometric
titration. If it is known that there are n o ions other than protons (hydroxyl ions)
that can react with the surface t o any appreciable extent-a rather exceptional
situation, as will be shown later-the number, of reacted surface groups per

GLOBULA
Aggregated structure of silica gel of 84 spherical

each a radius of and the structure of
unit area of the surface is equal to the of protons (hydroxyl ions) bonded or
released at that area
where is the Avogadro number, V the volume of the solution, the mass of the
dispersed solid, the water dissociation constant, the initial concentration
of protons determined by the amount of added, and the equilibrium
concentration due for measurement. For insoluble solids in the absence of concur-
rent ion adsorption, can be identified as the number of elementary charges, per
unit surface area: that is, can be treated as the surface charge density.

As follows from the above definition, can be either positive or negative. As
water is normally present in abundance, the negative adsorption of protons is
thermodynamically equivalent to their desorption or to the adsorption of hydroxyl
ions.
2. Electrokinetic Measurements
In principle, both the potential and charge density at the surface of a porous solid
can be calculated from electrokinetic data such as the electro-osmotic transfer rate
or the conductivity of a porous plug. Interpretation of this type of experimental
data is based on the solution of the so-called electrokinetic equations
and
Equations and describe the dynamics of a viscous incompressible

charged liquid in the external electric potential field E = under the pressure
gradient Vp. Equation expresses the continuity requirement for ion fluxes, and
is the Poisson equation for the potential Here, is the number density
of the ith ion species, its diffusion coefficient, the charge number, the
viscosity of the liquid, and the volume charge density; the are assumed to
meet some mass balance requirements, and a possible departure from the equili-
brium values is neglected. After imposing certain boundary conditions specifying
the geometry of the system under study and the behavior of the solution at the
surface and in the bulk, system (10) becomes determinate and allows the velocity
vector field and the electric field to be calculated. These quantities can then be used
to calculate the electro-osmotic velocity and streaming current, the quantities that
can be measured experimentally. Since both the electro-osmotic velocity and
streaming current are related to the surface potential, the latter can be found as
well.
Unfortunately, an explicit solution to Eqs. (10) can be obtained for only a very
few simple geometries. In most cases of practical interest, a numerical solution is
required. T o simplify the problem, some additional assumptions are involved: The
potential is usually considered to be uniform and the double layer to be thin.
However, introduction of the charge regulation, with its inherent
potential distributions and overlapping double layers, would mean avoidance of the
above simplifications, thus putting the problem in the range of formidable and not
vet solved ones.
Another problem rests in the difference between the surface potential that
appears in a charge regulation model and the experimentally determined

kinetic potential, or zeta potential. This is a of uncertainty in the
location of the hydrodynamic shear plane
These difficulties explain why u p to now the most valuable practical result is that
obtained by at the beginning of the century. As applied to the
electro-osmotic transfer rate, it reads
where is the zeta potential, which is thought to be equal to the value of at

the shear plane. Applicability of the above equation is limited to thin and weakly
conducting double layers. The superscript is used to emphasize that the
corresponding quantities from an electro-osmotic experiment, for the results
obtained with different electrokinetic techniques may differ appreciably.
Once the zeta potential has been found, it can be used to calculate the surface
charge density using the Grahame equation,
But again, the obtained in way cannot be directly compared to the charge
density that appears in the charge regulation model.
Notwithstanding the existing limitations of the interpretation of experimental
results, electrokinetic measurements are a valuable tool for locating, paradoxically
in the present context, the uncharged state of ionizable surfaces. As shown later,
no proper interpretation of pH potentiometric charge data is possible
without knowledge of the conditions under which the net surface charge becomes
zero, commonly specified by indicating the point of zero charge or the isoelectric
point.
Classical thermodynamics offers a unique tool for analysis of basic concepts of

charge regulation from quite a general point of view. Consider the course of events
occurring when an insoluble and undeformable solid is set in contact with a liquid
solution. After a relatively short relaxation period needed for reorientation of
solvent molecules in the potential field of the solid surface, one or more
steady physicochemical processes, such as adsorption or surface hydroxylation,
may still be in progress until global equilibrium is achieved. Each process can be
depicted as motion along the corresponding reaction coordinate, with speed
proportional to the affinity, A, At constant temperature, pressure. and system
dimensions-rather limiting constraints chosen here for the sake of simplicity-the
change in the free energy of the whole system is given by

where is the area of the interface, usually taken equal to the net
surface area of the dispersed solid measured in dry conditions, and is the inter-
facial tension. The physical meaning of the latter quantity is stipulated later.
In equilibrium, = 0; hence
where is the number of moles of the component and its stoichiometric

coefficient in the process. Then, by using the standard relations between the
affinity, A,, and the chemical potentials, of components involved in the process,
and between the chemical potentials and the mole fractions, of the components,
the following generalized of the mass action law is easily obtained:
where is the so-called intrinsic equilibrium constant of the process, expressed

through the standard chemical potentials, by
Although preferable from a theoretical viewpoint, the use of mole fractions as unis
of concentration may be inconvenient in the case of heterogeneous systems. Since,
however, mass balance is usually implied, concentration units can be changed as
appropriate.
It is expedient to express via the increments,
The same result can be arrived at in another way. Introduce a so-called adsorp-
tion phase, which is composed of adsorbed components and the solid itself, referred
to as the adsorbent. Next, it is necessary to define excess quantities. To d o so, a
reference system is required Normally, the latter is taken to be composed of
the same components, and under the same conditions, as the system in study
but without adsorption. Then the excess quantities, Qf, are defined as
= -
where is some quantity measured in the main system and is the same quantity
measured in the reference system. Chemical potentials of all components except
adsorbent can be chosen to be the in both systems. Then, according to general

thermodynamic principles, the change, in the total energy of the adsorption
phase is
where
and, for # a, by definition of the reference system. The subscript a indi-

cates the adsorbent. Here is an ill-defined quantity (referring to the surface area
of 1 of adsorbent) whose exact value is immaterial.
By introduction of the corresponding thermodynamic potential,
differentiation, and comparison with the following result is obtained:
which is nothing other than a generalization of the equation, trans-

forming itself into the adsorption on substituting =
By introducing another well-known thermodynamic potential,
= - =
i
writing the equilibrium condition as = + = 0, and taking into account

that = +
and so dn, = one gets
from which. after comparison with Eq. the same result as expressed in (19)
follows at once.
Now that Eq. (17) is available and well substantiated, it is possible to proceed
with its application to the problem of charge regulation. The underlying idea here is
that for solids, especially those containing ionizable surface groups, brought
into contact with an aqueous electrolytic solution, the main contribution to 7 is
made by the free electrostatic energy, of the interface,
=
where can be calculated according to the formula
or using the thermodynamically equivalent charging procedure

In the former case, the knowledge of the spatial distribution of potential is required,
whereas in the latter case, knowledge of the surface potential will suffice.
After substitution of Eqs. (27) and (28) for (17) turns into
where the apparent equilibrium constants, K,, have been introduced. Then, making
use of the evident identity,
which follows from (29) on noting that
one gets a more familiar expression for Eq. (30):
where the prime reminds us that the sum should be taken over surface-attached
components only. The latter equation together with the mass conservation principle
constitute the basis of charge regulation.
Before proceeding further, it is of concern to demonstrate the logic of the basic
equation (30) in explaining equilibrium properties of a few real processes. Consider,
for instance, two proteolytic reactions that occur at the surface of amine-modified
silica in contact with aqueous electrolytes:
Example 1. Protonation of surface amino groups,
Here and hence K, is expected to be a decreasing

function of the degree of or the amount of protonated amino groups.
Example 2. The reverse reaction is
Here =
3 .
and therefore this time is expected to increase
together with This somewhat trivial, through, for the equilibrium
constant of the reverse reaction is just the reciprocal of that for the direct reaction.
Of course, the above conclusions are so evident that there would not probably
be any need to resort to Eq. (30). They are intended to demonstrate its general
consistency.

Perhaps the easiest way to understand how charge regulation works lies through
consideration of a particular system. Take partly oxidixed silicon nitride, for exam-
ple. In contact with an aqueous solution, its surface acquires an electric charge on
account of the protolytic reactions
A',
SiOH
Such a set of reactions responsible for charging the surface constitutes a charge
regulation model. Writing down the mass action law for each of the reactions,
where braces are used to designate the so-called surface concentrations

expressing the surface charge density a via and
a= -
and imposing the mass balance requirements

+ = + =
one gets after rearrangement
Then, on adding a potential-to-charge relation,
is some, in general function of a whose mathematical expression

depends on the choice of the electrical double layer model and the geometry of the
system, a determinate system of equations results. substituting
(39) into (38) yields a transcendental equation for a that can be solved numerically,
0 tot tot
provided that and are known. Practically, this is rarely
so; hence another approach proves to be more usable: and are
fitted to approximate some experimental data, typically a dependence of a or zeta
potential on pH = -
Particular implementation of a charge regulation model depends largely on the
form of (39). In the so-called low potential, or approximation, the latter

assumes the simplest linear form. A few explicit formulas are given below.
Henceforth, for the sake of convenience, the reduced potential measured in
units of and reduced linear dimensions measured in units of are used.
Plane:
Long cylindrical cavity of radius
Long empty cylinder with internal radius and external radius
Here, and = 0, 1) are modified functions, is the dielectric constant

of the bulk solution, and is that of the solid. In the surface charge
density on the internal and external walls of the cylinder may not be the same,
which is emphasized by the subscripts.
Theoretically, approximation is valid only for potentials so low that the
exponentials in (38) can be expanded in power series retaining terms up to and
including quadratics. Without such an expansion, it will still work for high poten-
tials; however, in this case the magnitude of charge density is severely underesti-
mated, whereas the magnitude of the surface potential is overestimated.
In general, large potentials necessitate the use of a nonlinear potential-to-charge
relation. Unfortunately, the latter can be written explicitly for a plane geometry
only, in which case
For more complex geometries, it is sometimes possible to obtain approximate

analytic expressions However, in general, the desired potential-to-charge rela-
tionship must be obtained via numerical solution of the Poisson-Boltzmann equa-
tion. In particular, for a long cylindrical cavity, the case most relevant to this study,
the desired relation can be obtained by solving the system

sinh =0
where is the Poisson-Boltzmann equation for potential, and is just a

mathematical expression of the electroneutrality condition.
O n substituting for to and integrating by parts, another
equivalent formulation is obtained:
which is, indeed, a direct consequence of Gauss's theorem.

T o give some more ideas of how the choice of a potential-to-charge relation can
affect the behavior of a charge regulation model, Fig. 4 compares spatial distribu-
tions of the potential within cylindrical cavities of different sizes under different
Radial distribution of potential inside cylindrical pores with charge-regulated

walls. The charge-regulation model used in the calculation was based on the reaction set
(58) with the following model parameters: 4.6 x = 1.5
and Curves and 8 correspond to = 0.5, and
curves and 6 correspond to 2.0 (radius is in units of and the ionic strength
is 0.01 M). ) Linear potential-to-charge relation (- - -)
nonlinear potential-to-charge relation (Poisson-Boltzmann).

conditions, calculated using the charge regulation model based on reaction set
(34) with potential-to-charge relations established by Eqs. (41) and respec-
tively. The parameters were taken to simulate the properties of amine-coated
silica gel with equivalent amounts of and amino groups.
As can be seen, the use of approximation beyond the limits of its
applicability results in serious errors. Another important conclusion that can be
drawn from Fig. 4 is that when double layers overlap, both the surface charge
density and surface potential become dependent on pore size.
1. Relation with the Porous Structure Parameters

As follows from Eq. (29). represents the density of the free electrostatic energy, a
quantity that depends on pore size. Consequently, the equilibrium state of surface
reactions accompanied by charge transfer also becomes pore size dependent. The
same concerns, of course, all the "charge-regulated" quantities, including a, and
surface concentrations of reaction components For instance, by substitut-
ing the equation
which describes the radial distribution of potential within a cylindrical pore of

radius into formula, and making simple quadratures, one
gets
where the subscript is added to emphasize the pore size dependency of the corre-
sponding quantities. This result can also be obtained by using charging
process, (29).
Now let be the pore size distribution normalized as follows:
where is the specific pore volume. Then it is easy to see that
determines the average free electrostatic energy of the solid whose specific surface
area is given by

.5 (a, c) Surface charge density and (b, d) surface potential as functions of pH and
pore size. (a. b) Linear potential-to-charge relation (c, d) nonlinear
potential-to-charge relation The charge regulation model used in
the calculation was based on reaction set (34) with the following model parameters:
= 2.4 = 5.1 = = 2.3
and ionic strength 0.01

Charge Regulation at a Porous Solid

The averages, local equilibrium constants K, are determined by
By solving the charge regulation problem for different r values, distributions

of surface charge density and surface potential over pore sizes are obtained (see
Fig. 5 ) .
2. Asymptotic Behavior of Surface Charge Density and

Surface Potential in Small Pores
The results in Fig. 5 show that as the pore size tends to zero, the surface charge
density also approaches zero, whereas the surface potential takes a finite non-
zero value. This can be proven analytically From Eq. it follows that
which, by virtue of Eq. can be rewritten as
The right-hand side of Eq. (53) is a function of as can be seen after retrieving the
concentrations of protonated amino groups and deprotonated from
( 3 5 ) . By taking into account the evident relations,
and the fact that sinh = and sinh = the existence of a finite
real root of the transcendental equation sinh = f ( u ) is established. The finiteness
of the root necessitates equality of to thus simplifying the task of
finding that root to solving the quadratic equation
where
3. Influence of Electrostatic Saturation on Charge Regulation

It is known that the dielectric constant of the solution close to the surface of a
charged colloidal particle is lower than that in the bulk because of the electrostatic
saturation effect Therefore, the hydration energy is also lower, and ions tend
to leave this region and move to a region with a higher dielectric constant

General thermodynamic principles declare that the above-mentioned effect is not
only limited to the redistribution of the ions in the electrical double layer but also
exerts an influence on the acid-base and ion-exchange properties of
able groups and thereby on the entire process of formation of the surface charge
When the hydration effects are taken into account, the

equation governing the potential distribution takes on the form
where is the potential, is the coordinate-dependent dielectric constant of the

solution close to the surface, is the dielectric constant of the bulk solution, a, is
the radius of the ion, is the refractive index of water, and = where
= 1.41 x Other symbols have their conventional meanings. The above
equations together with the electroneutrality condition
establish a one-to-one potential-to-charge relation that can be used in place of
Correspondingly (33) is replaced by
where the primes over the summations recall as usual that the summation is limited
to surface-attached components.
Figure 6 compares the pH dependencies of surface charge density and surface
potential calculated either neglecting or allowing for the above-mentioned effect.
The charge regulation model can be used successfully for interpretating pH

tiometric titration data, thus enabling the determination of intrinsic equilibrium
constants of surface reactions and the prediction of surface composition under
various conditions. The following example demonstrates this.

Influence of electrostatic saturation on the p H dependencies of (a) surface
charge density and (b) surface potential for different degrees of overlapping of the
electrical double layers. ) Electrostatic saturation allowed for; - -)
electrostatic saturation neglected. The charge regulation model used in this
calculation was based on reaction set (34) with the following model parameters:
= 2.4 x = 5.1 x = tot = 2.3
= = 2.371 A , and ionic strength 0.01 The
potential was calculated from either (44) or (56). Pore radii are given in units
of and are inversely proportional to the degree of overlapping.

Consider the titration of reactive at the surface of a porous silica gel
Typical titration data are shown in Fig. 7 by discrete points. One may want to
approximate these data by a titration curve calculated according to the charge
regulation model based on the minimal reaction set
+ +
F o r a given pore size the solution to this charge regulation problem is found from
the system
where
with the potential-to-charge relation established by (44). Since protons are in

fact being removed from the surface, it is possible to speak of their negative adsorp-
tion,
FIG. 7 Titration data for mesoporous silica gel. The characteristics of the sample are
specified in the caption to Fig. 1 . Ionic strength: 0.001 0.01 and (A)
0.1 Discrete points are the experimental results; solid lines are the best-fit
curves calculated according to the charge regulation model based on reaction set (59)
6
with the following parameters: = 4.6 = 1.05 x
= 1.0 = = 2.371 A. (From Ref. 28.)

However, what can be determined experimentally is only a n average of A,. over all
pore sizes,
where is the pore size distribution defined in (48). The above equation can
be rewritten in a finite-difference representation reflecting the discreteness of experi-
mental data,
Ill
40
Ill
30
0)
20
0)
10
Surface speciation of phosphate-doped silica gel loading of

phosphate 1.2 retrieved from titration data by applying the charge regulation
model based on the reaction set
+ K=1 x mol/dm 3
+ 5x mol/dm 3
+ K=3 x
(From Ref. 29.)

Finally, to compare ( A ) with the amount, of reacted silanols determined experi-
mentally a correction for dissolution of silica gel at high p H should be
made:
= (A) -
where v is the volume of the solution, is the of the sample, and is the
product of the solubility of orthosilicic acid and its dissociation constant. Now,
and as well as some other parameters of interest, can be used as fitting para-
meters and evaluated by minimizing the difference, usually expressed as the squared
deviation, between the theoretical and experimental results
Knowledge of intrinsic equilibrium constants of surface reactions allows
lation of surface specification as a function of pH and ionic strength of the solution,
sometimes leading to quite unexpected results. As a n example, Fig. 8 demonstrates
the percentage of different species a t the surface of phosphate-doped silica gel
It is the charge regulation that allows protonated silanols to exist in appreciable
amounts u p t o p H as high as 6-7 (for pure silica gel, the amount of protonated
silanols becomes negligible a t a p H of about 4 a t the same time dramatically
reducing the dissociation ability of phosphate groups. Thus, the charge regulation
model proves to be effective in the analysis and modeling of protolytic and ionic
equilibria a t charged interfaces.
Bockris., and Yeager Treatise of

Electrochemistry. Vol. 1, The Double Plenum, New York. 1980.
W. Healy and R. White, Colloid Interface (1978).
0. James and A. Parks, in Surface and Colloid Science
New York, 1982, p. 119.
and Adv. Colloid Interface (1980).
Reiner and Radke, Adv. Colloid Interface Sci. (1993).
S. J. Gregg and Sing, Adsorption, Surface Area and 2nd
Academic, London. 1982.
K. Unger, Porous Silica, Amsterdam, 1988.
A. V. Neimark, in Characterization of Porous Solids
Sing, and Unger, Elsevier, Amsterdam, 1991,
p. 67.
V. Zhmud and Sonnefeld, Faraday Trans. (1995).
P. Wiersema, A. Loeb, and Colloid Interface Sci.
(1966).
and V. Electrophoresis, Moscow, 1979.
R. Colloid Interface (1983).
Hunter, Zeta Potential in Colloid Science, Academic, London, 1981.
I. Prigogine and Chemical Green, London,
1954.

A. A. Lopatkin, Theoretical Foundations Physical Adsorption, Moscow
Press, Moscow, 1983.
V. Theory of the Stability of Colloids and Thin Films,
Moscow, 1996.
and Owen, The Physical Chemistry of Electrolytic Solutions,
3rd Reinhold, New York, 1958.
L. Bergstrom and Colloids Surf. (1990).
A. A. Golub, A. I. Zubenko, and V. Zhmud, J. Colloid Interface
(1996).
Sonnefeld, Colloid Surf. A (1996).
V. Zhmud and A. A. Golub, Colloid Surf. A. (1995).
Hunter, Foundations of Colloid Science, Clarendon Press, Oxford, 1987.
Panjukov, Colloid Interface Sci. (1986).
B. V. Zhmud, Colloid Interface 183:11 1 (1996).
F. Chem. Phys. (1951).
Gur, I. Ravina, and A, Babchin, Colloid Interface (1978).
B. V. A. House, and Sonnefeld, Chem. Faraday Trans.
(1997).
B. V. Zhmud, E. Sevastyanova, and A. A. Golub, (1997).
S. Can. Chem. (1966).

Department of Chemical and Materials Engineering, University
, Alberta, Canada
Hewlett Packard, Alto, California
epartment of Mining and Metallurgical Engineering,
University, Montreal, Quebec, Canada
I. Introduction
I I . Surface Charging Mechanisms
A. Unequal hydration
B. Surface hydrolysis
C. Adsorption
D. Isomorphous replacement
I I I . A Hybrid Triple-Layer Model
IV. Site-Binding Model
V. Balance Equations
VI. Determination of Site-Binding Constants
VII. Prediction of Surface Charge Density
VIII. Effect of Ion Adsorption on the Zeta Potential
IX. Prediction of Surface Species
X. Current Status
XI. Summary
References
Nanosized particles and porous materials play an important role in many techno-
logical applications. They have been used widely as biological sensors, integrated
electronic devices, catalyst supports, and adsorbents for detoxification of industrial
effluents and domestic water, decoloring of processing water, and purification of
pharmaceuticals and proteins. In Chapter 3, Xu showed some typical applica-
tions of functionalized superparamagnetic particles in biological cell separation and
industrial effluent detoxification. Porous media. on the other hand, constitute an

important class in catalyst, membrane, and supported adsorbent technology. For
example, a tailored pore size in a membrane allows selective binding of specific
targets or the separation of biological cells based on geometric fit, cell size.
It is well known that a solid surface of either nanosized particles or porous
materials. often becomes charged when brought into contact with a polar medium
such as an aqueous electrolyte solution, although the charging mechanism varies
according to the system preferential dissolution to ionization, hydrolysis,
specific adsorption, isomorphous replacement, tempering, and surface polarization
The charging process is a complex process and involves, in
general, more than one charging mechanism. Regardless of the mechanism, the
charges developed on a solid surface impose an electrostatic force on nearby
charged species electrolytes or biological cells). As a result, an electrical
double layer is developed, usually consisting of a layer of compact surface charge
and a diffuse region of charged species in solution. The surface charge along with
the electrical double layer controls the interactions and behavior of these particles
and porous media with targets or ligands.
The surface charge density and hence the surface electrical potential are deter-
mined by the nature of the surface and solution physicochemistry. As shown in Fig.
1, the surface potential [measured by the zeta potential, the potential at the slip
plane] of corundum depends on the solution pH and the type and
..
Concentration of electrolyte (equivalents per
. Zeta potential of corundum in solutions of various electrolytes. (After Ref. 7.)

Ionization an
tration of metal ions present in addition to the nature of the surface itself As a
result, the interaction of and porous media with targets such as metal
ions is anticipated to be affected by changing either surface properties or solution
conditions. It is therefore important to understand the charging mechanism and to
be able to predict the variations of surface charge (the sign and magnitude) in
response to changes in solution.
In this chapter, a brief summary is presented of various charging mechanisms,
each with a typical example. The progress on the development of a surface
plexation model to describe quantitatively the adsorption of or anionic
species is then reviewed, with emphasis on contributions from our research labora-
tory. A general approach to modeling the charged particles and predicting surface
species distribution is described. Finally, a few examples are given to illustrate the
applications of the approach to an understanding of molecular adsorp-
tion at a mineral surface.
Hydration
When an ionic solid crystal, such as is brought into contact with pure water, it
often becomes negatively charged although the solid particle is electrically neutral.
The negative surface charge indicates the existence of excess surface anions (I-).
The asymmetrical distribution of surface cations and anions may be attributed to
unequal hydration between cations and anions. A stronger hydration of a cation
results in a lower activity of that cation in the solution phase and hence a
surface deficiency. The surface charge can in this case be predicted by comparing
the hydration free energy of cations and anions. Provided that the concentration of
is increased to above the surface positively charged
because of the increased activity of in the solution and hence excess on
the surface. I n this example, and I are referred to as potential-determining
ions (pdi's). Most inorganic solids suspended in an aqueous solution in which they
have some solubility are charged according to this mechanism. Other examples
include and
It is well known [3] that natural silica particles are negatively charged in pure water,
which can be attributed to surface hydrolysis followed by the pH-dependent dis-
sociation of surface hydroxyl groups. When a silica matrix is broken in air, the
surface concentration of and are the same to meet the requirement of
electrical neutrality. When the fractured surface is brought into contact with water,
these charged sites, called unsaturated sites, are readily hydrolyzed to form neutral
surface species of releasing either O H or Adjusting the pH, chan-
ging and O H activity, will cause the neutral surface species to dissociate,
giving rise to a net surface charge. In this case, and O H are pdi's and the
pH at which the particle is neutral surface charge density a = 0) is called the
point of zero charge (pzc). Negative surface charge is created by an acidic

of surface hydroxyl groups, which occurs a t pH above its pzc, while positive
surface charge results from protonation of neutral groups at pH below its
For most oxides, the surface charge originates from this charging mechanism.
Unoxidized sulfide minerals are charged by this mechanism, although mineral
oxidation may involve more than one charging mechanism. The other examples
in this category include surfaces containing groups such as sulfates, and
carboxyl hydroxyl groups that are readily charged when the free energy of
the system favors their dissociation
In some cases, the adsorption of specific site-binding ions, such as and

is responsible for the surface charge of a previously uncharged surface. The surface
charge of lipid bilayers of zwitterionic head groups in the presence of is due to
the adsorption of on - C O Osites vacated by or Obviously, the
surface charge due to this mechanism requires specifically adsorbing ions in aqu-
eous solutions. As a result. this charging mechanism can be readily tested. The
preferential adsorption of charged species in some cases is responsible for surface
charge. The surface charge at and interfaces in the
presence of simple electrolytes has been attributed to the preferential adsorption of
anions at these interfaces [lo-121. It is interesting to note that the adsorption of
specifically adsorbing ions not only makes certain neutral surfaces charged but can
also change the sign of the surface charge for a charged surface. A well-known
example is hematite. In the presence of the originally negative! charged
hematite becomes positively charged as as the concentration of is
ciently high. The charging mechanism of salt-type minerals, such as apatite and
calcite, is a combination of surface hydrolysis followed by preferential dissociation
and adsorption of specifically adsorbing ions.
The surface charge of clay particles serves as an example of a charging mechanism

of this kind. T o illustrate, consider the charging process of mica, a layered structure
mineral in which of the ions are substituted by ions, resulting in a
positive charge deficiency. The excess negative charge in the matrix is compensated
for by interstitial cations, such as or in order to keep the mineral neutral.
These interstitial ions are usually small, weakly bonded to the matrix, and mobile.
When mica is placed in water, the interstitial ions within the surface regions transfer
into the water through a strong hydration force, leaving a net negative charge on
the basal plane of the mica, which is pH-independent
Ã
The hydrolysis of unsa-
turated bonds such as broken S i - 0and A I ‘ bonds on the edges, followed by
the dissociation of hydrated species creates. on the other hand, a pH-dependent
surface charge. Therefore, and O H may also act as For a clay mineral
system, the surface charge density cannot be reduced to zero, although an apparent
point of zero charge may exist. In this case, the basal plane and edge may still carry
the charge of opposite signs although the overall effect shows a pzc.

ace Ionization an 7
The charging mechanisms described above do not cover a complete spectrum

but rather focus on mechanisms often encountered in practice. A number of the-
oretical models have been developed to predict the sign of surface charge and point
of zero charge. Examples are the original minimum solubility treat-
ment Parks' adsorption model the Lai-Fuerstenau site distribution model
the fractional charge model and, more recently,
Miller's hydration model With these models, a variety of interfacial phenom-
ena can be readily explained. In the following section, we demonstrate the use of a
site-binding model incorporated with electrical double layer theories in predicting
electrokinetics and understanding adsorption phenomena of ionic species on a
charged surface.
A
To use the surface complexing phenomenon in various technological applications,
it is essential to develop a predictive tool using a surface complexation model. For a
charged surface, the electrostatic field of the particle is known to affect the dis-
tribution of ions in the interfacial region and the probability of a given species
reacting with the surface. Therefore, the contribution of electrostatic free energy to
the surface binding of various species must be considered. To incorporate the sur-
face charge effect on the surface complexing process, a few models have been used
including the diffuse double-layer the constant-capacitance model
(CCM), the basic Stern model (BSM), the triple-layer model (TLM) and
recently the four-layer model (FLM) With these models, the effect of surface
charge on surface complexation has two aspects: (1) ion distribution (usually
distribution of ionic species given by exp where is the
charge of the adsorbing ions; elemental charge; electrical potential at the
location of adsorbing ion; the Boltzmann constant; and temperature), and
(2) electrostatic contribution = to the intrinsic (chemical) free energy
of adsorption. The TLM in combination with surface complexation has been used
extensively in developing a predictive tool for understanding the acid-base and
metal-binding properties of oxide colloids and is considered to be satisfactory for
most metal oxide and hydroxide systems It has been shown that only the
triple-layer model incorporated with proper surface complex reactions can repro-
duce both the surface charge and surface potential relationships
However, a recent comparison among various models showed that the
TLM cannot give a satisfactory description of all the major adsorption phenomena
such as charging curves, pH and salt dependency of cation adsorption, and
exchange simultaneously. In this study, we describe a different
approach based on, but not limited by, the existing triple-layer model shown sche-
matically in Fig. 2.
In this model, there exist two charge-free layers and a diffuse double layer. The
potential determining ions are located on a surface, designated by subscript 0, and
are responsible for developing surface charge The surface potential at this
plane drops linearly across the inner Helmholtz layer to at the inner
Helmholtz plane (denoted by where most specifically adsorbing ions along

inner outer
Surface Helmholtz Helmholtz
plane plane plane
charge d
absorbed
ions
und a sphalerite particle used in

layer model.
with pair formers are located. (Note: A fraction of specifically adsorbing ions, such
as zinc ions, may locate on the surface and hence contribute to surface charge.)
Further out in the solution phase, a second charge-free layer is present due to the
adsorption of specific adsorbing ions along with ion pairs, with a further linear
drop in the potential across this layer but at a reduced gradient. The plane dividing
the compact layer and diffuse layer is referred to as the outer Helmholtz plane,
where the diffuse layer starts (denoted by d). The potential at the outer Helmholtz
plane coincides with the potential at the end of the diffuse double layer
Based on the classical electrical double layer theory, the potential drop across
the inner Helmholtz layer is related to the surface charge by
= -
where is the integral capacitance of the inner Helmholtz layer, given by
in which and are the dielectric constant in a vacuum and relative dielectric
constant, respectively, and is the thickness of the layer. The potential drop across
the outer Helmholtz layer, on the other hand, is given by
= - -
in which is the integral capacitance of the outer Helmholtz layer. Finally,
according to the Gouy-Chapman-Stern diffuse layer theory, the potential at the
outer Helmholtz plane is related to the charge density by

ace Ionization and
in which and are the bulk concentration and charge, respectively, of the symme-
trical electrolyte counterions in the diffuse layer. Based on the definition of differ-
ential capacitance of the diffuse double layer, in (2) can be expressed as
Equations (1)-(4) are the foundations of electrical double layer theory and are
often used in modeling the adsorption of metal ions at interfaces of charged solid
and electrolyte solutions. In a typical TLM, the outer layer capacitance is often
fixed at a lower value = 0.2 whereas inner layer capacitance can
be adjusted to between 1.0 and 1.4 It should be noted that the
plane model (TPM) is a variation of the classical triple-layer model, in which the
outer layer capacitance is not fixed. Although the physical presentations of the
TLM and TPM are identical as shown in Fig. 2, both involve a surface layer
an inner Helmholtz plane and an outer Helmholtz plane where the
diffuse double layer starts, a one-step protonation process 1 approach) is,
in general, assumed in the TPM, in contrast to a two-step protonation process
2 approach) in the Another distinct difference is that pair-forming ions
are assumed to be on the outer Helmholtz plane in the TPM but on the inner
Helmholtz plane in the TLM. In our study, the outer layer capacitance is allowed
to vary while the pair-forming ions are placed on the inner Helmholtz plane with a
complete set of surface complexation reactions being considered. Therefore, our
approach represents a of the TPM and
T o successfully model the surface charge and surface potential requires an accurate
description of surface complexing reactions. For illustration purposes, we use
lerite, a sulfide mineral, as an to show the surface site-binding model
The first step in the model development is to identify the surface species.
In the present case these are and with
senting surface. The second step involves considerations of all species present in
the inner Helmholtz plane for the system of interest. For sphalerite in a simple
electrolyte solution as in our example, the species includes -
and With all the species being identi-
fied, the complexing reactions associated with these species need to be specified and
the mass action law applied. These are summarized in Table 1 for the current example.
The constants in Table 1 are site-binding constants, and the subscript
represents concentration of surface species, which is related to the corresponding
bulk concentration through = distri-
bution. Replacing all surface concentrations in Eqs. (1)-(6) by this relation, and
rearranging them, the concentration of individual species can then be expressed in
terms of the concentrations of two basic surface species and the
bulk concentrations of ionic species and and the electrical poten-
tial at the locations of ions through corresponding site-binding constants. These
relations are summarized in Table 2 for ease of reference.

II
+ +
t
0 0
+ +
Ill
Ill t
Ill
t
3:
+
3:
+ + 0 0
a
Ill Ill
+ +
2
+ 3:
33 0
N
0
N
Ill Ill Ill
ace
an
Ill
After establishing the surface reactions (5)-(lo), Tables 1 and
the charge densities at arious planes can be derived by using speciation concentra-
tions. The surface charge, for example, is given by the summation of all ionic
species on the surface,
where B is a conversion factor from surface concentration to surface charge

density and is given by B = in which and are the Faraday
constant, the specific surface area and the solid concentration respec-
tively. Similarly, at the plane, the specifically adsorbed charge is
The corresponding species concentrations in Eqs. (1 1) and (12) are given by

(Table 2). Considering electrical neutrality of the system, the accumu-
lated charges in the compact layers must be compensated for by an equivalent
amount of opposite charges in the diffuse layer,
Assuming that surface species are distributed among the total number of avail-
able surface sites, one final condition that has to be satisfied is the
surface mass balance given by
+ + +
+ - + -
One of the major uses of setting up these fundamental relations and constraints is
to determine site-binding constants of surface complexing reactions These
constants are the foundation for various applications, such as predicting the surface
charge density and potentials, identifying adsorption characteristic of metal ions on
solids, and calculating the distribution of surface species. In our hybrid TLM
approach, the basic information required to solve this complex problem is a set
of experimental or (at least as many data points as the parameters
to be estimated minus balance constraints imposed, five for the above example).
In this chapter we use as our input variables. Although it is possible to
measure using ion-sensitive field effect transistors a common approach is to
use the zeta potential as an approximation of due to its access
experimentally. The possible error involved with this approximation was addressed
in detail earlier The effect of shifting the slip plane, where is measured, a
known distance Z into the solution on the fitted site-binding constants can be
examined using the exact solution of the Poisson equation

ace Ionization and 3
The measured zeta potentials of sphalerite a t different electrolyte concentrations

were fitted with the above model using a numerical approach, and the fitted
binding constants are summarized in Table 3. It is evident that the fitted equili-
brium constants are independent of electrolyte concentrations, as expected for a
thermodynamic variable. Also shown in this table is a relatively constant integral
capacitance of the inner Helmholtz layer (0.68 0.01 This value is in the
range of reported values (0.2-2.4 for various sulfide, oxide, and hydroxide
materials A slightly larger pair formation constant was obtained for
sodium ions = -3.45) than for chloride ions = although both
bond to the surface by electrostatic interactions. Similar asymmetrical pair forma-
tion constants of and were also reported A larger binding constant
representing stronger binding appears to be associated with smaller cations
than with anions 0.181 nm) This difference in pair
binding constants causes a shift of iep from pzc, and the self-consistency,
= + as often assumed in the conventional TLM, breaks
This accounts, partially, for the observed poor description of the pH dependency of
metal ion binding encountered with the conventional
With these site-binding constants being determined, the value of sphalerite in
electrolyte solutions can be calculated as a function of pH, and the results are
shown in Fig. 3 by solid lines (only three concentrations are shown for
clarity). The excellent fit of the prediction with the experimental results is illustrated
by all the points lying on the predicted curve. These results testify to the significant
improvement achieved by our hybrid T L M over the conventional which
cannot predict accurately both pH and ionic strength dependence of surface poten-
tials. The improvement is due in part to allowing the variations of differential
capacitance across the diffuse double layer, given by in contrast to the
conventional in which the capacitance of the outer Helmholtz layer is
assumed to be independent of electrolyte concentrations.
In the above example, a hybrid TLM was developed for a sphalerite in NaCl
solutions. This approach can be readily applied to other systems. T o illustrate its
versatility, experimental data on the boehmite water interface, generously provided
Best-Fit Parameters for Sphalerite in NaCl Solutions
-1.45 4.53 -1.16 -3.25 -3.51 -3.05 0.70

-1.45 4.52 -1.15 -3.24 -3.49 -3.05 0.68
-1.43 4.50 -1.13 -3.23 -3.49 -3.05 0.68
1 -1.43 4.51 -1.14 -3.22 -3.48 -3.06 0.67
1 -1.44 4.50 -1.13 -3.24 -3.48 -3.05 0.66
Average -1.44 4.51 -1.14 -3.24 -3.49 -3.05 0.68
Std 0.01 0.01 0.01 0.01 0.01 0.01 0.01

FIG. 3 between the predicted using hybrid triple-layer model and
measured zeta potentials of particles in solutions.
by Wood et were examined. The fitted parameters are given in Table 4,

along with those obtained by Wood et with the conventional TLM. Although
the fitted parameters from both approaches are comparable as shown in Table 4, a
better fitting by our hybrid TLM is seen clearly in Fig. 4. An important improve-
ment in the numerical procedure used here over those published previously is that
experimental data on surface charge density are not required in the fitting process.
With the fitted site-binding constants, on the other hand, can be
calculated. For boehmite in solutions, the calculated is also
shown in Fig. 4. The excellent agreement between the measured and calculated
seen in this figure further supports our hybrid TLM.
4 Best-Fit Parameters for in Solutions
1 6.4 11.5 7.2 9.8 0.60

6.3 10.9 7.8 9.5 0.55
Wood et 6.3 11.9 7.5 9.6 0.59

Ionization
FIG. 4 Comparison between the predicted. using triple-layer and hybrid triple-layer
models, and measured surface charge density and zeta potential of in
solutions.
The hybrid T L M described above can be further extended to the adsorption of

metal ions on metal oxide, hydroxide, or sulfide surfaces. As an example, the
adsorption of ferrous ions on sphalerite, which causes unintentional activation of
sphalerite is examined. Additional species to be considered include
and while and are not considered because of their
relatively low concentrations in the pH range of current interest (pH 8). It has
been noted that the ferrous species on the sphalerite surface oxidizes to the corre-
sponding ferric species. Therefore, the reaction on a sphalerite surface needs
to be considered by including the corresponding ferric hydroxy species in surface
complexing reactions. The reactions considered in addition to those shown in Table
1 are given in Table 5.
Similarly, the surface binding constants can be obtained numerically from the
above surface model in combination with the electrical double layer
theories, including one additional constraint, total iron balance Assuming that
all the binding constants are unknown in our hybrid the best-fit
parameters to the experimental (pH) relation are given in Table 6. Also shown in
this table are the fitted parameters for sphalerite in solutions. It is interesting
to note the excellent agreement between the binding constants obtained in
the presence and absence of ferrous ions, further reinforcing the fidelity of our
hybrid TLM. The excellent fit between the predicted and measured (pH) relations
has been reported elsewhere

an 07
Best-Fit Surface Binding Constants and Integral Capacitance Using the

Hybrid TLM for Sphalerite in 0.01 M Solutions with (I ppm) and Without
Ferrous Ions
With the surface binding constants being determined the adsorption of ferrous
ions on sphalerite can be predicted with the use of commercial packages such as
MINEQL, developed by et o r S O I L C H M , developed by Sposito
and Coves Here we focus o n the effect of ferrous ion adsorption o n the
sphalerite zeta potential. Using the best-fit surface binding constants the
zeta potentials of sphalerite as a function of ferrous ion concentration at various
pH values are calculated and the results are given as solid lines in Fig. 5 , along with
the measured values under the same conditions. Here, the excellent agreement
between the two is evident. F o r all p H values studied, a similar effect of ferrous ions
o n of sphalerite was observed, the zeta potential increases initially with ferrous
.5Effect of ferrous ion addition on zeta potentials of sphalerite: prediction versus

measurement.

ion concentration, reaching a maximum at 2 ppm, and followed by a slight decrease
with a further increase in ferrous ion concentration. This type of calculation pro-
vided valuable insight into the role of ferrous ions in sphalerite activation
The species generated on nanoparticles or porous media play an important role in a

variety of technological applications. For instance, some biological enzymes will
react only with (protonated) surface sites, while other metal ions may
react with anionic surface ligands only. Changing the pH of the system often transfers
these sites from one molecular form to another. It is therefore crucial to be able to
predict and identify the surface species. With the hybrid TLM, reliable surface bind-
ing constants can be obtained, from which the surface species concentration can be
calculated readily. As an example, we calculated the surface concentration of various
iron-containing species on a sphalerite surface as a function of pH, and the results are
given in Fig. 6. Also calculated are the ferrous species distributions in solution with
(Fig. 7a) and without (Fig. 7b) sphalerite. Figure 6 shows that the concentration of
ferrous species on sphalerite is relatively low compared with that in solution. The
main species on the surface at pH 8-10 is which may be
responsible for unintentional activation of sphalerite by iron. Also shown in Fig. 6 is
that reaches a maximum at 10, as does the zeta poten-
tial of sphalerite These findings correspond to the sphalerite flotation observed
around pH 10 with collector in the presence of ferrous ions strongly
Predicted surface distribution of iron-containing species relevant to sphalerite

flota-ion.

7Predicted effect of sphalerite on distribution of iron-containing species in
solution (a) with (b) without sphalerite.
suggesting that is the principal species responsible. This con-

clusion is in general agreement with that from the previous work by X-ray photo-
electron spectroscopy and Fourier transform infrared spectroscopy
Comparing the ferrous species distributions in Fig. 7a with those in Fig. we
see the similar main features. However, species concentration changes gradually in
the presence of sphalerite, suggesting that the concentrations of ferrous species are
buffered by sphalerite. This finding may provide an alternative avenue to buffer
species by adding fine solids.
In the previous sections, we have demonstrated that a hybrid approach integrating

the useful features of the triple-layer model and the three-plane model allows
accurate description of the zeta potential (pH) for a range of electrolyte concentra-
tions and excellent prediction of the surface charge density vs. pH relationship. As
noted, the effect of replacing by in the T L M needs to be further examined by

investigating the effect of shifting the slip plane on the fitted parameters. A sensi-
tivity study using the relation proposed by and will be useful.
The current hybrid TLM remains based on all the ions in the compact layers
being located at a single plane. However, due to the variations in size of various
ions, such as and Zn, in our case the assumption needs to be further
justified by using a recently developed four-layer or yet to be developed five-layer
model. Introducing more layers for ions of various sizes is equivalent to redistri-
buting the charges at various planes and hence involves more capacitors in series.
The integral capacitance in our model is actually an equivalent of two capacitors
(described by the four-layer model) in series. The pH dependence of adsorption can
be examined by varying the position of the planes in which various ions adsorb.
Due to the practical importance of surfactant adsorption in many disciplines, the
development of a new model is urged to include the adsorption of ionic species such
as at both the inner and outer (d) Helmholtz planes where activated
sites are present. In this case, the hybrid triple-layer model needs to be further
refined to account for the contributions of charges from the adsorbed ionic
tant species.
I. SUMMARY
In this chapter, a novel hybrid triple-layer model was described. Three significant
differences between the hybrid and conventional triple-layer models were noted:
1. N o assumption was made with regard to 1 or 2 approach in contrast to
the conventional triple-layer model in which an assumption of iep
+ valid only for a system of symmetrical binding of
formation ions, is often involved,
2 The capacitance of the outer Helmholtz layer was allowed to vary (following
the differential capacitance of the diffuse double layer), in contrast to being
fixed in the conventional
3. The surface charge density was not required as input; rather it was calculated
from the fitted equilibrium constants of surface complexing reactions. An
improved fit was clearly demonstrated with the hybrid triple-layer model
compared to the conventional one.
The sphalerite-ferrous surface binding constants in solutions were deter-
mined. Applications of the hybrid triple-layer model in predicting electrokinetics,
surface charge density, ion adsorption, and surface solution speciation were
described and illustrated with examples.
1.P. De Bruyn and Agar. in Froth Flotation (D. W Fuerstenau, AIME,

New York, 1962, pp.
2. A. and W. Fuerstenau, Colloid Interface (1964).
3. P. Somasumdaran and P. Trans. Indian Metals
(1979).

ace an 61
Fuerstenau, J. D. Miller, and C. Chemistry of Flotation,

AIME, New York, 1985.
Myers, Surface, Interfaces, and Colloids-Principles and Applications, VCH,
New York, 1991.
E. P. Honig and J. Th. Colloid Interface (1969).
Modi and W Fuerstenau, Phys. (1957).
E. Colloids Surf. (1988).
Israelachvili, and Surface Forces, 2nd Academic, San
1992.
A. Frumkin, Phys. Chem. (1924).
and Yordan, Colloid Interface (1986).
A. (1961).
van An Introduction to Clay Colloid Chemistry, Interscience, New
York, 1963.
A. Parks and P. Phys. Chem. (1962).
A. Parks, Chem. Rev. (1965).
Lai and Fierstenau. AIME Trans. (1976).
Salman, and Colloid Interface (1979).
Miller, R. Yalamanchili, and Keller, Langmuir (1992).
J. A. Davis and J. 0. Leckie, Colloid Interface Sci. (1978).
Piasecki, and Rudzinski, Langmuir, (1995).
A . P. Robertson and 0 . Leckie, Colloid Interface (1997).
Katz and Hayes, Colloid Interface Sci, (1995).
Wright and Hunter, (1973).
and Adv. Colloid Interface (1980).
T. and H. Colloid Interface
(1996).
J. Hunter, Foundations of Colloid Science, Oxford. New York, 1986.
Stumm and Morgan, Aquatic Chemistry, 3rd New
York. 1996.
Zhang, Xu, and A. Finch, Colloid Interface Sci. (1995).
Zhang. Xu and A. Finch, Colloid Interface (1995).
P. Bergveld, IEEE Trans.
Fornasiero and Ralston, Colloid Interface (1992).
Rudzinski, S. Partyka, Thomas, and Bottero.
(1992).
Piasecki, and Rudzinski. Langmuir (1995).
Fornasiero, V. and J. Colloid Interface (1992).
Wood, Fornasiero, and Colloid Surf.
Koopal, H. van Riemsdijk, and Bolt, Colloid Interface Sci.
118: (1987).
A, Gibb and Koopal, Colloid Interface (1990).
Crawford, I, H . Harding, and Mainwaring, Colloid Interface
(1996).
Zhang, Rao, and A. Finch, Colloids Surf. (1995).
Zhang, I. Butler, Xu, and A, Finch,
(1994).

41. Zachary, and Morel, Technical Note 18, Water Quality
Laboratory. Department of Engineering, MIT, 1976.
42. Sposito and Coves, Manual. University of California, Berkeley,
1988.
43. A. and Adv. Colloid Interface 28:111 (1988).

Department of
Metallurgical Engineering, University of Utah, Salt Lake City, Utah
I. Introduction
II. Simplified Lattice Ion Hydration Theory
I I I Nonequilibrium Electrophoretic Measurements
A. Electrokinetic phenomena-electrophoresis
B. Laser-Doppler electrophoresis
C. Results and discussion
IV. Extended Lattice Ion Hydration Theory
A. Case A (KF, RbF, CsF, CsCI, CsBr, and CsI)
B. Case B (LiF, NaF, LiCl, NaCI, LiBr, NaBr, Lil, Nal, and Kl)
C. Case C (KCl, RbCl, KBr, RbBr, and Rbl)
D . The sign of the surface charge
V. Oxygen Defect States
VI. Conclusions
References
The importance of surface charge in understanding important phenomena such as

adsorption, crystallization, and has been well established
for many different systems Until recently the significance of surface charge
at high ionic strengths was questioned. It is now evident that in such systems it is
very important. For example, the flotation of alkali halide particles from their
saturated solution is controlled in many instances by the sign of the surface charge
of the alkali halide particles.
The significance of charge in alkali flotation systems was very contro-
versial when it was first introduced in 1967 by Roman et Critics argued that

14 Miller and
in such systems the double layer is collapsed and electrostatic particle interaction
forces do not become significant except at very close separation distances.
Furthermore, Roman et [4] failed to show any direct experimental evidence to
validate the hypothesis. Conventional electrophoresis could not be used to study
soluble salts because of the difficulty in obtaining colloidal sized particles and also
because of the high conductivity of the solutions In the streaming
potential technique was evaluated without success. Later, Pizarro [6] had similar
problems in providing experimental evidence regarding the role of surface charge in
the flotation of alkali salt systems. Since then, the advent of the laser-
Doppler electrophoretic technique has made possible the characterization of certain
interesting systems including soluble salt systems that were hitherto not investigated
due to the lack of an appropriate experimental technique.
Based on nonequilibrium electrokinetic measurements for alkali halides by laser-
Doppler electrophoresis reported by Miller et al. the sign of the surface charge
of alkali halides in their saturated brines has been established. The surface charge
results thus determined are generally as expected from simplified lattice ion hydra-
tion theory. However, there are seven
KI, and RbI-when the results from nonequilibrium electrophoretic measurements
by laser-Doppler electrophoresis are compared with the predictions from lattice ion
hydration theory. Previously it was observed that in almost each case inaccu-
rate predictions occur when the difference between and anionic gaseous
hydration energies is small. However, more recent theoretical analysis has shown
that when the surface is considered to be partially hydrated, dependent on the
properties of the lattice ions such as the ionic radius (steric effect) and the hydration
number (hydration effect), the sign of the surface charge of all alkali halides except
for KC1 can be accurately predicted by this extended lattice ion hydration theory.
Finally, it has been shown by Yalamanchili and Miller that oxygen defects
present in the KC1 lattice can strongly influence its surface charge characteristics
and the negative deviation of KC1 from theory can be explained by the presence of
oxygen defects. Further, Veeramasuneni et showed that oxygen defects
deliberately introduced into positively charged can reverse the surface charge.
The fundamental aspects of, and some recent contributions to, the analysis of
surface charge of alkali halides in their saturated solutions include simplified lattice
ion hydration theory, nonequilibrium electrophoresis extended lattice ion
hydration theory and oxygen defect states and these issues are discussed
in the following sections.
Surface charge at mineral surfaces can be generated by one or more mechanisms.

The most commonly encountered charging mechanisms include dissociation of
surface acid groups lattice substitution preferential hydration of
surface lattice ions, and preferential adsorption phenomena In some cases,
a surface may be charged by the orientation of dipoles or by surface polarization.
Of particular concern in the case of alkali halides is the preferential hydration of

surface lattice ions. Bryn and Agar first described this mechanism for
nonreactive ionic solids in terms of lattice ion hydration theory.
The simplified lattice ion hydration theory describes surface charge development
by nonreactive ionic solids placed in water. This theory applies to ionic solids that
do not react with water to form weak surface acid groups (hydrolysis) and to ionic
solids that do not undergo surface oxidation reactions. Either reaction type would
modulate the surface charge that the solid develops in water. The only reaction of
consequence then is the hydration of lattice ions, and the differential hydration of
these lattice ions at the surface of the solid determines the sign of the surface
charge.
The theory as first proposed by and Agar was for simple
univalent salts in which the cation and anion are a t interchangeable lattice posi-
tions, which is to say that their lattice energies are equivalent. In this elementary
case, the surface charge can be estimated simply from a comparison of the hydra-
tion free energies of the corresponding gaseous ions. Until recently, this concept has
been demonstrated only for certain silver halides as shown in Table 1, where the
surface charge of the silver halides predicted by lattice ion hydration theory based
on gaseous ion hydration free energies is compared with experimentally deter-
mined surface charges. It can be noted Table 1 that there is excellent agree-
ment between theoretical predictions and experimental results. The sign of the
surface charge of silver halides, except for has been measured experimentally
by many researchers. However, recently, for the first time, nonequilibrium electro-
kinetic measurements for made with the laser-Doppler electrophoretic tech-
nique were reported Now, with the electrophoretic mobility measurements for
the lattice ion hydration theory has been substantiated for all silver halides
(see 1). It can be noted from Table 1 that, unlike the other silver halides,
silver fluoride is positively charged in its saturated solution as determined by
quilibrium electrophoretic measurements and as expected from simplified lattice ion
hydration theory. Until recently, the analysis of other halide salts was not possible.
The process of charge generation by hydration of lattice ions should be repre-
sented more rigorously by considering the lattice energy of surface ions, which in
turn is dependent on the crystal structure and preferential plane of cleavage
1 Sign of Surface Charge for Silver Halides

Sign of surface charge
Salt Predicted
458.6 441 Positive Positive
347.3 441 Negative Negative
Source: Ref. 22

For uni-univalent ionic solids such as the alkali halides the energetics
can be represented, for the solid MX, as follows:
If the hydration energy of the surface lattice cation is more negative than the
hydration energy of the surface lattice anion, as described by reactions then the
surface of M X will carry a negative charge. The converse is also true. The complete
analysis requires knowledge of the hydration free energies for gaseous ions [reac-
tions and the lattice energy for surface ions [reactions The hydration free
energy, for individual ions can be determined from the Born expres-
sion given by
where
N = Avogadro's number
z valence of the ion
= charge on the ion
= radius of the ion
dielectric constant of the water
The lattice energy for surface ions can be determined from Born-Mayer theory
as expressed by the equation
where
= lattice energy
constant for the crystal
= characteristic unit cell dimension
= a parameter related to the repulsive forces in the crystal that is evaluated
from compressibility measurements
Equations (4) and (5) can be used to obtain the hydration free energy of the
surface cation or anion [Reaction as follows:
Depending on whether the hydration free energy of the cation or the hydration free
energy of the anion is more negative, the corresponding surface ion will hydrate to
a greater extent, leaving the surface either positively or negatively charged as men-
tioned above. Of course, for symmetric uni-univalent nonreactive ionic solids hav-
ing a structure in which cation and anion are interchangeable, the terms are

equivalent, and the free energy of hydration for the gaseous ions can be used to
determine the sign of the surface charge. This is the essence of the simplified lattice
ion hydration theory.
For valence-asymmetric nonreactive ionic solids such as calcium fluoride, the
analysis of the sign of the surface charge becomes tedious and cannot be done
simply by comparing the gaseous ion hydration free energies of anion and cation
for a particular solid. In this regard, Miller and Calara extended the
simplified lattice ion hydration theory to more complex ionic solids by considering
the lattice energy effects. These authors developed a computational technique for
the calculation of surface constants and binding energies of surface ions
for semi-infinite solids. The calculated binding energy of surface ions together with
the gaseous ion hydration free energy allows for the calculation of the hydration
energy of surface ions. Thus, the sign of the surface charge can be predicted for
complex ionic solids from the relative magnitudes of the ionic surface hydration
energies. The positive surface charge for in its saturated solution was
explained on this basis. In summary, the utility of the lattice ion hydration theory
has been demonstrated both for simple ionic solids and for complex ionic
solids when the lattice energy must be taken into consideration by calculation of
surface constants
In view of the above, the lattice ion hydration theory has been shown to be useful
to predict the sign of the surface charge of alkali halides in their saturated solutions.
There are four basic types of electrokinetic techniques that can be used to evaluate
the zeta potential under a given set of conditions. The common feature of these
electrokinetic measurements is the relative motion between a charged surface and
its surrounding fluid, with an electric field either applied or generated from the
movement. Perhaps the most familiar type and most widely used electrokinetic
technique is electrophoresis, which involves the movement of charged particles
under the influence of an applied electric field
The second technique, known as electro-osmosis, causes a fluid in contact with a
stationary charged surface to move in response to an applied electric field. The third
technique, which can be regarded as the converse of electro-osmosis. is known as
the streaming potential technique. The streaming potential involves the measure-
ment of the electric potential generated when a liquid is forced under pressure to
move in contact with a stationary charged surface. Finally, the sedimentation
potential technique is a method that is, in a way, the converse of electrophoresis.
A sedimentation potential arises across the suspension when charged particles settle
in a stationary liquid. (29) explained the basic principles of these electro-
kinetic techniques in a nutshell as shown in Table 2.
The electrophoretic technique is of particular interest because it has been used
for measuring the electrophoretic mobilities of alkali halide particles. In this case,
charged particles suspended in the electrolyte are transported toward the electrode
of opposite charge when an electric field is applied across the electrolyte. However,

2 Electrokinetic Phenomena
Nature of solid surface
Potential Stationary surface Moving surface
Applied Electro-osmosis Electrophoresis
Induced potential Sedimentation potential
Source: Ref. 29
viscous forces oppose this movement, and when an equilibrium is reached between
electrical attraction and viscous drag, the particle moves with a constant velocity.
This velocity depends on the voltage gradient, the dielectric constant, and the
particle's zeta potential. The particle's velocity divided by the electric field is gen-
erally known as its electrophoretic mobility.
The zeta potential can be determined from the electrophoretic mobility by
using the Henry equation in which the electrophoretic mobility is seen to be a
function of the double layer thickness,
where
= electrophoretic mobility
= dielectric constant
= viscosity of the medium
a = particle radius
= double layer thickness
For large values of for high electrolyte concentrations, Smoluchowski
modified the Henry equation as
=
This equation can be used to simply calculate the zeta potential from the measured
electrophoretic mobility. For small values of Eq. (7) becomes
For many years, electrophoretic measurements have been carried out using a
simple optical microscope to observe the individual particles moving between two
electrodes when a potential gradient is imposed. The actual electrophoretic mobi-
lities are determined by manually timing the movement of particles over a fixed
measuring distance. This microscopic method, which is still being used by some
researchers. suffers from several limitations:
. First of all, the method is very slow and tedious. For this reason, a trained
operator must devote considerable time and effort to accumulate a meaningful
set of data from which a mean mobility can be calculated. Operator fatigue and
eyestrain are real factors that could affect the outcome of these

2. The technique is particularly strenuous and is limited to particles that are
visible. Particles with an unfavorable refractive index may be virtually invisible.
3. Electro-osmosis is always a problem in classical microscopic methods, and
extreme care must be taken to precisely locate the stationary layer before
each measurement.
4. Electrophoretic mobility measurements are limited to low ionic strength solu-
tions due to electrode polarization and conductivity limitations, which, if not
satisifed, can lead to sample heating and the generation of thermally induced
convective flow in the sample cell.
5 . Finally, only dilute samples of suspended particles can be used, in order to
allow individual particles to be clearly observed during the measurement.
In spite of some improvements made in the classical optical microscopic method,
such as the use of video cameras to relieve eyestrain and the use of laser light
illumination for better particle visibility, the disadvantages mentioned above
remain in large part.
The availability of low power lasers that provide bright and coherent light sources
together with sophisticated signal analysis instrumentation have led to the devel-
opment of an electrophoretic light scattering method to measure electrophoretic
mobilities of particles in suspension. This method combines the principles of elec-
trophoresis with laser-Doppler velocimetry, a method used for measuring the speed
of moving particles by analyzing the Doppler shifts of scattered light such
moving particles. The new generation of laser-Doppler instruments have trans-
formed electrophoresis a tedious technique to one that can be applied
routinely to obtain fast, accurate data without extensive operator training.
Doppler electrophoretic offer several advantages over classical micro-
scopic methods.
Advantages of Laser-Doppler Electrophoresis
Laser-Doppler electrophoretic measurements can be made very rapidly, in a matter
of few seconds rather than a few hours, and over a wide range of particle concen-
trations. Systems suitable for these measurements typically have particle concen-
trations of up to 10,000 ppm. These features are very important in systems that are
either chemically or physically unstable or settle rapidly.
Small values of the electrophoretic mobility [even smaller than 0.1
can be measured using the laser-Doppler electrophoretic technique. This aspect is
particularly important when the mobilities involved are small in magnitude and are
in the region close to the point of zero charge.
The traditional microscopic method is complicated by the electro-osmotic velo-
city due to the negatively charged glass walls of the cell. There is only one particular
region between the wall and the center of the cell, which is called the stationary
layer, where the particles move with a velocity that is due solely to their charge. It is
difficult to consistently focus the microscope in this layer with classical methods.
However, the laser-Doppler electrophoretic technique overcomes this difficulty by
accurately measuring the particle's velocity in the stationary layer.

Miller and
Sample heating and electrode polarization are greatly reduced in laser-Doppler

electrophoresis instruments due to certain features incorporated in the technique.
These features facilitate electrophoretic measurements even in solutions with high
salt concentrations.
Laser-Doppler electrophoretic measurements are made on a much larger num-
ber of particles compared to other techniques, resulting in the determination of
mobility distributions rather than single-point mobilities. Operator bias is comple-
tely eliminated because the operator does not select the particles to be measured.
2. Alkali Halide Particles
The advent of laser-Doppler electrophoresis instrumentation allows for the char-
acterization of the electrophoretic properties of many interesting systems that were
previously experimentally inaccessible. In particular, that change with time
can now be studied. In view of this, laser-Doppler electrophoresis was used to
determine the electrophoretic mobilities of soluble salt particles (alkali halides
and other soluble salts) and thus predict the sign of their surface charge in saturated
brines. Of course, zeta potential measurements in a saturated brine are impossible
because the double layer is collapsed and consequently a zeta potential does not
exist. However, a surface charge may still be present. T o describe the electrokinetic
behavior of soluble salt particles, nonequilibrium electrophoretic mobility measure-
ments of the alkali halide salt were made after injection of the particles into
fresh water. Such measurements are dynamic in the sense that the salt particles are
dissolving in the aqueous phase during the course of the electrophoretic mobility
measurement. It is important, however, that these measurements should give a
relative indication of the sign of the surface charge of the salt particles in their
saturated brine provided the rate of release is the same for both cation and anion.
3. Principles of Operation
Zetasizer 3 Instruments Ltd., London, UK) with the AZ4 standard cell
was used for measuring electrophoretic mobilities by laser-Doppler electrophoresis.
The schematic of the Zetasizer 3 is shown in Fig. 1. Two coherent laser beams of
red light produced by splitting the output of a low-power He-Ne laser are focused
and made to intersect within the quartz capillary cell at a point of zero convective
flow. As a result, a pattern of interference fringes is formed, and the particles scatter
light as they move across the fringes under the influence of the applied electric field.
The intensity of the scattered light varies with a frequency that is related to the
velocity of the particles. A fast photomultiplier tube together with a digital
is used to the signals, and the distribution of zeta potentials and/or
distribution of electrophoretic mobilities is thus determined. The sign of the elec-
trophoretic mobility and thus the sign of the surface charge is determined by
referencing the observed Doppler frequency of the light scattered by the particles
moving through the fringes to the modulation frequency applied to one of the laser
beams.
The accurate measurement of low electrophoretic mobilities of particles is thus
accomplished by modulating one of the laser beams. This is done by reflecting one
of the beams off a mirror that is attached to a piezoelectric crystal. A varying
voltage is applied to the crystal, causing it to move. The modulation frequency,

Beam
Beam Selection
Shutter
\ Intensity M o n i t o r
. Schematic arrangement of Zetasizer 3. (From Ref. 7.)
which is between 100 and 1000 Hz, is always present in the signal, so that if the
particles have no net velocity the scattered light collected by the photomultiplier
will have exactly that frequency. In this way, particles with low or zero mobility can
be accurately measured.
The nonequilibrium measurements for alkali halide particles take only a few
seconds. Since the measurement is done on a large number of dissolving particles,
a distribution of electrophoretic mobilities is obtained, and a distinct eletrokinetic
characterization of the alkali halide particles is expected.
Figure 2 presents a schematic diagram of the AZ4 standard cell. This cell con-
sists of a 4 m m diameter quartz capillary. The platinum electrodes are in compart-
ments at each end of the cell. A semipermeable membrane separates the electrodes
from the suspension sample to prevent contamination of the electrodes by the
sample. The electrode chambers are filled with an electrolyte that is at least as
conducting as the sample itself. Polarization of the electrodes is reduced by simply
reversing the direction of the applied electric field a t about 2 Hz. The measurement
of particle velocity from many field reversals is added to obtain a good
noise ratio. A technique called "duty cycling", which allows the voltage to be
switched off for a short time between cycles is used to avoid heating of the suspen-
sion sample, especially for high conductivity solutions Switching off the
applied voltage allows the heat in the sample to dissipate. The problem of
tro-osmosis is overcome by measuring the velocity of particles at a point in the cell
where the electro-osmotic velocity is zero.
4. Experimental Procedure for Laser-Doppler Electrophoresis
Electrophoretic mobility measurements for various soluble salts were performed by
the laser-Doppler electrophoretic technique. The Zetasizer 3 with AZ4 standard cell
was used for measuring the electrophoretic mobilities.

,Main Electrodes
Membrane
Electrodes
.2
Schematic diagram of the capillary electrophoresis cell for the Zetasizer 3.
(From Ref.
The soluble salts were sized by dry sieving, and the 65 x 100 fraction was
used in this study for nonequilibrium electrophoretic mobility measurements. The
electrode chambers were filled with an appropriate electrolyte of required ionic
strength. The electrophoretic mobilities of the alkali halides were measured in the
following manner:
1 . The salt particles were added to of high purity water in a beaker.
2. The of salt added (from 0.5 to 2.5 g depending on density and solu-
bility) was such that the Zetasizer 3 system would signal when an ideal level of
particles was present in the solution after injection.
3. The suspension was stirred gently for a moment 1 then quickly injected
into the cell, and the electrophoretic mobility was measured.
Each measurement required approximately 15 The starting size of the salt
particles is a critical factor. If the salt particles are too small, they dissolve before
the measurement can be completed. On the other hand, if they are too large,
sedimentation occurs in the capillary. Great care was taken to avoid the injection
of air bubbles into the cell. The experiment was performed at least three times for
each salt under the same conditions to determine the reproducibility of the mea-
surements.
In the case of reference equilibrium measurements which were possible for
and the salts (-400 mesh) were allowed to equilibrate at natural pH for at
least 60 min prior to electrophoretic measurements.
The nonequilibrium electrophoretic mobilities of selected alkali halides were mea-

sured during dissolution by laser-Doppler electrophoresis. As stated, the measure-
ments were made during a time interval of 10-15 s. The results were found to be
quite reproducible with respect to the sign of the electrophoretic mobility.

The laser-Doppler measurements give a distribution of electro-
phoretic mobilities. For example, typical electrophoretic mobility distributions for
and KC1 are shown in Fig. 3. Note that. although small in magnitude, NaCl
exhibits an electrophoretic mobility of 0.14 which is dis-
tinctly positive, while the electrophoretic for -0.47 0.26
cm), is distinctly negative.
The distributions of electorphoretic mobilities are determined for a sample con-
taining a large number of particles, which is quite impossible by optical microscopic
methods. The mobility distributions obtained by laser-Doppler electrophoresis
represent a comprehensive and accurate assessment of the mobility spectrum of
all the constituent particles in the sample. More important. the sensitivity of the
technique is such that particles ranging in size from a few nanometers to several
micrometers can be measured. Table 3 presents experimental results for 21
different alkali halides. In all cases excellent reproducibility was achieved with
respect to the sign of the electrophoretic mobility, when ultrapure 99.9%)
salts were used. As many as nine replicate measurements were made, and it should
be emphasized that for a given salt the measured electrophoretic mobility was
consistently of the same sign. The electrophoretic mobilities reported in Table 3
were determined by taking into account the composite data from all the replicate
measurements. For example, in the case of the experiment was performed nine
times, resulting in nine different electrophoretic mobility distributions. The data
from these nine measurements were combined, and a grand distribution for the
electrophoretic mobility of was calculated.
Of course. the ultimate demonstration of the validity of these
measurements is the measurement of the corresponding equilibrium electrophoretic
.3 electrophoretic mobility distributions for KC1 and

(From Ref. 7.)

Results from Laser-Doppler Electrophoretic Experiments--
Conditions
Electrophoretic mobility
Alkali halide No. of observations
Fluoride
LIF 5 0.15
4 16 0.12
KF 5 0.07
9 +O. 10 0.06
3 0.19
Chloride
4 -0.64 0.24
5 0.10
6 -0.43 0.19
3 0.16
4 Â 0.24
Bromide
7 -0.33 0.17
3 -0.32 0.16
3 -0.47 0.24
5 -0.09 0.06
3 0.05
Iodide
4 -0.76 0.24
9 -0.72 0.19
4 0.06
4 -0.25 0.14
3 0.04
4 0.03
Source: Ref. 7.
mobilities that was possible for and The results are presented in Table 4
and show the nonequilibrium electrophoretic mobilities to be equivalent in sign to
the equilibrium electrophoretic mobilities. Again, the electrophoretic mobilities
reported in Table 4 were determined from the composite electrophoretic mobility
data.
It should be mentioned that diffusiophoresis, the movement of particles in sus-
pension under the influence of an externally imposed solution concentration gra-
dient, can affect the nonequilibrium measurements of alkali halide particles. When
a particle of an alkali halide salt is dissolving in solution, it is possible that a
concentration gradient will be established in the volume adjacent to the outer
boundary of the double layer. The concentration gradient thus established can
influence the measured electrophoretic mobilities by laser-Doppler electrophoresis.
However, in the case of alkali halide particles, the slope of the concentration

Comparison of Nonequilibrium Electrophoretic Mobilities with
Equilibrium Electrophoretic Mobilities for Two Alkali Halide Salts
Electrophoretic mobility"
Salt Solubility Nonequilibrium Equilibrium

0.15 ( 5 ) 0.12 (3)
4.22 100 0.12 (4) 0.08 (4)
"Number of observations reported in parentheses
Source: Ref. 7.
gradient varies with time because of the rapid dissolution of these particles in
solution. Therefore, due to the small concentration gradient close to the particle's
surface, the influence of diffusiophoresis on measured electrophoretic mobilities is
expected to be of little significance. Furthermore, the results from equilibrium
electrophoretic mobility measurements for and (no diffusiophoresis) sug-
gest that at least the sign of the surface charge, if not the magnitude, as measured
by laser-Doppler electrophoresis is not affected by diffusiophoresis due to concen-
tration gradients in the outer layers of the double layer. Further, it is interesting
that the measured electrophoretic mobility for is negative, whereas that for the
thermodynamically stable is positive. It would appear that the hydration
of lattice ions does have a significant influence on the surface charge of alkali
halides and other soluble salts
As described in the previous section, electrokinetic measurements

using laser-Doppler electrophoresis have established the sign of the surface charge
for 21 alkali halide salts. These surface charge results for alkali halides are generally
as expected from simplified lattice ion hydration theory with some exceptions,
specifically and as shown in Table 5 .
Thus the sign of the surface cannot be predicted solely from consideration of the
hydration energies of the gaseous ions that constitute the ionic lattice. Accurate
prediction of the surface charge must include consideration of the partial hydration
of surface lattice ions. In this regard, interaction energies have been
taken into consideration for three cases, based on and hydration effects, in
order to predict more accurately the surface charge of alkali halides. The calcula-
tions indicate that the sign of the surface charge of alkali halides can be accurately
predicted from an extended lattice ion hydration theory that involves the partial
hydration of surface lattice ions
One deficiency in the simplified lattice ion hydration theory is that the surface
state, indicated by in the equation is an anhydrous state in a
vacuum, and, in fact, this surface ion should be considered to be partially hydrated
in order to more accurately represent the system. The additional reaction that needs

Miller and
Sign of the Surface Charge for Selected Halides

Negative gaseous ion hydration
free energies Sign of surface charge
Cation Anion
Fluoride
KF
Chloride
KC1
Bromide
Iodide
KI
A = Difference between cation and anion gaseous hydration free energies.

lattice ion hydration theory.
Electrophoretic mobility measurements by laser-Doppler electrophoresis.
Source: Ref. 7.
to be considered is + and the overall reaction that more

accurately represents the system of interest would be
The results from this current theoretical analysis suggest that the surface ions
must be considered to be partially hydrated, dependent on the properties of the
lattice ions such as ionic radius (steric effect) and hydration number (hydration
effect), and on this basis the sign of the surface charge of all alkali halides except for
KC1 can be accurately predicted. This analysis, which involves the inclusion of
dipole interactions together with the simplified lattice ion hydration theory, is
referred to as the extended lattice ion hydration theory. The physical process of
charge generation as described by the extended lattice ion hydration theory can be

represented by considering the relative energies of the following reactions for a
univalent ionic solid, MX:
where subscripts g, and represent partially hydrated surface state,

free (vacuum) surface state, gaseous state, and aqueous state, respectively.
and are the surface binding energy, the gaseous
hydration free energy, and ion-dipole interaction energy, respectively, of the cation
Further the hydration free energy of the surface ion,
Similarly, for the anion
Depending on whether or is more negative, the corresponding

ion will hydrate to a greater extent, and the surface will acquire the charge of the
other ion. Here, as in the case of simplified lattice ion hydration theory, the cation
and anion are considered to be interchangeable, and thus the lattice energies of
surface cation and anion are equivalent. Finally, the surface charge is established
according to the extended lattice ion hydration theory by comparing both the
gaseous ion hydration free energies and ion-dipole interaction energies for partial
hydration of surface lattice ions.
Many molecules possess an electric dipole. For in the molecule the
chlorine atom tends to draw the hydrogen's electron toward itself, and this mole-
cule therefore has a permanent dipole. Such molecules are called polar molecules.
Permanent dipole moments occur only in asymmetric molecules and arise the
unequal distribution of electrons in the covalent bonds. Thus, it is not surprising
that a characteristic dipole moment can be assigned to each type of covalent bond.
The dipole moment of such a polar molecule is defined as
where is the distance between the centers of two charges, and The electro-
static pair interaction (interaction energy) between a charged atom and a polar
molecule presented by Israelachvili is
where is the dipole is the distance between the centers of charge of the
ion and a polar molecule, is the angle subtended by the dipole and the line joining
the centers of charge of the ion and the polar molecule, is valence, is electronic
charge, is permittivity of free space, and is dielectric constant As men-
tioned above, Eq. (17) gives the free energy for the interaction of a charge and a

point dipole in some medium of known dielectric properties. Then, by definition,
when a cation is near a dipolar molecule, maximum attraction
negative energy) will occur when the dipole points away from the ion (9 =
whereas if the dipole points toward the ion (9 = the interaction energy is
positive and the force is repulsive.
In the present work, (17) was used to calculate the ion-dipole interaction
energies for the alkali halide water dipole Here, water was treated as a
simple spherical of radius with a point dipole of moment 1.85
Also, based on steric and hydration effects, three cases-A, and C-were used to
calculate the ion-dipole interaction energies. (See Table 6.) Case A considers the
alkali halides K F , and which have an alkali ion radius
close to that of the water molecule and a ratio of hydration numbers of alkali ion to
halide ion of less than 2. Case B considers the alkali halides
and KI, which have an alkali ion radius less than that of the water
molecule and a ratio of hydration numbers of alkali ion to halide ion of more than
2. Finally, case C considers the alkali halides and
which have an alkali ion radius close to that of the water molecule and a ratio of
6 Steric and Hydration Effects for and Halide Ions

Properties of alkali halides
Steric Hydration Alkali
Case effect effect halide (nm)
KF
B Significant; No; F
behavior
1.4 limited by
effect
NaBr
Lit
C Small Significant KC1

- 1.4
2 RbCl
Source: Ref. 9

hydration numbers of alkali ion to halide ion of more than 2. A detailed description
of these three cases showing the ion-dipole interaction energy calculations is pre-
sented in the following sections.
Six alkali halide salts K F , and an alkali ion radius

close to that of the water molecule and a ratio of hydration numbers of alkali ion to
halide ion of less than 2. These alkali halides were classified into one group, and
calculations based on steric and hydration effects were made to analyze their sur-
face charge. The steric effect, which is related to the size of the alkali ion, halide ion,
and water molecule, has little influence on the partial hydration of these six alkali
halides due to the fact that the radii of alkali ion, halide ion, and water molecule are
very close to one another. The other effect, the hydration effect, mainly depends on
the ratio of alkali ion hydration number to halide ion hydration number. In this
case, since the ratio is less than 2, the influence of hydration is also negligible. The
situation for case A is summarized in Table 6.
Figure 4 shows a schematic representation of case A. The angles and in
Fig. 4 represent the angles subtended by the dipole to the line joining the ion and
the polar molecule. The angles and used to calculate the interaction energy
were obtained by taking the inverse cosine of the ratio of the sides of the respective
triangles as shown in Fig. 4. The lengths of these sides were obtained from the
available radius values for alkali ion, halide ion, and water molecule. These angles
were then used in (17) to calculate the ion-dipole interaction energy values.
Both attractive and repulsive components of interaction energy were calculated
using Eq. (17). For example, an attractive interaction energy was calculated for
an alkali ion-water dipole with an angle = and a repulsive interaction energy
was calculated for the same dipole with an angle 9 = The total ion-dipole
interaction energy for alkali ion-water dipole is obtained simply by adding
the attractive and repulsive components. Similarly, an attractive interaction energy
was calculated for halide ion-water dipole with an angle 0 = and a repulsive
interaction energy was calculated for the same system with an angle 0 = The
total ion-dipole interaction energy for halide ion-water dipole is obtained
Alkali ion
Attractionat distance CD and
Repulsion at distance and
ion
Attractionat distance and
at distance AD and
4 Case A, suggested for and alkali halides

having an alkali ion radius close to that of the water molecule and a ratio of
hydration numbers of alkali ion to halide ion of less than 2. (From Ref. 9.)

7 Ion-Dipole Interaction Energies for Alkali Halides KF, and According to Case A
Ion-dipole interaction energy
=
Alkali dipole Halide ion water dipole
Alkali -
halides Attraction Repulsion Attraction Repulsion
KF -62.61 23.67 -38.94 -71.13 22.91 -48.23 9.28

-58.52 22.43 -36.09 -71.13 21.4 -49.73 13.64
-53.77 20.88 -32.89 -71.13 19.66 -51.47 18.58
-53.77 15.54 -38.23 -51.45 15.64 -35.81 -2.42
-53.77 14.26 -39.51 -46.96 14.58 -32.38 -7.13
-53.77 12.54 -41.23 -40.91 13.0 -27.91 - 13.32
--
Source:

simply by adding the attractive and repulsive components. Table 7 presents the
results for interaction energies calculated for alkali ion-water dipole and halide
ion-water dipole systems. Finally, the ion-dipole interaction energy for an alkali
system can be estimated simply from a comparison of the
dipole interaction energies of the corresponding alkali dipole and halide
dipole systems. For example, if the interaction energy of the alkali
water dipole system is more negative than the interaction energy of the halide
water dipole system, then the partial hydration of surface lattice ions will cause the
surface to be more negative. The converse is also true. However, for case A, as is
evident from Table 7 , the difference between the ion-dipole interation energies of
the alkali dipole and halide dipole systems is So the
dipole interactions will have little influence on the sign of the surface charge as
predicted by the simplified lattice ion hydration theory.
Nine alkali halide and

have an alkali ion radius less than that of the water molecule and a ratio of
hydration numbers of alkali ion to halide ion of more than 2. These alkali halides
were classified into one group, and calculations based on steric and hydration
effects were made to analyze their surface charge. The steric effect, which is related
to the sizes of the alkali ion, halide ion, and water molecule, has a remarkable
influence on the partial hydration of these nine alkali halides. The other effect, the
hydration effect, which depends mainly on the ratio of alkali ion hydration number
to halide ion hydration number, has limited significance in this case due to the very
high steric effect. These characteristic features are presented in Table 6.
Figure 5 shows a schematic representation of case Table 8 presents the results
from the calculation of interaction energies for the alkali dipole and
halide dipole systems. Both attractive and repulsive components of inter-
action energy were calculated using Eq. (17 ) . The ion-dipole interaction energy for
any alkali system can be estimated simply from a comparison of the
ion-dipole interaction energies of the corresponding alkali and halide
dipole systems. As is evident from Table 8, the difference in ion-dipole
interaction energies between alkali dipole and halide dipole
Alkali ion
Attraction at distance CD and
at distance C
no
.5 Case suggested for and

alkali halides having an alkali ion radius less than that of the water molecule and a ratio
of hydration numbers of alkali ion to halide ion of more than 2. (From Ref. 9.)

Ion--Dipole Interaction Energies for Halides According to Case
interaction energy
Alkali dipole Halide dipole -
Alkali
halides Attraction Repulsion Attraction Repulsion
NaCI
Source: Ref. 9.

systems is positive. These calculations suggest that the halide ion has a more
attractive (more negative) ion-dipole interaction energy value than the alkali ion
due to the steric effect;the halide ion is much larger than the alkali ion. Finally, the
ion-dipole interaction energy values calculated for the nine alkali halide salts were
found to have a significant influence on the relative hydration tendency and hence
on the sign o f the surface charge predicted by the simplified lattice ion hydration
theory.
Five alkali halide and an alkali ion

radius close to that o f a water molecule and a ratio o f hydration numbers o f alkali
ion to halide ion o f more than 2. These alkali halides were classified into one group,
and calculations based on steric and hydration effects were made to analyze their
surface charge. The steric effect,which is related to the sizes o f the alkali ion, and
water molecule, has very little influence on the hydration o f these five alkali halides.
The other effect,the hydration effect,which depends mainly on the ratio o f alkali
ion hydration number to halide ion hydration number, is significant in this case.
These properties are presented in Table 6 .
Figure 6 shows a schematic representation o f case C . Table 9 presents the results
from the calculation o f interaction energies for the alkali dipole and
halide dipole systems. Both attractive and repulsive components o f inter-
action energy were calculated using (17). As discussed before, the ion-dipole
interaction energy for an alkali system can be estimated simply from a
comparison o f the ion-dipole interaction energies o f the corresponding alkali
water dipole and halide dipole systems. As is evident from Table 9 , the
difference in ion-dipole interaction energies between alkali dipole and
halide dipole systems is negative. These results suggest that the alkali ion
has more attractive (more negative) ion-dipole interaction energy values than the
halide ion since the hydration number o f the alkali ion is higher than that o f the
halide ion. The higher hydration number o f the alkali ion is responsible for the
accommodation o f a greater number o f water dipoles in the absence o f any steric
2 at distance and
Case C , suggested for and alkali halides having

an alkali ion radius close to that o f the water molecule and a ratio o f hydration numbers
of alkali ion to halide ion o f more than 2. (From Ref. 9.)

Ion--Dipole Interaction Energies for Alkali and Rbl According to Case C
Ion-dipole interaction energy
Alkali dipole Halide dipole
Alkali
halide 2 Rep. Rep. Rep. 2
KC1 -62.61 -23.8 16.87 -69.54 -51.45 17.13 10.27 -24.05 -45.49
-58.52 -19.41 15.94 -6 1.99 -51.45 16.15 19.74 - 15.56 -46.43
-62.61 -20.8 15.3 -68. I -46.96 15.73 21.48 -9.75 -58.36
-58.52 - 19.95 14.54 -63.93 -46.96 14.91 20.6 -1 -52.48
-58.52 -15.59 12.58 -61.53 -46.96 13.19 16.54 - 17.23 -44.3
Att = attraction; Rep = repulsion.
Source: 9.

effect. Finally, the ion-dipole interaction energy values calculated according to this
case for these five alkali halide salts were found to have a significant influence on
the relative hydration tendency and hence on the sign of the surface charge pre-
dicted by the simplified lattice ion hydration theory.
Table 10 compares the results obtained for the extended lattice ion hydration
theory with the results for the nonequilibrium electrophoretic mobility measure-
ments. It is evident that the extended lattice ion hydration theory provides for an
10 Sign of Surface Charge of Alkali Halides-Extended Lattice Ion

Hydration Theory
Sign of surface charge
Alkali +
halide Predicteda
Fluoride
KF
Chloride
Bromide
Iodide
'Extended lattice ion hydration theory.

electrophoretic mobility measurements by laser-Doppler electrophoresis
Source: Ref.

Miller and
excellent explanation of the experimental results. In particular, the addition of the

interaction energy values to account for the partial hydration of surface lattice ions
showed a significant effect on the total energy (gaseous ion hydration free energy
plus ion-dipole interaction energy) values of the seven exceptions
KI, and that, according to simplified lattice ion hydration theory,
had deviated from the results of electrophoretic measurements by
laser-Doppler electrophoresis. For example, which has a difference of -9.62
moi between cation and anion gaseous hydration free energies, was predicted to
have a negative sign of surface charge by the simplified lattice ion hydration theory.
However, by considering partial hydration of surface ions with an interaction energy
of according to surface ion-dipole interactions, the total energy is found
to be 51.08 and a positive sign is predicted for the surface charge, which is in
excellent agreement with the experimental data. These calculations indicate that the
sign of the surface charge of alkali halides except for KC1 can be accurately predicted
by the extended lattice ion hydration theory, which involves the additional con-
sideration of surface ion-dipole interactions.
T o further examine the validity of the above calculations, a sensitivity analysis
was carried out, particularly for the seven exceptions that had deviated from the
simplified lattice ion hydration theory. Slight changes in the angles and the con-
sideration of other possible interactions for the calculations of ion-dipole interac-
tion energies were the main features of the sensitivity analysis. In almost all cases
only a very small change in the energy values was observed, and, most important,
there was no change in the sign of the surface charge of the alkali halides except
that of From Table 10 it is evident that the energy value of KC1 is very small,
3.46 and this might account for the change in the sign of the surface charge
of KC1 from negative to positive during the sensitivity analysis. Further, it is
interesting to note that Yalamanchili and Miller showed that oxygen defects
present in crystal lattice can strongly influence the surface charge characteristics
of At a low oxygen concentration KC1 is positively charged, and
as the oxygen content increases, the sign of the surface charge changes from posi-
tive to negative. Therefore, the small energy values (gaseous ion hydation free
energy plus ion-dipole interaction energy) may allow for the energy associated
with oxygen defects to account for the anomalous behavior of general,
excellent agreement in the sign of surface charge is now observed between theore-
tical predictions and experimental data for the alkali halide salts in their saturated
brines. Further refinement in the calculation of ion-dipole interaction energies for
the partial hydration of surface ions may include consideration of different orienta-
tions for water dipoles, dielectric boundary effects, and dipole-dipole interactions.
In order to understand the anomalous behavior of the effect of oxygen defect

states on its surface charge was studied in detail by Yalamanchili and Miller
The possible types of defects that can be present in ionic crystals are many As
is well known, the most common types of defects that are always present in ionic
crystals are Frenkel and Schottky defects. A Frenkel defect forms when an ion

leaves the normal lattice site and occupies some interstitial position. A Schottky
defect forms when an ion leaves the lattice site and migrates to the surface or, more
simply, when a lattice site is vacant. These defects can be significant even in pure
crystals of ideal composition. It has been demonstrated that these defect states
exhibit substantial mobility in the crystal lattice. Two other types of defects that
can be present are dislocations and stacking faults, which are important from the
viewpoint of mechanical properties and crystal growth.
The presence of defects has been found to influence the bulk as well as surface
characteristics of alkali halides. Basically, these defects alter the charge neutrality
the cleavage planes, leading to a change in the surface properties of
valent alkali halides. In this regard, oxygen defects have been either introduced into
or removed from the crystal lattice in order to examine their influence on the
surface properties of KC1 and The KC1 and salts were selected for
these studies because they have been shown to be very stable hosts for various types
of defects and also represent the most important alkali halide salts that deviate
from the simplified lattice ion hydration theory.
As a first step, an ultrapure laser quality KC1 crystal was grown at the University
of Utah crystal growth laboratory. The Czochralski method was employed to
grow the KC1 crystal from a starting material with five nines purity. The properties
of the University of Utah laser quality KC1 crystal were compared to those of the
optical quality KC1 crystal obtained from Optovac, Inc. and the KC1 reagent grade
shown in Table It can be seen from this table that the University of
Utah KC1 crystal has a positive charge as predicted by lattice ion hydration theory,
whereas the Optovac KC1 crystal and KC1 reagent grade powder exhibit a negative
surface charge. The oxygen levels as determined by neutron activation indicate that
the Optovac KC1 has a higher oxygen content than the University of Utah as
Electrokinetic Properties of KC1 from Various Sources

KC1 Sign of surface Oxygena
source Crystal charge
U of crystal method; Positive 60 8.6
laser quality;
99.999% KC1 starting
material;
chlorine purging
Optovac crystal Not known; Negative 345 13
optical quality;
99.5% KC1 starting
material;
no purging
Reagent grade powder Not applicable Negative Not measured
Corporation, San
Source: Ref.

Miller and
shown in Table 11. The higher oxygen levels in Optovac KC1 appear to be due to
the use of relatively less pure KC1 as a starting material for crystal growth and the
crystal growth procedure employed. For example, if the crystal is grown under
ambient atmosphere, oxygen in the atmosphere could react with the KC1 melt,
forming various defect states that are discussed in the following paragraphs.
Unfortunately, Optovac, Inc. was unwilling to share information regarding their
crystal growth procedures. However, it is most likely that one or all the reasons
mentioned above could account for the presence of higher levels of oxygen in
Optovac KC1 crystals. Nevertheless, it seems that the difference in oxygen levels
in the University of Utah and Optovac crystals is responsible for the difference in
the electrokinetic behavior of these KC1 crystals.
Oxygen can be present in the crystal in two different forms, namely,
and centers If the crystals are grown in an ambient atmosphere, KC1 melt
reacts with oxygen to form centers) as follows:
When excess potassium is introduced into present within the lattice is

converted to centers) according to the reaction
Equation (19) can also be represented as
The symbol in (20) represents a vacancy in the crystal lattice. The introduc-
tion of excess potassium ions into the KC1 lattice creates F centers ( e at anion
vacancies) in the crystal lattice as mentioned in the previous section. Therefore
(20) can be rewritten as
It is clear from (21) that the number of defects will be reduced by the
presence of F centers in the crystal lattice, thereby forming the defect pairs.
The presence of and defects can be characterized by their typical
absorption frequencies in the region Figures 7 and 8 show the
absorption spectra of the Optovac KC1 and the University of Utah respec-
tively, in the region. The presence of defects in the Optovac KC1 can be
clearly seen from Fig. 7, as indicated by the band positions at 205 and 244 nm. On
the other hand, Fig. 8 shows that the University of Utah KC1 crystal is free of
defects, as indicated by the absence of any absorption bands in the region of
Fisher et reported that both and defects in KC1 have a
characteristic absorption band in the range of 200-250 It should be mentioned
that defects could easily be converted into defects if the crystal were
exposed to prolonged radiation even in the natural environment. However, neither
the University of Utah nor Optovac crystals show any characteristic absorption due
to defects. Typically, defects have multiple absorption bands a t 440, 286,
215, and and clearly these multiple bands are absent in Figs. 7 and 8,

.7Absorption spectrum of Optovac KC1 crystal showing the absorption due to
oxygen defects. (From Ref. 10.)
which indicate that the University of U t a h and Optovac crystals had not been
exposed to long-term radiation, as they are free of defects.
Therefore it can be concluded that the absorption bands in Fig. 7 for the
Optovac KC1 are due to the presence of defects. while Fig. 8 indicates that
the University of Utah KC1 crystal is free of any oxygen defects, which was
expected, as the University of U t a h crystal was grown with extreme care to
avoid oxygen contamination.
In any case, the spectra for Optovac KC1 and University of U t a h KC1
and their nonequilibrium electrophoretic mobility results suggest that defects
Absorption spectrum of the University of Utah KC1 crystal showing the

absence of absorption due to oxygen defects. Ref. 10.)

Miller and
strongly influence the surface properties and are responsible for the deviation of
normal from predictions of lattice ion hydration theory.
The concentration of oxygen defects was varied to examine their influence on the
electrokinetic properties of Optovac and reagent grade KC1 particles were
thermally treated in a nitrogen atmosphere containing less than 1 ppm oxygen in a
tube furnace. Similarly, the KC1 crystal grown at the University of Utah was
thermally treated in an oxygen atmosphere. Table 12 presents the results from
these thermal treatment experiments. It can be noted this table that when
the University of Utah particles were thermally treated at for about 27 h
in an oxygen atmosphere, the sign of the surface charge was reversed from positive
to negative due to oxygen incorporation into the crystal lattice. In the case of
Optovac and reagent grade KC!, the sign of the surface charge reversed from
negative to positive when these particles were thermally treated in the presence
of nitrogen having less than 1 oxygen. The sign of surface charge reversal in
the case of the University of Utah KC1 indicates that oxygen has diffused into the
crystal, as suggested by the increase in oxygen level from 61 to
147 9 ppm, thus accounting for the change in the electrokinetic behavior. These
results show that when KC1 is relatively free of oxygen defects, its electrokinetic
behavior is predicted according to lattice ion hydration theory. However, it is clear
from these studies that KC1 deviates from lattice ion hydration theory when oxygen
defects are introduced into the crystal lattice.
Similar observations can be made for both Optovac and reagent grade KC1
particles. Both Optovac and reagent grade KC1 (samples with high oxygen content)
exhibited a negative surface charge before thermal treatment. When these KC1
particles were thermally treated in the presence of nitrogen having less than
1 ppm oxygen, the sign of the surface charge of these particles reversed from nega-
tive to positive as expected from lattice ion hydration theory. Presumably, oxygen
present in the crystal has diffused out during the thermal treamtment due to low
oxygen pressure in the atmosphere above the bed of particles. In fact. the
Vis spectrum of a KC1 plate (Optovac, 5 x 5 x 1 mm), thermally treated at about
in presence of a nitrogen (oxygen 1 atmosphere, revealed that
oxygen indeed been removed from the KC1 plate. as indicated by the absence of
12 Thermal Treatment Experiments-Effect of Oxygen

Electrophoretic Oxygen
KC1 source mobility
As prepared 60k8.6
27 -0.26 0.17 147 9.0
As received -0.46 0.20 345 13.0
650, 27 0.16 Not
Reagent As received -0.42 0.19 Not
27 0.16 Not measured
Source: Ref. 10

characteristic absorption bands associated with defects in the range of
as shown in Fig. 9.
The change in the electrokinetic behavior of Optovac KC1 has been monitored
with respect to the thermal treatment time. It took approximately 30 h of thermal
treatment 1 ppm) for the change in the electrokinetic behavior of
to occur as shown in Fig. 10. These observations further substantiate the fact that
oxygen defect states are responsible for the deviation of KC1 from behavior
expected according to lattice ion hydration theory.
Based on these results, it seems that the deviation of KC1 from the sign of the
surface charge predicted according to theory can be explained in terms of defect
states present in the crystal lattice, which would account for the anomalous elec-
trokinetic behavior of In this regard, to further understand the effect of
oxygen defect states on the surface charge of alkali halides, NaCl was studied in
detail by Veeramasuneni et al. I]. Oxygen imperfections were intentionally cre-
ated in an NaCI single crystal in order to examine the influence of oxygen defects on
its electrokinetic properties. The surface charge of the NaCI crystal with oxygen
4000 ppm) grown at the University of Utah crystal growth laboratory was found
to be negative. On the other hand, optical quality NaCI crystal without oxygen was
found to be positively charged. These results suggest that oxygen defects deliber-
ately introduced into positively charged NaCl can reverse the surface charge.
Bubble attachment studies also show that the results for the Optovac NaCI
(positively charged) were quite opposite to those observed for the University of
Utah NaCI (negatively charged).
200
WAVELENGTH
Absorption spectrum of Optovac KC1 thermally treated at in a nitrogen

(oxygen 1 ppm) atmosphere for 30 (From Ref. 10.)

42 Miller and
FIG. 10 electrophoretic mobility of Optovac thermally treated at

650Â° in a nitrogen (oxygen 1 atmosphere as a function of time. (From Ref. 10.)
In any case, spectra for University of Utah NaCI and Optovac NaCl
and the results from nonequilibrium electrophoretic mobility measurements suggest
that the negative surface charge of the University of Utah NaCI is due to the
presence of defects. It has been demonstrated that oxygen defects can be
created in alkali halides to them negatively charged, thus confirming the
previously reported results for oxygen defects in These results demonstrate
that oxygen defect states in NaCI and KC1 crystal lattices can change the sign of the
surface charge and alter the flotation response of these salts and that of other alkali
halide salts is expected.
Recent contributions have clearly improved our understanding of the fundamentals

of various phenomena related to the surface charge of alkali halides in their
rated solutions. For the first time, nonequilibrium electrophoretic mobility mea-
surements for alkali halide salts have been made by laser-Doppler electrophoresis.
The measurements, which were made during the initial stages of salt dissolution,
were found to be reproducible, and their utility was confirmed by measurement of
the corresponding equilibrium electrophoretic mobilities for and More
important, the nonequilibrium electrophoretic mobility measurements allow for the
sign of surface charge to be predicted for alkali halides in their saturated solutions.
In general, the results can be explained with the simplified lattice ion hydration
theory.

The necessity of considering the partial hydration of surface lattice ions in the
analysis of the surface charge of alkali halides has been demonstrated. The seven
exceptions KI, and whose surface charge, based
on simplified lattice ion hydration theory, deviated from experimental measure-
ments were found to be in good agreement with the extended lattice ion hydration
theory, which accounts for partial hydration of surface lattice ions by consideration
of ion-dipole interaction energies. Sensitivity analysis further confirmed the validity
of this Thus, the sign of the surface charge of alkali halides predicted
from extended lattice ion hydration theory, which considers partial hydration of
surface lattice ions. together with the simplified lattice ion hydration theory was
found to be in excellent agreement with electrophoretic measure-
ments by laser-Doppler electrophoresis.
The deviation of the electrokinetic behavior of KC1 from that predicted by
lattice ion hydration theory has been found to be due to the presence of oxygen
defect states in The surface charge of a laser quality KC1 crystal with very low
oxygen content is positive and conforms to the behavior expected according to
lattice ion hydration theory. On the other hand, optical quality and reagent
grade particulate KC1 crystals with relatively high oxygen content are negatively
charged and d o not conform to the behavior expected according to lattice ion
hydration theory. Thermal treatment of these KC1 crystals in appropriate atmo-
spheres showed that the sign of the surface charge is controlled by the oxygen
content of the KC1 crystal. Experimental results suggest that for a n oxygen content
in excess of about 150 ppm, deviation of from behavior expected according to
lattice ion hydration theory should be anticipated. Finally, the results suggest that
oxygen defects can be created in alkali halides to make them negatively charged as
demonstrated for NaCI and KC1 salts.
Support by the DOE Basic Science Division, grant DE-FG-03-93ER14315, for this
work is gratefully acknowledged. Thanks are also extended to Dr.
Yalamanchili and D r . Yuehua H u for their suggestions and helpful discussions.
1. Miller, and C. Chemistry of Flotation,

AIME, CO, 1985.
2. Verwey and Overbeek, Theory of the Stability of
Colloids, Amsterdam, 1948.
3. Pugh, in Colloid Chemistry in Mineral Processing (J. Laskowski and J.
Amsterdam, 1992, p. 243.
4. Roman, Fuerstenau, and Trans. AIME (1968).
5. J. Roman. Ph. D. Thesis, Colorado School of Mines, Golden, CO, 1956.
6. Pizarro, Thesis, Colorado School of Mines, Golden, CO, 1967.
7. Miller, Yalamanchili, and Kellar, Langmuir (1992).
8. Yalamanchili, Ph. Thesis, University of Utah, Salt Lake City, 1993.

Veeramasuneni, and Miller, Surf, (1997).
Yalamanchili and Miller, Colloid Interface Science (1994).
Veeramasuneni, M. Yalamanchili, and Miller, J. Colloid Interface
(1996).
and Readett, Adv. Colloids Interface (1987).
Salman, and Colloid Interface (1979).
Colloids Surf. (1988).
and G . Agar, in Froth Flotation-50th Anniversary Volume (M.
Fuerstenau, AIME, New York, 1962, p. 91.
Aplan and Fuerstenau, in Froth Flotation-50th Anniversary Volume
van Olphen, An Introduction to Clay Colloid Chemistry, Interscience, New
York, 1963, p. 90.
I, Iwasaki, Cooke, and A. Colombo, U.S. Bureau of Mines, RI 5593
Department of the Interior, 1960, p.
I. Iwasaki, Cooke, and S. Choi, Trans. AIME (1960).
I. Iwasaki, Cooke, and S. Kim, Trans. AIME
Miller and J. V. Calara, in Flotation, A. Memorial Volume (M. C.
S. Veeramasuneni, and Miller. Colloid Interface Sci.
(1997).
Calara and Miller, Phys. (1976).
V. Calara and Miller. in Colloid and Interface Science, Proceedings of
International Conference-50th Colloid and Surface Science
Academic, New York, 1976, 157.
P. Hunt, Metal Ions in Solutions, A. Benjamin, New York, 1965.
Marcus, Ion Salvation, Wiley, New York, 1985, p. 123.
A. Thermodynamics of Solvation: Solution and Dissolution, Ions and
Solvents, Structure and Energetics, Ellis Horwood Phys. Ellis
New York, 1991, p. 106.
Hunter, Zeta Potential in Colloids Academic, New York, 1981.
A. Physical Chemistry of Surfaces, 4th Wiley. New York, 1982.
Anon., Instruments User's Manual, 1990, Issue 2.
Hancer, Fuerstenau. and Miller, Proceedings of the X X
International Mineral Processing Congress GDMB,
Germany, 1997, p. 715.
J. and Surface Forces, 2nd Academic, New
York, 1992, p. 235.
W. Fyfe, Geochemistry of Solids, New York, 1964, p. 151.
in Crystal Growth: A Tutorial Approach Bardsley,
and J. eds.), North-Holland. New York, 1979, pp. 1-16.
and Z. Phys. (1966).
Gellerinan, and Luty, Appl. Phys. (1987).
Luty, in Physics of Color Centers (W. Fowler, Academic, New York,
1968, p. 182.

Department of Chemical Engineering, Washington
State University, Pullman, Washington
I. Introduction
I I . Historical Perspective
III. Structure Within the Adsorbed Layer
A. Experimental verification
B. Observations of hydrophobic interactions
IV. Adsorption with Attractive Electrostatic Interactions
V. Adsorption with Repulsive Electrostatic Interactions
VI. Adsorption of Mixed Surfactants
VII. Predictive Models of Surfactant Adsorption
VIII. Summary
References
Surfactants are used in a wide variety of applications such as ore flotation, cleaning,
polymerization processes, and pharmaceuticals and agriculture. The usual role for
the surfactant is to modify interfacial properties, whether they be
or interfaces. T o be effective in any of these applications,
however, the surfactant must adsorb strongly at the interface. In addition to its
concentration at the interface, the conformation of the surfactant at the interface is
also an important factor. While the influence of solution properties such as con-
centration, ionic strength, and pH on adsorption are well known, the
properties of the other phase also exert a significant influence.
Perhaps the most varied combinations occur in solid-liquid systems because of
the variety of properties possessed by the solid phase. The most obvious property of
the solid phase to affect surfactant adsorption is the or hydrophobic

character of the surface. But even within either of these groups, variations in the
polarity of the surface have also been found to exert a significant influence on
adsorption. Since solid surfaces may also possess an electric charge, the electrostatic
interactions between the surface and the surfactant are a third major factor influen-
cing both the amount absorbed and the conformation of the absorbed material.
This chapter reviews results for the adsorption of ionic surfactants onto hydro-
phobic surfaces. Even with this limitation there are many possible combinations.
Both anionic and surfactants are considered. The hydrophobic surfaces
may be either uncharged or positively or negatively charged. Thus, the electrostatic
interactions between the adsorbing surface and the be either non-
existent, attractive, or repulsive. Finally. the adsorbing materials may be either a
single species or a combination of surfactants. This review is not a comprehensive
overview of all of the published material on this topic. Rather it presents represen-
tative results concerning the adsorption of ionic surfactants onto hydrophobic
surfaces.
An early investigation of the adsorption of ionic surfactants onto hydrophobic

surfaces is the work by et on soap titration Although conceived as
a method for the measurement of the average particle size of a synthetic latex, soap
titration begins to shed some light on adsorption phenomena. The surface tension
of a of the suspension to be analyzed is measured as a function of the
amount of surfactant added. This analysis is repeated for a second sample that
contains a different concentration of suspended solids. The difference between the
amounts of surfactant needed to reach saturation, as measured by reaching the
critical concentration (cmc), for the two samples must be the amount of
surfactant needed to saturate the additional surface area present in the sample with
the higher suspended solids concentration. If the surface area occupied by a single
surfactant molecule at saturation is a constant, this amount of surfactant can be
converted to latex surface area and ultimately to an average particle size.
Following the work by and coworkers, a number of publications
appeared investigating various aspects of the technique and therefore also of
adsorption. Brodnyan and Brown [2] used surface coverage areas for both
anionic and nonionic surfactants measured independently of the soap titration to
determine the average particle size of a latex. This was compared with the average
particle size measured by ultracentrifugation and found to agree. Brodnyan and
Brown took this as a verification of the technique and also as proof that the
assumption of monolayer coverage by the at saturation was valid. This
investigation also demonstrated that nonionic surfactants could be used as well as
anionic A number of other investigations also appeared seeking
improvements in the technique (improved detection of the cmc corrections
for the of surfactant dissolved in the aqueous phase but none ques-
tioned the basic underlying assumptions.
Although other investigators had verified that the adsorption area of the
tants was a constant independent of the adsorbing surface this assumption

began to be questioned. Paxton [6] reported that the adsorption area for sodium
dodecylbenzene sulfonate varied depending not only on the ionic strength
of the solution but also on the polymer forming the latex particles. Piirma and Chen
also observed a variation in the adsorption area of two anionic surfactants
adsorbed onto polystyrene or methacrylate) latices. In addition, they
also observed a steady increase in the adsorption area for sodium dodecyl sulfate
(SDS) as the composition of a copolymer latex went from 100% styrene to
methyl methacrylate (from less to polar). Another important observation was
the effect of adding solvents for the polymer to the system. When various solvents
for the polymer forming the latex particles were added, the adsorption area for
both potassium and sodium dodecyl sulfate increased, leading to the sugges-
tion that a solvent-swollen particle surface allowed greater penetration of the
tail into the particles.
The observation that the adsorption area for a depended not only on
the surfactant but also on the adsorbing surface was further confirmed in a number
of investigations. Vijayendran [8] studied the adsorption of sodium dodecyl sulfate
onto a number of different polymeric latices and onto various
interfaces. In so doing he was able to demonstrate the impact of the polarity of the
interface in determining the adsorption area of the surfactant. By
assuming a Langmuir type of adsorption isotherm, Vijayendran was able to sepa-
rate the polar from the dispersive contributions to the polymer surface energy.
In a series of articles, Kronberg and coworkers studied the adsorption of
ionic surfactants onto a polymeric latex. A approach was used to
analyze the adsorption of a series of
surfactants onto a polystyrene latex that indicated that the driving force for
adsorption was primarily due to the surfactant structure (EO chain length)
Extending this work, Kronberg and Stenius used the thermodynamic
model they had developed to analyze the adsorption of
lene surfactants of varying chain length. From this analysis they determined
that adsorption measurements could be used to determine the latex polarity and
that for this situation, 80% of the adsorption free energy was due to the structure of
the surfactant, with the remaining 20% arising the characteristics of the
adsorbing surface. They also postulated that the adsorbed surfactant would be
oriented with the hydrophobic moiety pointed toward the polymeric latex surface.
This allowed surfactant-surfactant interactions to replace some of the
hydrophobic moiety interactions, thus reducing the overall free energy and leading
to adsorption. Finally, Kronberg examined the applicability of the Langmuir
adsorption isotherm for surfactant adsorption and found that two fundamental
assumptions in the Langmuir isotherm were violated: (1) There were significant
interactions between adsorbed species and (2) it is not possible to have a
surface contact replaced by a surfactant-surface contact due to the differences in
the sizes of the two species. However, the effects of these two differences work in
opposite directions so that typical isotherms do indeed appear to follow
the Langmuir model.
The result of these investigations leaves the following view of adsorption of ionic
surfactants onto hydrophobic surfaces. Measured adsorption isotherms (Langmuir
type) and comparisons between particle sizes as measured by soap titration and

other techniques led to the conclusion that adsorption leads to monolayer coverage
by the surfactant. This monolayer is adsorbed on the surface with the hydrophobic
moiety of the surfactant pointed toward the hydrophobic surface and the
philic moiety pointed toward the aqueous phase. While the structure of the surfac-
plays a major role in determining the amount adsorbed (or adsorption area per
molecule) the characteristics of the adsorbing surface also play a signifi-
cant role in the adsorption process. Various investigators have noted the possibility
of cooperative adsorption phenomena resulting from hydrophobe-hydrophobe
interactions between adsorbed surfactant molecules. Many questions remain unan-
swered, however. Some of these are: What is the structure of the adsorbed surfac-
tant layer? Can this structure be explained by known colloidal phenomena? What is
the effect of electrostatic interactions between the adsorbing surface and the
What happens when there is more than one surfactant present? Can the
adsorption phenomena be modeled for all of the possible sys-
tems?
STRUCTURE WITHIN
Intuitively one would expect that a surfactant would adsorb onto a hydrophobic
surface with its hydrophobic portion pointed toward the hydrophobic surface and
the hydrophilic portion pointed toward the aqueous phase. This adsorption is
driven by the "hydrophobic attraction" between these two hydrophobic moieties.
For a hydrophilic surface, adsorption should take place with the opposite orienta-
tion, as observed by Furlong and
When ionic surfactants adsorb onto an uncharged, hydrophobic surface, the
charge at the solid surface increases, giving the surface a more hydrophilic char-
acter. As the adsorption continues, the concentration of charged groups a t the
outer edge of the adsorbed layer becomes great enough to block any further
adsorption due to the electrostatic repulsions between the adsorbed surfactant
and the surfactant molecules approaching the surface. This is opposite the expecta-
tion for adsorption onto hydrophilic surfaces In this latter situation. surfac-
are expected to first adsorb with the hydrophilic portion in contact with the
solid surface, frequently as a result of electrostatic attractions. This leaves the sur-
face covered with an adsorbed layer with its hydrophobic ends exposed and thus
looking very much like a hydrophobic surface. Subsequent adsorption will then
occur onto this adsorbed layer, with the hydrophobic ends of the next layer of
oriented toward the first layer. The final result, at saturation, is an
adsorbed bilayer at the surface. Further adsorption is blocked by electrostatic
interactions between the hydrophilic ends of the
While data many adsorption experiments would support the

assumption that adsorption occurs, direct measurements of the
adsorbed layer structure were not reported until recently. Herder used a com-
bination of contact angle measurements and surface force measurements to show
that this is the situation for the adsorption of oxide onto a mica surface

made hydrophobic by the deposition of dimethyldioctadecylammonium bromide
using Langmuir-Blodgett deposition. Using infrared-visible sum-frequency spec-
troscopy. Ward studied the adsorption of dodecanol onto an uncharged
hydrophobic surface (a monolayer of octadecanethiol on gold). By using an
uncharged surface-active material and a compact monolayer of hydrophobic mate-
rial, this experiment ensures the lack of any electrostatic interactions between the
adsorbing species and the surface. Their results show that the dodecanol is
adsorbed with the terminal methyl group of the dodecanol oriented toward the
hydrophobic surface. When similar experiments are conducted using dodecanol
dissolved in an organic solvent, the dodecanol adsorbs onto the surface with the
alkyl chain lying parallel to the surface.
In a study of the adsorption of sodium sulfate onto graphite,
and Ducker found that the situation was more complex. Using atomic force
microscopy, they were able to observe periodic structures developing on the surface
with increasing surfactant concentration. For low bulk concentrations of surfactant
(0.35 cmc), the adsorbed material was observed to be At higher bulk
concentrations, periodic ridges, assumed to be hemicylinders, were observed. As the
surfactant concentration increased still further, no change in the morphology was
observed, although the spacing between the ridges decreased. These observations
were counter to those observed for an analogous double-chain surfactant where the
adsorbed layer remained homogeneous at all bulk surfactant concentrations.
Using ellipsometry, was able to measure adsorbed film thickness and
use these measurements to deduce orientations for a series of
glycol alkyl ethers adsorbed onto either a clean silicon surface (hydrophilic) or a
silicon surface treated with dimethyloctylchlorosilane (hydrophobic). When the
ethers were adsorbed onto the hydrophilic surface, surface micelles and
were observed. On the uncharged hydrophobic surface, however, submonolayer
coverage increasing to coverage was observed. At low coverage the
adsorbed surfactants were observed to tilt toward the surface, a tilt that decreased
continuously with increasing surface coverage. The claim was also made that the
adsorption onto a hydrophobic surface appeared to be a noncooperative process.
The observation that monolayer coverage results when surfactants are adsorbed
onto a hydrophobic surface was also confirmed by Waltermo who
studied the adsorption of alkyl glucosides onto a mica surface that had been ren-
dered hydrophobic by using a Langmuir-Blodgett technique to deposit a layer of
dimethyldioctadecylammonium bromide onto the surface.
The tendency for a hydrophobic material to adsorb onto a hydrophobic surface is

frequently described in terms of a "hydrophobic interaction (attraction)." While
arguments for the presence of such a phenomenon are easy to make, the
fact that long-range attractive forces between hydrophobic surfaces have been
observed cannot be accounted for by typical DLVO theory A number of
explanations of phenomenon have been advanced, such as enhanced hydrogen
bonding, electrostatic correlation forces, film stability, dissolved gases, and density
fluctuations, but none has been successful in explaining all of the features of this

force that have been observed. Kekicheff and conducted a series of
experiments in a surface force apparatus using carefully degassed samples but
still observed a long-range attractive force, thus indicating that dissolved gases
are not the underlying cause of the attractive force. They also measured the refrac-
tive index between the surfaces in the same apparatus using multiple-beam inter-
ferometry and found an absence of density fluctuations, indicating that these too
are not the source of the attractive hydrophobic forces.
Of interest in adsorption phenomena, however, is the work of Tsao et In
their paper the hypothesis is made that the long-range attractive force arises from
electric fields generated by lateral structures present on the hydrophobic surfaces.
Furthermore, such a field is generated on one surface, with the second surface
responding to the field of the first surface through its Lateral inter-
actions are also important when considering adsorption phenomena, since they are
often cited as giving rise to enhanced adsorption through associations between the
hydrophobic portions of the adsorbed surfactant molecules. These interactions may
become stronger when a molecule is adsorbed, as van Lent et reported that
the interactions between polymeric species within grafted polymer layers are stron-
ger than the interactions that occur between molecules within a grafted layer and
free polymer molecules.
These results indicate that the presence of mono- or bilayer coverage depends on
both the surfactant and the adsorbing surface. Within these layers there is an
ordering of the adsorbed molecules due to lateral hydrophobic interactions. The
presence of the adsorbed layer gives rise to a long-range attractive force. This force
not only affects other nearby surfaces but could also attract surfactant molecules.
In all of these studies the adsorbing surface has been electrically neutral. In the
following sections, electrostatic effects on adsorption are considered.
A number of adsorption studies have been conducted in which an ionic surfactant

has been adsorbed onto a hydrophobic surface containing bound electric charges of
the opposite sign. In this situation the first surfactant molecules adsorbed are
assumed to adsorb via the attractive electrostatic forces present and thus to adsorb
with their hydrophilic head groups pointed toward the surface. Since the hydro-
phobic tails will thus extend toward the aqueous phase, the particle becomes pro-
gressively more and more hydrophobic, as a result of both the hydrophobic tail
orientation and the neutralization of the surface charges. Ultimately a second layer
of surfactant starts to adsorb, this time with the hydrophobic portion of the mole-
cule interacting with the hydrophobic portion of the electrostatically bound
and the head groups pointing toward the aqueous phase. This type of
behavior was observed by and Ottewill for the adsorption of
surfactants onto a polystyrene latex containing carboxyl groups on the surface and
by for a series of trimethylammonium bromides onto polystyrene latices.
The amount of adsorbed was determined by measuring the amount
remaining in solution after equilibration with the latex, using liquid scintillation

counting. The electrophoretic mobility of the particles was also measured at the
same time. When the surfactants were adsorbed onto an uncharged poly-
styrene latex, a Langmuir-type isotherm was observed. When adsorbed onto a
charged surface, however, a well-defined "knee" was observed in the isotherm. A
comparison of the adsorption results with the electrophoretic mobility measure-
ments showed that the knee occurred at the point where the surface charge of the
particles reversed. Thus they concluded that up to the knee in the isotherm, adsorp-
tion occurred via ionic interactions between the surfactant head group and the
bound surface charges. Above the knee the adsorption isotherm closely resembled
that observed for adsorption onto an uncharged latex surface. The fact that the
adsorbed concentration was affected by the ionic strength of the aqueous phase
also led and Ottewill to suggest that lateral interactions between the
adsorbed surfactant molecules were present and that the adsorbed molecules
were not in a vertical orientation at low ionic strength even when the
surface was saturated.
An investigation similar to that reported by and Ottewill was recently
published by Gonzalez et al. except that in this latter investigation anionic
surfactants (alkyl were adsorbed onto a polystyrene latex.
Again the amount of adsorbed was determined by measuring the differ-
ence in aqueous phase surfactant concentration before and after equilibration with
the latex (via the methylene blue method) along with measurements of the electro-
phoretic mobility of the latex particles. Again a jump in the adsorption isotherm
was observed at the point of charge reversal for the latex particles, which was
attributed to a changeover in adsorption from a head-down orientation due to
head group-surface group electrostatic interactions to a head-outward orientation
due to hydrophobic interactions between the surfactant tail group and either the
particle surface or previously adsorbed material. Unlike the investigation by
and Ottewill, however, this investigation found higher concentrations of
the anionic surfactants adsorbed at the latex particle surface. The investigators
explained this by hypothesizing a mechanism for adsorption. They
hypothesized that the first molecules to adsorb may d o so either by electrostatic or
hydrophobic interactions and by lying flat on the latex surface. At higher surface
coverages the adsorbed surfactant molecules begin to move to a partially vertical
orientation with the hydrophilic groups pointing toward the aqueous phase. As the
cmc was approached there appeared to be interactions between the adsorbed
molecules, finally leading to a leveling off of the adsorbed amount.
Zhao and Brown in a more recent pair of articles studied the adsorption
of both (alkyltrimethylammonium bromide) and anionic (sodium
sulfate) surfactants onto polystyrene latex particles modified to have aromatic
amino groups on the surface. Conductimetric titrations of the latex particles indi-
cated that they possessed both groups (the amino groups) and anionic
groups (sulfate residues from the initiator). The amount of surfactant adsorbed
was determined by surface tension measurements (drop weight method), while the
conformation of the adsorbed surfactant was inferred from measurements of the
hydrodynamic diameter by dynamic light scattering. both cases, arguments are
made for including the effect of the roughness of the particle surface in considering
adsorption phenomena. The ionic groups on the polystyrene chains are known to

impart a surface-active character to the chains that results in a "hairy" surface
rather than a hard, billiard ball-like surface. After examining the changes in the
hydrodynamic diameter of the latex particles, Zhao and Brown postulate that the
initial adsorption is a combination of both a hydrophobic interaction between the
polystyrene chains in the hairy layer and the hydrophobic tails of the surfactants,
and either a hydrogen bonding interaction (for or an electrostatic interaction
(for the surfactants) between the surfactant head group and the surface
group of the opposite charge. At higher surfactant concentrations, the adsorption
process then becomes dominated by hydrophobic associations, ultimately resulting
in globular micelles strung along polystyrene chains that extend out from the sur-
face due to the electrostatic repulsions between head groups of the surfactants
adsorbed along the chain.
The assertion that both hydrophobic and electrostatic (or hydrogen bonding)
interactions occur during the initial phases of adsorption appears to arise because
the amount of surfactant adsorbed at the point where their adsorption isotherms
show a break is higher than the number of oppositely charged groups at the particle
surface. This does not agree with the observation by and Ottewill of a one-
to-one correlation between initial surfactant adsorption and surface groups of the
opposite sign. Zhao and Brown also claim that the adsorbed amount continued to
increase even after the cmc had been reached, which could be explained by having
globular micelles attached along an extended polymer chain. Again, this counter
to many other investigations where the final plateau adsorption level is usually
observed to occur very near the cmc. It is also interesting that the maximum
amount of surfactant adsorbed in any of the results shown in the two papers by
Zhao and Brown corresponds to an adsorption area of between 0.20 and 1.0
per molecule. These values are not significantly different from the adsorption areas
at saturation reported for SDS and ions, which are typi-
cally in the 0.20-0.40 range. Considering that the cross-sectional area
of a chain has been reported to be 0.23 nm 2, the adsorption area reported
by Zhao and Brown would appear to argue more for monolayer coverage rather
than for such structures as globular micelles strung along an extended polymer
chain.
The potential for tail-to-tail interactions in the adsorbed surfactant was clearly
demonstrated in an article by Kandori et They studied the adsorption of
sodium dodecyl sulfate, bromide (CTAB), and sodium
(Aerosol onto both an anionic and a
ionic latex. They found, as did and Ottewill, that a one-to-one relationship
existed between adsorbed amount and surface charge for the case of adsorption of
an ionic surfactant onto a surface of opposite charge. They also determined,
through zeta potential measurements, that for both CTAB and Aerosol OT there
appeared to be two surfactant molecules adsorbed to the tail of each molecule
adsorbed in the head-down conformation. This is interesting in view of the fact
that Aerosol OT is a two-tailed thus giving rise to twice as many tails
for subsequent adsorption when it is adsorbed in the head-down conformation. The
result by Kandori would thus indicate that the limiting factor for the second layer
of adsorbed material is not the number of tails available but is perhaps the inter-
actions between the head groups on this second adsorbed layer.

For the situation where both the ionic surfactant and the adsorbing surface have
charged groups of the same sign, there will be a repulsive electrostatic force tending
to prevent the surfactant from adsorbing onto the surface. Any adsorption must
then take place via hydrophobic interactions between the surface and the hydro-
phobic portion of the surfactant. Since adsorption must occur through the hydro-
phobic portion of the surfactant. the expected conformation of the adsorbed
surfactant molecule would have the head group pointing toward the aqueous
phase. This conformation would seem to preclude multilayer coverage, giving
rise instead to an adsorbed monolayer. This is the conclusion reached by
Keesom et as a result of studying the adsorption of sodium dodecyl sulfate
onto a polycarbonate surface. The results of this investigation show that the
adsorbed materials are oriented with the hydrophobic end toward the surface
and the head groups oriented toward the aqueous phase with a surface coverage
of less than a full monolayer. The observation that less than monolayer coverage
may occur was also made by Becher et al. for the adsorption of sodium
dodecylbenzene sulfonate onto atrazine and cyanazine, two hydrophobic crystalline
herbicides.
In contrast to these observations are the results reported by Brown and Zhao
for the adsorption of sodium dodecyl sulfate onto polystyrene and
styrene-butadiene latex particles. In these investigations the amount of
adsorbed is determined by measuring the surfactant concentration in the aqueous
phase before and after equilibration with the latex via surface tension measure-
ments (drop weight method), while changes in the hydrodynamic diameter were
measured via dynamic light scattering. They continue to report two-stage adsorp-
tion isotherms, although this result has not been reported by other investigators. As
with their reports for adsorption of a surfactant onto a surface containing a group
of opposite charge, Brown and Zhao report that the initial adsorption is via hydro-
phobic interactions between the surfactant tail group and the latex particle surface.
As adsorption proceeds, however, Brown and Zhao report that the adsorbed
molecules begin to form micelles and hemimicelles at the sites of the pre-
viously adsorbed species. Computing the adsorption area at saturation from their
results yields values between 0.30 and Since this is very close to
the results for monolayer coverage by sodium dodecyl sulfate reported by many
other investigators, the claim of multilayer adsorption seems unlikely.
The most complete studies of surfactant adsorption onto hydrophobic surfaces
containing the same charge are those by et Hoeft and Zollars and
Brouwer and Zsom In all of these studies a series of polystyrene latices were
used with varying surface charge densities. The adsorption of sulfates, usually
sodium dodecyl sulfate, onto these latices then allowed the direct study of the effect
of the electrostatic interactions between the surfactant and the surface. In these
studies the amount of sodium dodecyl sulfate adsorbed at saturation was found to
decrease with increasing surface charge density of the same sign, as shown in Fig. 1.
Unlike the investigation reported by Brown and Zhao, the adsorption isotherms
were all Langmuir-type isotherms, indicating the presence of only monolayer

-Surface Charge Density
FIG. 1 Amount of sodium dodecyl sulfate adsorbed onto polystyrene latex surfaces as
a function of surface charge density. Hoeft et Hoeft and
Zollars Brouwer and Zsom Gwin
age. In this situation the decrease in adsorbed amount is easy to explain. As

is adsorbed, the surface charge density of the particle will increase due to a
combination of the groups bound to the surface and the head groups on the
adsorbed surfactant molecules. As the surface charge of the same sign increases
there are increasingly stronger electrostatic repulsions between the surfactant
and the adsorbing surface acting to prevent the adsorption from occurring. In
fact, as observed by when the surface charge on the latex particle surface
becomes large enough, no observable adsorption takes place. Similar changes in
adsorbed amount with surface charge density were also reported for octyl
sulfate and sodium sulfate in the investigation by Hoeft and
study also showed that the amount of surfactant adsorbed at a constant surface
charge density increased with increasing alkyl chain length of the surfactant as
would be expected for increasingly more hydrophobic species.
When all of the data for the adsorption of sodium dodecyl sulfate onto a poly-
styrene latex particle containing bound sulfate groups is examined (Fig. the
relationship between adsorbed amount and surface charge appears to be more
complex than described above. Except for an offset between the data from
Brouwer and Zsom (possibly due to inaccuracies in the measurement of the surface
charge density resulting from not pretreating the latex with a dilute acid), the
amount of sodium dodecyl sulfate adsorbed appears to first increase with increasing
surface charge density followed by a decrease, a region of relatively little change

from -0.03 to followed by a further decrease with higher surface
charge densities. and Stein also observed that surface charge density
had little effect on the adsorption of sodium dodecylbenzene sulfonate onto poly-
styrene latex particles for surface charge densities between 0.02 and
These changes have been explained as arising from changes in the adsorbed
tant conformation resulting from the electrostatic interactions.
groups are first introduced onto the latex surface, the electrostatic repulsions will
cause the adsorbed molecules to change to a more extended conforma-
tion to reduce the electrostatic interactions with the particle surface. This results in
two changes. First, more of the polystyrene surface will be exposed, allowing for
further surfactant adsorption. Second, with fewer surfactant hydrophobe-particle
surface contacts, the potential for more hydrophobic interactions between the
molecules is increased. which can also increase surfactant adsorption. These
rearrangements of the adsorbed surfactant molecules could give rise to increases in
the amount adsorbed or act to counter the tendency for reduced adsorption arising
from increased electrostatic repulsions.
F MIX
Since hydrophobic interactions appear to play a significant role in determining
surfactant adsorption it is possible that these interactions may also alter the manner
in which a mixture of surface-active materials might adsorb. In a very early inves-
tigation, Orr and found that when mixtures of anionic and nonionic
surfactants were adsorbed onto a polymeric latex particle there was a decrease in
the adsorption area per molecule for the anionic surfactants. Hay et how-
ever, observed a decease in total adsorption for sodium dodecyl sulfate and
octyltetraethylene glycol onto graphitized carbon.
In general, however, it is observed that the presence of a nonionic surface-active
material will increase the adsorption of ionic surfactants. This was directly
observed using in situ sum-frequency spectroscopy for the adsorption of sodium
dodecyl sulfate and by et where more highly ordered films
are observed for dodecanol-sodium dodecyi sulfate mixtures than for pure sodium
dodecyl sulfate. The reason for this is usually given as a screening of the electro-
static interactions between the head groups on the sodium dodecyi sulfate by the
nonionic material, allowing for a closer approach by the head group while still
maintaining strong hydrophobic interactions within the adsorbed layer. This, in
turn, allows more compact adsorbed layers and greater adsorption.
A possible reason for the decreased adsorption of some nonionic materials has
been suggested by et In a study of the adsorption of sodium
dodecyl sulfate and cellulose onto hydrophobized silica sur-
faces, they found that the amount of adsorbed cellulosic derivative was reduced
when sodium dodecyl sulfate is also adsorbed. A possible reason for this is again
the effect of hydrophobic associations. The sodium dodecyl sulfate, being more
surface-active, will tend to adsorb onto the hydrophobic surface more strongly. It
will also associate with the cellulose in solution, rendering this
nonionic material more With both the adsorbing surface and the

nonionic material becoming more hydrophilic in presence of the sodium
sulfate, the driving force for adsorption is reduced.
For much bulkier surfactants the orderly arrangements observed for simpler
surfactants appear to disappear. Johal et used sum-frequency spectroscopy
to observe the coadsorption of a series of trichain surfactants and dodecanol. Their
data indicated that the best model for the coadsorption was one consisting of an
ordered dodecanol with scattered monomers or small clusters of dis-
ordered trichain molecules.
Hoeft also studied the cooperative and competitive adsorption of ionic
surfactant mixtures onto hydrophobic surfaces. When shorter chain
tants (sodium octyl sulfonate and sodium decyl sulfonate) are adsorbed, the decyl
will displace the octyl surfactant. For mixtures of sodium dodecyi sulfonate and
sodium octyl sulfonate, however, there appears to be an association between the
surfactant molecules leading to enhanced adsorption of the sodium dodecyl
nate with no depletion of the octyl sulfonate adsorption. This is shown in Fig, 2,
where the lines indicate the expected adsorption determined using a two-component
Langmuir adsorption isotherm with the adsorption parameters determined by ana-
lyzing the data from adsorption of each species individually. Also shown in Fig. 2 is
the concentration of surface-active materials in the aqueous phase at equilibrium.
In each of these experiments the total molar concentration and amount of
tant solution added to the latex was a constant, as was the amount of latex. Thus a
lower value for the bulk concentration corresponds to greater adsorption.
10
.
5
-0
0)
Adsorption of sodium octyl sulfonate and sodium dodecyl sulfonate

onto polystyrene latex particles, bulk surfactant concentration, , (- - -) value
predicted Langmuir model.

The investigations reported above generally rely on established models, such as the
isotherm, for the analysis of their data. The difficulty with this approach
is that each model contains parameters that are specific to the
solution-temperature combination used during the experimentation. This makes it
virtually to develop a model capable of predicting adsorption
because of the large number of possible combinations of surfactants, surfaces, and
solutions. An exception to this are the self-consistent field models proposed by
Scheutjens and various coworkers. The models all started with a common approach
a type of analysis of the solution phase in contact with the
surface. In this approach the solution is modeled as a quasi-lattice, with each site
in the lattice being occupied either by a solvent molecule or by a segment of the
surface-active material. By developing interaction parameters for segment-solvent,
segment-surface, and solvent-surface interactions, the free energy of the system can
be calculated and minimized to yield the concentration profile of the surfactant
segments as a function of distance from the surface. This profile can be used to
predict both the amount of surfactant adsorbed and the average conformation of
the adsorbed material.
As first developed the model allowed only a single surface-act'
rial composed of only one type of In subsequent publications the limita-
tions of the original model were relaxed to allow for the presence of more than one
type of segment within the surfactant chain the presence of more than one
surface the formation of lipid and vesicles the formation of
micelles and the inclusion of charged groups The predictions show the
generally expected behavior for adsorbing species with the hydrophobic moieties
predominantly toward a hydrophobic surface. the hydrophilic groups generally
concentrated toward the aqueous phase, and, for larger chains, adsorption via
either loops, trains, or tails. This model has also been used to predict that the
absorption of a nonionic surfactant onto a hydrophobic surface will occur as a
monolayer
Koopal and after incorporating electrostatic effects, were unable to
solve the general problem but instead presented solutions to two extremes of beha-
vior, adsorption where the molecules were adsorbed either flat on the surface or
perpendicular to it. Hassan and Zollars were able to overcome this difficulty
and present predictions for the adsorption of an ionic surfactant onto a surface
with charges of the same sign. Using this model it is possible to predict conditions
where local maxima and minima do appear in the adsorption isotherm, as observed
by Hoeft and Zollars. When the adsorbed amount, as predicted by the model, was
compared with experimental data, however, the predicted was low by
about an order of magnitude. One of the key parameters affecting the prediction
of the amount adsorbed is the lattice spacing. In their calculations Hassan and
Zollars used a spacing of 0.4 nm. Reducing this value would give greater values for
the adsorbed amount but would also magnify the electrostatic effects resulting in
numerical instabilities in the problem solution. Therefore, although the work by
Hassan and Zollars indicates the potential for the interplay of electrostatics with

adsorbed conformation and adsorbed amount, a definitive answer will have to
await more robust numerical techniques.
From reviewing all of these sources a picture begins to emerge about the current
state of knowledge about the adsorption of ionic surfactants onto hydrophobic
surfaces. For adsorption onto an uncharged surface the adsorbed associ-
ate with the surface via a hydrophobic interaction. This results in molecules being
adsorbed in a tail-down conformation. At low concentrations the molecules appear
to lie flat on the surface but become progressively more vertical as the concentra-
tion increases. In this vertical position significant lateral interactions between the
adsorbed molecules occur, leading to increased adsorption and, ultimately,
layer formation. On surfaces that themselves possess a structure, such as graphite;
this may also induce a structure in the adsorbed layer (adsorption epitaxy?) along
the individual grains of the surface as observed by and Ducker
When ionic adsorb onto a hydrophobic surface that also contains
charged groups of the same sign as the surfactant, electrostatic interactions begin to
play a role. At low surface charge densities the initial adsorption will take place as
in the case of adsorption onto an uncharged surface, with the surfactant lying
nearly flat on the surface. As the surface charge density is increased, the adsorbed
surfactant will be forced into a more vertical orientation. In this orientation the
potential exists for greater adsorption due to the opportunity for more hydrophobic
associations via either surface-surfactant or surfactant-surfactant contacts. At still
higher surface charge densities the electrostatic repulsions so strong that no
observable adsorption occurs.
For the adsorption of ionic onto hydrophobic surfaces of opposite
charge, a couple of different scenarios are possible. At low surface charge densities,
the initial adsorption may take place primarily through hydrophobic associations
as in the case of an uncharged surface, with an occasional adsorbed molecule
occurring in a head-down orientation due to electrostatic attractions between the
surface group and the surfactant head group. Once the surface charge has been
neutralized, subsequent adsorption should occur as if on an uncharged surface.
This should result in coverage with the potential for some structuring
of the layer around the sites where a molecule has adsorbed in the
down orientation. At high surface charge densities, the initial absorption should
occur predominantly via electrostatic attractions between the surface groups and
the surfactant head groups. As the adsorption proceeds, the surface will become
more and more hydrophobic due to the tails of the adsorbed extending
toward the aqueous phase and charge neutralization. Ultimately the surface will
become so hydrophobic that another adsorbed layer will start to form due to
hydrophobic associations between the surfactant tails. This ultimately leads to
the formation of an adsorbed bilayer,
When more than a single surfactant species is present, the situation becomes
more complex again. It seems clear, however, that whether the presence of other
surface-active species increases or decreases the tendency for adsorption depends

very strongly o n the structure of the materials. More strongly hydrophobic materi-
als, especially nonionic materials, appear to cause greater adsorption of a n ionic
species. This is probably due to hydrophobic interactions between the species and
screening of the ionic groups o n the surfactant by the nonionic material. As Hoeft
observed, however, synergistic effects may also be present between two ionic
species.
Elder, and I. Ulevitch, Colloid (1954).

Brodnyan and Brown, Colloid (1960).
A. Maurice, Colloid Interface (1985).
Abbey, J. Erickson, and Seidewand, Colloid Interface Sci.
(1978).
R. J. Orr and Breitman, Can, Chem. (1960).
Paxton, Colloid Interface (1969).
I. Piirma and Chen, Colloid Interface Sci, (1980).
R. Sci. (1979).
B. Kronberg, P. Stenius, and Y. Thorsell, Colloids Surf. (1984).
Kronberg, and P. Stenius, Colloid Interface (1984).
Kronberg and P. Stenius, Colloid Interface (1984).
Kronberg, J. Colloid Interface (1983).
Furlong and Colloids Surf. (1982).
A. and Fuerstenau, Trans AIME (1955).
E. Herder, Colloid Interface (1991).
Ward, P. and D. Phys. Chem. (1993).
and A. Ducker, Phys. Chem. (1996).
Waltermo, P. and I. Colloid Interface
(1996).
A. Ducker, Xu, and J. Langmuir, (1994).
P. Kekicheff and 0. Langmuir (1994).
Evans, and Wennerstrom, Langmuir (1993).
van Lent, Israels. Scheutjens. and Fleer, Colloid
Interface Sci. (1990).
P. and H. Ottewill, Colloid Interface (1971).
Kayes, Colloid Interface Sci, (1976).
Gonzalez. A. Vilchez, and Hidalgo-Alverez, Colloid
(1991).
Zhao and Brown, J. Phys. (1995).
Zhao and Brown, Langmuir (1996).
Kandori, Kon-no, A. Kitahara. Langmuir (1989).
W. R. Zelenka, and J. Colloid Interface
(1988).
P. Becher, and P. Petty, Dispersion Sci. Tech. (1981).
Brown and Zhao, Macromolecules (1993).

Brown and Zhao, (1994).
S. I. J. Steach, and Zollars, Colloids Surf (1987).
C. E. Hoeft and Zollars, Colloid Interface (1996).
Brouwer and Zsom, Colloids Surf, (1987).
L. Gwin. M.S. Thesis, Washington State University, 1988.
and N. Stein, Langmuir (1994).
Orr and Can, (1960).
J. Hay, Mactaggart, and Rochester, Chem. Faraday
82:1805 (1986).
Bain, P. B. Davis, and Ward, Langmuir (1994).
P. and Langmuir (1991).
Johal. Ward, and P. B. Davies, Phys. Chem. (1996).
Hoeft. Thesis, Washington State University, 1993.
H. Scheutjens and Fleer, Chem. Phys. (1979).
H. Scheutjens and Fleer, Chem. Phys. (1980).
van Lent and J. M. H. Scheutjens, (1989).
Fleer and J. M. Scheutjens, Colloid Interface (1986).
A. Leermakers and H. Scheutjens, J. Chem. Phys. (1988).
A. Leermakers and H. Scheutjens, Phys. Chem. (1989).
A. and H. Scheutjens, Chem, Phys. (1988).
A. Leermakers and Scheutjens, Colloid Interface Sci.
(1990).
Koopal and Ralston, Colloid Interface (1986).
and L. Koopal, Langmuir (1990).
Hassan and R. Zollars, Colloid Interface Sci (1991).

Department of Marine
Chemistry, Hokkaido University, Hakodate, Japan
I. Introduction
II. Metal Complexing Ability of Humic Acid
A. Acid-base properties of humic acid
B. Metal-humic acid complexation
I I I . Adsorption of Metal Ions by Immobilized Humic Acid
A. Preparation of composite biopolymer adsorbent
B. Metal adsorption ability of HA-M
IV. Adsorption of Metal Ions onto Insolubilized Humic Acid
A. Preparation of insolubilized H A
B. Effect of insolubilization on metal-HA complexation
V. Practical Application of IHA
A. Cadmium-midgut gland complexation
B. Removal of cadmium from M G by competitive adsorption
References
Metal recovery by adsorption from dilute aqueous solutions is an emerging field of

interest from the standpoints of both resource conservation and environmental
remediation. Many biopolymers derived from microbes and plants are known to
bind metals strongly. The use of biopolymers as adsorbents for the recovery or
removal of metals from aqueous environments has been a topic of extensive
research in recent years such as cellulosics,
proteins, and chitin derivatives are industrially attractive for a num-
ber of reasons. Biopolymers are capable of removing heavy metal ions at parts per

billion concentrations Moreover, they are readily available and ecologically
acceptable.
substances, which are biopolymers widely and abundantly present in
natural waters and soils, also have a high complexing ability with various heavy
metal ions. These compounds are formed by the random condensation of break-
down products of terrestrial and aquatic plants and extracellular metabolites of
phytoplankton. Concentrations of metals in marine and fresh waters are often
higher than predicted from the solubility products of corresponding hydroxide
and carbonate compounds. The complexation of metal ions with dissolved humic
substances is responsible for the apparent of metals in natural
waters Water-soluble humic substances are usually divided into two frac-
tions, humic acid (HA) and fulvic acid (FA). H A is defined in operational terms as
the fraction of humic substance soluble in alkaline solutions and insoluble in acidic
solutions, while F A is the fraction soluble in both alkaline and acidic solutions. A
general method for the fractionation of humic substances is illustrated in Fig.
HA is easily obtained as a precipitate in acidic solution (pH 1.5). Although H A
appears to be an attractive adsorbent for the recovery of heavy metal ions, there is
little information on its practical application as adsorbent. It is difficult to use
humic acid as adsorbent because of its high solubility water.
We have studied the use of HA as adsorbent. A composite adsor-
bent in which HA was immobilized by a combination of calcium alginate gel
membrane and activated carbon powder was developed Furthermore, another
HA was made by a thermal process Based on a metal complexation model
the effect of immobilization and on the metal complexing ability of
HA is discussed in the following sections.
. Method for fractionating humic substances.

Acidic groups on HA, which serve as effective electron donors in the binding with
metal ions, also bind protons in aqueous environments. Therefore, determining the
equilibrium parameters of the acid dissociation the number and dissociation
constants) is important to estimating the metal complexing ability of HA.
Total Number of Acidic Groups on Humic Acid
The total number of acidic groups on HA was determined by a conductometric
titration A typical conductometric titration curve of commercial HA
Chemical Co.) with is shown in Fig. 2. The intersection of two straight lines
defines the end point of the titration. The total number of acidic groups was
determined from the titration curve to be 3.40 Since the HA used
here was in the form of sodium salt. no end point was obtained from the titration
with
2. Dissociation Constants of Acidic Groups on HA
To determine the dissociation constants of the acidic groups, an adsorption iso-
therm for the adsorption of protons by HA was obtained from the potentiometric
titration of H A as shown in Fig. 3. The ordinate, represents the equilibrium
number of protons bound to acidic groups of 1 g of HA. The number of protons
bound to acidic groups on HA was determined from the difference between the
bulk proton concentrations in the absence and presence of HA. The ionic strength
of the system was adjusted to 0.1 by the addition of
The potentiometric titration curves of HA are usually broad and ill-defined,
reflecting the varying properties of acidic groups on HA Wilson and
investigated the acid-base properties of lake water and marine
humic substances and determined a two-step dissociation process. Schnitzer and
0 0.5 1.0
Added amount of
Conductometric titration curve for 300cm3 of a solution containing 0.2 g of

HA with 0.1 at (From Ref.

type 1 4.17
type 2 6.34
. 3 Adsorption isotherm of proton adsorption to HA at Initial

concentrations of HA and were and 0.1 respectively.
The solid line represents the theoretical curve calculated with (3).
Khan accounted for the acidic properties of compounds in terms of a

range of aromatic carboxyl groups and to a lesser extent phenolic hydroxyl groups.
In this case, the acid dissociation reactions of H A can be written as
where represents acidic groups of type on HA. The acid dissociation con-
stant is defined as
where a, is the degree of dissociation of type acidic groups. can then be

expressed as
where and are the numbers of type 1 and 2 acidic groups, respectively. Their
sum should be equal to the total number of acidic groups determined from the
conductometric titration,
We applied a nonlinear least squares method to determined the four constants

and in ( 3 ) . The constants that gave the best fit with the experi-
mental data are listed in Fig. 3. The solid line in Fig. 3 represents the theoretical
curve calculated from ( 3 ) using the constants. Figure 4 shows a comparison of
the proton-binding capacity of HA with that of some inorganic adsorbents such as
silica gel and Aerosol hydrous titanium oxide
hydrous throium oxide and magnetite. It is evident that the number of

of Metal Ions
Magnetite
.4 Comparison of the proton-binding capacity of HA and inorganic adsorbents.
acidic groups on HA, which serve as the metal complexing sites, is higher than
that of most conventional inorganic adsorbents.
The acid dissociation constants of type 1 and 2 acidic groups differ by about two
orders of magnitude. Therefore, the complexation of metal ions with HA
can be written as
+ (- i= or 2
where represents the metal ion. The complexation constant is
defined as
where
where represents the fraction of type i acidic groups complexed with metal ions.
The number of metal ions complexed with 1 of HA, is expressed as
where and represent the numbers of type 1 and 2 acidic groups available
for metal complexation, respectively.

The pH dependence of the number of cadmium and lead ions complexed with
HA is presented in Figs. 5 and 6, respectively. A nonlinear least squares method
was applied to find and The constants that gave the best fit
with the experimental data are listed in the figures. The solid lines in the figures
represent the theoretical curves calculated Eqs. (7) and (8) using the fitted
constants. The number of acidic groups available for metal complexation, is in
reasonable agreement with the number of acidic groups on HA determined by
titration, This means that all the acidic groups on H A are
available for metal complexation. The metal complexing capacity of HA,
+ 1.7 x is large enough to use HA as an adsorbent
material.
ADSORPTION OF METAL IONS BY IMM ILIZED HUMIC

As mentioned above, humic acid is an attractive adsorbent material for the recov-
ery of valuable and toxic heavy metals from aqueous environments. However, it is
difficult to use H A itself as the adsorbent because of its high solubility in water. Ho
and Miller investigated interaction of uranium ions and HA at the
interface. They demonstrated that the presence of HA enhanced the
adsorption of uranium on hematite to a remarkable extent. In their study, HA
was adsorbed on hematite particles prior to the uranium adsorption experiments.
The hematite particles can therefore be regarded as HA-immobilized adsorbent.
and Wagener immobilized HA on anion-exchange resins as carriers,
and they tried to use them for the recovery of uranium from seawater. Since the
carriers are usually inert materials in adsorption, the ratio is very
important. In those studies, the percentage of H A in the total adsorbent was
pH dependence of complexation at Initial concentrations of HA,

and were, respectively, and 1 and 3.0
dm3. The solid lines represent the theoretical curves calculated with Eqs. ( 7 ) and
(From Ref. 7.)

Adsorption of Metal Ions onto Humic Acid
FIG.6 dependence of Pb-HA complexation at 30Â°CInitial concentrations of HA,

and were, respectively,0.2 and 1 x 10 and 4.0
The solid lines represent the theoretical curves calculated with Eqs. ( 7 ) and (8).
Ref, 7 )
3.5% 25% on a dry weight We developed a composite

biopolymer adsorbent in which HA was almost completely immobilized by a com-
bination of calcium alginate gel membrane and activated carbon powder. The
adsorbent comprised as much as 40% HA
. Preparation of Composite
An aqueous solution containing HA, acid (AA), and activated carbon pow-
der (ACP) was well mixed. It was extruded, dropwise, onto a plastic and dried
at Then it was in an excess of solution for 1 The membranes
of HA-entrapped calcium alginate had a diameter of about 5 mm and a thickness of
about 0.3 mm. Using this immobilizing method, we obtained a membranous adsor-
bent (HA-M) consisting of 40% HA, 20% AA, and 40% on a dry weight
basis.
1. Metal Complexing Ability of

Many arc known to bind metal ions strongly. Alginic acid used as the
agent also has a complexing ability with metal ions. Many
studies have been carried out on the application of AA to the aqueous phase
separation of heavy metals, and the possibility of AA as the adsorbent
has been suggested HA-M comprises 20% AA on a dry weight basis.
Therefore, it can be expected that AA plays an important role in adsorption
by HA-M.

The metal complexing ability of AA was evaluated in the same manner as that of
HA AA, which is a main cell-boundary constituent of brown algae, is a linear
polysaccharide composed of acid and acid. The
acidic property of AA is due to the presence of carboxylic groups
Therefore, can be regarded as an adsorbent containing only one type of
metal-binding site. The total number and the acid dissociation constant of acidic
(carboxylic) groups on AA were determined from the conductometric and
tiometric titrations to be = 4.08 x and = 6.15 x
respectively. The subscript A is used to designate the parameters for AA. The acid
dissociation constant of carboxylic groups on AA was about one order of magni-
tude larger than that of type 1 acidic (carboxylic) groups on HA.
The complexation of metal ions with biopolymers that contain only one
type of metal binding site can be written as
The fraction of acidic groups complexed with metal ions, 0, and the number of
metal ions complexed with 1 g of biopolymer, can be expressed as
where
and
where represents the number of acidic groups available for metal complexation
on 1 g of biopolymer. The complexation experiments were carried out
using the membranous calcium alginate gel (containing no H A or ACP) as the
adsorbent. The equilibrium parameters for the metal-AA complexation,
and were determined in the same manner as those for HA, and they listed
in Table 1 . The number of carboxylic groups available for metal complexation on
AA, is about 30% smaller than the total number of carboxylic groups on AA
determined from the conductometric titration, et al. reported that
the primary mechanism of of alginate was the dimerization of
sequences, with specific site binding of calcium ions between alginate chains.
This process reached completion at a calcium level of about 30% of the total cation
requirement of the alginate. Therefore, it can be considered that about 30% of the
1 Equilibrium Parameters for Metal-AA Complexation
3.04 -3.17
Lead 3.02 -5.13"
Data from Ref. 8.

of Metal Ions onto
groups on the membranous calcium gel were occupied by calcium

ions during the process. The metal complexing capacity of AA,
= 1.5 x is comparable to that of HA (1.7
2. Adsorption of Metal Ions on HA-M

The number of metal ions adsorbed to 1 g of HA-M, can be expressed as
where represent the weight fractions of AA and HA, respectively, in the

adsorbent. Since the equilibrium parameters for metal-AA and metal-HA
plexation were already obtained, we can calculate by using and
(10)-(12). Figure 7 shows a comparison between the experimentally found number
of metal ions adsorbed onto 1 g of HA-M, and the number calculated from
the equations, The data agree well with those calculated. This means that
the complexation constants and the numbers of available acidic groups for
plexation of metal-HA and metal-AA systems were scarcely influenced by immo-
bilization. The metal adsorption capacity of HA-M, which comprises as much as
40% HA and 20% AA, was about 1 x on a dry weight basis.
The heating of raw humic acid above a certain temperature causes dehydration
(condensation) between the acidic groups and lowers its solubility in water (Fig. 8).
Therefore we tried to prepare an HA by heating (IHA).
.7Comparison of the experimental amount of metal adsorption onto HA-M and

the amounts calculated with (1 and (12).

Heating Temperature
Effect of heating temperature on the percentage of dissolution in distilled

water at 30
reparation of
acid was pretreated with excess solution and converted to a calcium
salt it was made insolubilized by heating at for 1 Heating at
higher temperature or for a longer period reduced the acidic groups on HA to a
considerable extent The HA thus obtained (IHA) was ground in an agate
mortar with a pestle and washed repeatedly with 1 solution and hot
distilled water. Finally, IHA was converted from the hydrogen form to the sodium
by soaking in 1 solution, rinsed with distilled water, and dried at
IHA prepared in this way was virtually insoluble in water up to pH 10 at
and there was only 7% dissolution even in boiling water
Humic acid is by condensation of acidic groups such as aromatic

carboxyl groups and phenolic hydroxyl groups, which are probably the sites of
metal complexation. Therefore, it may be presumed that the number of acidic
groups and the acid dissociation constant of IHA are lowered by heating compared
with those of HA. The number and acid dissociation constants of type 1 and 2
acidic groups on IHA were determined by potentiometric titration Surprisingly,
the acid dissociation constants were scarcely influenced by insolubilization.
However, of the type 1 acidic groups and 22% of the type 2 acidic groups
on HA were lost by the condensation reaction 1.45 and
1.15
The equilibrium parameters for the metal-IHA complexation were determined
in the same manner as those for HA. The parameters obtained are listed in Table 2
The insolubilization process scarcely influenced the complexation constants,
and The number of type 1 acidic groups available for metal
almost coincided with the number of acidic groups determined by

Equilibrium Parameters for Metal-IHA Complexation
Cadmium 1.45 x 0.90 x -3.70 -5.40

Lead 1.45 0.96 x -5.80 -7.64
Source Ref. 7
titration, However. 20% of the type 2 acidic groups were unavailable

for metal complexation. The metal complexation capacity of IHA was about
1.2 x on a dry weight basis.
The internal organs of most marine organisms contain valuable materials such as
DHA, proteins, and essential acids However, one of the "big
three" toxic heavy metals, cadmium. is concentrated in the internal organs of
organisms throughout the food chain. Many studies have been carried
out to study the accumulation of cadmium in the internal organs of edible marine
organisms such as mussels snails crabs whelk scallops
and squid For example, cadmium concentration in the mid gut gland
(MG) of the scallop which is one of the most popular
edible shellfish, is as high as 200 on a dry weight basis For the effective
use of these valuable materials, the cadmium removal process should be estab-
lished. Cadmium in the internal organs can be removed by treatment in acidic
solutions. Cadmium ions are released as a result of the protonation of
ing proteins in the organs. Therefore, the proton concentration of the acidic solu-
tion must be kept at a high level, and the subsequent treatment of
containing, low pH wastewater is required. Moreover, treatment in a strongly
acidic solution may denature these valuable biomaterials.
In a recent study, we proposed a new method (competitive adsorption method)
for the removal of toxic heavy metal ions from acid-sensitive and
applied it to the removal of cadmium from the M G of the scallop
In this method, a competitive adsorbent (IHA) is added to a weakly
acidic solution containing M G to lower the cadmium concentration in the liquid
phase. The lowering of the cadmium concentration in the liquid phase promotes the
release of cadmium ions from M G , and the released ions are immediately
adsorbed by IHA. In this manner the competitive adsorption method enables the
simultaneous removal of cadmium from the scallop and from the liquid
phase under mild acidic conditions.
Accumulation of heavy metals by several organisms has been shown to be asso-

ciated with the production of special proteins. The internal organs of marine organ-
isms accumulate not only cadmium but also copper and zinc. Fractionation of
metal-binding proteins of squid liver by gel permeation chromatography revealed

that the cytosolic cadmium, copper, and zinc are bound to different proteins
Assuming that a similar relation holds in the present MG, M G can be regarded
as an adsorbent containing only one type of cadmium-binding site. Therefore we
carried out cadmium adsorption experiments on M G in an attempt to determine
the acid dissociation constant the complexation constant and the number
of complexing sites, for the Cd-MG system in the manner as
those for The parameters that gave the best fit with the experimental data were
obtained as = 8.47 x = 8.89 and
= 5.21 x respectively. We obtained good agreement between the
experimental amounts of cadmium adsorbed onto M G and the values predicted
with Eqs. (10) and ( 1 1). The correlation coefficient between the experimental and
predicted values was 0.972.
Competitive adsorption experiments were carried out in the following procedure. A

polyethylene mesh bag (7 x 9.5 cm) containing IHA and a certain amount of M G
were added to an (0.1 solution. The pH of the solution was
adjusted to the desired value with The solution was stirred for the time
necessary to attain equilibrium, and then the polyethylene mesh bag containing
IHA was removed from the M G was separated from the liquid phase in a
centrifuge at for 10 and the cadmium concentrations of the super-
natant and M G were measured. The M G used in these experiments contained
about 1.8 200 ppm) of cadmium on a dry weight basis.
From the material balance on cadmium, the equilibrium concentration of free
cadmium ion in the competitive adsorption system can be expressed as
where and represent the amounts of M G and IHA added to the system.
respectively, and are the amounts of cadmium ions compiexed with M G
and IHA, respectively, and is the total cadmium concentration in the system.
Sliced M G was used in these competitive adsorption experiments. Since 46% of the
cadmium binding protein in M G was constantly dissolved in the liquid phase, we
modified Eq. (13) to
where represents the cadmium concentration (the sum of the free cadmium ions
and the cadmium ions complexed with the dissolved protein) in the liquid phase.
The pH dependence of in the competitive adsorption systems containing 30
3
dm of M G and 1.0, 2.0, or of IHA is presented in Fig. 9. For compar-
ison, the results of acid treatment (in the absence of IHA) are also plotted in Fig. 9.
A successive approximation method was applied to numerically estimate from
(1 and (14). The experimental data (symbols) agreed well with
the predicted values (solid lines). The pH dependence curves of have minimum
values at pH 5. The competitive adsorption method enabled the simultaneous
removal toxic cadmium from both the liquid and M G phases at pH 5. The
increase in in the pH 5 region was attributed to the increase of free cadmium

of Metal Ions onto
pH dependence of cadmium concentration in the liquid phase in the

competitive adsorption system at Initial concentrations of MG, and
3 3
IHA were 30 0.1 mol/dm , and /dm , respectively. The solid lines represent
the theoretical curves calculated from (1 and (14). (From Ref. 54.)
ion, while the increase in the p H 5 region was attributed to the increase in the
amount of cadmium- protein complex dissolved in the liquid phase.
Roe, and Biotechnol. Bioeng.

(1993).
Deans and G . Dixon, Water (1992).
K. Jang, Brand and W. Environ. Prog. (1990).
Jang, Geesey, L. Lopez, L. and P. Wichlacz, Water
Res. (1990).
Jang, S. Lopez, and Biotechnol. Bioeng. (1991).
Seki, A. Suzuki, and I. J. Colloid Interface (1990).
Seki and A. Suzuki, Colloid Interface Sci. (1995).
Seki and A Suzuki, Eng. Chem. Res. (1996).
Sugimura, Suzuki, and Y. Miyake, Oceanogr. Jpn. (1978).
Y. Sugimura, Suzuki, and Miyake, Deep-sea Res. (1978).
Environ. Sci. Technol. (1981).
A. Saar and Weber, Anal. Chem. (1980).
A. Saar and J. Weber, Environ, Technol. (1982).
J. Greter, and Haerdi, Anal. Chem. (1977).
Ramamoorthy and D. Kushner, Nature (1975).
A. Torres and R. Radiochim. Acta (1984).
Caceci. Radiochim. Acta (1985).
Cacheris and Radiochim. Acta (1987).
G. Radiochim Acta (1988).

A. De Brabandere, and A. Acta (1988).
V. P. Vitorge, and Acta
(1988).
0. James and A. Parks, in Surface and Colloid Science, Vol. 12 ( E .
Plenum. New York. 1982, pp.
A. and I . A. Acta (1960).
D . Gamble, Can. Chem. (1972).
D. Wilson and P. (1977).
Schnitzer and Khan, Substances in the Environment, Marcel
Dekker. New York, 1972. p.
P. Dick. and P. Wolf, Colloid Interface Sci.
(1976).
A. R. Gupta and Bull, Jpn. (1988).
H. Hohl and W. Colloid Interface (1976).
Ho and Miller, Colloid Interface (1985).
Heitkamp and Wagener. Chem, Process (1982).
Jang. N. Harpt. Uyen, and Geesey,
Part B (1989).
J. Stevens, A. Morris, and A. Int.
Biol. (1980).
(1973). [In Japanese]
Seki, Thesis, Hokkaido University, Hokkaido,
Seki and A. Suzuki, Bull. Fish. Hokkaido (1992). [In
Japanese]
Bull. Fish. (1986).
Nippon Gakkaishi (1988).
K. Nippon Suisan Gakkaishi (1989).
Frankenne, Noel-Lambot, and A. Comp. Biochem. Physiol.
(1980).
M. Frazier, George, Overnell, and
Biochem. (1985).
F. Noel-Lambot, Experientia (1976).
A . Viarengo, Moore, G. A. Pipe, and
Farrar, Mar. Biol, (1987).
Carpene, Via, and Cortesi. Arch. Environ. Contain.
Toxicol. (1989).
Langston and Mar. Biol. (1986).
K. Lerch, and R. W. Biol. Chem. (1982).
Overnell, Comp. Biochem. Physiol. (1984).
J. Overnell, Comp. Physiol, (1984).
Overnell and Biochem. Physiol. (1979).
Ohba. and Biochim. Acta (1983).
Stone, Wilson, and Overnell, Comp. Biochem. Physiol.
(1986).
V. and Bull. Fish. Hokkaido Univ. (1991).
Tanaka, and Experientia (1983).
Seki and A. Colloid Interface (1997).

eta
Centre for Applied Colloid and
Science, Swinburne University of Technology, Melbourne, Australia
Department of Applied Chemistry, Royal Melbourne
Institute of Technology. Melbourne, Australia
I. Introduction
A. The use of hydrous metal oxides as adsorbents
B. Surface characterization
C. Removal of heavy metals from solution
II. Experimental Techniques
A. Reagents
B. Background electrolyte
C. Metal ion solutions
D. Colloidal substrates
E. Precipitation, adsorption, and coprecipitation studies
F. Determination of metal ion removal
111. Precipitation, Adsorption, and Coprecipitation
IV. Comparison of H C O and H F O as Substrates
A. Adsorption
B. Coprecipitation
V. The Influence of Competing Metal Ions
A. Cr(III) with H F O
B. Zn(II) with H F O
C. Ni(II) with H F O
D. Zn(II) with HCO
E. Ni(II) with HCO
VI. The Influence of a Complexing Ligand
A. The influence of ammonia on the solution chemistry of aqueous
heavy metals
B. The influence of ammonia on the adsorption of Zn(II)
C. The influence of ammonia on the coprecipitation of Zn(II)
VII. Modeling of Adsorption and Coprecipitation
A. Introduction

B. The James-Healy model for metal ion adsorption
C. The significance of the chemical free energy contribution to
adsorption
VIII, Conclusions
A. Precipitation, adsorption, and coprecipitation
B. Comparison of H C O and H F O as substrates
C. The influence of competing metal ions
D. The influence of a complexing ligand
E. Modeling of adsorption and coprecipitation
References
The removal of heavy metal ions from both natural water supplies and industrial
wastewater streams is becoming increasingly important as awareness of the envir-
onmental impact of such pollutants is fully realized. In particular, the likelihood of
such metal ions precipitating out of solution and/or coating other materials can
have a profound effect on both aqueous and nonaqueous environments. There is
considerable evidence in the literature that the primary mechanism for transporta-
tion of metal contaminants in aquatic systems is the movement of suspended par-
ticulate material containing the adsorbed pollutant metals It is also known
that a strong correlation exists between the concentration of trace metals in the
(aquatic) environment and the extent to which those metal ions adsorb onto col-
loidal substrates present in the environment A similar correlation between the
concentration of trace metals in the (aquatic) environment and their precipitation
behavior is not so clear. There is, then, a well-founded need to study
related phenomena in order to understand and predict the behavior of toxic metals
in the environment.
Industrial companies are increasingly being required to account for the fate of
all chemical species, whether deliberately added or present as by-products, at all
stages of industrial, mining, and manufacturing processes. The processes of pre-
cipitation, adsorption, and coprecipitation. apart from directly controlling the eco-
nomics of many chemical processes, are also often involved in the cleanup of
industrial wastewater Thus there is also a well-founded need to study adsorp-
tion-related phenomena in order to understand and predict the behavior of toxic
metals in industry.
Metal ions that are prevalent in industrial effluent and have a considerable
environmental impact include and Waste solutions of
these metals are routinely found at levels that are in excess of acceptable disposal
limits. Such waste solutions are produced, for example, by electroplating and metal
pretreatment printed board manufacture
leather tanning and wool scouring and anodizing among
others These three metal ions are studied in detail in this chapter.

As will be discussed shortly, the structure of amorphous hydrous metal oxides is
open to conjecture. They almost certainly comprise submicrometer porous particles
and/or small primary particles that aggregate into larger submicrometer
sized structures The influence this porosity and/or ultrasmall particle size has
on adsorption-related phenomena has not been well documented but should be
investigated
Amorphous hydrous metal oxides not only represent a class of adsorbents that
are commonly encountered in the environment they are also commonly used as
adsorbents in industrial wastewater treatment They are easy to prepare,
inexpensive, and common in nature. As such, they are appropriate substrates to
choose for fundamental studies of adsorption phenomena.
Two examples are the amorphous hydrous oxides of and
For simplicity, these are referred to hereafter as H F O and HCO,
respectively.
H F O and HCO are used as the substrates in this study.
The amorphous hydrous oxide of (HFO) has been studied extensively as
a substrate for metal ion adsorption, including the metal ions of interest here
Pertinent observations relevant to H F O as an
adsorbent for and/or include the following:
The localized structure (determined by X-ray absorption fine structure) of
adsorbed depends on the history of the ion, on whether it
was adsorbed onto or coprecipitated with the H F O surface
Metal ions present in an industrial waste were successfully treated using
adsorbing colloid flotation, dependent on the pH. Some metal ions could be
treated at lower p H values than others, with the order being
This is the same order found for their hydrolysis in solution and
reflects a well-known pattern for adsorption phenomena (see I.C.2).
Coprecipitates of chromium and iron hydroxides appear to enhance the
removal of metal ions
Adsorbed metal ions such as and may inhibit
the crystallization of H F O into crystalline forms such as goethite and hematite
Other metal ions, such as apparently do not.
The amorphous hydrous oxide of (HCO) has been studied extensively as
a source of sols of uniform particle size and shape and as an electrophoresis
standard however, it has not been studied extensively as an adsorbent of metal
ions. Sen found that chromium hydroxide generally has a larger adsorp-
tion capacity than the hydroxides of other triple-charged metals. Simon et al.
reported that coprecipitation of with hydroxide removed
from solution at a p H that was two units lower than that obtained using
hydroxide as the coprecipitating hydroxide. Similarly, Packter and
Panesar used and mixtures to show enhanced removal for both
ions compared with the individual metal systems. HCO has also been used as an
adsorbing surface to effectively adsorb the proteins ovalbumin, y-globulin, and
lysozyme as a separation technique

The surface characteristics of H F O and HCO would be expected to play a domi-
nant role in their use as adsorbents for metal ions. In particular, one would expect
the surface area of the substrate to play a dominant role in determining the extent
of adsorption. Intuitively, one would expect the adsorption to be enhanced if either
the porosity of the substrate was increased (and hence its available surface area
increased) or the particle size was decreased (and hence the available surface area
per mass increased). Electrokinetic properties would also be expected to be impor-
tant, with a negative surface providing the best substrate for positively charged
metal ions.
1. Particle Size and Morphology
The average particle sizes of the H F O and HCO used here were measured using
dynamic light scattering and found to be 420 and 260 nm, respectively It
should be remembered, however, that these sizes probably reflect the average
sizes of aggregates composed of nanosized particles and/or porous material. In
terms of adsorption phenomena, the significance of particle size lies in its relation-
ship to surface area. The smaller particles (HCO) are assumed to have a larger
surface area (per mass) and therefore a greater capacity to adsorb (all other things
being equal). As we show in this chapter, to a first approximation this is the case.
2. Surface Area
For metal ion adsorption studies, the surface area of the adsorbent that is of
interest is the area that contains ionizable surface sites that are capable of inter-
acting reversibly with (metal) ions in solution. It is usually assumed to be the
exterior surface and/or the surface of large macropores. Ionizable surface sites
may also, however, exist in the interior of a porous adsorbent solid or may not
be immediately available due to the coagulation of primary particles. Such sites
may interact at a slower rate than readily accessible surface sites, giving rise to slow
adsorption kinetics Measurement of surface area by the BET method is
generally considered suitable for measuring the exterior surface area of a substrate.
BET surface areas require drying samples in a vacuum oven at elevated tem-
peratures, and therefore it is probable that some fine structure that is present in
aqueous solution is lost The BET surface area should therefore be considered a
lower estimate of the surface area of porous or amorphous materials. Nevertheless,
a simple BET surface area should give an indicative estimate of the surface
area available for adsorption.
The surface areas of the two substrates used in this study were measured using
the nitrogen BET method on a Micromeretics ASAP 2000 and found to be 225 and
for H F O and HCO, respectively
As expected, the BET surface areas are much larger than would be indicated by
their particle size if they were hard sphere particles. Thus the surface areas indicate
either a porous particle and/or an aggregate comprising ultrasmall, nanosized
tides. As also expected, the BET surface area is greater for HCO than for HFO.
This difference in surface area must be taken into consideration when discussing
any intrinsic difference in the efficiency of adsorption between the two substrates.

Metal Oxides as 67
3. Electrokinetic Properties
The points of the two substrates used in this study were
by using a DELSA 440 and found to be at pH 8.2 for H F O and at pH 9.2 for
HCO. Both and are completely adsorbed at pH values considerably
less than either of these values (see Figs. 3-51, indicating that the surface is
positively charged during adsorption. Electrostatic forces, in this case, are unfavor-
able to adsorption, and there must be a strong specific interaction for adsorption to
occur.
1. Interdependence of pH, Metal Ion, and Substrate

(a) Role of p H. One common method of obtaining an adsorption isotherm is to
keep the total amounts of adsorbent and adsorbate constant and vary the pH
(typically from a low pH to a high pH). The results of such an experiment are
shown in Fig. 1 for the adsorption of several aqueous heavy metals onto silica
this figure, the influence of pH is apparent. At low pH values,
there is little adsorption. At intermediate pH values, which depend on the parti-
cular metal ion being adsorbed, adsorption increases from near zero to near com-
plete adsorption over a relatively small pH range. At high pH values, the metal ion
is
There are two possible reasons for this trend. As the pH is increased, the substrate
surface becomes more negative (or less positive) and thus more favorable for
. Adsorption curves at for as a function of pH for metal ions at

various concentrations of 6.7 ppm 10.4 19.9 ppm 2.5
and (From Ref.

of a positively charged metal ion. At the same time, however, the metal ion itself
undergoes hydrolysis as the pH is increased, and it is often argued that hydrolysis is
required before a metal ion can become "activated" to adsorption
( b ) Role o f the Metal Ion. One of the strongest pieces of supporting evidence to
suggest that hydrolysis is required for a metal ion to undergo adsorption is given by
the data in Fig. 1. If hydrolysis is required, then the minimum pH at which adsorp-
tion takes place (for a given metal ion) should reflect the hydrolytic behavior of that
metal ion. In Fig. 1 , the order of adsorption is
Corresponding values (the onset of hydrolysis in solution) for the
metal ions are also, in order, (1.71,
3.94, 9.25, 9.92, and 12.78, respectively A similar relationship between
metal ion adsorption and the hydrolytic characteristics of the metals was observed,
for example, by Benjamin and Leckie and Rodda et
however, observed several discrepancies in this trend when using birnessite,
tite and goethite substrates, as did Kinniburgh and Jackson when using iron
hydroxide and hydroxide gel substrates,
There is, then, to be a strong correlation between adsorption and
hydrolysis. This does not, however, prove that hydrolysis is required for adsorption
to occur. Rather, the correlation may simply reflect the similarity of binding a metal
ion to a hydroxide surface functional group and binding a metal ion to a hydroxide
ion in solution.
( c ) Role of the Substrate. Many inorganic oxides have been used in adsorption
trials for the purposes of assessing their effectiveness. Some oxides d o appear to
exhibit a greater affinity for particular metal ions over others however,
this effect is, in general, minimal.
For example, Harding and and showed for
and respectively, that the characteristics of adsorption of a given metal ion
onto a wide range of oxide substrates can be described by a general curve when the
results are normalized by surface area. Oxides that are exceptions are solids with
high dielectric constants, such as and solids that undergo reactions,
such as The roles of porosity and/or ultrasmall particle size were not inves-
tigated, and it remains to be shown whether or not H F O and HCO will prove to
have specific affinity for metal ion adsorption.
The importance of surface area should not be overlooked. Kooner reported
that by increasing the surface area of goethite from 30 to a corresponding
increase (from 40% to of the percent removal of an initially 5 ppm
solution could be achieved (at pH 5.5). Benjamin and Leckie reported that
adsorption of onto H F O also increased with surface area; in this case from
0% to 90% (at pH 6.5) by increasing the total iron concentration (and hence the
surface area) from 1.3 to 1.3
Although these effects appear dramatic, they should be kept in perspective by
considering the importance of p H . From Fig. 1, for example, it can be seen that an
increase of 50% in ion adsorption can also be achieved by increasing the p H
value by as little as 0.25 pH unit (if the right initial pH is chosen to demonstrate the
point). The choices of pH and metal ion remain dominant factors over the surface
area of the substrate

2. Effect of Agents
The presence of the O H group on metal ion hydrolysis products probably allows
hydrogen bonding to occur between the hydrolysis species and the substrate sur-
face. This, then, gives rise to the aforementioned correlation between adsorption
and hydrolysis. Substitution of the O H groups with ligands that are not capable of
forming hydrogen bond structures allows the contribution of these groups to the
adsorption process to be studied. The role of complexed metal ions, which involve
hydroxide and another ligand, is not clear and requires further investigation.
James and found that hydrolysis species adsorb strongly onto
yet Stumm et found that did not. Fuerstenau and
studied the adsorption of and onto titania,
hematite, alumina, and quartz in solutions and found that the con-
ventional sigmoidal adsorption curve was replaced by an adsorption profile that
increased initially with increased pH, declined in adsorption as the com-
plexes formed, and then increased at high p H as the hydroxide ligands replaced the
ligands. This effect was also reported by and Huang who
studied adsorption onto and oxides in
ammonia solution for the pH range
The effects of ligands other than ammonia have also been extensively studied.
Music and showed that adsorption in the presence of EDTA or
glycine decreases with increasing concentration. Rauf et observed similar
results for adsorption onto from binary mixtures in
the presence of EDTA. Bowers and Huang reported similar results for
EDTA adsorption onto and mordenite but found that the presence of a
second metal ion [such as introduced competition for the complexing agent.
In general, species complexed with multidentate ligands may exhibit enhanced
adsorption due to multipoint anchoring or surface precipitation of the complexed
metal or may still inhibit adsorption. et reported that
adsorption in the presence of fulvic acid was either was either enhanced or inhibited
depending on the substrate, indicating that the ligand-substrate interaction is also
important. The role of a porous and/or ultrafine substrate such as H F O or HCO in
the presence of a ligand such as ammonia is not clear and will be investigated here.
Apart from using complexes as a probe for adsorption mechanisms,
there is a need to develop treatment procedures for ammonia-containing metal
wastes. Ammonia-containing metal wastes are encountered in, for example, mining
by-products, which may be subject to leaching either directly or after
treatment Leather tannery effluent also contains high levels of
often in the presence of ammonia, the pollutant that causes the most difficulty in
effluent treatment
3. Precipitation, Adsorption, and Coprecipitation

The distinction between precipitation, coprecipitation, and adsorption of
aqueous heavy metal ions is not always clear, and in particular the terms "adsorp-
tion" and "coprecipitation" are often used interchangeably. Figure 2a is a sche-
matic diagram illustrating the differences among the three mechanisms, and Fig. 2b
illustrates their relative effectiveness.

Aqueous heavy metal ion
(a) (1) Precipitation (2) Adsorption (3) Coprecipitation
Schematic representation of the three processes responsible for the removal of

heavy metal ions from solution.
(a) Precipitation. It is well known that increasing the p H of a solution of

heavy ions will eventually result in the formation of an insoluble metal
hydroxide precipitate. This simple mechanism is commonly used for removing
ions in waste treatment processes with no further sophistication
required. If the removal solution is presented on a percent removal versus
pH graph, simple hydroxide precipitation typically results in a sharp sigmoidal
removal profile as depicted by curve 1 of Fig. It is the least efficient
of the three mechanisms and requires a high p H to effect removal.
At very high pH levels, redissolution of the metal hydroxide may occur
However, this is not often considered, as it is the initial removal that is of interest,
and even when thermodynamically favored, redissolution of the precipitate is often
slow to occur.
( b ) Adsorption Aqueous heavy metals can be removed via adsorption processes if
a solid substrate surface is present and, when presented on a percent removal versus
pH graph, also typically results in a curve, commonly not as steep as that
observed for simple hydroxide precipitation This is schematically
depicted by curve 2 in Fig. 2b. Adsorption processes show significant metal ion
at a lower pH than those obtained by direct precipitation

Coprecipitation. Metal ion coprecipitation processes are different adsorp-
tion processes in that the solid substrate is formed by simultaneously increasing the
pH of both the precipitating hydroxide and the ion to be removed from
solution. In this fashion, the colloid forms in the presence of the adsorbate rather
than being preformed in a separate step. The percent removal versus pH curve for
coprecipitation studies is similar in shape to that of adsorption curves, again com-
monly not as steep as that observed for simple hydroxide precipitation This
is schematically depicted by curve 3 in Fig. 2b.
Coprecipitation processes tend to be more effective than adsorption processes at
removing aqueous heavy metal ions from solution. This could be due to a much
larger effective surface area becoming available as various "layers" of the colloid
form. Alternatively, it may result from the formation of mixed metal hydroxide
complexes (either in solution or as a solid) that exhibit greater adsorption char-
acteristics than those of the individual ions.
All reagents used to prepare solutions were of analytical reagent (AR) quality
unless otherwise specified.
The electrolyte concentration used in adsorption and coprecipitation experiments

was dominated by the high concentration of nitrate required to prepare the colloi-
dal substrates (precipitated from a solution of the appropriate metal nitrate,
with respect to the metal), the concentration of the adsorbing or
pitating metal ion nitrate solutions with respect to the metal), and the base
required for adjustment of the pH. It is difficult to estimate the resultant electrolyte
concentration, because an unknown amount of the electrolyte would be occluded
into and/or adsorbed onto the colloidal phase and thus lost to the solution. The
total background electrolyte concentration used was determined to be the equiva-
lent of 0.0175 M by calibration of the colloidal suspension's conductivity
with standard solutions. It was found that the conductance of these solu-
tions did not change over the range 3-1 1.
All solutions were prepared to give the equivalent of back-
ground electrolyte solution, except for those experiments involving adsorption
and coprecipitation of metal ions from arnmoniacal solutions, where a 0.5 M
background electrolyte was used.
. Metal lo
Metal ion solutions were prepared from the nitrates to provide stock solutions that,
when diluted, produced concentrations of 50 ppm with respect to the adsorbing or
coprecipitating metal under investigation and with respect to the metal ion

used to prepare the colloidal substrate [either or The total volume
of each batch precipitation, adsorption, or coprecipitation experiment was 400
The concentration of each metal ion was chosen to be
50 ppm, as this was the approximate concentration found in industrial effluent with
significant metal ion contamination
Two types of solid substrate surfaces were used in this study: amorphous hydrous
oxide (HFO) and amorphous hydrous oxide (HCO). Both
substrates were prepared by slowly increasing the acidic pH of either an
nitrate or nitrate solution of concentration with respect to
the metal ion in question. Metal ion adsorption or coprecipitation experiments
using amorphous hydrous metal oxide substrates are generally described in terms
of the concentration (ppm) of metal ion that was used to form the colloidal adsor-
bent rather than its corresponding specific surface area
The extent to which metal ions were removed from solution was assessed by per-
forming experiments carried out in three modes: direct precipitation, adsorption,
and coprecipitation. In all cases, the initial p H of the starting metal nitrate solu-
t i o n ( ~w
) as low pH 3.5 or less). Atomic absorption analysis confirmed that no
appreciable quantities of insoluble hydroxide precipitates were present in the metal
nitrate solutions at these low pH values.
Direct precipitation profiles of single metal ions were obtained by increasing the
pH of nitrate solutions of the metal ion in question. The pH and extent of pre-
cipitation were measured after equilibration at the final pH for 40 The experi-
ment was repeated with a fresh sample for each p H measured.
In the adsorption experiments, the colloidal substrate (either H F O or HCO) was
prepared by increasing the pH of an or nitrate solution of
concentration with respect to metal ion, and allowed to equilibrate for a
period of 5 When studying the removal behavior of single metal ions, the
metal ion of interest was then added to the pH-adjusted colloidal suspension and
equilibration was allowed to take place, after which time the equilibrium pH of the
suspension and extent of adsorption were measured.
When studying the removal behavior of the various metals in multiple
systems, the metal ions under investigation were premixed in a separate container
before being added to the reaction vessel containing the pH-adjusted colloidal
substrate. This procedure was followed in order to avoid the possibility of prefer-
ential adsorption of any of the metal types that may have occurred if the metal ions
were added to the suspensions separately. Again, the experiment was repeated with
a fresh sample for each pH measured; equilibration time was 40
In the coprecipitation experiments, the p H of the solution containing both the
or nitrate solution and the metal under investigation
was increased. Again, the pH and extent of coprecipitation were measured after

equilibration (40 min) had taken place and the experiment was repeated with a
fresh sample for each pH measured.
A series of adsorption and coprecipitation experiments were also performed in
the presence of an initially solution. The procedures adopted for
adsorption and coprecipitation of metal ions under these conditions were the same
for those performed in 0.0175 M background electrolyte.
The metal ion removal achieved during the various precipitation, adsorption, or
coprecipitation experiments was evaluated by determining the concentration of the
metal remaining in solution after equilibration had taken place. Each sample for
analysis was removed from the reaction vessel, filtered through a 0.22 Millipore
nitrocellulose filter into a glass vial containing two drops of concentrated nitric
acid, and analyzed using a Varian atomic absorption spectrometer
with a nitrous flame. The quantity of metal ion precipitated,
adsorbed, and/or coprecipitated was assumed to be given by a simple mass balance,
the mass added less the mass found in solution.
Figures 3-5 show the uptake of and respectively, using H F O

as the substrate (when required). The symbols marked on this graph, as with all
subsequent graphs, refer to experimental data; however, the lines d o not. Rather,
the lines are given by the "best fit" of the theory to be discussed in VII.
Adsorption and coprecipitation of (50 ppm) with HFO ppm with

respect to compared to precipitation of (50 ppm) alone.

4 Adsorption and coprecipitation of with H F O with
respect to compared to precipitation of (50 ppm) alone,
Adsorption and coprecipitation of with HFO with

respect to compared to precipitation of (50 ppm) alone.

Metal 7
The concentration of in aqueous phase changes from the initial

concentration 0 % removal) to nearly zero 100% removal) over a narrow
pH range (Fig. 3). This pattern is characteristic for aqueous metal removal by
adsorption or precipitation mechanisms James and Healy
for example, showed that the adsorption of onto changes from
approximately zero at pH 3.5 to 50% at pH 4.5 and 100% at pH 6.1. The pH
range for adsorption of onto HFO, given in Fig. 3, is very similar.
Comparable results were obtained by Chang et al. for adsorption
onto
The removal of (Fig. 4) follows the same qualitative pattern as the one
seen for The adsorption curve shows an increase in adsorption ranging
from zero at pH 6.0 to nearly 100% at pH 7.2. This pH range is consistent with the
results obtained by Duker Benjamin and Benjamin and Leckie
for adsorption onto H F O and by Rodda for adsorption
onto synthetic goethite. The results are not, however, consistent with those of
Kinniburgh et The p H range over which adsorption took place for
onto H F O oxide gel was found by the latter authors to be approximately
one pH unit lower than the results presented here.
The pH range over which was removed by adsorption in this study (Fig.5)
is 6.8-8.2, which is similar to that found by Barrow et for adsorption
onto goethite, but again dissimilar to that obtained by Kinniburgh for
adsorption onto H F O gel.
For both adsorption and coprecipitation mechanisms, the expected correlation
between hydrolysis and adsorption (or coprecipitation) is seen in Figs.
which hydrolyzes at the lowest pH of the three metal ions, was also removed by
adsorption and coprecipitation at the lowest pH. Conversely, which hydro-
lyzes at the highest pH, was removed by adsorption and coprecipitation a t the
highest pH.
In all three cases, the expected trend is observed: Coprecipitation results in
removal at a lower pH than does adsorption, which in turn results in removal at
a lower pH than simple precipitation. In the case of however, the difference
between adsorption and coprecipitation is barely perceptible. For the cases of
and the enhancement of coprecipitation over the adsorption is
very significant and is worth discussing further.
argued that the main difference between adsorption and
lies in the geometry of the adsorbent surface, that coprecipitation should be
considered a three-dimensional process leading to a greater effective removal capa-
city than simple adsorption. Such an argument would predict considerably
increased removal, at any given pH, for coprecipitation over adsorption.
Alternatively, the increased removal may simply reflect an increase in available
surface area as the H F O is formed. However, it was earlier argued that such an
increase would not necessarily result in a significant shift along the pH axis. For this
to occur, the increase in effective surface area probably needs to be of two or three
orders of magnitude.
Meng and Letterman and Taylor among others, pre-
sented evidence for the formation of mixed hydroxides of well-defined
try and solution behavior that differs from that of either hydroxide alone. One may

thus speculate that the existence of such mixed hydroxide species during
leads to enhanced removal compared to adsorption. Such species may also
help to explain one of the unusual aspects of the data presented in Figs. 3-5, this
being the degree of enhancement of coprecipitation over adsorption.
For the case of the enhancement is minimal, yet falls between
and both in terms of the pH at which hydrolysis occurs and in
terms of the pH at which adsorption occurs. One would expect to show
the greatest enhancement (as it does), because its removal is closest to the precipi-
tation edge for HFO. Therefore, one would expect. a large amount of to be
still in solution with is being coprecipitated. According to this argument,
however, one would expect to show the least enhancement because its
removal is at the highest furthest away) pH. This trend may support the
concept of mixed hydroxide species forming during coprecipitation in that such
mixed hydroxide species would probably differ in their affinity for one another.
Thus in the case of and there may not be such a complex, or rather
the complex formed may not have any special affinity for and hence there is
minimal enhancement of removal.
Experiments to remove metal ions from solution similar to those discussed in

were performed using and with HCO as the adsorbing or
cipitating colloid. It is not possible to perform adsorption studies of with
HCO, because the H F O forms at a lower pH than the HCO.
Adsorption or coprecipitation studies do not appear to have been performed
using HCO as the substrate, and hence comparisons between the data obtained in
this study with previous literature values are not possible.
The adsorption of and onto HCO is compared with that for
in Fig. 6. Again, the lines in Fig. 6 refer to the "best fit" of the model to be
discussed in VII. This is the case with all remaining graphs in this chapter.
Amorphous hydrous oxide (HFO) has often been used as an adsorbing
substrate not simply because of the ease with which it can be prepared but also
because of its recognized efficiency as an adsorbent The above comparison
shows that is an even more efficient adsorbing substrate than HFO.
Moreover, the difference between and HCO is significantly greater than the
minor variation that would be predicted by, for example, Harding and Healy on
the basis of their slightly different surface areas. This indicates that HCO is atypical
of hydroxides and oxides and has a specific affinity, at least for and
The origin of any such special affinity is not clear; however, it may relate to the
solution chemistry of has an unusual solution chemistry in
that it has a slower exchange rate than the other transition metals due to its
stable electron configuration This could well result in structure at the
HCO surface that enhances adsorption by, for example, preventing adsorbed ions
from readily desorbing and thus effectively shifting the equilibria to favor greater
adsorption.

FIG. 6 Adsorption of or with HFO with
respect to or HCO with respect to
The coprecipitation of and with HCO is compared with that for H F O

in Fig. 7. Again, HCO is seen as a far stronger adsorbent (or more correctly a
coprecipitant) than HFO. For the case of both and the
curves initially reflect the formation of HCO and may thus indicate a simple
entrainment process as the removal mechanism. This pattern continues at high pH
for removal. but removal is shifted slightly to higher pH values. Of the
two metal ions, is removed at slightly higher pH values than and this
seems to be enough (at high pH values) to shift the mechanism from entertainment
(or at least a mechanism that is dependent on the formation of HCO) to genuine
coprecipitation.
The ability of to remove and more efficiently (at any given
pH) than HFO has considerable implications for the application of adsorption
phenomena in the presence of Sanciolo et for example, showed
that metal ion removal using adsorbing colloid flotation is more efficient in
effluent samples containing the efficiency depending on the
concentration. The results obtained in this study suggest that this may be due to the
enhancement of adsorption and coprecipitation by the formation of HCO.
The influence of a second or third metal ion on the adsorption and coprecipitation
removal of metal ions is given in this section, initially using HFO, and then HCO,
as the adsorbing or coprecipitating colloid.

FIG. 7 of (50 ppm) or (50 with HFO with
respect to or with respect to
In single-metal systems, the adsorption or coprecipitation of the metal ion

alters the surface chemical characteristics of the substrate. In multiple sys-
tems, then, the surface seen by a second or third adsorption metal ion is different
from the original surface. This one might expect to see significant changes in the
removal behavior. Furthermore, it the metal ions have similar hydrolytic behavior,
they may interact in solution prior to, or in combination with, adsorption. Thus
simple adsorption experiments may involve coprecipitation among the adsorbing
metal ions, even if the substrate is added preformed.
The following sections detail adsorption and coprecipitation experiments invol-
ving H F O and HCO as the substrate and and in their
various permutations, as the competing metal ions.
The removal, by adsorption, of onto HFO in different solution environ-

ments is illustrated in Fig. 8. The four adsorption curves follow the same removal
profile, designated as curve 1 , indicating that a similar mechanism is operating in all
cases. The presence of and/or has little influence over the adsorption
profile of onto HFO. This presumably is because the adsorbs at a
lower p H than either or and thus the substrate surface, at the pH
values of interest, is unaffected.
The removal by coprecipitation of with H F O in different solution envir-
onments is illustrated in Fig. 9. The four coprecipitation curves again follow a

100
90 HFO
alone
with Ni
with Zn
with Ni, Zn
Adsorption of alone with H F O with respect to

and with either or or both.
Coprecipitation of (50 ppm) alone with H F O ppm with respect to

and with either 50 or or both.

single profile, curve 1, again indicating that a similar mechanism is operating in all
cases. The presence of and/or has little influence over the
profile of onto HFO.
The removal by adsorption of onto H F O in different solution environments

is illustrated in Fig. 10. The presence of has little effect on the adsorption of
however, the presence of considerably enhances the adsorption of
Enhancement of adsorption is rarely seen in metal ion adsorption experi-
ments, and even in this study the enhancement may reflect coprecipitation (which
has been shown to be efficient than adsorption) rather than adsorption as the
removal mechanism. The four sets of adsorption data follow two distinct trends,
one in the presence (curve 1) and one in the absence (curve 2) of Again, the
curves represent the best fit model using to be discussed in In this case,
since there are two curves, there are also two sets of modeling parameters.
Closer inspection of the data in Fig, 10 shows that there is a slight decrease in
in the presence of probably simply due to competition for
surface sites and subsequent decreasing of the surface area available for adsorption.
Experimental accuracy, however, does not allow a genuine distinction to be made.
onments is illustrated in Fig. 11. It can be seen that at pH values greater than about
7, the four curves are approximately equivalent and are designated curve 1. There
Adsorption of alone with HFO with respect to

and with either 50 ppm or or both.

Metal
alone
with
with Ni
with Cr, Ni
11 Coprecipitation of ppm) alone with HFO ppm with respect to

is, however, a slight deviation at pH values less than 7. Again. the presence of
enhances metal ion removal; however, the enhancement is small and occurs
only at low degrees of metal ion removal.
The removal by adsorption of onto H F O in different solution environments

is illustrated in Fig. 12. Two distinct curves are obtained, one in the presence (curve
1) and one in the absence (curve 2) of The presence of has little effect
on the adsorption of despite the fact that has almost completed its
adsorption at the onset of adsorption. There is a common argument (see,
and Hughes that ions compete for surface sites and that
adsorption of one metal ion should inhibit the adsorption of a second. Benjamin
and Leckie on the other hand, argued that different metals may adsorb onto
different surface sites and that the adsorption of one metal can have little influence
over later adsorption of a second metal. The data shown in Fig. 12 support this
latter idea, although the lack of sensitivity to the presence of could be simply
due to the lack of sensitivity to changes in surface area. Whatever the case, adsorp-
tion of appears to have little effect on the subsequent adsorption of
onto HFO.
The presence of however, considerably enhances the adsorption of
whether is present or not. As shown in
tates with to remove it a t a lower pH than by adsorption onto HFO. The
data shown in Fig. 12 probably reflect coprecipitation of with

HFO
12 Adsorption of alone with H F O with respect to

and with either 50 or 50 ppm or both.
13 of alone with H F O with respect to


onments is illustrated in Fig. 13. Two distinct curves are obtained; however, the
difference in this case is not simply the presence or absence of In contrast to
the adsorption results, coprecipitation with H F O in the presence of both
and displays a profile that is similar to that observed with H F O in
the absence of these ions, yet the presence of without still enhances
adsorption. The most likely explanation for this latter result is that the
of is enhanced by the presence of due to
coprecipitation.
The data in Fig. 12 suggest that the coprecipitates with at a lower
p H than does the effectively removing the from solution. The con-
sequence of this is that the enhanced removal of the by coprecipitation with
is not observed. This result was not observed in the earlier adsorption
experiments because in those cases sufficient was present to coprecipitate
with both and In this coprecipitation experiment, however, a much
smaller quantity of was available for coprecipitation with and
as a greater proportion had already coprecipitated with
The removal by adsorption of onto H C O in different solution environments

is illustrated in Fig. 14. The presence of does not affect the removal of
by adsorption onto HCO, and a single curve (curve 2) is obtained. It is important to
. 14 Adsorption of alone with with respect to

or with 50 ppm

note that at pH 5.5, where adsorption has commenced, is not fully
precipitated, and this would inhibit any enhanced removal at low pH. When mod-
eling the data, the reduction in overall surface area caused by incomplete formation
of the HCO was taken into account. The precipitation curve appropriate to
(curve 1) is included for comparison in Fig. 14.
The removal by coprecipitation of with HCO in different solution envir-
onments is illustrated in Fig. 15. The presence of has no effect on the removal
of by coprecipitation with HCO, and again a single curve results (curve 2).
Again, however, any enhanced removal at low pH would be by the incom-
plete precipitation of The precipitation curve appropriate to 250 ppm
(curve 1) is also included for comparison in Fig. 15.
The adsorption of onto HCO in different solution environments is illustrated

in Fig. 16. adsorption in the absence of follows the expected shape
(modeled by curve 2); however, the curve corresponding to removal in the
presence of is of an unusual shape. At a pH less than approximately 6.5,
removal is enhanced by the presence of and can be modeled by
assuming that there is interaction between and (curve 1). At higher
pH values, removal is, if anything, slightly inhibited and can be modeled by
assuming such inhibition (curve 3).
alone
with
Coprecipitation of (50 ppm) alone with HCO ppm with respect to

or with 50 ppm

. 16 Adsorption of alone with HCO respect to
or with
The enhancement seen at low pH is very unusual and is difficult to attribute to

coprecipitation mechanisms operating in tandem with adsorption.
could not coprecipitate with because the H C O was preformed. Any effects
due to remaining in solution or being stripped from the surface to form a
coprecipitate with would be seen in the data for adsorption onto HCO
[in the absence of and there would be no enhancement in the presence of
If were to coprecipitate with then one would expect to observe
enhanced removal of in the presence of onto HFO, which was not the
case.
It is possible, however, that a mixed and precipitate forms
and that this mixed precipitate is responsible for the enhanced adsorption. It was
earlier argued that the enhanced removal of in the presence of and
by H F O could be due to coprecipitation of a mixed hydroxy species,
although it was not strictly necessary to invoke such a phenomenon. In this case
adsorption onto in the presence of the concept of a mixed
hydroxide precipitate is more likely.
The removal by coprecipitation of with in different solution envir-
onments is illustrated in Fig. 17. In both cases the removal of is similar (curve
2) and initially follows the precipitation curve for the alone (curve
1). Subsequent metal ion removal follows a curve possessing a shape characteristic
of an adsorption or coprecipitation process. has little, if any, effect on the
percentage removal by coprecipitation of with HCO.

FIG. 17 Coprecipitation of alone with with respect to
or with
The influence of a complexing ligand (ammonia) on the adsorption and

of using H F O and is reported here. Before discussing the results,
however, it should be noted that direct comparison between the results presented
here and those of earlier sections is made difficult due to changes in the ionic
strength. It is necessary to use a high concentration of ammonia (added
as ammonium nitrate) to effect significant complexing, particularly to effect com-
petitive complexing with the hydroxide ligand. This necessitates the use of a much
higher ionic strength than was used in the studies reported in the previous sections.
When discussing the significance of results, the effect of ionic strength should
therefore be considered.
Hayes and Leckie reported that changing the ionic strength of the back-
ground electrolyte has a negligible effect on the adsorption of specifically adsorbed
ions such as and However, for alkaline earth
cations such as and the ionic strength had a dramatic effect.
Ionic strength effects have been shown to be negligible in the case of the adsorp-
tion of 5 and onto goethite in 0.01 and 0.1 solutions
and that of onto Chang et al. however, reported
decreased adsorption of 1 x onto as the concentration of the
background electrolyte was increased from 0.002 to 0.1 Schindler et
reported a similar decrease in adsorption with increase in electrolyte concen-
tration for and onto kaolinite. Barrow et also reported
a decrease in the adsorption of and onto goethite with increasing

background electrolyte concentration f0.01-1.0 M however, a negligible
effect was noted for under the same conditions. is an unusual
choice for a "background" electrolyte because, being nonsymmetrical, it has a
different effect on the surface of goethite on either side of the isoelectric point,
although of the adsorption data reported were obtained at p H levels lower
than the isoelectric point of goethite.
The influence of ionic strength on adsorption of metal ions is attributed to
two main effects, the effect on the solubility of the ions in solution and the effect on
the electrical potential of the surface Baes and described methods
to adjust dissociation constants to allow for differing ionic strength, and generally
the activity of heavy metal hydrolysis products decreases with increasing ionic
strength. Increasing the ionic strength also decreases the absolute values of the
electrical potential in the plane at which adsorption takes place For metal
ions such as that adsorb at relatively low pH onto oxide surfaces, the
decrease in positive potential of the surface increases adsorption by an amount
that approximately offsets the decreased activity of the hydrolysis species of
for example For such metals, changes in electrolyte concen-
tration have little effect on the adsorption profile. For metals that adsorb at higher
pH, the pH range usually extends into a region where the substrate surface becomes
negative, and as increasing the ionic strength decreases the negative surface poten-
tial, the adsorption of the ion also decreases. The extent to which this occurs,
however, is subject to debate.
The concentration of the background electrolyte has also been shown to have an
effect on the degree to which metal ions are included in coprecipitate structures.
Pingitore showed. by performing coprecipitation studies in both the presence
and absence of that the presence of at this concentration
reduces substantially the amount of that can be incorporated into a
coprecipitate.
To help understand the influence of ammonia on adsorption and coprecipitation, it

is useful to first look at the influence of ammonia on the solution chemistry of the
metal ions to be removed. The concentration of the ammonia ligand is pH-depen-
dent, forming the ammonium ion at low pH. It is assumed that the ammo-
nium ion does not, itself, complex with positively charged metal ions and serves
mainly to limit the amount of free ammonia available at any given pH.
Ammonia may form a series of with any given metal ion that have a
variable number of ammonia ligands per metal ion. Each of these complexes may
further interact with the hydroxide ion to form a series of complexes having a
variable number of ammonia ligands and a variable number of hydroxide ligands
per metal ion. The metal ion itself will interact with hydroxide ions and any elec-
trolyte that may be present to form a suite of these complexes as well.
Fortunately, a complete set of equilibrium constants have been obtained for the
system, and these are given in Table 1.

Hydrolysis and Ammonia Complexation in 0.5
Solution at
Equilibrium Constant
Annotation as used by and
The easiest way to use the constants presented in Table 1 is to graphically plot
concentrations of the various species as a function of pH. Given the number of
species involved, this is best performed by using three separate graphs, as shown in
Fig. 18. This figure will be referred to when discussing the species likely to be
present at any given pH value. For example, at 6, the species in highest con-
centration are and (Fig and these species are relatively inde-
pendent of pH near that The species (Fig. 18b) and
(Fig. are not present in high concentration, but the shapes
of their curves indicate that they become more and more important as the is
increased.
The data obtained from the adsorption of onto H F O in the presence of

0.5 are given in Fig. 19. It can be seen that the shape of the removal
curve is complex. showing several reversals of direction with respect to In this
respect, the results are similar to those obtained by and Fuerstenau

FIG. 19 Adsorption of (50 ppm) onto with respect to in
0.5 M
for adsorption of several aqueous metals onto a variety of oxide substrates.

et al. performed a similar experiment using adsorption onto
but did not see such reversals of direction. In their case, however, this may reflect a
difference in the way the results were obtained and, in particular, the way they were
plotted. They obtained equilibrium concentration-dependent, rather than pH-
dependent, adsorption curves and so may have missed this trend.
The results given in Fig. 19 show that the adsorption removal begins at
approximately pH 5 [as was observed for adsorption onto H F O in 0.0175
solution] and increases with pH until approximately 7.5. This cor-
responds to the point at which sufficient ammonia is present to form
complexes with the aqueous metal ion. Inspection of the speciation diagram
(Fig. 18) shows that the concentrations of complexes become domi-
nant at this pH. At pH levels corresponding to high levels of species,
the removal of decreases. At a pH of 9.8, however, the concentrations of
hydrolysis species and mixed liydroxammine species become signifi-
cant. At this point, an increase in the removal is observed.
The removal of by adsorption onto HCO in solutions is
illustrated in Fig. 20. The data around pH 6-7 do not follow the characteristic
curve, with the percent removal decreasing sharply followed by an
immediate sharp increase. The data are as two distinct curves (see
VI) to highlight the probable cause of this apparent deviation from normal adsorp-
tion behavior. Curve 1 represents the removal assuming that a
mechanism is operating. Curve 2 represents the removal assuming that an adsorp-
tion mechanism is operating.

FIG. 20 Adsorption of (50 onto HCO with respect to in
0.5 M
The removal of by coprecipitation with HFO in ammoniacal solutions is

illustrated in Fig. 21, and its removal by coprecipitation with in ammoniacal
solution, in Fig. 22. In the case of the pattern is similar to that observed for
adsorption. Coprecipitation, in this case, results in only a slight increase in removal.
In the case of HCO, however, the pattern is quite different, with removal consider-
ably enhanced at all pH values. These results indicate that the coprecipitation
mechanism is very effective and can overcome the effect of a complexing ligand
to promote heavy metal removal over a wide range of pH values and conditions.
A large number of models exist for modeling the adsorption of metal ions onto
colloidal substrates, and most of them can be adapted to include coprecipitation.
They all involve "fitting" parameters the equilibrium constant for adsorption
or the chemical specific free energy for adsorption) in order to match experimental
and theoretical results. Therefore, there remains some skepticism over the ability of
current theories to fully account for the adsorption process. Nevertheless, these
"fitting" parameters are generally held constant as a function of pH. Thus the trend
in adsorption (or coprecipitation) with pH is a guide to the success of the model. It
is not the purpose of this chapter to detail all the adsorption models or to compare
them. Rather, one model is used to test its robustness in modeling precipitation,

. 21 of onto H F O with respect to
in 0 5
Coprecipitation of onto with respect to

in 0.5

adsorption, and coprecipitation under conditions of simple metal ion removal and
metal ion removal in the presence of competing metal ions and a complexing agent.
The model chosen is a modified form of the James-Healy model for metal
ion adsorption.
The James-Healy model has been used extensively to model the adsorption
behavior of a series of aqueous heavy metals onto a variety of substrates. However,
there is still considerable uncertainty in many of its underlying concepts. The model
is based on determining the free energy of adsorption for all potential adsorbing
species i For example, in the study of adsorption in the presence
of a background electrolyte, such species are the free metal ion, its hydro-
lysis species, and its nitrate complexes. For each species, is divided into
three components: and The first component,
is associated with the replacement of part of the solvation sheath that surrounds a
charged ion in solution, with a solid surface. The second component, is
associated with bringing a charged ion into close proximity to a charged surface.
The third component, is associated with specific chemical interactions
not accounted for in the first two terms. Such specific interactions may include
hydrogen bonding, crystal field stabilization, hydration forces, and chemical bond-
ing. It was also originally envisaged that it could account for errors in the calcula-
tion of and but aspects of all three terms remain open to further
definition.
The James-Healy has been used to model the adsorption of
and onto of onto of onto
amphoteric polystyrene latexes of onto and goethite of
and onto synthetic goethite of and
complexes onto and of
onto and of mercury complexes onto In each of these cases, the
model was used in the original form proposed by James and Healy although
recent advances in physical chemistry can improve the algorithms used to calculate
the various free energy components.
There has been little improvement to the calculation of , by James
and Healy despite the simplicity of the approach, which involves a discrete,
rather than a continuous, change in solvation energy. Although this may be con-
sidered a weakness, it is appropriate because there is increasingly more evidence
that interfacial properties are of a discrete nature
By contrast, there is considerable evidence to suggest that the Nernst equation,
used to calculate the surface potential and thus is inappropriate in the case
of hydrous metal oxides et al. proposed the use of zeta
potentials rather than surface potential to calculate (and also
and this forms the basis of the modified model used throughout this chapter.
Of the greatest concern is the uncertainty associated with the origin of
This term is used largely as a fitting parameter, although there are some indications
of its phenomenological basis. and James for example, showed
that complexation of a metal ion with a unidentate such as chloride,

vents the adsorption of that metal ion; however, use of the James-Healy model can
predict this only if for the complex is set at, or near, zero. Fuerstenau and
showed that optimal modeling results are obtained if different
values are assigned to the different hydrolysis products of metal ions,
aqueous metal hydroxide species), although this is not necessary to obtain
reasonable agreement between theory and data. Importantly, Rodda
suggested that the use of differing terms lessens the credibility of a the-
oretical model.
In general, a significant (negative) indicates that adsorption is greater
than that predicted on the basis of electrostatic and solvation forces. Species that
exhibit this are the hydrolysis products of the metal ion. Species complexed with
simple other than hydroxide, d o not adsorb and cannot be given a
significant value.
nificance of the Chemical Free

to Adsorption
For each species involved in each experiment given in this chapter, a
value was determined by modeling the data with the modified James-Healy
model. In some cases, several sets of data could be used, although in most cases
the choice of was clear. Details of this can be found elsewhere
The philosophy behind modeling is to assume that independent of
pH, concentration, and the presence of other metal ions. It was, however, depen-
dent on the particular metal species and the particular substrate. Where possible,
kept constant for a given suite of complexes.
Several interesting features have arisen from the modeling process. for
zinc nitrate complexes, for example, need to be considerably smaller than for the
free metal ion and could be set to zero. for zinc
(mixed) complexes, however, must be smaller than for zinc complexes but
cannot be set to zero. All of the modeling shown in this chapter is consistent with
the general rule that hydroxide must be present to have significant specific interac-
tion and that any other complex (such as ammonia) diminishes this interaction
without stopping it.
The agreement between experimental data and modeled curves shown through-
out this chapter is very good, indicating that the model is quite robust.
The adsorption and coprecipitation profiles for removal using are very
similar. There is a small but significant difference between adsorption and
cipitation for and a large difference for using the same substrate. In
all cases, coprecipitation was as efficient as or efficient than adsorption, which
was in turn more efficient than precipitation alone.
In all cases, was removed at a lower pH than which was removed
at a lower pH than consistent with their hydrolytic behavior.

In all cases, H C O was a more efficient adsorbent than HFO.
Adsorption and coprecipitation of with H F O was independent of the pre-

sence of and/or because precipitates at a lower pH than either
or and apparently at a p H lower than or similar to that of any mixed
precipitate involving any combination of these three metals.
The adsorption and coprecipitation of with either H F O or H C O was
relatively unaffected by the presence of although there may be a slight
suppression of removal, possibly due to competition for surface sites.
The adsorption of onto HCO, however, provided an unusual profile in the
presence of Adsorption was enhanced at low pH but not at high pH.
adsorption onto H F O in the presence of did not show this unusual removal
profile.
The adsorption and coprecipitation removal profiles for with H F O and

HCO in the presence of 0.5 M solution were measured. The onset
of complex formation inhibited the adsorption and coprecipitation of
onto both H F O and HCO.
The modified James-Healy model for metal ion adsorption could be to

all the results (precipitation, coprecipitation. adsorption, single metals, multiple
complexing ion) as illustrated throughout this chapter.
Hart, (1982).
A. Jenne, Trace element sorption by sediments in soils: sites and processes, in
Symposium on Molybdenum on the Environment and Peterson,
Marcel Dekker, New York, 1977, Vol. 2, 425.
Krauskopf, Geochim. 9:1 (1956).
P. Hartz, and Environ. (1987).
Waste Treatment with and Other
Data Corporation, Park Ridge. 1977.
Introduction to Waste Water Treatment Processes, Academic, New
York, 1977.
P. I. H. Harding, and D. Sep. Technol.
(1992).
C. A. Money and Cranston, Reductions in Tannery Effluent by
Implementation of the Process with Liquor Recycling and a Rapid

Processing Technology, Melbourne Board of Works Industry Waste Grants
Scheme Report, CSIRO Division of Wool Technology, Melbourne, 1990, 9.
M. Wastewater sources, in Electroplating and Related Metal Finishing:
Pollutant and Toxic Materials Control, Pollution Technol. Rev. No. 46,
Data Corporation, Park Ridge, 136.
A. Davis and 0. Leckie, Colloid Interface (1978).
L. and A. A. J. Colloid Interface (1992).
Namasivayam and Ranganathan, Environ. Pollut. (1993).
Namasivayam and K. Ranganathan, Ind. Chem. Res. (1995).
Sass and Rai, Inorg. (1987).
Mustafa and I. Haq, Environ. Technol. Lett.
M. and W. J. Chem. Technol.
(1992).
Jackson. and K. Soil Sci. Am.
(1976).
M. Benjamin and Leckie, Colloid Interface Sci. (1981).
Benjamin and 0. Leckie, Colloid Interface Sci. (1981).
Bruninx and Van Anal. Chim. (1975).
J. I. Harding, and E. Mainwaring, Langmuir (1993).
Crawford, I. H. Harding, and Mainwaring, Langmuir (1993).
C. Ainsworth, Pilon, Gassman, and Van der Soil
Am. (1994).
A. Duker, A. and B. Appl. Geochem. (1995).
J. I. Harding, and Mainwaring, Colloid Interface
(1996).
J. Crawford, Mainwaring, and I. Harding, Colloids Surf.
(1997).
Guan, J. and Han, Min. Met. August 1994, p. 130.
Wiese and Healy, J. Colloid Interface Sci. (1975).
Houchin and J. Warren, Colloid Interface (1984).
A. Bell and J. Inorg. Chem. (1975).
Sprycha and Colloids Surf. (1990).
Sen, Chem. (1929).
Simon, Schulze, and R. Reinke, Anal. Chem. (1973).
A. Packter and Res. Technol,
Johnson and E. Colloid Polym. (1992).
0. James and A. Parks, in Surface and Vol. 12
Wiley-Interscience, New York, 1982, p. 119.
I. Harding and Healy, Prog. Water Technol. (1979).
0. James. and Healy, Faraday Discuss, Chem.
(1975).
0. James and Healy, J. Colloid Interface (1972).
and A. E. Stability Constants, Chem. (London) Spec.
Publ. 17. Metcalfe and Cooper. London, 1964.
L. and A. Stability Constants, Suppl. 1, Chem. (London)
Spec, Publ. 25, Metcalfe and Cooper, London, 1971.

Metal Oxides as 7
P. Rodda, Johnson, and D. Wells, J. Colloid Interface

(1993).
Soil Res. (1980).
and Jackson, Cation adsorption by hydrous metal oxides
and clay, in Adsorption of at Solid-Liquid A. Anderson
and A. eds.), Ann Arbor Science, Ann Arbor, MI, 1981, 91.
S. Kooner, Environ. (1993).
P. Huang, and in Adsorption from Solution (P.
New York, 1981, p. 61.
and 0. James, J. Colloid Interface (1974).
0. James and W. Healy, Colloid Interface Sci. (1972).
Hohl, and F. Dalang, Croat. Chem. (1976).
W. Fuerstenau and Colloid Interface Sci. (1987).
Osseo-Asare and Fuerstenau, Int. Miner. Process. (1979).
Osseo-Asare and Fuerstenau, Int. Miner. Process. (1980).
Luo and Huang, Sep. Technol. (1993).
Music and Radioanal. Chem, (1992).
A. Rauf, and M. Sep. Technol. (1993).
A. Bowers and C. P. Huang, Water Res. (1987).
A. Beveridge and F. Pickering, Water Res. (1983).
W. Fuerstenau and K. Han, Miner. Process. Technol. Rev. (1983).
C. Nathsarma and P. V. Bhaskara Sarma, Hydrometallurgy (1993).
Baes and The Hydrolysis of Cations, 1986.
M. Benjamin, Environ., Technol. (1983).
I. Harding and Healy, Colloid Interface (1985).
0. James and W. Healy, Colloid Interface (1972).
Benjamin and J. 0 . Leckie, Environ. Technol. (1981).
Benjamin and 0. Leckie, Environ. Sci. Technol. (1982).
Aoki and Munemori, Bull. Osaka Prefect., Ser. A. (1984).
I. Harding and Healy, Colloid Interface Sci. (1985).
K. Chang, Lin, Lee, Lo, and Colloid Interface Sci.
165:169 (1994).
J. Barrow, J. Gerth, and Soil. Sci. (1989).
B. Corey, in Adsorption of Inorganics at Solid-Liquid Interfaces A.
Anderson and A. Ann Arbor Ann Arbor, MI, 1981, p. 161.
X. Meng and Letterman, Environ. Technol. (1993).
and Voltz, Anal. Chem, (1971).
F. Taylor, Mag. (1973).
Benson, Mechanisms of Inorganic Reactions in Solution, McGraw-Hill, London,
1968.
Fendorf, Fendorf, Sparks, and Gronsky, J. Colloid Interface
(1992).
A. Netzer and Hughes, Water Res. (1984).
Hayes and 0. Leckie, Colloid Interface (1987).
C. Chang, A. Davis, and S. Kuwabara. Estuarine, Coastal Shelf
(1987).
P. Schindler, P. Liechti, and (1987).

80. Pingitore, Models of coprecipitation of and with calcite, in
Surface and Colloid Science (ACS Symp. 323) A. Davis and K. F.
Hayes, American Chemical Society, Washington, DC, 1986, 577.
81. Crawford, Thesis. University of Melbourne, 1995.
82. Moignard, 0. and Healy, J. Chem. (1977).
83. J. and Kott, Colloid Interface (1989).
84. Healy, D. Chan, and White, Pure Chem. (1980).
W. Healy, White, and Chan. Electroanal. Chem.
(1977).

Institut fur Chemie und
Dusseldorf, Germany
I.Introduction
II.Theoretical Considerations and Modeling
III.Experimental Considerations
IV. Investigations of Adsorption of Various Ions
A. Adsorption of H+ and OH-
B. Adsorption of cations
C . Elemental anions and o x o anions
D. Complex ions
E. Organic ions
V. Concluding Discussion
References
Adsorption of ions onto alumina is a problem of great interest in very different

fields. As can be seen from the journals mentioned in the reference list, it is impor-
tant in geochemistry as well as in limnology, in pollution control as well as in
catalyst preparation or soil chemistry, and even in unexpected disci-
plines such as nuclear techniques. The different applications lead to different ques-
tions and experimental conditions. For example, the "normal" concentrations are
low in and the p H is restricted to nearly neutral conditions. In contrast,
in catalysis there are rather high concentrations and the pH, in principle, is vari-
able, though sometimes not even discussed, probably as a consequence of the use of
older technical impregnation techniques in which conditions were
applied.

As a consequence we must expect to find a mosaic, or perhaps a puzzle, of
information, a puzzle with many missing elements. In order to get an overview
of ion adsorption on alumina in spite of the difficulties, in we discuss the
basic physicochemical principles, which are necessarily the basis for any serious
model calculations. As modeling of the adsorption process is treated in a separate
chapter of this book, it is not treated in great detail here.
A small section is dedicated to the experimental methods of special
importance. Various adsorption systems are treated in IV. Section deals
with proton and hydroxyl adsorption onto and also onto aluminum
hydroxides, as the principles turned out to be the same for any of these hydrated
surfaces. In the adsorption of "normal" cations is discussed, generally
following the position in the periodic table but mentioning other ions used earlier in
the same investigation. A discussion of "normal" anions (elemental ions and
nions) follows in Complex ions cations and anions are discussed
together as some cations may be converted into anions and vice versa
by changes in pH. Aquo complexes are treated as "normal" ions. A short review of
the adsorption of some organic ions follows in In some general
aspects of ions adsorption onto alumina, as obtained from the puzzle, the
chapter.
THEORETIC
In quantitative descriptions of adsorption it is usually expected that equilibrium is
reached, and thus the laws of thermodynamics may be used to describe the system,
so the equation
0
where is the free enthalpy of reaction, must be valid for any reaction, leading
to
where
standard free enthalpy of reaction
standard enthalpy of reaction
standard entropy of reaction
R gas constant
reaction temperature
equilibrium constant
product of activities of substances i
The activities of the substances are related to the concentrations by
where/, is the activation coefficient (in the case of ions, for ion pairs).
The following relation between the activity coefficient and ionic strength J is
derived from the extended theory for dissolved ions.

of Ions onto 71
where = is the ionic strength and is the charge of ion

For surface phases the electrostatic effects are usually considered directly, using
models with different numbers of layers for the positions of different ions relative to
the solid surface leading to the capacitance parameter. A description of
face activity coefficients is given by Liu et
For a complete model all reactions have to be considered:
1. Speciation reactions within the liquid phase
2. Speciation reactions in the surface phase
3. Equilibria between the phases
It follows that the number of parameters is great, and the information obtainable
from experiments is usually insufficient for a determination of all of them.
The problem may get even more complicated if equilibrium is not reached
quickly enough for all partial reactions. On the other hand, kinetics measurements
may help to construct a realistic model.
In the case of the alumina adsorption system, one may investigate the oxide/
acid-base system separately. This is necessary because are amphoteric
oxides of remarkable solubility with a number of different species in solution,
and the possible adsorption of protons and hydroxyls at the surface leads to the
development of surface charges.
Each surface site may be taken to consist of an unit with one replaced by
surface in this case the adsorption enthalpy will depend on coverage for high
loads, especially if the aluminum atom carries two hydroxyl groups. Alternatively,
the completely coordinated aluminum ions may be seen as centers, and these may
be compared with dissolved species dissociating stepwise.
Besides the mononuclear and partially hydrolyzed species
the solution phase also contains polynuclear species.
The latter may be formed by condensation of some of the mononuclear species,
especially the nearly neutral ones, to form intermediates to precipitation On the
other hand, the charged surface species may be seen as intermediates to dissolved
species.
In a similar way, the speciation and possible precipitation of adsorptive ions
have to be discussed before adsorption itself may be investigated. In practical
systems other equilibria may also play a notable role. for example, that of
and forming and and its consequential effect on acid-base
balance and precipitation.
Finally, the adsorption will depend on the standard free enthalpy of adsorption:
will cause the greatest differences between different adsorbed ions, leading to
physisorption in the case of small values and to chemisorption, with the formation
of chemical bonds, for large values. Partial models may be constructed for weak or
strong interactions.

is influenced most strongly by the changing structure of the hydrate
spheres of adsorbed and desorbed ions during the adsorption process.
Any real model calculations require more or less simplification. As only a very
few measurements were made at different temperatures, only equilibrium constants
or the corresponding values can be derived. Activity coefficients are often
disregarded by measuring at nearly constant ionic strength obtained by the addition
of a sufficient number of background ions. Here the equilibrium constants will
contain an "activity factor." Further simplifications will depend on the particular
system.
The typical curve for the adsorbed part of an ion as a function of pH
is often used to define the the pH value at which 50% of the total amount
of any ion is adsorbed. This value helpful for a comparison of different ions
depends on the initial concentration or on the concentration at called
In contrast to investigations of adsorption the gas phase, the number of

methods applicable to adsorption from the liquid phase is very small. On the one
hand this is caused by the fact that not all methods using either electrons or ions
can be applied in situ. In addition the adsorbents are normally powders with no
plane surfaces. As a consequence the results of quantitative adsorption measure-
ments are usually calculated from the difference between the liquid concentrations
before and after the adsorption process. In principle, any analytical method may be
used provided it has sufficient sensitivity; p H measurements with a glass electrode
and atomic adsorption spectroscopy (AAS) are standard, but and
ion-selective electrodes can also be used. methods are useful in the
case of small final concentrations. If electrochemical methods are used, one has to
consider that activities, not concentrations, are obtained. In the case of partially
soluble adsorbents, such as transition aluminas, their concentration should also be
determined, as well as those of all other constituents of the solution,
A detailed review of the spectroscopic methods applied to investigate ion
adsorption was given by Scheidegger and Sparks Electron spin resonance
(ESR) is one of the situ methods that can give qualitative and quantitative infor-
mation about metal ions with unpaired electrons. Complexation at the surface may
cause line shifts.
Electrokinetic measurements, observation of the motion of the adsorbent
particles in an electric field, depending on pH, provide independent information on
the point of zero charge (pzc) related to the adsorption constants for and OH- .
Ex situ investigations can be carried out with all the methods usually applied to
gas-phase adsorption problems, provided they do not require flat surfaces. These
methods are of limited value, as separation from the liquid phase will generally
influence the composition of the surface unless the adsorption enthalpy is very
great, which means that a stable surface compound may be formed. Under this
condition all additional information may be useful.

of Ions onto
The adsorption of protons and hydroxyl ions is of fundamental importance, as it

greatly influences the properties of the adsorption system. Moreover, the investiga-
tion of this "simple" system may provide information about the distribution (num-
bers and adsorption enthalpies and entropies) of the surface sites. In order to
describe the we first need some values determined independently: the con-
centration of the alumina suspension and the specific surface of the solid. Both
values will vary to a certain degree with dissolution of the oxide, an effect that will
be small in concentrated suspension. The dissolution process leads to a "living"
surface, one on which the adsorption centers are continuously formed and
destroyed; this makes rather doubtful the information about the energy of surface
centers taken from investigations of oxides dehydrated at higher temperatures. It
can be expected that the high energy centers will be destroyed first by the dissolu-
tion process.
The parameters describing the system, in addition to suspension concen-
tration and specific surface area, are the concentration of adsorption sites on the
solid surface and the equilibrium constants in the adsorption equations, provided
Langmuir-type adsorption, only one type of adsorption centers with constant
adsorption enthalpy, is assumed:
+ OH- + (8)
(The formulas represent the formal addition of one or two protons or

not necessarily the correct structures! The water activity is not included in the
constants.)
In principle, the surface speciation of alumina may be compared with that of the
species in water solution (here the lattice structure is replaced by O H in
the liquid phase) (Fig. 1) or the surface hydroxyl groups may be compared with
those of water (the surface aluminum ion is replaced by a proton).
The solubility of alumina strongly increases with decreasing pH. so dissolved
and will tend to stay dissolved because of the high charge, whereas
neutral and weakly charged species tend to condense. Thus in the pH region where
most other ions are adsorbed, and are present in low concentrations
only Taking water as a model substance for =A\- -OH, only one type of
center can be present, but the adsorption enthalpy of such a center should depend
on the coverage if neighboring charged centers are present on the surface. Under
these conditions only one positive and one negative species can exist, in contrast to
Fig. la.
The treatment of the surface reaction gets more difficult if centers of different
acidity or basicity are considered. Contescu et al. discuss experimental results

1
0.8
0.6
0.4
0.2
0.0
0 2 4 6 8 1 0 1 2 1 4
value
.1 Relative concentrations of species vs. (A) Speciation of aluminum in

liquid-phase (water) (from Ref. 5). are valences of species.
species = (B) Speciation of aluminum in the surface phase (from
Ref. 7). Speciation of liquid phase.
indicating heterogeneous surfaces, and et developed a surface

model (MUSIC) to describe such systems.
The basis for the investigation of equilibria is generally the titration of a suspen-
sion of the adsorbent with acid or base. Figure 2 shows some examples that also
give an idea about the reproducibility obtained with of different ori-
gins. A problem may arise from the kinetics of the processes. observed
concentration changes for 24 h when adding so different
rates may influence the result.
Schulthess and Sparks pointed out that erroneous results were often
obtained because the consumption of acid or base as a consequence of dissolution
processes was not considered. In the case of alumina this seems to be of great
importance and will vary with the solubility and the speciation of the dissolved
ions, depending on pH. They proposed a of the
liquid to correct values. In Fig. 3 a comparison is given of the difference
between proton and hydroxyi surface concentrations and respectively,

3 A 0.001
A 0.01 NaCI
E 0.001 M
0.01
+ 0.001
1 C 0.01 NaCI
-1
1 2 3 4 5 6 7 8 9 1 0 1 1 1 2
value
Difference of and O H coverage as a function pH. Measurements at

different NaCI concentrations by various authors. A, Ref 14; Ref 15; Ref 7.
obtained directly by the titration method (Fig. 3a) and from model calculations
(Fig. 3b)
reports that surface charge pH, and its derivative with
respect to pH vs, pH can often be described on basis of pseudo.
homogeneity. In the case of transition (and of gibbsite) this may, at least
partially, be caused by the fact that these oxides are remarkably soluble, which will
lead to more homogeneous surfaces, as the most active centers, having the highest
energies, will dissolve first.
The adsorption of protons and hydroxyls necessarily leads to charges on the
surface, which are balanced by counterions of opposite charge in the solution.
Assumptions about the structure of the charged layers, which also contain other
adsorbed species, lead to different models double- or triple-layer models).
The point of zero charge (solution pH at which the surface is uncharged) is
related to the constants H I ) and by
pzc = - HI) +
where
provided that no surface sites are occupied by other ions. In the presence of back-
ground ions, an extrapolation to zero concentration is possible. H2) and
can be disregarded at this pH; in most publications they are not used
at all, but they are introduced here as they are to be expected theoretically. The pzc
can be determined independently from electrokinetic measurements provided the
isoelectric point (iep) is equal to the pzc. This reduces the number of constants to be
derived from the titration experiments, or it gives a tool to test the reliability of the
titration provided the pzc results are trustworthy themselves.

model calculation
A
Difference of and O H coverage as a function of All measurements

with prepared according to Torkar BET areas A. 180
(from Ref. 18); B. 244 Ref. 19); i70 (From Ref. 18).
A t this point one may ask whether a second constant can be defined in analogy
to the ion product in the liquid water phase:
Such a product was indeed used in logarithmic form as D P K but in con-

trast to water, which mainly consists of neutral water species even in acidic and
basic solution the neutral alumina surface phase is only stable
near p H 7 This shown in Fig. 4, in addition to Fig. 1 , for systems differing
mainly in the BET surfaces. S o D P K is a constant for p H = pzc only. But, as even
under these conditions it is not certain that = 1, it is more reasonable to
describe the system by using the equilibrium constants.

Speciation of in the liquid phase, obtained by model calculations from
experiments with prepared according to Torkar A, Ref. 18;
from Ref. 19. Numbers are valences of species.
Lewis-Russ reviewed methods for determining pzc values, and Table 1

contains values for different from the literature. It can be seen that the
pzc values vary by about 5 units, and thus they cannot be used to replace direct
methods of obtaining equilibrium constants. The pzc values, if obtained by
base titration, reflect uncertainties even for the corresponding adsorption constants.
Therefore constants should be determined with the same material and under the
same experimental conditions if the values are to be used in model calculations for

Point of Zero Charge (pzc) of Determined by Various Authors
Oxide Method Background ion
Titsation 0.1 M
Titration
Hydroxide Electroacoustics 3M
Electrophoresis
Titration, ion
adsorption,
electrophoresis
None
0.3 M
--
r
Triple-layer model None
a Triple-layer None
Nat. prod.' Electrophoresis
prod.' potential
prod, = product
adsorption of other ions. Table 2 contains some adsorption constants for protons
and hydroxyl ions. Other cannot be compared directly, as they d o not
contain electrostatic terms, they are not intrinsic values.
The concentration of surface centers is taken to be equal to the number of
hydroxyl groups in a monolayer o n the surface other values obtained
by acid-base titration tritium exchange and equilibration experiments
combined with calculations ("long time Table 3 gives
a n overview of numbers of centers determined by various researchers. Most values
are found between 1.0 x and 3.2 only that of [7]
(1.7 x clearly lower, possibly because of the back-titration tech-
nique. So the assumption that the number of centers is equal to the number of
surface O H groups seems to be acceptable, though in principle different of
Values for and of

Ref. Oxide type
7.2 9.5 Intrinsic
4.45 8.45 Intrinsic
5.7 7.9 Intrinsic
5.2 7.9 Intrinsic
6.90 10.90
4.47 11.5
5.13

of
Number of Centers
Ref. Oxide .Method
Titration
exchange
Exp + mod
Exp + mod
Exp + mod
OH groups
OH groups
OH groups
centers for and might be expected. Centers different in number

adsorption energy may exist for adsorption, as mentioned before.
In addition to Fig. 1, further surface speciation data are shown in Fig. 5 . Figure
5a shows the relative abundance obtained by et Fig. shows the
of centers obtained the same model calculation for all measurements
by Baumgarten et and Figure 6 adsorption isotherms for
and from data of Ref. 18 assuming isotherms and equal
numbers of centers for and OH- ions. The model used assumes
geneous The speciations in the liquid and the surface are calculated
the respective constants. Modified solubility products. with the
activity of the solid proportional to the coverage with neutral sites and possible
consideration of charges, serve to calculate the interphase
The numbers of publications about the adsorption of different cations onto alu-
mina are very different depending on the practical importance of the adsorption
process. A lot of is done with hazardous ions, or ions of known catalytic
activity, with the ions themselves or after their reduction to finely dispersed metal
particles. the other the literature about other ions, being of scientific
interest only. is limited.
Any specific adsorption of ions takes place in competition with that of or
provided it takes place at the same centers:
where denotes a free surface center. Here both surface and specia-
tion of the metal ions in solution must be known, with being equivalent to

surface speciation> assuming one positively and one
charged species each. (a) Relative amounts of species, data from (b)
coverages, of centers obtained by model calculation
from Ref. 18; from Ref. 19; Ref.
One adsorption equilibrium, according to reaction (1 must be given. As usually

background electrolytes (normally alkali metal nitrates o r halides) are used to
obtain constant ion strength, these ions may compete for centers because of
high concentrations. The possible formation of precipitates consisting of hydro-
xides; and, if is completely carbonates to be
considered. Often it is difficult to exclude these processes.
Alkali metals are examples of ions of limited direct interest. In most adsorp-
tion experiments their salts are used as " background electrolyte,'' to avoid greater
changes in the activity coefficient during the reactions by keeping the ion strength
nearly constant. Comparison of adsorption measurements of ions in the
presence of different amounts of alkali ions o r metal ions reveal that, indeed,
their influence is limited (Fig. 7). A direct measurement of the amount adsorbed,
as a difference between the added concentration and the concentration after the
adsorption process, is thus difficult> the difference is very small. As stronger

.
6 Adsorption isotherms of calculated from Ref. assuming
Langmuir adsorption and equal numbers of adsorption centers.
covalent forces are absent, electrostatic forces will be dominant. Another effect
may be of greater importance here, too: As there is no evidence of strong specific
interaction, one has to consider the corresponding anion also; if the anion is
adsorbed specifically, in the inner layer of the double- o r triple-layer system,
the corresponding cation will be enriched in the outer compensating for
the charge.
0 NaCl
0.001 NaCl
NaCl
0.05 M NaCl
. Influence of different concentrations of on and adsorption as a

function of (From Ref. 7.)

6 8 9
pH value
Adsorption of at different temperatures from a solution containing 0.1

as a function of (Calculated from Ref.
Figure 8 the adsorption of sodium ions at two temperatures from a 0.1

NaCI solution as a function of pH as calculated from Ref. 48. degree of
adsorption is very low compared to specifically adsorbed (chemisorbed) ions.
used various N M R techniques (MAS, CP-MAS, quadrupolar
NMR) to investigate adsorption on and reported the occurrence of
two adsorption species, Al-0-Na and = in addition to "adsorbed"
at coverages Some information on indicating adsorption com-
parable to that of is given in Ref. 48. An equilibration of Na-doped
with (Fig. 9) seems to indicate much higher adsorption,
but probably equilibrium was not attained. Shiao et al. discuss distribution
Surface coverage with and dissolved concentration after equilibrating

Na-containing alumina with solutions. (From Ref.

sorption of Ions onto 7
of ions between the surface and the liquid phase in 0.05 M Assuming a BET
surface of 100 this value on the order of looks reasonable
at 8 and a liquid concentration of 0.1 In contrast, shows a 100-fold
increased distribution coefficient, and a 10"-fold one.
The adsorption of alkaline earths onto was described by Huang and
Stumm Examples of adsorption depending on pH are shown in
Adsorption onto a slightly negative surface may lead to a reversal of the surface
charge, the surface speciation curves are shifted to higher p H values by
alkaline earth ions compared to those of pure alumina. From the corresponding
experiments it was concluded that ions are chemisorbed, or at least bound in
an inner surface layer. The logarithms of equilibrium constants for the
reaction
are: Mg, -5.43;Ca, -6.06; Sr, -6.27; Ba, -6.59. Adsorption on fresh alumina gel
was described by et As expected, the adsorption tendency is the
same as for alumina, though a direct comparison is not possible because no specific
surface can be defined. The results are discussed in terms of values.
Measurements with Ba and on show that is constant for
concentrations below 1 x while it increases linearly with log above
this concentration. At low concentration the value for both is 10, increasing
about one unit per decade of concentration for Ba above
. 10 Adsorption of main group ions. Mg, Sr. 2.5 x M in 1 M

on (0.093 M Al) (from Ref. 51). In: 3.0 in 0.01 M on
(from Ref. 52).

The adsorption of and In 3x on was investigated by
et Under these conditions the observed values were 2.65 (Ga) and
(Fig. 10). et derived the following equilibrium constants (in loga-
rithmic form):
SOM-OH
From kinetic results they conclude a two-step mechanism:

1. Proton release from the surface
2. Coordination of the metal ion
This description of the group ions of the first three columns of the periodic table
shows that with the exception of alkali metal ions all these ions are adsorbed speci-
fically; the same is true of other ions, especially the heavy metal ions, which generally
are of greatest practical interest. The other main group elements form anions or
oxoanions directly. In the case of equal valence, the transition metal ions resemble
each other more than the main group elements do, so more generalization should be
possible.
Adsorption kinetics will be discussed first. Figure 11A shows the equilibration of
the alumina system after the addition of Figure shows that the
adsorption of Cu, with liberation of protons takes about 24 h to reach equili-
brium. Figure gives kinetic data for the adsorption of Co, Fe, and ions
evidently in equilibrium after about 1 On the other hand, pressure jump experi-
ments of adsorption used a time scale of These results are
mentioned to demonstrate that reactions taking place in the adsorption system
may have rates that differ by several orders of magnitude.
Figure 12 gives an example of the partial irreversibility of the adsorption of
copper ions. and Evans give other examples of poor reversibility. A
reasonable choice of reaction time is a serious We will discuss the transi-
tion metal ions, beginning with the first column, containing Cu, and but, as
mentioned earlier, we will also discuss other ions treated in the same publication in
order to enable comparisons. Because of their revelance, measurements of interac-
tion with and are also discussed. Figure 13 shows the pH depen-
dence of transition metal ion adsorption on hydroxide and oxide.
Electron spin resonance studies reveal that Cu is adsorbed in dispersed
form in tetragonally distorted octahedral symmetry, with the signal intensity
increasing linearly until a is reached (6.3 at Cu
on the dehydration process changes the spectra but can be reversed
by adding water. This indicates a strong local fixation of Cu, which withstands
drying and subsequent wetting without forming clusters. Rudin and Motschi
used ESR and with and proposed a structure of Cu bound to two
atoms via oxygens and four ligands. et al. compared pressure

sorption of Ions o
. 11Kinetics of equilibration. (A) Dissolution of after addition of

(From Ref. 16.) Consumption of and pH change during adsorption on
(From Ref. 47.) (C) Adsorption of different metals at (From Ref. 54.)
jump experiments with sorption data and found log of deprotonated surface
groups with of 9.71 and 9.59 L
Hachiya et al. found a linear relationship between the logarithms of
intrinsic adsorption constants and the hydrolysis constant of the metal ions for
Cu, Pb, Co. and They described pressure jump experiments with
showing fast and a slow process attributed to two surfaces sites - with
and The relaxation times decreased with metal ion
and proton concentrations. The kinetic constants for adsorption and desorption are
(7.24 2.0) and (3.1 0.9) The order of the
adsorption rate constants corresponds to that of the water release reaction from the
aquo complex in the liquid phase.

amount of added
12 of the adsorption process. Desorption of Cu adsorbed on 0.1 g of

by addition of different amounts of At = (From Ref. 47.)
Baumgarten and precipitation and adsorp-

tion of Cu, Co, and Cr at and interpreted the results on the basis of a
model, considering dissolution of alumina. The adsorption is described by
The and values (negative logarithms of the adsorption constant and the
solubility product) are given in Table 4. In the case of Co and it is difficult to
separate the effects of precipitation and adsorption.
Further results were obtained with and et al.
discussed adsorption isotherms of and C o ions on aluminum
hydroxide at different temperatures and derived the thermodynamic values shown
in Table 5.
4 Negative Decadic Logarithms of Adsorption

Constants and Solubility Products
Source: Refs. 17 and 47

of Ions onto
Degree of adsorption of transition metals as a function of pH. (a)

1.25 x on alumina (From Ref. 51.) (b)
8.4) x on (2 BET area (From Refs. 19 and 47.)
Thermodynamic Data for Adsorption on

Ion AHo AS 3
K)]
Source: Ref. 62.

Pavlova and Sigg measured equilibrium constants for the reactions
2(= OH) + +
The constants are given in Table 6. Copper adsorption at "low pH" in contrast to
that of Ca, Mg, and is mentioned by Different values for
the adsorption of on hydroxide gibbsite and boehmite (5.7) may
be caused by the different numbers of centers.
The second transition ion group contains Zn, Cd, and which are known to
be strong poisons. Besides the works cited already. and give
curves for the removal of of different initial concentrations by different
amounts of Brady shows the temperature dependence of Cd and Pb
adsorption. The speciation of Zn and Cd in the liquid phase necessary for mode!
calculations is reported in Refs. 4 and 67 with remarkable differences, especially
for the species. Liu et [4] observed a linear relation of the logarithms of
adsorbed and dissolved amounts of and et al. report
extremely slow equilibration of with alumina in the presence of
and Evans studied the adsorption of Zn and Cd ions from very dilute solutions
on
hydroxide was used as an adsorbent in a of publications.
Coprecipitation of and was by Simon et al. in the
absence and presence of Shiao et demonstrate that cadmium
distribution coefficients between liquid and surface are practically uninfluenced by
the concentration, while has a remarkable influence. According to
et and Packter and Derry coprecipitation leads to the forma-
tion of Zn Besides investigating various ions and co-
workers describe sorption on especially that of Zn, at
different temperatures. They observed isotherms and derived thermody-
namic data. Kinetic studies were carried out by Simon et using
graphic techniques. In their experiments equilibrium was reached after 3 with
two consecutive first-order reactions, in contrast to the findings of Lo and
Negative Decadic Logarithms of Adsorption

Constants
Ion
- 2.1 -7.0
- 2.2 -8.1
-3.6 -9.2
-3.1 -7.9
-3.8 -10.0
-3.5
Source: Ref.

who also observed two reactions on amorphous alumina, but with much
slower equilibration, probably because of the pore system of the adsorbent. Even
desorption processes were slow, but they depended on the solution pH.
Of the next three groups (Sc, La), (Ti, Zr, and (V, yttrium
was investigated by in a work about adsorption thermodynamics
that also treated Ni and The adsorption behavior of one other rare earth ion
was measured: Adsorption of Eu turned out to be a function of pH and the con-
centration of One publication about the adsorption iso-
therms of Nd, and Er is of little value, as the pH was not measured
Of the members of Group VIB, M o and W were investigated only as
the adsorption of only is described. Static and kinetic measurements
done by and Chang et showed fast and slow reactions, the
slow one with a linear relation between log k and pH. measured polarographically
The pressure jump technique and equilibrium measurements with different
concentrations were evaluated with the triple-layer model, assuming var-
ious adsorption equilibria. Different assumptions on the adsorbed species gave
equally good fits.
Typical adsorption behavior was observed by Baumgarten and
for Cr (Fig. Here the rest concentration at a given pH is clearly
lower in the presence than in the absence of alumina (pure precipitation). ESR
measurements show a broad signal = which is interpreted as merely
adsorption onto the surface without incorporation into the surface; a change of
color from green to violet was observed as a consequence of exchange. Zaki
et used UV diffuse reflectance and X-ray diffraction and found highly
dispersed chromate species on
Some work has been done with Mn in addition to that mentioned already
The kinetics of adsorption on was measured for different initial
concentrations, showing equilibrium after 30 min A similar time to equili-
brium was found by Roy et 100 min. In static experiments they showed
redissolution at high pH, probably by oxidative formation of and
measurements indicate two forms, one reversibly adsorbed
and one irreversibly adsorbed. and Evans showed that the influence
of the anion on the adsorption of is almost negligible. The
desorption of adsorbed Mn brought about by lowering the pH value is incomplete.
In the Group VIIIB Fe attracted little interest, probably because
of the two possible valence states. The pH dependence of Ni adsorption by
was shown by de Bokx et they even stated that aluminate" is
formed in the case of low coverage, while higher coverage leads to the formation
of an phase. They give a diagram showing a linear relationship between the
logarithms of the intrinsic equilibrium constants and the hydrolysis constant. Shiao
show equal precipitation of C o on two different samples.
Isotherms obtained for the system at different pH values are shown
by and Lycourghiotis Adsorption isotherms of Co and Ni on a-
given by et al. were evaluated as isotherms but
cannot be compared directly with those mentioned before, as the number of centers
is unknown here.

and investigated Co and adsorption
onto (as well as that of and Cr) and evaluated the data on the basis of a
surface complexation model. and Lycourghiotis discuss the mechanism
of and deposition onto pure and F - d o p e d
al. interpret their EXAFS and TEM results of the (Linde
A alumina powder) system as a surface precipitation of with many vacan-
cies, while Scheidegger et al. report, also on the basis of EXAFS, the formation
of mixed cation phases. Finally, Brown et show in situ EXAFS and X-ray
absorption measurements with the result that Co (and Pb) adsorb at pH 6.8 on
as inner-sphere complexes without forming three-dimensional precipitates or
diffusing into the oxide, but species are formed on the oxide.
According to Mieth and Schwarz adsorption on is complete
after 60 min.
Anions have attracted much less interest than cations. Some of them, such as
nitrate, are not specifically adsorbed, they are not able to replace surface OH
groups. is adsorbed to a maximum of 2.7 at pH 2.5 from a
0.1 M solution Figure 14 shows the pH dependence of adsorption for
some anions. It must be stressed that, in contrast to the figures for cations, the
adsorption of anions cannot be compared directly, as the measuring conditions are
14 Adsorption of anions. From Ref. 95, conditions not given.

M; Ref. 96.) = x
0.1 M 1g Ref. 97.)

very different. However, it can be seen that the pH dependence differs from that of
cations, as expected: The greatest
p
is adsorbed at low pH. In contrast to Cl

mentioned before, and are adsorbed nearly completely at low
p
pH, indicating specific adsorption. Petkovic investigated adsorption of

B r , and I onto colloidal by titration in the respective
potassium salt solutions and obtained the following values of negative logarithms
of intrinsic surface complexation constants (TLM): 5.6 5.4 5.5
(Br ), and 5.5 (I ). They concluded, because of the equal values, that the adsorp-
tion is not specific. A comparison with other or instead of shows
remarkable differences.
Borate sorption investigated by Toner and Sparks shows the maximum
values at about pH 8.5 with a slow decrease to lower pH values and a strong
decrease to higher ones. Saeki and Matsumoto discuss selenite adsorption
onto various adsorbents. 5.3 evidently had too small a surface
to give meaningful results. with = 126 the maximum adsorption
(80%) was found at pH 6, decreasing to 20% at both pH 3 and p H 10.
et show adsorption isotherms of ammonium metavanadate for
different pH values. They also describe the desorption of permanganate, chromate,
and metavanadate at different pH values in the presence of different background
ions. EXAFS was used to investigate the adsorption of meal oxoanions (Re,
Cr, W) on y-alumina It shows that molybdate and tungstate are
adsorbed more strongly than the other oxoanions. Tungstate adsorption, especially
its mechanistic aspect, is treated by Karakonstantis et Adsorption of
molybdate was investigated by al. adsorption isotherms are
shown for different conditions. UV-Vis spectroscopy was used by et al.
to follow the adsorption of meta- and paratungstates on
In principle, all ions form complexes in solution, normally aquo complexes. Here
we discuss complexes with ligands other than They are treated together,
whether cations or anions, as their charge depends on the number and charge of
the ligands for a given central ion.
The inorganic ligands that were investigated are and Brunelle
gives an overview of the tendency of elements of Groups 8, and to form
or anionic complexes and uses the adsorption of or anionic com-
plexes for catalyst preparation
Amino ligands, not changing the charge, were used in the following investiga-
tions. Spielbauer al. described aquo complexes, using Pd speciation
in the liquid phase described by Rasmussen and Jorgensen containing
for comparison. They interpret their results on the basis of the triple-layer
model (TLM) of Davis unfortunately without experimental values
above p H 11.5, where the adsorbed amount should decrease according to the
model calculation.
Contescu et al. gave adsorption isotherms for for pH 8 and
and one for pH 10.6 was given by Schwarz et al. Fuerstenau and
showed the adsorption of Cu, and C o (1 x M) at

5 6 7 8 9 10 11
pH value
FIG. 15 Adsorption of complexes. Co: 1 on +
in the presence of 0.1 M + 160m 2 (From Ref. 113.)
different pH values by alumina in the presence of total ammonia.

Figure 15 gives results for alumina (Linde A, Union Carbide, The results
are interpreted on the basis of an extended double-layer model. Interaction of
alumina (7 and with complexes formed at different
p H values was investigated by Contescu and Vass Experimental isotherms are
compared with calculated Langmuir-type isotherms (Fig. 16). In another paper
Contescu and Vass gave kinetic data on adsorption of palladium tetrahalide
,
c( moles
16 Adsorption isotherms for Pt complexes on + depending on pH.

(Starting from pH 1.92, ammonia was added to reach higher pH values.) (From Ref.
114.)

anions Br) and proposed chemisorption by ligand exchange. Adsorption
kinetics for was shown by equilibrium was reached
after about 90 min. According to Mang et al. sorption on is
maximal at pH 4. A simulation assuming ligand exchange gives better results than a
purely electrostatic one. Cadmium coprecipitation with aluminum is reduced in the
presence of
Cheng and Huang discussed the adsorption of ferri- and ferrocyanide,
giving adsorption as a function of pH and distribution coefficients depending on
liquid concentration. studied the adsorption of
and on y-alumina. The kinetics of Co
fophthalocyanine was described by Belikhmaer and
Organic ions have attracted the most interest in the form of and anionic
surfactants, though other organic ions may be a problem too.
and Motekaitis reviewed the adsorption of aliphatic acids (maleic,
citric, and acid), salicylic acid, and a number of sulfone acids
acid)-
and nitrogen-containing complex ligands (NTA, EDTA, DTPA). Equilibrium con-
stants are given for the relevant equilibria. An example of the application of organic
acids (oxalic acid, salicylic acid, is aimed for better dispersion of Pd on a
catalyst carrier by coadsorption The adsorption of a surfactant
(TTAC = chloride) is shown in the form of an
adsorption isotherm for pH 10 by Huang and Somasundaran that of anionic
surfactants and sodium octylbenzenesulfonate) by
et al. and et (Fig. 17). The load of
2
2x molim shows that the whole surface is covered, assuming that
an area of 0.28-0.83 nm2 is covered by one molecule. The anionic
show the same normal pH dependence as an inorganic anion, with greater adsorp-
tion at lower pH values 10).
The adsorption of nitrophenols onto alumina pretreated with various acids is
shown in Refs. 124 and 125, unfortunately without giving the final pH values.
First, one point may be mentioned, which was not discussed before: If the intention
is to prepare a surface with an ionic adsorption layer, as for catalyst preparation,
then adsorption from the gas phase may be an alternative. In this way organic ions
may be adsorbed onto an alumina surface dehydrated at a higher temperature.
Infrared spectra reveal that a surface aluminum salt was formed under these con-
ditions Halogenated surfaces can be formed in the same way
Complexes can be adsorbed in this way too, though normally neutral complexes
are used here because of the better volatility.
Although it is not the intent of this chapter to thoroughly discuss model calcula-
tions, they also have to be regarded to a certain extent. In recent years a number of

-9
-5 -3 -2 -1
log (c) in moles
Adsorption of surfactants on a-alumina (0.03 M A,

paraxylene (pH 8-9; = Ref, TTAC
chloride; pH = 295 (From Ref.
p-octylbenzenesulfonate; pH T 323 (From Ref.
models have been developed that are broadly based on the surface
model of et which, in agreement with the spectroscopic results,
shows a chemical interaction at the surface. Some recent extensions of this model
try to close the gap between adsorption, precipitation, and coprecipitation with
different assumptions - surface precipitation, surface coprecipitation, the forma-
tion of a continuously variable surface layer or competitive precipitation
of pure or mixed oxides A revised physical
model based on very different assumptions was published recently by
and All models give fairly good fits to experimental values.
Robertson and Lecky demonstrate that equilibrium constants are "condi-
tional," they vary the experimental conditions, which means either that
not all experimental data were sufficiently considered or they were not correctly
implanted into the model.
The fact that different models lead to a sufficient fit of quantitative experimental
data shows that additional information be used for discrimination.
Spectroscopic data reveal the formation of chemical bonds between adsorbed
ions and the surface, which led to certain convergence to surface complex models,
and extended to considering adsorption in the case of high ion concen-
tration the solution.
A serious problem is the fact that for most models the number of model para-
meters is great and the information obtained the experiments comparatively
poor. Many parameters have to be taken from other sources or have to be esti-
mated (or even disregarded, because it is supposed that they have small influence
an estimation). Activity coefficients in the liquid phase are usually omitted by using
a "sufficient" concentration of background electrolyte. The surface activity coeffi-
cients are usually not mentioned at all. most only one type of adsorption

of onto
center is assumed, often with a maximal surface concentration equal to the

layer OH concentration. Some are disregarded, such as formation of
doubly charged surface species, in spite of the fact that analogous species are
in the liquid phase. The speciation in the liquid phase, if regarded at all, is taken
from the literature, even values may be erroneous. The
parameters, depending on the specific model and the choice of fixed parameters.
generally allow fitting the experimental data, as they are not rich in
Therefore even an incorrect may give a good fit.
As a consequence the experimental conditions should be standardized as much
as possible to allow the of data obtained by different groups. Thus, in
the case of equilibrium adsorption experiments as an example, the following con-
ditions might be standardized, with the proposed values in parentheses:
1. Background ions 0.1 M)
2. standard method of preparation, larger common stock,
constant area per volume
Constant conditioning time of oxide with all ingredients except for the metal
salt solution (1 h)
Adsorbent total 1
Analytical techniques.
a . pH types of electrodes, calibration
Metal analysis
6. Measuring temperature ( T = 298
7. Equilibration time (1 h)
bubbling during conditioning, at pH)
Only one condition should be varied at a time, to allow comparison. The primary
experimental might be published, on the Internet, for easier
exchange of data.
Some may also be proposed for the model calculations. A great
number of species are accepted in different models, but as it is irrelevant from
standpoint of thermodynamics reaction is chosen to calculate the concentra-
tion of one species from that of any other species, different reactions are used, and
different model parameters are obtained. A unified reaction scheme would
facilitate comparison of results.
valuable assistance of Mrs. Koehler and Mr. D.

Judat in preparing this manuscript,
A. and A. (1986).
and
K. Chattoraj, Surf. Tech.
Schulthess and Sparks, Soil (1988).
S. Howard, and Han, (1993).

Ch. and The of
A. Scheidegger and D. Sparks, Soil

Schulthess and D. Sparks, Soil Am. J.
J. R. White, and St. Hem, Colloid Interface
1 (1986).
Contescu, and A. (1993).
Contescu. A. Contescu, and A. Phys. (1994).
Hiemstra, van and H. Bolt, Colloid Interface
1 989).
J. van Colloid Interface
133: 105
Venema, and H. Colloid Interface
(1996).
and Stumni, Colloid Interface (1973).
W. and Mater.
J. Pure Chem. (1978).
Schulthess and Sparks, Soil Am.
173:104
Diploma Work, 1996.
Koopal, Acta
Torkar, Monat. Chem, (1960).
and White, Colloid Interface (1978).
Noh A. J. Colloid Interface Sci. (1989).
Hunter. Science, Oxford,
U K , 1987.
Contescu, and A. J. (1991).
A. 46: (1991).
and J. Colloid Interface
V. Brady, Acta (1994).
and Patrick, Colloid
Interface Sci.
Parks and de Phys. Chem. (1962).
Tewari and Ghosh, India
A. Yopps and Fuerstenau, Colloid (1964).
Koutsoukos and A. Colloids Surf. (1991).
J. M. A. Proctor, and Hercules, Phys. Chem.
(1995).
J. Lutzenkirchen, and Colloid Interface
Hayes, Redden, and Colloid Interface

(1991).
C. Thesis, University, 1971.
and A. Chem. (1957).
Modi and D. Fuerstenau, Phys.
J. Peri, Phys. (1965).

of onto Alumina
Knozinger and Ratnasamy, Cat. -

Dissertation, 1980.
Katayama, and Furuichi, Environ. Technol.
Sprycha and J. Colloid Interface

and A. (1986).
F. Dissertation, Dusseldorf,
Colloids Surf. A
C. Ye, Solid State
Shiao, and E. Meyer, (1981).
D. Jackson, and J. K. Soil Am, J.
(1976).
Chang, Lo, Yasunaga, Colloid
Interface (1997).
R. and Evans,
Parvulescu, and E. Rev.
35: (1990).
Chang, Lee, and Lo., (1993).
and Czech. Phys. B
Cozar, and Rev. Phys. (1982).
Y. Zeng, and React. Kinet. Lett, (1991).
Rudin and Motschi, Colloid Interface
Hachiya. Sasaki, Y. Saruta, Mikami, and Phys.
(1984).
Hachiya, Sasaki. Ikeda, Mikami, and
Mustafa, Safdar. and Khan, Phys. (1993).

Pavlova and Water 2211571 (1988).
Soil. Am.
Clays Clay
P. and A. Colloid Interface Sci. (1989).
P. and Soil. Am. J.
A. A. P. Dushina, and Zh.
SSSR 42:1883 (1972).
Schulze, and Anal, (1971).
Schulze, and Anal,
and Proc. Tokai
(1993).
Mustafa, and Yar Surf. Technol. (1989).
Mustafa, Safdar, and Nawab, Phys, (1990).
Simon, W. Schulze, and (1972).
L. Lo and Leckie, Proc, (1989).
Lo and Leckie.
Lo and 0. Leckie, Water Technol.

Mahmood, and Technol.
(1992).
and
Stumm, Colloids Surf.
Chang, Lee, Lo, and Colloid
Interface 169 (1994).
and Lehmann, Chem,
Zaki, A. A. and G . A. Mekhemes, Langmuir
(1992).
private communication.
and Branica, Colloid Interface (1976).
R. Roy, A. Cogneau, and
and
L. Villa, React, Catal. Lett. (1975).
de Bokx, A. Wassenberg, and W. Catal. (1987).
and A. Lycourghiotis, (1994).
and A. Lycourghiotis, Faraday Trans. (1993).
St. R. G. Brown, and A. Parks, Colloid Interface
(1997).
A. Scheidegger, and Sparks, Colloid
(1997).
E. Brown, A. and
(1989).
A. and A. Schwarz, Catal.
Stumm and Rev. Geophys. (1990).
He, , and Martens, Soil
Am.
A. C. Colombo, and Buondonno, Soil Sci. Ain. (1991).
Petkovic, and Technol.
Toner and Soil Sci. Am. (1995).

Saeki and Soil Plant Anal., (1994).
Mundhara, and J. Indian Chem.
M. Fay, A. Proctor, Houalla, and Hercules, Catal.

1
Ch. and A. Lycourghiotis, Langmuir 8:13 (1992).
and Lycourghiotis, Catal.
(1990).
N. and A. Lycourghiotis, Catal.
Caceres, and Adsorp. (1992).
and A. Acad. Paris (1973).
Langmuir
and K. Jorgensen, 2212313
A. Davis, and J. Colloid Interface (1978).
Contescu, and A. Chem. Fasaday 8914091
(1993).
A. Schwarz, and Zhang, Catal, (1992).

of Ions onto Alumina
W. Fuerstenau and J. Colloid Interface (1987).

Contescu and I. Vass, Appl. Catal. (1987).
C. Contescu and I. Vass. React, Kinet. Catal. Lett. (1991).
Th. Mang, Polanek, and Knozinger, Appl. Catal. A:
(1993).
W. Cheng and Huang, Colloid Interface Sci. (1996).
A. and 0. Coord. Chem. (1994).
A. and R . J. 194th Meeting Am. Chem.
(Envir. Chem. and of Aluminum) 1987 (T. Lewis,
Contescu and I. Vass, React, Kinet. Catal. Lett. (1991).
Huang and P. Somasundaran, Colloids Surf, (1996).
J. H. and Harwell, Res.
(1996).
Fu, P. Somasundaran, and Xu, ACS Symp, (1992).
Jha, Mundhara, R. Sharma, and Tiwari,
(1992).
M. Sharma, Mundhara. and S. Colloids Surf.
(1992).
Baumgarten, Wagner, and Lentes-Wagner, Catal. (1987).
Baumgarten and Dedek, Catal. (1990).
and A. Zachos, (1981).
Parley, A. Dzombak, and F. Morel, Colloid Interface
(1985).
Katz and Hayes, Colloid Interface (1995).
Katz and Hayes, J. Colloid Interface (1995).
Agashe and Regalbuto, Colloid Interface (1997).
A. Robertson and 0. Leckie, Colloid Interface (1997).

Molecular Recognition Group, Hannah Research Institute,
Ayr, Scotland
I.Introduction
II .
Protein Structure
III .
Polystyrene Latex Particles
IV. Interaction of Proteins with Polystyrene Latex
A. Adsorption
B. Conformation of adsorbed protein
C. Stability of adsorbed protein layers
D. Controlling protein adsorption
V. Conclusions
References
Proteins, like other surface-active macromolecules, readily adsorb onto interfaces,

particularly if the surfaces are hydrophobic. This property has important implica-
tions in a number of diverse technological areas. For example, within the food
processing industry, protein-stabilized emulsions, where separation of the immisc-
ible oil and water phases is prevented by stabilizing individual droplets with an
interfacial layer of protein and surfactants. are, in terms of their commercial value,
the most important class of food colloids. Similarly, food foams are usually stabi-
lized by a combination of proteins and low molecular weight surfactants adsorbed
at the interface. On the negative side, fouling of processing equipment,
particularly of heat exchangers, resulting from the deposition of proteins on solid
surfaces decreases efficiency and increases the need for frequent cleaning. In bio-
medicine, fouling and rejection of implants, membranes, and contact lenses are
problems, but immunoassays based on immobilized antibodies and antigens are

744
of increasing importance. In dentistry, control of plaque formation is a major

interest, as is targeted drug delivery within the pharmaceutical industry. Protein
purification via chromatography where protein molecules adsorb to, and are spe-
cifically released from, solid surfaces and the manufacture of specific enzyme-based
bioelectrodes are of great interest in the bibprocessing industry.
In order to control protein adsorption, to enhance it in some cases and prevent it
in others, it is necessary to understand the various stages involved in the process.
The interaction of protein molecules with polystyrene (PS) latex particles having a
well-defined surface has proved to be a very useful model system with which to
study the interfacial behavior of proteins. Other colloidal systems, including silica
and metal particles, have also been used in these investigations, and although this
review concentrates mainly on interactions between proteins and latex particles,
other systems are also mentioned where appropriate. Before looking at the inter-
actions of proteins with PS latex particles in detail, it is worthwhile to take a brief
overview of the two major components in the system.
Proteins are linear chains of amino acids in which individual residues are linked by
bonds between the carboxyl group of one acid and the amino group
of the next. This is the primary sequence of the protein. The side chains of some of
the amino acids undergo enzyme-catalyzed modifications such as glycosylation and
phosphorylation, which serve to alter the functionality of the protein. Folding of
the polypeptide chain as a result of hydrogen bonding between C 0 and N-
groups of the amino acids gives rise to a-helices and and sheets, with
structured regions being linked by flexible loops. This is the secondary structure.
Other types of secondary structure have also been identified, but in globular pro-
teins these forms predominate. Compared to bond energies ( > 200
these hydrogen bonds are relatively weak (around
The extent of ordered structure in different protein species is very variable.
Some proteins, and in the context of this chapter the caseins are a prime example,
adopt a relatively open, expanded-coil conformation. In contrast, globular pro-
teins are almost spherical, with a packing density at the interior of the molecule
being equal to that found in crystals and close to the theoretical value for
packed spheres. This high packing density means that short-range interactions
such as hydrogen bonding and van der forces are important in stabilizing
the secondary structure. In general, hydrophobic amino acid residues tend to be
buried in the interior of the protein, whereas hydrophilic charged groups are
located on the outside of the protein in contact with the aqueous phase.
However, polar and amino acids are fairly equally distributed at the sur-
face of globular proteins.
An additional type of covalent bond, the disulfide bridge, can form between
individual cysteine residues and is also important in determining the stability of
many proteins. Proteins with disulfide bridges usually contain at least two, proteins
containing single disulfide bridges being relatively rare. Both inter- and intramole-
cular disulfide bridges are found; the addition of thiol-containing reagents such as

2-mercaptoethanol and dithiothreitol reduces disulfide bridges and
bilizes the protein.
Denaturation, or unfolding, of globular proteins can be induced by exposing the
protein to high temperature or pressure, extreme pH, high concentrations of salts,
organic solvents, surfactants, urea, or guanidine hydrochloride. In addition, as
discussed later, some proteins are also denatured by adsorption to interfaces. In
general, globular proteins behave as random coils when dissolved in 6 M guanidine
hydrochloride.
Since polystyrene latex particles are generally hard spheres, protein adsorption to
them is solely a surface phenomenon. Methods for the production of polystyrene
latex particles have been reviewed by et
The diameter of polystyrene particles can range from less than 0.1 to more
than 1000 By varying the polymerization conditions, it is possible to produce
polystyrene latex particles in a broad range of sizes and degrees of monodispersity.
Larger beads are usually produced by suspension polymerization. In this process,
agitation breaks the monomer phase into individual droplets suspended within the
dispersion phase, which also contains a surfactant to stabilize the droplets. An
initiator is also incorporated into the phase, and by varying both the
rate of agitation and the temperature, polymerization is induced. The polystyrene
particles produced by this technique typically have mean diameters between 50 and
1000 with a relatively broad distribution. Emulsion polymerization, where
surfactant-stabilized monomer micelles are polymerized by initiator molecules dis-
solved in the dispersion phase, is used to produce extremely uniform, very
spherical particles with diameters up to 0.1 In phase inversion polymerization,
the monomer, initiator, and stabilizer are all dissolved in an inert solvent. When the
growing polymer chains reach a certain critical size, phase inversion occurs, and the
chains aggregate to form stable monodisperse particles. The size range of particles
produced by this method is between those of particles produced by the other tech-
niques.
The polarity and density of charged groups on the surface of latex particles
can be varied by altering the chemical composition of the monomer mix.
Alternatively, it is also possible to change the surface by allowing acrylate mono-
mers to adsorb to preformed particles and subsequently polymerizing these in
The surface charge density of beads can be determined by measuring their
average zeta potentials and electrophoretic mobilities or by conductometric
Size, and smoothness are usually determined by microscopy or
photon correlation spectroscopy.
The various stages in the interaction of a protein molecule with a surface are shown
schematically in Fig. 1. Initially, protein molecules in the bulk phase approach the
latex surface. Adsorption may then occur, and in some instances this is followed by

.1 Schematic picture illustrating the various stages in the interaction of a protein
molecule with the surface of a PS latex particle.
rearrangement of the protein structure; both of these steps are controlled through a
combination of ionic and hydrophobic interactions and hydrogen bonding. Finally,
in some systems the structurally modified protein molecules be released back
into the bulk phase. These stages are now considered in more detail.
Dimensions of Adsorbed Protein Layers

The two parameters of the adsorption process that are the easiest to quantify and
are therefore most frequently reported are the thickness of the adsorbed protein
layer and the density of the protein on the surface. The increase in the thickness of
the adsorbed protein layer and in surface coverage as a function of added protein
are shown schematically in Figs. 2A and respectively.
( a ) Protein Layer Thickness. The thickness of the adsorbed layer is listed for a
variety of proteins on PS latex in Table 1. Although protein layer thickness on
polystyrene latex particles has most frequently been determined by light scattering
using photon correlation spectroscopy (PCS), small-angle X-ray scattering (SAXS)
has also been used. Random-coil proteins such as the caseins give relatively thick
layers (10-15.5 whereas compact, globular proteins such as 3-lactoglobulin
produce much thinner layers. The actual thickness of the adsorbed layer is depen-
dent on a number of variables, including pH, surface charge on the particles, charge
on the protein, and extent of denaturation of the protein. These are discussed in
more detail later.
Surface Loading. Maximum values of the protein surface coverage, for a
variety of proteins are listed in Table 2. is usually expressed in milligrams of
protein adsorbed per square meter of surface. By measuring the proportion of the
protein adsorbed at increasing concentrations of protein, the adsorption isotherm
and hence the affinity of a protein for the surface can be determined. The amount of
adsorbed protein is most commonly determined by the solution depletion tech-
nique, in which protein adsorbed onto the latex proteins is separated from

protein
Schematic diagram showing (A) the increase in layer thickness as a function of

added protein and (B) the adsorption and desorption isotherms.
adsorbed protein by either centrifugation or filtration and the concentration of

protein in the liquid phase is determined spectrophotometrically. An alternative
method in which the amount of protein bound to the latex particles was determined
directly using a modified procedure has been reported As with protein
layer thickness, values vary widely depending on the same variables. Plateau
values of for the adsorption of from various species onto a wide variety
of surfaces under various conditions are listed by Hidalgo-Alvarez and
Gonzalez
Protein adsorption to PS latex has also been determined by sedimentation field
flow fractionation The maximum surface coverages of and
lactoglobulin on negatively charged PS latex calculated using method were
similar at around 1 This figure, which was confirmed by amino acid
analysis of the material irreversibly bound to the surface, was significantly lower

Layer Thickness of Proteins on Polystyrene Latex Particles
Surface Thickness'
Protein charge Additions Ref.
7.0
Urea
dephosphorylated
-Casein Urea
K-Casein,
nonglycosylated
K-Casein,
glycosylated
K-Casein
7.0 2-ME
Phosvitin
Urea
thicknesses determined by PCS except for those marked*. where SAXS was
than that measured by solution depletion techniques, the difference being explained
by the authors as being due to loosely associated protein that is not detected by the
method. Adsorption of to latex particles at p H 7.0 determined using
showed that adsorption times were less than 10 min and the maximum
value of was 4.75
2. Influence of pH and Charge on Protein Adsorption
Proteins are electrolytes with the side chains of their acid,
acid, tyrosine, histidine, and arginine residues being
involved in acid-base equilibria. The net charge on a protein is therefore dependent
on pH. At the point (pi), the net charge on the molecule is zero, and at
pH values below and above their proteins are positively and negatively charged,
respectively. The interaction of a protein with a surface will therefore vary with
and, in general, the maximum value of is attained at or near the of the protein.
Obviously, electrostatic interactions between the surface and the protein and
between adjacent adsorbed protein molecules are therefore important.

Surface Coverage (T) of Proteins on Polystyrene Latex Particles
Surface Max coverage
Protein charge Ref.
0-Casein,
dephosphorylated
nonglycosylated
K-Casein,
glycosylated
Ase
a-Lactalbumin
Lysozyme
Fibrinonectin
Fibrinogen
Immunoglobulin
Lipolase
Savinase
Bovine serum
One way to demonstrate the influence of electrostatic interactions on the inter-

action of proteins with surfaces is to use different forms of the same protein mod-
ified so as to change the net charge on the molecules. contains five
phosphoserine residues, each of which increases the net negative charge on the
molecule by one unit. Enzymatic dephosphorylation cleaves off these phosphate
residues and reduces the electronegativity of the protein. As shown in Tables 1 and
2, reduced repulsion between the protein and the negatively charged latex and
between adjacent protein chains resulted in the formation of a thinner protein
layer and a slight increase in the packing density of the adsorbed dephosphorylated
protein.
K-Casein, which has a net negative charge at 7.0, formed a thinner layer on
a positively charged latex than on a negatively charged one. Naturally occurring

7
glycosylation of this protein increases its net negative charge by one unit for each
sialic residue attached to the polypeptide chain and, as expected, also increases
the thickness of adsorbed protein layer on a negatively charged surface
(Table 1).
Another way to investigate electrostatic interactions is to change the surface
characteristics. al. prepared a series of latices of the same
particle size but different surface structures and measured the adsorption of human
serum albumin (HSA), cycochrome, myoglobin, and horseradish
oxidase. In general, hydrophilic surfaces had little affinity for proteins. Generally,
increasing the hydrophobicity or the surface charge increased the affinity of pro-
teins for the surface. Electrostatic interactions were found to be more important in
determining the adsorption of polar and relatively small protein molecules. If
adsorption of a protein onto a surface was largely determined by electrostatic
interactions, it would be expected that these should be maximal in the region
where the protein and the surface are oppositely charged. However, measurements
of the adsorption of HSA and y-globulin onto carboxylated surfaces showed that
this was not the case This may be due either to conformational changes in the
protein structure induced by adsorption to the surface, which weakens electrostatic
interactions between the protein and the surface, or to deionization of groups on
the surface of the protein, or to both.
In addition to the polarity of the surface, the density of charged groups on the
surface also affects protein adsorption At carboxyl group densities less than
2
1 n m , adsorption of bovine serum albumin to latex was both dominated by
hydrophobic interactions and relatively sensitive to pH. At densities above two
carboxyl groups per square nanometer, hydrogen bonding predominated irrespec-
tive of pH. The surface coverage of lysozyme and (a-lactalbumin
from which the calcium ion bound at the high affinity binding site has been
removed and which is therefore less stable than the native protein) on negatively
charged latices has also been shown to be dependent on the density of charged
groups on the surface of the support. At pH 7, the adsorption of lysozyme, which is
positively charged at this pH, is higher on a more negatively charged latex whereas
that of apolactalbumin, which is negatively charged, is lower, demonstrating the
importance of electrostatic interactions to protein adsorption Gallisteo and
Norde also showed that the maximum value of was pH-dependent, being
highest at a pH value just below the p i of the individual protein and decreasing at
pH values to either side of the As expected, the influence of the density of
negative charge on the surface loading of these two proteins was also pH-depen-
dent, with loading on the more highly negatively charged surface being greater at
pH values below the of each protein.
As part of a study aimed at the rational development of the
influence of surface characteristics on the interaction of fragments
obtained by peptic digest of rabbit polyclonal has been investigated
The latex particles differed in their hydrophilic character and in the nature and
density of their surface ionic groups. The most important force governing the
adsorption of the protein onto the surface was shown to be hydrophobic inter-
actions, although in low ionic strength buffers electrostatic interactions were also
found to play a part. On anionic latices, maximum values were obtained at p H

5, which was very close to the of these fragments (pH 5.3). At this pH, where
the molecules have no net charge, molecular folding is at its most compact, and
they can therefore pack closer together on the surface. On latex, max-
imum adsorption was shifted to pH 7.0, showing the importance of electrostatic
interactions at low ionic strength.
3. Effect of Inorganic Ions

According to the (DLVO) theory, the stabi-
lity of colloidal dispersions is determined by the surface potential, which is the sum
of the attractive van der forces and the repulsive electrostatic force arising
from the diffuse electrical double layer around the surface of charged particles.
Decreasing the surface potential by increasing the ionic strength reduces electro-
static repulsion and promotes aggregation.
The adsorption of and apolactalbumin was enhanced by low ionic
strength, particularly around the of the protein and the surface coverage
of HSA and on amphoteric polystyrene latex was lower at higher ionic
strength irrespective of pH The influence of ionic strength on the colloidal
stability of PS latex coated with either fragments or bovine serum
albumin has been investigated At low surface coatings of protein, the
particles behaved as expected, with increasing ionic strength promoting aggrega-
tion. At high surface concentrations of although aggregation occurred at
moderate ionic strength, at high ionic strength the coated particles did not aggre-
gate. The explanation offered for these observations was that the hydration repul-
sion force arising from counterions adsorbed onto the protein dominated
interactions between particles at short range.
Adding to negatively charged latex precoated with resulted in
a 4-5nm reduction in the thickness of the protein layer This was due to sup-
pression of ionization of the negatively charged groups located in the hydrophilic
tail region of the protein and on the surface of the latex, reducing electrostatic
repulsion. Increasing the concentration further caused aggregation of the
coated particles. When low concentrations of were added to latex in the
presence of 50mM aggregation occurred This was due to bridging
flocculation, where suppression of ionization of the charged groups on the protein
and the surface permitted protein molecules to span between partially coated latex
particles.
Calcium ions have also been shown to decrease the thickness of preadsorbed
casein layers on negatively charged PS latex The maximum decrease in layer
thickness was 6nm. The concentration of calcium ions required to achieve this
reduction in adsorbed protein layer thickness was around 1 of the NaCl concen-
tration. Comparison with the behavior of preadsorbed layers of dephosphorylated
where the reduction in layer thickness was less, indicated that the calcium
ions were largely binding to negatively charged phosphate groups in the tail region,
reducing repulsion between this portion of the molecule and the surface and also
between adjacent protein tails. Higher concentrations of calcium ions caused aggre-
gation of the coated particles.

The conformation that protein molecules adopt at an interface is not necessarily the
same as that which they possess in solution. A variety of physical and biochemical
methods have therefore been employed to determine the conformation of both
random coil and globular proteins when adsorbed to latex particles. Since they
are relatively lacking in extensive secondary structure, the results obtained with
the former are simpler to interpret, and these are therefore considered first.
Random Coil Proteins

The caseins, the major milk proteins, are classic examples of random coil proteins.
They are generally good emulsifiers, and the interfacial conformation of in
particular has been studied in some detail both in model emulsions and on poly-
styrene latex. p-Casein has a high content of residues, which serve to disrupt
the formation of ordered regions of secondary structure, and the molecule has a
relatively open random coil conformation. In addition, it has a distinctly
philic character. The N-terminal 21 amino acid region, which includes a cluster of
four phosphoserine residues, has a net charge of -12 at pH 6.6 and is therefore
markedly The remaining large hydrophobic domain has no net charge
at this pH. In solution, at temperatures above about the molecules undergo a
highly cooperative reversible self-association to form polymers with a structure
similar to that of a detergent micelle. As measured by PCS, at pH 7.0 this protein
formed a monolayer about thick on negatively charged PS latex (Table 1).
Dephosphorylation using acid phosphatase reduced the net negative charge on the
protein, specifically in the N-terminal region, and resulted in the formation of an
adsorbed protein layer only 11 nm thick. Addition of the proteolytic enzyme trypsin
to latex particles precoated with either form of the protein resulted in a biphasic
decrease in the layer thickness as the adsorbed protein was hydrolyzed. The rate of
the initial rapid decrease in layer thickness was greater with the dephosphorylated
form of the protein, although the rate of the second, slower phase and the thickness
of the residual layer (about 5 were the same It appears that the nega-
tively charged N-terminal region of both types of the protein forms a tail that
projects from the negatively charged surface, with the remaining hydrophobic por-
tion of the molecule lying close to the surface (Fig. 3A). Removal of the phosphate
groups reduces the electrostatic repulsion between the tail and the surface and also
between adjacent protein tails, decreasing layer thickness and coincidentally per-
mitting easier access of trypsin molecules to susceptible bonds in this region. The
same conformation has been proposed for this protein at the interface on
the basis of proteolytic experiments and also at the interface on the
basis of small-angle neutron scattering measurements Small-angle X-ray
scattering (SAXS) of the protein on PS latex showed that most of the protein lies
close to the surface, with part of the polypeptide chain extending out into the
aqueous phase Neutron reflection studies of this protein adsorbed onto a che-
mically modified silicon surface also indicated a two-layer conformation with a
dense layer 2.3 nm thick and a surface coverage of 1.9m g /m 2 lying close to the
surface and an external layer with a thickness of and a protein volume
fraction of 12% extending into the aqueous phase

The probable conformation adopted by (A) and K-casein at pH 7.0
on the surface of a negatively charged PS latex particle. N and C indicate the terminals
of the proteins.
The conformation of another milk protein, K-casein, on PS latex particles has

also been investigated Much of the surface of the casein micelles, the
submicrometer-sized particles, which contain approximately 80% of the proteins
present in milk, is thought to be formed by the K-casein component. This protein
has a highly electronegatively charged C-terminal tail that projects into the aqueous
phase in the form of a "hairy layer." Through a combination of electrostatic and
steric repulsion, this layer prevents individual micelles from aggregating. In the
manufacture of cheese. by the enzyme chymosin at a specific
lanine-methionine bond leads to the release of the charged tail, the
This reduces repulsion between individual micelles, allowing
aggregation to occur. K-Casein adsorbs readily onto both positively and negatively
charged latices, the surface coverage and layer thickness being dependent on the
surface charge and the form of the protein (Tables 1 and 2). Addition of 2-mer-
captoethanol to K-casein prior to adsorption onto latex particles increased the
thickness of the adsorbed layer as a result of reduction of the intermolecular
ulfide bridges (Table 1). Addition of chymosin to K-casein-coated latex particles
caused a decrease in the thickness of the adsorbed protein layer This
was similar to the 7-8 nm chymosin-catalyzed decrease in the hydrodynamic radius
of casein micelles determined using the same PCS technique The
released from the surface of the coated latex particles was shown to be CMP,
indicating that K-casein binds to negatively charged PS latex particles with the
same orientation it adopts in the micelles, with the C-terminal tail projecting
into the aqueous phase (Fig. 3B). In the presence of calcium ions, the proteolyzed
particles aggregated, the rate of aggregation being dependent on both the concen-
tration of calcium ions and the degree of glycosylation of the protein. Again, this is
very similar to the behavior of micelles in milk, and P S latex aggregation appears to
be due to unhydrolyzed K-casein molecules forming bridges between individual P S
particles.
2. Globular Proteins
Due to their more highly folded structure, globular proteins can potentially
undergo more extreme conformational changes when adsorbed to surfaces, and
interpretation of these structural changes is consequently more At the

754 Leaver
simplest level, if the molecular dimensions of a protein are known, then, purely on
the basis of the measurements of the surface coverage, information on
the orientation of the molecules on the surface can be obtained (Fig. 4). and
their fragments apparently adsorb onto hydrophobic surfaces in an end-on
orientation but provided the proteins have a high charge density opposite that of
the surface, the orientation on negatively charged hydrophilic latex switches to
on in order to maximize electrostatic interactions
More detailed information on the conformation of adsorbed proteins has been
obtained by using a variety of physical techniques. Although fluorescently labeled
HSA and have been shown to undergo conformational
changes consistent with the loss of internal fluidity when adsorbed onto
latexes in general the application of spectroscopic techniques to
the analysis of protein conformation at solid interfaces is complicated by the scat-
tering and adsorption of light caused by the particles themselves. This can be
reduced either by using very small particles that adsorb and/or scatter light to a
much smaller degree or by using particles that display negligible adsorption and
scattering. Using the latter approach, circular dichroism (CD) and fluorescence
anisotropy were used to detect changes in the conformation of the proteolytic
enzyme 309 resulting from its adsorption to a novel hydrophobic
fluoroalkoxy fluorocarbon latex At = the fluorescence of the
adsorbed protein was quenched and an increase in the a-helix content was detected
by This finding was in contrast to the decrease in a-helix content that is usually
detected when proteins adsorb onto hydrophilic silica particles. At =
which represents full surface coverage, the fluorescence and a-helix content of the
adsorbed protein were essentially the as in solution, indicating that in this
instance the extent of conformational changes in the structure of adsorbed proteins
is dependent on the density of packing of the molecules on the surface.
In addition to spectroscopic techniques, microcalorimetry, which measures tran-
sition enthalpies (AH) and transition temperatures (To.) as proteins undergo coop-
erative unfolding, has also been applied to the study of conformational changes
resulting from protein adsorption. Since PS latex particles show no thermal transi-
tions over the temperature range of interest for protein structural analysis, their
presence does not interfere with this technique. HSA, a three-domain protein whose
HIGH
. 4 Schematic showing a protein molecule adsorbing to the surface of a

latex particle in the side-on conformation at low protein loading and the end-on
at high protein loading.

structure unfolds readily at low concentrations o f the denaturant guanidine hydro-
chloride (GuHCl),also unfolds readily on the surface o f PS latex, as evidenced by
the complete absence o f thermal transitions as determined by microcalorimetry o f
HSA-coated polystyrene latex (Fig. 5 ) Similarly, hen lysozyme, which is struc-
turally a more stable protein, unfolding at much higher concentrations o f GuHCl,
does not undergo a thermal transition when adsorbed onto PS latex. However,
when this protein was adsorbed onto contact lenses made from anionic
a thermal transition was clearly seen. Even here there were differences compared
with the behavior o f the protein in solution. The was lower, the peak was
broader, and the transition enthalpy was smaller than that measured in solution,
indicating that some change had occurred in the structure. Adsorption o f
vidin, which is stable to denaturation even at very high concentrations o f GuHCl,
onto PS latex particles had little effect on the appearance o f the thermograms
compared with that o f the protein in solution, indicating that association with
the surface had little effect on protein structure. Egg white lysozyme and bovine
milk both lost most o f their ordered secondary structure upon
adsorbing onto hydrophobic negatively charged polystyrene latex
adsorbed onto negatively charged latex particles did not undergo
the native-to-apo transition that the soluble protein undergoes at low pH.
Adsorption o f a-lactalbumin onto the hydrophilic surface o f hematite dispersions
also resulted in almost complete loss o f secondary structure, again indicating the
relative instability o f this protein. In contrast, most o f the secondary structure o f
lysozyme, which is a more stable protein, was retained on this surface. Proton
Temperature
. 5 Idealized thermograms o f ( A ) a globular protein in solution, ( B ) the same

protein adsorbed onto PS latex particles with partial retention o f secondary folding,
and the same protein adsorbed onto latex particles with extensive loss o f secondary
folding.

measurements showed an increase in the average of carboxyl groups in
both of these proteins, suggesting that some carboxyl groups are relatively close to
the latex surface and/or that unfolding of the protein structure during adsorption
diminishes electrostatic repulsion between the remaining groups.
Microcalorimetric measurements of the globular milk protein
showed that a significant amount of secondary structure was retained after adsorp-
tion onto negatively charged PS latex
C-reactive protein (CRP), which is thought to play an important role in
nomodulation, has been adsorbed onto latex beads and shown to bind a variety of
plasma proteins, including some, such as the immunoglobulins, that are not bound
by the soluble form of the protein This alteration in specificity may be caused
either by conformational changes in the adsorbed protein or by an increase in the
number of binding events due to an increase in the density of the reacting species,
or by a combination of the two. The results obtained with the PS latex-immobilized
model system strongly support the hypothesis that in vivo possesses lectin
function only when it is immobilized on the surface of inflamed tissue cell mem-
branes, allowing it to bind bacterial etc., rendering them suscep-
tible to phagocytosis.
Although almost all of these studies have dealt with the adsorption of single
protein species onto surfaces, in biological systems surfaces are usually exposed to
mixtures of proteins. Little work has been reported on whether different species of
proteins segregate and form patches on the surface. Brynda et measured
excitation energy transfer between fluorescein donors attached to bovine serum
albumin molecules and tctrainethylrhodainine acceptors attached to other
cules of the same protein when the molecules were adsorbed onto hydrophobized
silica slides, which would be expected to have surface properties similar to those of
polystyrene latex particles. Unlike the situation in solution, when denaturation
resulted in aggregation, the surface-adsorbed protein molecules did not appear to
aggregate as a result of the loss of their secondary structure. When the adsorption
of mixtures of BSA with was studied, there was no evidence of patch forma-
tion. Instead, molecules of the different protein species appeared to adopt a random
distribution on the surface.
The more stable globular proteins such as lysozyme, a-lactalbumin, and

which appear to retain most or all of their native structure when adsorbed
onto PS latex particles, would be expected to be desorbed more readily by dilution
than proteins such as serum albumin, and hemoglobin, which appear to lose
most of their secondary structure. Haynes and Norde investigated the
of lysozyme and a-lactalbumin from negatively charged PS latex. At pH 7.0,
dilution did not result in any detectable change in the concentration of bound
protein, indicating that binding under these conditions is irreversible by dilution.
Adsorbed lysozyme was also not displaced by changing pH over the range 2-12 or
by increasing the temperature to The behavior of a-lactalbumin was similar,
except that above pH 11.5, where both the protein and the surface have a large

negative charge, displacement did occur. Adsorption of which belongs to
latter group of proteins, was also not reversed by dilution
measurements of the desorption of as a result of dilution indicated
that slow desorption of bound protein to a level of about 20% of the original
loading occurred over the first 2 h after dilution. Negligible desorption occurred
over the following 6 h Dilution has also been reported to cause significant
release of radiolabeled from negatively charged polystyrene latex The
homogeneous exchange reaction between radiolabeled molecules in the bulk
solution and those adsorbed onto the surface of negatively charged polystyrene
latex was found to follow first-order kinetics, with an apparent exchange rate
constant of (2.3 0.4) has also been shown to desorb from
negatively charged PS latex on redispersion at pH values between 5 and 8 In
addition, a significant amount of the preadsorbed was displaced by incubation
with monomeric bovine serum albumin.
Adsorbed proteins can be displaced by small surface-active molecules such as
detergents. Although this has been widely reported in publications dealing with
protein-stabilized emulsions, little appears to have been published dealing with the
same proteins adsorbed onto latex particles. Figure 6 shows the desorption of
casein from negatively charged PS latex particles and the accompanying decrease in
protein layer thickness as a result of the addition of increasing concentrations of the
nonionic detergent Tween 20 (unpublished data).
FIG. 6 The displacement of bovine from the surface of a negatively charged

PS latex by the nonionic detergent Tween 20, showing that protein displacement is
accompanied by a decrease in the thickness of the adsorbed protein layer.
Protein coverage; ( hydrodynamic layer thickness. R is the molar ratio of
detergent to protein.

Those conformational changes that labeled human serum albumin
and undergo when adsorbed onto latexes were not
reversed on desorption This therefore acts as a useful marker for the exchange
of adsorbed molecules with nonadsorbed molecules. Even when the amount of
latex in a sample was very low compared to the total amount of protein, the
percentage of adsorbed protein was small at any one time, eventually all of the
protein was found to have undergone this conformational change, indicative of a
relatively active exchange process.
1. Decreasing Protein Adsorption

The adsorption of serum proteins in general, and of fibrinogen in particular, onto
surfactant-treated polystyrene latex is accompanied by displacement of surfactant
molecules Using a series of polyoxyethylene-containing block copolymer
factants, the stability of the adsorbed layer to displacement by protein
was shown to be dependent on a combination of its surface concentration, layer
thickness, and surface dynamics. (polyoxyethylene-containing)
adsorbed onto the surface of PS particles made the hydrophobic surface
hydrophilic, minimizing protein adsorption and increasing the blood circulation
half-life from a few minutes to more than 13 h The most stable coating
was not displaced by incubation with either HSA or human fibrinogen. However,
incubation with whole resulted in displacement of 30% of the surfactant
Coating PS latex beads with alcohol) also decreased nonspecific
protein adsorption
2. Increasing the Specificity of Protein Adsorption
Various methods have been adopted to alter the selectivity of the relatively non-
specific adsorption of protein molecules onto polystyrene latex. By activating the
end of the surfactant molecules and forming, for example, the 2-pyridyl
disulfide derivative, it was possible to make the surface specific for thiol-containing
proteins and peptides, which could subsequently be desorbed by addition of low
molecular weight thiol-containing Blue cova-
lently attached to alcohol)-coated microspheres specifically adsorbed
BSA up to a concentration of 60mg of protein per gram of latex beads
Phospholipids adsorb onto positively charged amidine PS latex initially as a
monolayer but subsequently as lipid bilayers By incorporating the cholera
toxin receptor monosialoganglioside into these bilayers, specific binding
of cholera toxin has been achieved
Proteins adsorbed onto polystyrene latex particles have proved to be good models
for the interfacial behavior of proteins in processed foods. The behavior of caseins
on latex particles appears to be very similar to that at the interfaces of
emulsions, at the interface of foams, and at the surface of casein micelles.

Investigations into the behavior of mixtures of proteins together with other low
molecular weight food grade emulsifiers to move model system closer to real
processed foods would be a n obvious development of these studies.
In the area of fouling and implant rejection, progress has been made in under-
standing how proteins interact with these PS latex surfaces and how this interaction
can be controlled. Again, the model will be improved as the complexity of the
system is increased. The demonstration that the surface of latex particles can be
modified either chemically o r by adsorption of cell surface constituents opens up
new possibilities for the use of such systems in understanding a host of complicated
cellular events including the development of colloid-based drug delivery systems.
Latex agglutination kits are becoming increasingly important in their own right
as diagnostic tools. It is therefore very likely that in the future much more effort
will be expended in optimizing the interaction of antibodies and antigens with the
support in order to improve their range and effectiveness.
Core funding for the Hannah Research Institute is provided by The Scottish Office
Agriculture, Environment and Fisheries Department.
A. A. and Ed.
(1994).
Dalgleish and J. Leaver, in Food Polymers,Gels and Colloids Dickinson,
Royal Society of Chemistry, Cambridge, UK, 1991. pp. 1
G. Dalgleish, Colloids Surf. B: Biointerfaces (1993).
Dalgleish, Colloids Surf. (1990).
V. Brooksbank, Davidson, S. Horne, and Leaver. Chem.
A. Mackie, and A. North, Faraday Trans.
(1991).
J. Leaver, Brooksbank, and Colloid Interface
(1994).
A. Mackie. and Dann, in Food Polymers,Gels and Colloids
Dickinson, Royal Society of Chemistry, Cambridge, UK, 1991, pp.
A, Kondo and K. Colloid Interface Sci. (1992).
Norde, Cells Mater. (1995).
Bale. and Sutton, Colloid Interface Sci.
(1988).
Ortega Galvez and
(1996).
Duinhoven, Poort, van der Voert, Agterof, Norde, and
Lyklema, Colloid Interface (1995).
Duinhoven, Poort, van der Voert, Agterof, W Norde, and
J. Colloid Interface Sci. (1995).

A. P. Carriere. and Colloid Interface
(1996).
16. Vinuesa and R. Hildago-Alvarez, Sci. Ed.
(1994).
17. and Basinska, ACS Symp. (1992).
18. and Galisteo-Gonzalez, Rev. (1995).
19. K. Caldwell, and G. Dalgleish, Chromatogr. (1992).
20. Y. Jiang, J. Giddings, and Beckett, in Proteins at Interfaces 11: Fundamentals
and Applications A. Horbett and J. American Chemical Society,
Washington, DC, 1995, pp.
21. Kawaguchi, Hagiya, N. and Ohtsuka, Colloids
Surf. (1985).
22. J.-Y. Park, Kim, and Kim, Colloid Interface Sci.
(1996).
23. and W. Norde. Colloids Surf. B: Biointerfaces (1995).
24. A. Molina-Bolivar, Galisteo-Gonzalez, and Hidalgo-Alvarez, Colloids
Surf. B: Biointerfaces (1996).
25. A, Molina-Bolivar, and Hidalgo-Alvarez, Rev.
(1996).
26. Leaver and in Food Colloids and Polymers: Stability and
Mechanical Properties Dickinson and Walstra, Royal Society of
Chemistry. Cambridge, UK, 1993, pp. 332--340.
27. D. Dalgleish and Leaver, J. Colloid Interface (1991).
28. Leaver and Dalgleish, Biochem. Biophys. Acta (1990).
29. E. Dickinson, D . S. and Richardson, in Food Colloids
and Polymers: Stability and Mechanical Properties Dickinson and Walstra,
Royal Society of Chemistry, Cambridge, UK, 1993, pp.
30. Dickinson. Home, and Richardson, Food Hydrocolloids
(1993).
3 Fragneto, R. Thomas, A. Rennie, and Science (1995).
32. Leaver and Colloid Interface (1996).
33. Walstra, V. A. Bloomfield, Wei, and R. Jenness, Biochim. Biophys. Acta
(1977).
Th. W. Norde. and J. Colloids Surf. B:
Biointerfaces, 5:1 1 (1995).
35. Slomkowski, Kowalczyk, Basinska, and Wang, ACS Symp.
(1994).
36. C. Pap, A. van Hoek, W. Norde, and A. J. Visser,
37. and Caldwell, ACS Symp, Ser.
(1995).
38. A. Haynes and Norde, J. Colloid Interface Sci. (1995).
A. Haynes, and Norde, J. Colloid Interface
(1994).
40. Kottgen, B. A. Kage, and (1992).
41. Brynda, V. and Andrade, Colloid Interface (1990).

V. Ball, P. Huetz, A, Cazenave, and P. Proc.
Acad, USA (1994).
Hidalgo-Alvarez, and las Nieves, Sci.
Ed. (1995).
and Caldwell, Colloids Surf. Biointerfaces 7:9 (1996).
Norman, P. Williams, and (1992).
T. I. Armstrong, and Drug Targeting (1997).
A. Tuncel, A. C. and J.
634: (1993).
Li, J.-N. Lin, and Caldwell,
(1996).
A. Carmona-Ribeiro and M. Herrington, Colloid Interface
(1 993).
Sicchierolli and A. Carmona-Ribeiro, (1996).

Department of Pharmaceutical Technology, Chiba
University, Chiba, Japan
Department of Pharmaceutical Technology, Chiang
University, Chiang Thailand
I. Introduction
II. Adsorption onto Porous Materials
I I I . Improving Dissolution
IV. Effect of Adsorption on Chemical Stability
V. Fluorescence Spectroscopic Studies of Interaction
VI. Therapeutic Use of Porous Materials
VII. Conclusions
References
When poorly water soluble drugs are administered in solid dosage forms, the dis-
solution rate is often considered a rate-limiting factor in the adsorption process.
The rate of dissolution of these drugs depends on the effective surface area and the
energy state within the drug crystals The amorphous form of the drug can
exhibit a much higher dissolution rate and better bioavailability than the crystalline
form. Various techniques have been used to reduce the crystallinity or to obtain the
amorphous form of a crystalline drug The formation of an amorphous solid is
usually achieved by freeze drying, spray drying, solid dispersion, condensation from
the vapor, and milling. The rapid cooling of indomethacin liquid causes the glass
formation of the drug et [4] reported that amorphous
prepared by pulverization and lyophilization showed different dissolution behavior.
According to Yamaguchi the amorphous 16-membered macrolide com-
pound prepared by different conditions of spray drying showed different degrees of
solubility and stability.

When a system consists of multiple components, as pharmaceutical formulations
do, it is feasible that amorphous solid-state solutions can form in analogy to liquid
solutions. It has been reported that some crystalline medicinal compounds in mix-
tures with adsorbents or porous powders gradually became amorphous during
storage. These phenomena are believed to be caused by the adsorption of organic
medicinal molecules onto the surface of porous powders, as organic compounds
have relatively higher vapor pressure than inorganic compounds.
In this chapter we review the adsorption phenomena that can be commonly
encountered in pharmaceutics. As our interest was focused on the changes in
molecular states and pharmaceutical properties before and after adsorption, the
techniques used were a little different from those of mainstream adsorption
research works. The purpose of this chapter is to report general information
concerning the gaseous adsorption of pharmaceutical organic compounds onto
porous materials. The adsorption technique may have numerous pharmaceutical
applications that remain to be explored.
The changes in the X-ray diffraction patterns of mixtures of 30% flufenamic acid
(FFA) and 70% of either florite R or florite S after storage at 0% RH and are
shown in Fig. 1 In the mixture of F F A and florite the intensities of X-ray
diffraction peaks due to FFA crystals gradually decreased with period of storage,
and a halo pattern was observed after 5 weeks. In contrast, these X-ray diffraction
peaks were still observed in the FFA-florite mixture even after storage for 8
Changes in X-ray diffraction patterns of mixtures of 30% FFA and 70% of

either florite R (A) or florite after storage at 0% RH and Florite R: (a) Fresh
mixture; stored for 4 weeks; (c) stored for 5 weeks. S: (d) Fresh mixture; (e)
stored for 4 weeks; stored for 8 weeks. (From Ref.

weeks, though the crystallinity of the FFA was reduced. The X-ray diffraction data
indicated that florite has a greater capability to induce the transformation of
FFA into the amorphous state than does florite S. This was attributed to the
difference in the chemical compositions of florite and florite florite R being
composed primarily of porous calcium silicate and florite S of porous silicon diox-
ide,
The differential scanning calorimetric (DSC) curves of freshly prepared mixtures
of FFA and either florite R or florite S in various mixing ratios are shown in Fig. 2.
In the first run of the mixtures of FFA and florite those with 30% FFA showed a
broad endothermic peak between 90 and The mixtures containing 60%
F F A showed two overlapping endothermic peaks; the first peak was broad and
observed at low temperature, and the second peak was due to the melting of FFA
crystals. By increasing the mixing ratio of FFA, the area of the broad peak was
reduced while the second peak became more predominant. None of the mixtures
showed any peak in the second run. In the first run of the mixtures of F F A and
florite the mixtures that were 30-60% F F A showed a sharp endothermic peak
due to the melting of FFA at In the second run, they showed two endother-
mic peaks, the first peak attributable to the disordered F F A in the pores and the
second to melting of the crystalline
The DSC results confirmed that florite R had the greatest ability to induce the
transformation of FFA into the amorphous state. They also indicated that FFA
40 80 40 120
Temperature Temperature
curves of mixtures of FFA and either florite R or florite S in various

mixing ratios. (A) First run; (B) second run. (a) 30% FFA in florite (b) 60% FFA in
florite (c) 30% in florite (d) FFA in florite S. (From Ref. 6.)

had the following three phases in the mixture with porous powders: a crystalline
state that gave a sharp endothermic peak, a disordered state that gave a broad
endothermic peak at low temperature, and an amorphous state that gave no peak.
The spectroscopic results indicated that in the mixtures of FFA and florite
storage or grinding induced the transformation of F F A into the amorphous ionic
state, while in the corresponding mixtures of FFA and florite there was no change
in the molecular state of the drug.
The powder X-ray diffraction patterns of porous crystalline cellulose (PCC)
10% (EZ) mixtures before and after storage of the mixtures for 1
month at and 0, 40.0, and 97.0% relative humidity are shown in Fig. 3
In the freshly prepared mixture (A), X-ray diffraction peaks were observed at
= 14.5, 19.3, and that were attributable to crystals. Following storage
at 0 and 40.0% R H (represented by patterns B and C in Fig. 3), the X-ray diffrac-
tion peaks of EZ crystals disappeared. It was found that the mixing of EZ with
PCC under dry conditions led to the transformation of crystalline EZ into the
amorphous state. EZ molecules would be adsorbed physically onto the pore surface
of PCC. In the case of 97.0% (Fig. X-ray diffraction peaks of EZ crystals
were still observed; EZ remained in the crystalline state under this condition.
Matsumura et reported that coexisting water vapor caused a decrease in
the adsorption of methanol onto porous materials. At 97.0% the maximum
pore diameter for water condensation was calculated as All capillaries of
PCC were filled with water at 97.0% RH, and molecules of EZ had little chance to
adsorb onto the surface of PCC. These results indicated that the indispensable
condition for amorphization of by mixing with PCC was storage under dry
conditions.
Controlled pore glass (CPG) is used as a support for enzymes in size-exclusion
chromatography The adsorption of biological materials onto glass surfaces is a
Powder X-ray diffraction patterns of mixtures of 10% EZ and 90% after

storage at for 1 at various relative humidities. (A) Fresh mixture; 0%
(C) 40.0% RH; (D) 97.0% RH. (From Ref.

well-known phenomenon, the blood clotting reaction, macrophage adhesion,
an antigen-coated column, and low adhesion of transformed cells Surface
groups on and pore size distribution have received keen interest
because of their important roles in interface interaction in chromatography and
substrate adsorption in catalytic processes.
and coworkers investigated the molecular properties of benzoic acid,
ethyl and benzophenone mixed with porous glass powder by
DSC, powder X-ray diffraction, and gas adsorption measurements. The mixtures
had anomalous properties compared with the mixtures with glass beads; that is, (1)
a low concentration mixture (about did not show the melting peak and X-ray
diffraction peaks associated with the pharmaceutical crystals, (2) a high concentra-
tion mixture (about 20%) showed only a broad endothermic peak at a lower
temperature than the point, and (3) a higher concentration mixture
(more than 40%) showed an endothermic peak of the melting point together
with the broad peak at a lower temperature. The DSC peak area at the melting
point increased with an increase of drug concentration while the broad peak area
remained unchanged. Mixing with medicinal compounds caused a decrease of sur-
face area and a change in the pore distribution of the porous glass. From
these results it was concluded that the drugs took three phases in the phase
1, crystal structure; phase 2, disordered structure; and phase 3, probably adsorbed
on the pore walls. The amount of material in each of these phases was calculated
from DSC curves.
As crystalline drugs generally have very low vapor pressure at room tempera-
ture, almost no drugs noticeably sublime except for volatile drugs such as naphtha-
lene. However, when the drugs are mixed with porous powder, the capillaries can
provide a sink condition for the diffusion of the drug molecules in the mixture. The
rate of diffusion of benzoic acid was calculated for a mixture of 50 benzoic acid
and at The diffusion equation
is of the form
at steady-state under sink condition, where D is the diffusion constant of benzoic

acid in the air, N is the flux. C is the concentration of benzoic acid a t distance in
the diffusion layer, is the saturated concentration of benzoic acid in the air, and
X is the mean distance between the surfaces of benzoic acid crystals and the CPG
particles.
The following data were for the calculation of the diffusion rate. =
0.1 in air at 300K and 1 atm, X = 0.01 cm, and = 4.4 x
3
cm at The value was calculated from the vapor pressure of benzoic
acid of 8 x by assuming an ideal gas. Equation (2) gives

The specific surface area of benzoic acid was calculated to be 474 by using
the equation the value of 1.266 for the density, and 1 cm for the
mean particle diameter. Therefore, the surface area was 24cm 2 for the initial
amount of benzoic acid of mg. Thus, the diffusion rate of benzoic acid is calcu-
lated to be
Although this calculation was rather too simple, a considerably large value com-
parable to the total amount added was obtained, which could explain the rapid
disappearance of crystals of benzoic acid.
Drug molecules that penetrate into the capillaries have higher sticking probabil-
ities (adsorption rates) than those colliding with flat surfaces because they undergo
a great number of collisions with the capillary walls Thus, these
easily lose their kinetic energy during the collisions and rapidly become fixed on the
walls, which provide the sink condition at the surface of the porous powders.
T o verify that these phenomena take place via the gaseous phase, Konno et
studied the rate of amorphization using crystalline substances with various
vapor pressures. Benzoic acid and its derivatives were used as the crystalline sub-
stances, and activated carbon as the adsorbent. After each crystalline substance was
mixed with activated carbon in a porcelain mortar, the changes in the X-ray dif-
fraction patterns during storage were examined, and the time required for the
crystal peaks to disappear was determined. The relationship between the vapor
pressure of the crystalline substances and the time required to become amorphous,
plotted on a logarithmic scale, is shown in Fig. 4. At atmospheric pressure and
benzoic acid, which has the highest vapor pressure of the substances used,
became amorphous within only 1 whereas p-hydroxybenzoic acid, whose vapor
, , ,
10
Time
Plots of vapor pressure of the crystalline substances versus the time required to
become amorphous in mixtures with activated carbon. At and atmospheric
pressure; at and reduced pressure. 1, Benzoic acid; 2, o-hydroxybenzoic
acid; 3, acid; 4. p-hydroxybenzoic acid; 5, o-chlorobenzoic acid; 6,
acid; 7, p-chlorobenzoic acid. (From Ref. 13.)

pressure is the lowest, changed after 8600 h. Thus, it was observed that the higher
the vapor pressure, the faster the change to the amorphous state. A similar ten-
- -
dency was recognized at reduced pressure at The results suggest that the
phenomenon of the gradual loss at room temperature of the crystalline
properties of some organic pharmaceuticals mixed with adsorbents takes place
via the gaseous phase and is induced by the vapor pressure of the pharmaceuticals
themselves
Activated carbon, the residue from the destructive distillation of various organic
materials, was found to be capable of forming an amorphous solid. After storage of
a mixture of aspirin and activated carbon at for 7 days, the aspirin changed
from crystalline to amorphous. In the condition of reduced pressure, the
zation, the transfer of organic molecules to adsorbent surface took place more
rapidly as the mean free path of the organic molecules was much longer than at
pressure
In the presence of porous and nonporous silicon dioxides, organic crystallites
were converted into the amorphous state. Kim et al. reported the kinetic of the
transformation of the crystalline drug in the presence of various types of silicas
using DSC.
The relative enthalpy change obtained with the 1 : 1 drug-silica mixtures after
storage at is shown in Fig. 5. These thermal data clearly indicate that the rate
of crystalline transformation into the amorphous state was markedly dependent on
the types of silica employed. The mixtures with a mesoporous silica, such as silica
The change in endothermic transition energy with time for 1 : 1 mixtures of

drug and porous or nonporous silicon dioxide at 25Â°C Drug + silica 63 (porous);
drug + fumed silicon dioxide M-5 (nonporous); drug+ silica 244 (porous); (A)
drug + silica 74 (porous). (From Ref. 14.)

74 (mean pore diameter 15 nm) and silica 244 (mean pore diameter yielded
a fast rate of crystal transformation at In contrast, the crystalline drug mixed
with either a microporous silica such as silica 63 (mean pore diameter or a
nonporous fumed silicon dioxide remained intact and did not convert into the
amorphous state. Thus, the pore openings on the silica surface were considered
to be an important factor for the transformation of the crystalline drug into the
amorphous phase in the binary mixture systems. The crystal conversion process for
the drug in the silica 74 and silica 244 mixtures was biphasic; the initial process (up
to 6 generally followed first-order kinetics, and thereafter the slope changed. The
results indicate that the crystal modification in the initial stage occurred at a mark-
edly faster rate than that observed in the later phase of the conversion process.
Colloidal silicas are well known for their large surface area and highly polar
groups on the surface, which are capable of hydrogen bonding and electro-
static interactions with various organic compounds, in particular molecules that
contain hydroxyl groups such as alcohols. The adsorption interactions between the
surface hydroxyl groups of silica and drug molecules that also contain hydroxyl
groups are well documented For example, physical adsorption studies of
trimethyl on silica showed that the interaction is based on hydrogen
bonding with surface hydroxyl groups. Considering these physicochemical proper-
ties of colloidal silicon dioxide, it is quite reasonable to assume that drug molecules
having an or moiety can be hydrodgen-bonded to the silica sur-
face.
The primary use of the amorphous state in pharmaceutics is to increase

ability by increasing the rate and extent of drug dissolution. This relatively new
pharmaceutical technique plays an important role in increasing the dissolution and
absorption of drugs in dosage forms. An example of the use of porous materials to
improve the dissolution rate was reported by et The dissolution profiles
of FFA from newly prepared and stored mixtures with either florite or florite in
pH 1.2 solution at are shown in Fig. 6. The stored mixture of FFA and florite
exhibited a greater dissolution rate than the fresh mixture owing to the presence
of the drug in an amorphous state. The observed concentration of FFA after the
first 10 min is attributed to the stable crystalline state of FFA. However, when
florite was used, the stored mixture showed the greatest enhancement in the
dissolution rate of FFA due to the existence of FFA in an ionic
state. The decrease in the concentration of FFA after its initial fast dissolution
may be due to the conversion of the calcium flufenamate to the molecular form.
A clearer improvement of dissolution was shown in the mixtures with magne-
sium aluminum silicate (MAS) The dissolution profiles of FFA into various
dissolution media are shown in Fig. 7 . The dissolution profiles of F F A mixtures
with MAS stored at a reduced pressure were compared with those of fresh mixtures.
Since FFA was an acid with a value of 3.9, the amount of dissolved
changed markedly in this pH region. Furthermore, in mixtures with MAS stored at
a reduced pressure, the concentration of F F A became higher than in the case of

0 1 0 20 30 40 50 60
Time
Dissolution profiles of F F A from mixtures with either florite or florite in

pH 1.2 solution at (A) Mixture of 20% FFA and 80% florite mixture of
20% FFA and 80% florite Fresh mixture; mixture stored for 10 days at 0 %
and (From Ref. 6.)
Time
.7 Dissolution profiles of FFA from mixtures with MAS into various dissolution
media at 20% FFA, fresh; 20% stored for 5 h at a reduced
pressure at A, pH 1.2 solution (JP 1st fluid); 3.0 citrate buffer solution;
pH 4.0 acetate buffer solution; pH 5.0 acetate buffer solution. (From Ref. 16.)

freshly prepared mixtures, and the concentration then decreased gradually, showing
a typical supersaturation phenomenon. The longer the supersaturated state con-
tinued, the higher the pH value of the medium was. et reported
that the dissolution profiles of FFA from a solid dispersion system showed the same
supersaturation pattern. The decrease of FFA concentration was due to a phase
transition to stable forms accompanied by crystallization.
Naproxen, a nonsteroidal anti-inflammatory drug, is practically insoluble in
water. For pharmaceutical application, improvement in the dissolution properties
has been expected. Dissolution properties of naproxen in the first fluid (pH 1.2)
were studied according to the JP paddle method (Fig. 8) A mixture with
120 showed a dissolution curve similar to that of intact naproxen crystals. It is
noteworthy that the dissolution of naproxen was significantly improved in the
heated sample with CPG 120, especially in the initial stage of dissolution. In
T i m e (min)
Dissolution patterns of naproxen in JP first fluid at (a) Naproxen-CPG

120 system; naproxen-CPG 1000 system. Naproxen crystals; fresh mixture
of 10% naproxen crystal and 90% CPG; mixture after heating at for 3
Each point represents the mean value = 3). (From Ref. 18.)

contrast, only slight improvement of naproxen dissolution was observed in the
heated sample with CPG 1000. Consequently, the improved dissolution of
naproxen results from the molecular dispersion of naproxen on the CPG surface
and is affected by the pore structure of CPG.
The need to ensure product quality and hence stability of the drug concerned in the
product is obvious. Decomposition leads to a decrease in drug potency, and the
decomposition products may be unpleasant or toxic. Stability is therefore an
important issue with pharmaceutical products.
The hydrolytic decomposition of aspirin has been widely studied in aqueous
media. It has been shown that the rate of hydrolysis of solid aspirin is proportional
to the relative humidity of the atmosphere The effect of additives on this
hydrolysis has been investigated, and the presence of additives has been found to
affect the decomposition rate of aspirin.
The kinetics of aspirin hydrolysis were studied in mixtures containing 5%
aspirin. The pore diameters of varied from 75 to Plots of the
order rate constants at 50Â° under the two different relative humidities versus the
pore diameter of CPG are shown in Fig. 9 In the CPG mixtures where the pore
diameter was less than the rate constants of aspirin decomposition were
independent of CPG pore diameters in the cases of both RH and 79% RH.
However, the rate constant in the mixtures with CPG of large pore size decreased
with increasing CPG pore diameter. The pore diameter profiles of the rate constant
indicate that the stability of aspirin was closely related to changes in the crystal-
P o r e diameter
Effect of pore diameter (log scale) on the first-order rate constant for the
decomposition of aspirin at Aspirin concentration is 5%. 0% RH; 79%
(From Ref. 20.)

linity of aspirin in the mixtures. That is, greater and similar decomposition rate
constants were observed in the of small pore diameter (less than
where aspirin was in a completely amorphous state.
Figure 9 also reveals that the decomposition rate constants observed at 0%
were greater than those at 79% R H in samples from CPG 75 to C P G 350. In the
cases of the C P G 1000 and 3000 mixtures, the decomposition at 79% R H was
faster than that at 0 % RH. Although hydrolysis in the solid state was generally
accelerated by humidification, the opposite situation was observed in the mixtures
with small-pore The decomposition rate constant in the C P G 75 mixture at
RH and was determined as 4.90 indicating a half-life of 14.1
h for aspirin decomposition.
The R H dependency of the first-order rate constant of aspirin decomposition in
the C P G 170 at is illustrated in Fig. 10. An anomalous linear relation-
ship was obtained between the R H and the rate constants; as the RH was raised,
the rate constant decreased. and Mattocks proposed the mechanism of
aspirin decomposition in the solid state as follows. After the formation of a water
layer on aspirin particles, aspirin was dissolved into the water layer. The aspirin
then decomposed in the solution. In the crystalline state, the amount of adsorbed
water increased at a high RH; therefore, faster aspirin decomposition was observed.
On the other hand, the stability of aspirin in the solid dispersed system was affected
by many factors such as hygroscopicity of additives, on the surface, specific
surface area, and dispersed state. et reported that the aspirin in the
coprecipitated samples with povidone or urea had slightly higher degradation rates
due to its increased water sorption ability. In the case of a urea solid dispersion,
Relative
Effect of on the first-order rate constant for the decomposition of aspirin

mixed with CPG 170 at Aspirin concentration is 1%. (From Ref, 20.)

urea could also be dissolved into a thin water layer and imparted a slight alkalinity
to this film.
By way of contrast, Gore and Banker showed an enhanced aspirin stability
after the addition of colloidal silica at a high RH. They interpreted these results in
terms of the action as due to an internal moisture scavenger. The water molecules
were entrapped by the silanol groups of colloidal silica. The surface of CPG was
also covered by silanol groups as well as by colloidal silica; however, the concen-
tration of aspirin in these experiments was as low as differing the high
concentrations (85-99%) used by Gore and Banker. Therefore, the simple mechan-
ism of preferential water vapor adsorption of the surface should not be
applied to the explanation of anomalous behavior of aspirin decomposition in
the C P G 170 mixtures. The competitive adsorption of aspirin and water molecules
onto the CPG surface was considered a suitable concept to explain the R H depen-
dence of aspirin decomposition in the CPG mixture.
The aspirin concentration in the C P G 170 mixture varied from 0.3 to 30%. The
condition at and R H was used to study aspirin decomposition kinetics. In
the mixtures in which the concentration was less than the rate constants were
large and steady. In those with concentrations of more than increasing
concentrations of aspirin caused a decrease in the decomposition rate constant.
In the aspirin concentration range of both the independence of the
rate constant from the aspirin concentration and the good linearity of first-order
plots were observed. The results suggest that the aspirin molecules were dispersed
homogeneously in C P G 170 in low aspirin concentration mixtures and that a
sufficient amount of water was adsorbed to hydrolyze aspirin on the 170
surface. The decrease of the aspirin decomposition rate at concentrations greater
than 2.0% should be ascribed to the remaining crystalline portion of aspirin in the
mixture as well as to the time-consuming process of the change from crystalline to
amorphous aspirin.
Matsumoto investigated to the stability of aspirin in mixtures of
aspirin and at various temperatures and humidities. They calculated the
intrinsic rate constants of aspirin hydrolysis from the relationship between the
apparent first-order hydrolysis rate constant and the amount of water adsorbed
at each temperature. The apparent first-order rate constant of aspirin and the
amount of adsorbed water in the heated mixture of PCC and aspirin at different
temperatures are listed in Table 1 . It was suggested that the apparent first-order
decomposition rate constants of aspirin in the heated mixtures were closely related
to the amount of adsorbed water.
The relationship between the amount of adsorbed water and the apparent
order rate constant of aspirin in the heated mixtures at various temperatures is also
shown in Fig. 11. The amount of adsorbed water was defined as the number of
moles of adsorbed water per kilogram of the mixture. At each temperature, the
apparent first-order rate constant of aspirin increased with increases in the amount
of adsorbed water, and a linear relationship was observed. From Fig. 11, the
following equation can be derived.

1 Apparent First-Order Decomposition Rate Constants of Aspirin and
Amounts of Adsorbed Water in PCC-Aspirin Mixtures at 30, 40, and
Temp. RH Apparent rate Amount of adsorbed
constant ( d a y 1 ) water (%)
Source: Ref. 24
where is the apparent first-order rate constant of aspirin in the heated

aspirin mixture, k is the intrinsic rate constant of apsirin, is the amount of
adsorbed water, and a is the intercept on the abscissa. The values of k and a
obtained from (3) at each temperature are listed in Table 2. As the values of
k were independent of the amount of water, they are the intrinsic hydrolysis rate
constants of aspirin at each temperature. The values of a at 30, 40, and were
Amount of adsorbed water mixture)
. 11 Relationship between the amount of adsorbed water and apparent first-order

decomposition rate constants of aspirin in PCC-aspirin mixtures at various
temperatures. (From Ref. 24.)

Values of Intrinsic Hydrolysis Rate Constant of Aspirin (k) and Critical
Amount of Water (a) in PCC-Aspirin Mixtures at 30, 40, and
Temp. k 1 Amount of adsorbed
(x kg water
30 3.37 0.534 0.999
40 9.17 0.547 0.999
50 24.4 0.493 0.994
Source: Ref. 24.
in good agreement, and they can be considered the critical amounts of water when
aspirin hydrolysis takes place in the heated mixtures.
Edwards investigated the hydrolysis of aspirin in aqueous solution at 17Â°
and demonstrated that the aspirin is hydrolyzed by general acid-base catalysis and
water molecules for ionic and nonionic aspirin, comprising six simultaneous reac-
tions involving O H , and for ionic and nonionic aspirin. The intrinsic
hydrolysis rate constant in the heated mixture was comparable with the hydrolysis
rate constants of the two-element reactions of nonionic aspirin and or ionic
aspirin and in aqueous solution. As aspirin molecules would be adsorbed onto
the pore surface of PCC in the molecular state, as a possible mechanism of aspirin
hydrolysis in the mixture with PCC it was suggested that the aspirin is dispersed
monomolecularly in the heated mixture and reacts with water molecules rather than
by acid-base catalysis.
The photophysical behavior of molecules can provide information about molecular

orientation and association in the adsorbed state. The molecular state of drug
molecules should be related strongly to drug stability and The
solid-state emission and excitation spectra and the fluorescence decay of pyrene,
which was adsorbed onto porous additives, were investigated by Yamamoto
PCC and Inertsil were used as the porous additives. Mixtures of pyrene
and porous additive were prepared at definite mixing ratios (pyrene content
10%) by simple blending and were kept at 298 for various lengths of time. For a
1.0% pyrene-PCC system, the monomer emission peak was observed and the
intensity increased for 3 days, while the excimer intensity changed slightly (Fig.
12). The fluorescence patterns changed with duration of storage time. In the case of
the 1.0% system, the changes in monomer and excimer
occurred in the same way, but the rate of change was faster than that of the
PCC system. We propose two steps for this adsorption process: the adsorption
of molecules onto the surface of additives and the formation of the
ground-state dimer. The ability to form the ground-state dimer would be associated
with the surface hydrophobicity of the additives.

monomer
350 400 450 500 550 400 450 500 550

Wavelength
Wavelength
. 12
Change in fluorescence spectrum of 1% pyrene-PCC by storage at
= Left: Changes within 1 day. Right: Changes within 2 weeks.
(From Ref.
Drug delivery applications of drug-loaded porous materials are considered to be

advantageous for chemotherapy. Therapeutic agents incorporated into porous
materials include antibiotics, and hormones. Hydroxyapatite
is an important mineral component of hard tissue such as
teeth and bone. Therefore, when used in implanted artificial hard tissue,
xyapatite has a high affinity for natural hard tissue in situ and can be molded to fill
spaces created by physical damage to the bones teeth. Yamamura and
applied hydroxyapatite beads to bone filler with drug-delivering
ability and reported the release of adriamycin from the beads.
On the other hand, self-setting calcium phosphate cement is transformed into
hydroxyapatite and sets within 0.5 h in situ. The cement system is easy to handle
during clinical application and after setting. It possesses excellent bioaffinity for the
hard tissues and in effect serves as artificial bone. et investigated a
drug delivery system for antibiotics, polypeptides, and an anti-inflammatory drug
from the pharmaceutical perspective, using a self-setting calcium phosphate cement
as a new way to deliver drugs to bone. They demonstrated the controlled release of
6-mercaptopurine (6-MP), an anticancer agent, from a self-setting calcium phos-
phate cement over a period of 100 h. The results suggest that a skeletal drug
delivery system with self-setting calcium phosphate cement could be developed
into an effective treatment for localized bone cancers in patients following surgery.
Shimada et reported a new drug delivery system using small silica par-
ticles that can adsorb bleomycin and are easily transferred to the lymphatics to
deliver a large dose of anticancer agents to the regional lymph nodes. As shown in
Table 3, transplanted tumors had increased in size at the time of sacrifice in the

Anticancer Effect of Silica (BLM-SI) and Bleomycin
Solution (BLM-SOL) on Transplanted Tumor and Metastatic Lymph Nodes
Tumor size Tumor size (mm) Lymph node
at injection at sacrifice metastasis
BLM-SI = 11) 13.1 6.3 3.0a
BLM-SOL = 11) 11.8 19.6
Control = 11) 12.0 2.1 29.1 9.2
0.01 significantly different to control
Source: Ref. 28.
control and bleomycin solution groups. In contrast, the size of the tumor signifi-
cantly decreased in the bleomycin-small silica group 0.01, p 0.05). Lymph
node metastasis occurred in ail the animals of the control group, but the incidence
of such metastasis was 64% 1) in the bleomycin solution group and only 9%
11) in the bleomycin-small silica group. A new drug delivery system using silica
particles might be useful for the control of tumor and lymph node metastasis.
When mixed with porous materials, organic medicinal molecules changed their
arrangement in the crystal lattice from the crystalline state to a disordered phase,
accompanying changes in pharmaceutical properties. This transformation pro-
ceeded through the sublimation of organic from the crystals and adsorp-
tion onto the porous materials via the gaseous phase. The adsorption equilibrium
depended on the temperature, water content, and surface properties of the porous
materials. The resulting solid phase consisted of continuous disordered states invol-
ving a liquid-like molecular arrangement near the surface, an amorphous portion
including a vitreous state, and imperfect crystals where the crystal growth was
highly inhibited in the pore. Since solid dosage forms such as tablets, capsules,
and pills are muiticomponent systems containing the active ingredient, diluents,
binders, lubricants. and other pharmaceutical additives, it is important to recognize
that adsorption phenomena have a great influence on the effectiveness, stability,
and safety of dosage forms.
1. Byrn, Pfeiffer, Ganey, C. Hoiberg, and Poochikian, Pharm. Res.

(1995).
Hancock and G. Zografi, Sci. (1997).
3. Fukuoka, Makita, and S. Yainamura, Chem. Bull. (1986).
4. S. Maeda, Yonemochi, T. Oguchi, Yamamoto, and Nakai,
Chem. Bull. (1992).

Yamaguchi, Okamoto, Takeuchi, and K. Yamamoto,
Pharm. (1992).
A. S. Yamamoto, A. El-Sayed., and Nakai,
Pharm. Bull. 40:1289 (1992).
Matsumoto, Nakai, Yonemochi, Oguchi, and Yamamoto,
Pharm. 108: (1994).
Matsumura, Yamabe, and Takahashi, Carbon (1985).
Yazawa, Hyomen (1995).
Schnabel and P. Ber, (1989).
Yonemochi, Kojima, A. S. Okonogi, Oguchi, Nakai, and
K. Yamamoto, Colloid Interface (1995).
Adam, The Physics and Chemistry o f Surfaces. Oxford Press, Oxford,
1938, p. 274.
Konno, Kinuno, and Kataoka, Chem. Pharm, Bull. (1986).
Kin, M. Frank, and Henderson, Pharm. (1985).
Rupprecht, Biersack, Kindl. and Liebl, Pharm. Ind. (1976).
T. Konno and Kinuno. Pharm. Bull. (1989).
Imaizumi, N. Nambu. and Chem. Pharm. Bull.
(1982).
T. Oguchi, Tozuka, Okonogi, Yonemochi, and Yamamoto,
Yakuzaigaku 57:168 (1997).
Carstensen, F. Attarchi, and Hou, Pharm. (1985).
Yonemochi, Matsumura. Oguchi, Yamamoto, and Nakai, Chem.
Pharm, Bull. 39:1027 (1991).
J. and A. Mattocks, J. Am. Pharm, (1958).
El-Banna, A. and A. A, El-Fattah, Sci. (1982).
Y. Gore and Banker, Pharm. (1979).
Matsumoto, Nakai, Yonemochi, Oguchi, and Yamamoto, Drug
Stability (1996).
J. Edwards, Trans. Faraday (1950).
Yamamoto, Pacific Basin Workshop on Adsorption Science and Technology,
Kisarazu, Japan, 1997.
K. and Int. Pharm. (1992).
Matsuda, Fox, and W. I. Pharm.
(1995).
M. Shimada, S. Natsugoe, and Anticancer Res. (1995).

Untitled

Uploaded by

Untitled

Uploaded by

Copyright 1999 by Marcel Dekker. All Rights Reserved.

URFACTANT S IENCE SERIES

DANIEL BLANKSCHTEIN ERIC W. KALER

S. KARABORNI CLARENCE MILLER

LISA B. QUENCER DON RUBINGH

JOHN F. SCAMEHORN BEREND SMIT

Copyright 1999 by Marcel Dekker. All Rights Reserved.

8. Anionic Surfactants: Chemical Analysis, edited by John Cross (out of print)

21. Interfacial Phenomena in Apolar Media, edited by Hans-Friedrich Eicke and

Copyright 1999 by Marcel Dekker. All Rights Reserved.

48. Biosurfactants: Production Properties Applications, edited by Naim

49. Wettability, edited by John Berg

Copyright 1999 by Marcel Dekker. All Rights Reserved.

ADDITIONAL VOLUMES IN PREPARATION

Interfacial Phenomena in Chromatography, edited by Pefferkom

Solid-Liquid Dispersions, Xueping and Wolfgang von

Modern Characterization Methods of Surfactant Systems, edited by P.

Interfacial Forces and Fields, edited by Jyh-Ping

Copyright 1999 by Marcel Dekker. All Rights Reserved.

Surface Characterization Methods: Principles, Techniques, and Applications,

Copyright 1999 by Marcel Dekker. All Rights Reserved.

Copyright 1999 by Marcel Dekker. All Rights Reserved.

Copyright 1999 by Marcel Dekker. All Rights Reserved.

Copyright 1999 by Marcel Dekker. All Rights Reserved.

In the world of nanoparticles and microporous materials, the interaction forces

Copyright 1999 by Marcel Dekker. All Rights Reserved.

Copyright 1999 by Marcel Dekker. All Rights Reserved.

Copyright 1999 by Marcel Dekker. All Rights Reserved.

Part I Preparation, Characterization, and Transport Properties of Nanoparticles

Copyright 1999 by Marcel Dekker. All Rights Reserved.

8. Microporous Structure of Collagen Fibers 185

Copyright 1999 by Marcel Dekker. All Rights Reserved.

22. Surface Ionization and Complexation 593

24. Ionic Adsorbates on Hydrophobic Surfaces 645

25. Adsorption of Metal Ions onto Humic Acid 661

27. Adsorption of Ions onto Alumina 71 1

28. Protein Adsorption onto Latex Particles 743

Copyright 1999 by Marcel Dekker. All Rights Reserved.

Paul R. Anderson Department of Chemical and Environmental Engineering,

Lisa Axe Department of Civil and Environmental Engineering, New Jersey

aumgarten Institut fur Physikalische Chemie und Elektrochemie,

ergstrom Institute for Surface Chemistry, Stockholm, Sweden

oki Department of Pharmaceutical Sciences, Kinki University, Osaka,

. Carr Department of Chemistry, University of Minnesota, Minneapolis,

oelho Equipe Milieux Poreux, Institut de Physique du Globe de Paris,

Cristian I. Contescu Department of Chemical Engineering and Materials

Copyright 1999 by Marcel Dekker. All Rights Reserved.

ussell Crawford Centre for Applied Colloid and BioColloid Science,

D. D. Do Department of Chemical Engineering, University of Queensland, St.

dames A. Finch Department of Mining and Metallurgical Engineering, McGill

Mietek daroniec Department of Chemistry, Kent State University, Kent, Ohio

Soon-Yong deong Chemical Process and Engineering Center, Korea Research

Katsumi Kaneko Department of Chemistry, Chiba University, Chiba, Japan

Michal Kruk Department of Chemistry, Kent State University, Kent, Ohio

Jeffrey Leaver Molecular Recognition Group, Hannah Research Institute, Ayr,

Department of Applied Chemistry, Royal Melbourne

Ion V. McCormick Department of Chemical Engineering and Materials

onia Maria Cabral e Menezes Chemistry Division, Petrobras Research and

. Miller Department of Metallurgical Engineering, University of Utah, Salt

disson Morgado, Jr. Catalysts Division, Petrobras Research and Development

atya P. Moulik Center for Surface Science, Department of Chemistry, Jadavpur

Copyright 1999 by Marcel Dekker. All Rights Reserved.

konogi Department of Pharmaceutical Technology, Chiang Mai

Tsuneo Okubo Department of Applied Chemistry, Gifu University, Gifu, Japan

livier Micromeritics Instrument Corporation, Inc., Norcross,

Carlos Roberta Nogueira Pacheco Chemistry Division, Petrobras Research

anczyk Department of Theoretical Chemistry, Marie-Curie-

Joung-Man Park Department of Polymer Science and Engineering, Gyeongsang

idyut K. Paul Geological Studies Unit, Indian Statistical Institute, Calcutta,

Ljubisa R. Radovic Department of Materials Science and Engineering, The

ao Corporate Science and Technology Center, Air Products and

er Department of Chemical Engineering and Materials Science,

Wiadyslaw Rudzinski Department of Theoretical Chemistry. Marie-Curie-

chwarz Department of Chemical Engineering and Materials Science,

ideshi Seki Department of Marine Bioresources Chemistry, Hokkaido

hapiro Department of Mechanical Engineering, Technion-Israel