Professional Documents
Culture Documents
Colloid Science
SECOND EDITION
Robert J. Hunter
School of Chemisty
University of Sydney
OXFORD
UNIVERSITY PRESS
OXFORD
UNIVERSITY PRESS
Great Clarendon Street, Oxford 0x2 ~ D P
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide in
Oxford New York
Athens Auckland Bangkok Bogoti Buenos Aires Calcutta
Cape Town Chennai Dar es Salaam Delhi Florence Hong Kong Istanbul
Karachi Kuala Lumpur Madrid Melbourne Mexico City Mumbai
Nairobi Paris S5o Paulo Singapore Taipei Tokyo Toronto Warsaw
Oxford is a registered trade mark of Oxford University Press
in the U K and in certain other countries
Published in the United States
by Oxford University Press Inc., New York
0 Robert J. Hunter, 2001
The moral rights of the author have been asserted
Database right Oxford University Press (maker)
First published 2001
Reprinted 2002, 2004
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
without the prior permission in writing of Oxford University Press,
or as expressly permitted by law, or under terms agreed with the appropriate
reprographics rights organization. Enquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the address above.
You must not circulate this book in any other binding or cover
and you must impose this same condition on any acquirer
A catalogue record for this book is available from the British Library
Library of Congress Cataloging in Publication Data
(Data applied for)
ISBN 0 19 850502 7
Typeset by EXPO Holdings, Malaysia
Printed in Great Britain
on acid-free paper by Biddles Ltd., King’s Lynn, Norfolk
PREFACE TO THE SECOND EDITION
It is now over ten years since the second volume of the first edition went to press and
much has happened in colloid science in the interim. Such a short time frame has not,
however, significantly affected the foundations of the subject. Why then go to the
trouble of preparing a second edition? Whence comes the motivation and opportunity?
The motivation came from my feeling that the two volume work, as presently
constructed, is less effective as a teaching tool than it might be. Books of this kind are
seldom read from cover to cover but when an author takes sole responsibility for the
content of a book the natural tendency is to try to tell a coherent story and to do so in as
linear a fashion as possible. Because of the collaborative nature of the exercise, that was
not always possible with the first edition.
In 1998, through the good offices of my friend and colleague Professor Sture
Nordholm, I had the opportunity of trialing the entire scope of the book in a graduate
course in the University of Goteborg. A group of about 40 Scandinavian students from
various universities and industries submitted themselves to a 90 lecture presentation
over a period of about two months. In this I was helped by Professors John Gregory
(UC London), Roland Kjellander (Goteborg), Stjepan Marcelja (ANU), Ron Ottewill
(Bristol), Dimo Platikanov (Sofia), The0 van de Ven (McGill and PPRIC), and Lee
White (Carnegie-Mellon). Although the first edition was the reference text, the
lecturers were, of course, free to develop their material in whatever way they chose. I
have taken the opportunity to incorporate some of their many ideas into this revised
version of the text.
The editors of Oxford University Press were enthusiastic about the possibility of a
second edition, especially if it could be confined to a single volume. I therefore
undertook to revise the manuscript completely and to pare it down to around 900
pages, whilst at the same time updating the sections where change had been most
significant. Inevitably some material had to go and rather than try to make small
excisions everywhere I opted to simply remove a whole section from the scope: the
material on thin films and emulsions. That is by no means to minimize its importance
but rather to acknowledge that it is in many ways a separate study. Certainly those
subjects draw on common material, from thermodynamics and double layer overlap to
name just two, but there is also a good deal of material which is unique to the behaviour
of emulsions. I was also influenced by the appearance of new texts which provide a
much more detailed treatment of these subjects than I could hope to do in the few
pages that could be made available to them. I refer to the books by D. Exerowa and
P.M. Kruglyakov Foam and foamJilms (Elsevier 1998) and by B.P. Binks (ed.) Modern
aspects of emulsion science (Royal SOC.Chem. (London) 1998).
Those familiar with the first edition will notice the extensive changes in the order of
treatment of the subject. I have used one rule throughout: treat each subject as soon as
possible in the context of what has already been covered, with a minimum of forward
vii
viii I P R E F A C E TO THE SECOND EDITION
hunterr@chem.usyd.edu.au
Sydney,Janua y 2000
Solutions to the problems are available by email from the author at the above address.
2 THERMODYNAMICS OF SURFACES
2.1 Introduction 45
2.2 Surface energy and i t s consequences 45
2.3 Thermodynamics of surfaces 56
2.4 The Gibbs adsorption equation 63
2.5 Thermodynamic behaviour of small particles 72
2.6 Equilibrium shape of a crystal 81
2.7 Behaviour of liquids in capillaries 84
2.8 Homogeneous nucleation 93
2.9 Limits of applicability of the Kelvin and Young-Laplace
equations 97
2.10 Contact angle and wetting behaviour 100
2.11 Measurement of surface tension and contact angle 112
IX
X I CONTENTS
9 ASSOCIATION COLLOIDS
9.1 The critical micellization concentration (c.m.c) 435
9.2 Factors affecting the c.m.c. 438
9.3 Equilibrium constant treatment of micelle formation 443
9.4 Thermodynamics of micelle formation 450
9.5 Spectroscopic techniques for investigating micelle structure 460
9.6 Micellar dynamics 466
9.7 Molecular packing and i t s effect on aggregate formation 472
9.8 Statistical thermodynamics of chain packing in micelles 476
APPENDICES
Appendix A1 Calculation of the allowed surface interaction modes
in modern Dispersion Force Theory 767
Appendix A2 Evaluation of the sum of the roots of the dispersion
relation 768
Appendix A3 Vector calculus and Poisson’s equation 770
Appendix A4 Fourier transforms 777
Appendix A5 Elementary thermodynamic relationships in the
absence of surface contributions 780
Appendix A6 Electrical units 784
INDEX 787
Nature of Colloidal Dispersions
1.IIntroduction
1.2 Technological and biological significance of colloidal dispersions
1.3 Classification of colloids
1.4 Some typical colloidal dispersions
1.4.1 Preparation of colloidal dispersions
1.4.2 Monodisperse sols
1.4.3 Association colloids
1.4.4 Emulsions
1.4.5 Clay minerals
1.5 Brownian motion and diffusion
1.5.1 The one-dimensional random walk
1.5.2 The phenomenology of diffusion
1.5.3 Time-dependent diffusion processes
1.5.4 The Einstein-Smoluchowski equation
1.6 Electrical charge and colloid stability
1.6.1 The electrical charge a t a surface
1.6.2 Observation of coagulation behaviour
1.6.3 Coagulation by potential control
1.6.4 Coagulation by electrolyte addition
1.6.5 The critical coagulation concentration
1.6.6 Forces between colloidal particles
1.7 Effect of polymers on colloid stability
1.7.1 Steric stabilization
1.7.2 Polymer flocculation
1.IIntroduction
In the normal solutions encountered in chemical situations, the solute and solvent
molecules are of comparable size and we normally assume that the solute molecules
are, on average, dispersed uniformly throughout the (continuous) solvent. There is
an important class of materials, however, in which the kinetic units that are
dispersed through the solvent are very much larger than the molecules of the
1
2I 1: NATURE OF COLLOIDAL DISPERSIONS
solvent. Such systems are called colloidal dispersions and they may arise in a variety
of ways.
If a substance, A, is insoluble in substance B it will usually be possible to break A
down into very small particles that can be distributed more or less uniformly through
the substance B. Substance A is then called the disperse phase and substance B, the
dispersion medium. In general, A and B may be either solids, liquids, or gases so the
dispersion could be regarded as a state of matter, accessible to any substance A, given
the appropriate temperature and pressure and a means of producing and maintaining
small discrete lumps of A distributed throughout B. If A is a solid, the particles may be
produced by crushing and grinding a macroscopic piece of pure A, by growing small
crystals of A by some chemical reaction, or by controlled crystallization of A from some
solvent. Just how one goes about distributing A through the medium B (which might
be a solid, a liquid, or a gas) and maintaining the discrete nature of the A particles (i.e.
preventing aggregation) forms a considerable part of the theory and practice of colloid
science.
The lower limit of size for dispersions of this kind is around 1 nm. Smaller particles
would ultimately become indistinguishable from true solutions. The upper limit is
normally set at a radius of 1 p m but there is no clear distinction between the behaviour
of particles of 1 p m and the somewhat larger particles often encountered in emulsions,
in mineral separation processes, and in ceramic engineering.
There are, of course, some molecules that are individually larger than 1 nm in
size. These ‘macromolecules’ can often be uniformly dispersed through a fluid
medium and they then form a colloidal solution or dispersion. Proteins,
polysaccharides (such as starch), and many synthetic polymers fall into this
category. It was just such a substance (a naturally occurring gum) that suggested
the name ‘colloid’ (from the Greek word for glue) to the pioneer investigator in the
field, Thomas Graham, in the 1860s. Th e field of polymer science has now
developed into an entirely separate discipline with a vast specialized literature of its
own. We will, therefore, treat only those few parts of it that are of most significance
in colloid science.
A third class of colloidal dispersions arises when a number of molecules of normal
size associate together to form an aggregate. Soap molecules, for example, if they are at
a sufficiently high concentration in a suitable solvent, can associate together to form
micelles. These structures are of colloidal dimensions and the resulting system is
referred to as an association colloid. (The term colloidal electrolyte is also used for ionic
soap and detergent systems.)
The distinguishing feature of all colloidal systems is that the area of contact between
the disperse particles and the dispersion medium is relatively large. The energy
associated with creating and maintaining that interface is significant, so the study of
interfacial (or surface) chemistry is an integral part of the study of colloids. We will,
however, limit ourselves to those aspects of surface chemistry of direct relevance to
colloids. It is, incidentally, not necessary that all of the dimensions of the disperse
particle be very small for the system to be of interest to colloid scientists. Some
important colloidal systems have particles that can readily be seen with the aid of an
ordinary microscope, such as textile fibres or the cellulose fibres used in making paper.
It is sufficient if one of the characteristic dimensions of the particle (in this case the
INTRODUCTION 13
diameter) is small. The gas bubbles that make up a foam are usually larger than colloid
size but in that case the thin lamellae of liquid between the bubbles have a thickness of
the order of colloidal dimensions. They, therefore, also fall within the purview of
colloid science.
Most of the examples quoted above refer to dispersions of solids in liquids and
this is the area with which we will be principally concerned. Many of the
conclusions are, however, applicable to liquids in liquids (emulsions) and gases in
liquids (foams). Dispersions in which the continuous (dispersion) phase is solid are
becoming increasingly important in the field of material science where many new
composites are being developed, but this area remains outside the scope of the
present volume (see, for example, Evans and Langdon 1976 and the monographs by
Schwartz 1996, 1997). Likewise, dispersions of solids and liquids in gases will not
be treated in any detail. Such aerosols as they are called, are certainly colloidal
systems but the theoretical concepts required to describe their behaviour, although
similar in some respects, differ quite significantly from those with a liquid
dispersion medium. There are, in addition, a number of recent works concerned
with the solid-gas interface (e.g. Hercules 1992, Somorjai 1994) and the most
recent edition of Adamson’s long-standing text (Adamson and Gast 1997) gives a
strong coverage of that area.
Table 1.1 lists the various types of colloidal system with common examples of each.
Note that all possible combinations can be realized except that of gas in gas.
Table 1.7
The various types of colloidal dispersion with some common examples.
The nomenclature is adapted fiom Ostwald (1 907)
Exercises
1.1 . I Give further examples of each of the colloidal systems listed in Table 1.1 and
state their approximate composition.
1.1.2 Starting with a cube of solid, 1 cm along each edge, what is the total surface area
when the solid is subdivided into cubes lop4 cm on each edge? Repeat the
calculation for lop5 cm and lop6 cm cubes. Calculate the surface energy per
particle in each case, assuming the surface energy is 70 mJ m-' and compare this
with thermal energy ( k T )at room temperature (25 "C). What is the total surface
energy for each system?
1 . I .3 Show that the surface area per unit mass, A,, of particles of density p is given by:
A, = k'/pr where Y is some characteristic dimension, and k' = 3 for spheres,
2 for thin cylindrical discs and long rods, and 4 for long square prisms. (Am is
called the specific surface area and typical values for a colloidal material fall in the
range 1-lo3 m2 g-I.1
1.1.4 Chemical bonding energies are commonly of the order 100 kJ. Show that surface
energies of particles will approach this value for sizes of about 1 nm. (Assume
the surface energy is 0.1 J m-2 and take reasonable values for the density and
molar mass.)
1.1.5 A mineral oxide of density 2.8 g cmP3is broken up into colloid sized particles in a
ball mill. Calculate the total surface area of the crushed material (m2 g-') when
the average particle radius is 5 x loP4 cm. What is the total area when the
particle radius is 500 A?
Apart from the widely recognized colloidal nature of protein and polysaccharide
solutions there are many other biological systems that have been studied using the
methods of colloid science. The flow properties of blood are best understood in terms
of its being a colloidal dispersion of (deformable) flat plates (the red corpuscles) in a
liquid (Goldsmith and Mason 1975). The flow properties of faecal material must
sometimes be modified by colloid chemical techniques to avoid unpleasant
physiological consequences. The greater part of the food processing, preserving,
and packaging industry rests heavily on colloid chemistry (see, for example, Ritsch and
Reineccius 1995), and agricultural scientists require a knowledge of the colloidal
properties of soils in order to induce optimum plant growth.
Medical practitioners have used a colloidal dispersion of gold (potable gold) since
the Middle Ages for treating a variety of ailments. Modern colloidal microcapsule
techniques allow controlled release of a drug (Lee and Good 1987, De and Maitra
1997) and, in some cases, accurate targeting onto a particular organ. More routine
applications of colloid chemical principles crop up in the preparation of emulsions and
suspensions that must remain homogeneous for long periods on the shelf (or at least be
readily redispersed on shaking). Aqueous emulsions of perfluoro-hydrocarbons have
also been developed as (temporary) blood substitutes (Riess and Le Blanc 1978).
Apart from its contributions to engineering, agriculture, biology, and medicine,
colloid science also has an important role to play in reducing the harmful effects of
technological development. Many pollution problems are due to the presence of
unwanted colloidal materials, and their removal (from air or waterways) calls for the
application of colloid chemical techniques. The specific adsorptive properties of
colloids can also be used to remove, to concentrate, and possibly to recover, industrial
products (especially metal ions) from air and water.
come into contact with all or most of those atoms then solvation energy will be
important and the colloid should be lyophilic (reversible) in some suitable solvent. If
the solvent is prevented, by the structure of the suspended particles (i.e. the disperse
phase), from coming into contact with any but a small fraction of the atoms of those
particles then the colloid will almost certainly be lyophobic (i.e. irreversible) in its
behaviour, even if the surface atoms interact strongly with the solvent.
When the dispersion medium is water, the terms hydrophilic and hydrophobic are
used; the great majority of the present work is devoted to the examination of
hydrophobic sols.
The lyophilic colloid solution is thermodynamically stable since there is a reduction
in the Gibbs free energy when the ‘solute’ is dispersed. The strong interaction between
‘solute’ and solvent usually supplies sufficient energy to break up the disperse phase
( A H < 0) and there is often an increase in entropy as well; any reduction in solvent
entropy due to the interaction with ‘solute’ is usually more than compensated by the
entropy increase of the ‘solute’. For the lyophobic colloid, the Gibbs free energy
increases when the disperse phase is distributed through the dispersion medium so
that it is a minimum when the disperse phase remains in the form of a single lump. A
lyophobic colloid can, therefore, only be dispersed if its surface is treated in some way
that causes a strong repulsion to exist between the particles. In this way the particles
can be prevented from aggregating (or coagulating) for long periods, although it must
be emphasized that they are still thermodynamically unstable and the barrier to
coagulation is merely a kinetic one+. Given enough time they will ultimately form an
aggregate.
There is a well developed theory to describe the interaction between particles of a
lyophobic colloid but the behaviour of lyophilic colloids is more difficult to describe.
The reason for this is that all of the forces involved in lyophobic systems are also
important for lyophilic systems but in addition, for the lyophilics, there are very strong
specific solvent effects that are difficult to predict. We will, therefore, spend quite
some time developing the theory of lyophobic colloids, choosing as typical examples
the silver halide, clay mineral, metal oxide, and polymer latex sols, since these have
been well characterized and much studied. The principles that emerge from those
studies will then be applied to a number of more practical systems.
+Strictly speaking one should distinguish the aggregate of particles from the bulk
solid. The dispersion is metastable with respect to the bulk solid.
SOME TYPICAL COLLOIDAL DISPERSIONS 17
of dispersions in which the particles all have almost the same size and shape. Such
systems are ideal for testing aspects of the theory of dispersions, they have some
interesting properties in their own right, and they may even offer some special
advantages in certain technological processes.
We will also examine some naturally occurring colloidal systems that offer special
advantages for testing various theoretical analyses. It will come as no surprise to learn
that the theoretical description of the equilibrium, kinetic, and transport properties of
colloidal systems is almost always confined to certain simple geometric shapes; usually
the sphere or infinite flat plate but sometimes the cylindrical rod or disc and more
rarely the spheroid. T o adequately test the validity of these descriptions one must have
available colloidal dispersions in which the particle shapes conform as accurately as
possible to these simple geometric types and that has only recently become possible.
Only by improving our understanding of such model systems can we hope to improve
our descriptions of the behaviour of real colloidal systems.
Svedberg (1928) divides the preparation of colloidal dispersions into two categories:
dispersion and condensation, of which the latter is probably more important for fine
material. In the dispersion methods, a sample of bulk material is broken down to
colloidal dimensions by some kind of mechanical process. The most direct method is
by grinding in a colloid mill. This device subjects a coarse suspension of particles to a
very high shear field by forcing it into a narrow gap between two surfaces that are
rotating rapidly with respect to one another. The particles are then torn apart by the
shearing process and a colloidal dispersion results, provided that the solution contains
a suitable dispersing agent to prevent the small particles from aggregating together (see
Section 1.4.3 below). A similar effect can be achieved, especially with liquid-in-liquid
dispersions (emulsions), by subjecting a mixture of the two phases to a high frequency
sound wave (-20 kHz). This process, known as ultrasonication, also requires the
presence of a dispersing agent if a stable sol is to result.
Sols can also be formed by passing an electric arc between two wires placed under
the surface of a liquid (Bredig’s procedure). This is also an example of a dispersion
method in that some of the sol almost certainly results from pieces of metal being torn
from the surface of the wire; there is, however, some condensation from the vapour
also involved.
Condensation methods are much more numerous and more diverse. They may
involve dissolution and reprecipitation, condensation from the vapour or chemical
reaction. In the first category is the formation of a solid paraffin sol in water. This can
be done by dissolving paraffin wax in ethanol and pouring a little into a large volume of
boiling water. The ethanol rapidly boils off leaving an opalescent dispersion of the
paraffin. The second type is exemplified by the spontaneous formation of a mist or fog
from a supersaturated vapour; provided the degree of supersaturation is sufficiently
high (Svedberg suggests a vapour pressure more than eight times the equilibrium
value) the formation of many droplets of very small size (< 1 pm) is assured.
The chemical methods may involve reduction, oxidation, or double decomposition.
Metal sols can be produced by reduction (gold by reducing chlorauric acid, HAuC4,
with hydrogen peroxide or red phosphorus; silver from silver nitrate and ferrous
citrate). The particles are usually very small (<< 100 nm) and with some recipes are so
well stabilized that they behave as though they were lyophilic (Frens and Overbeek
1969).
8I I : N A T U R E OF COLLOIDAL DISPERSIONS
Oxidation can be used to produce sulphur sols either from hydrogen sulphide:
Nucleation period
----
Nucleation concentration
Time
Fig. 1.4.1 Illustrating the production of a monodispersesol by confining the formation of nuclei to
a very short period, so that the particle number remains constant and all grow together to the same
size. (After Overbeek 1981.)
10 I 1: NATURE OF COLLOIDAL DISPERSIONS
Fig. 1.4.2 Monodisperse inorganic colloids: (a) zinc sulphide (sphalerite);(b) cadmium carbonate;
(c) a-Ferric oxide (haematite); (d) basic ferric sulphate (alunite).(Photographscourtesy of Professor
E. Matijevic, Clarkson University, N.Y.)
12 I 1: NATURE OF COLLOIDAL DISPERSIONS
Air
Aqueous solution
Micelle
Fig. 1.4.3 (a) Conventional representations of a surfactant molecule as a rod or a flexible tail (the
hydrocarbon chain) and a head group. The chain is not infinitely flexible but is limited by the C-C-C
bond angle. Fairly free rotation can occur except at double bonds. The cross-sectional area of the
paraffin chain is about 0.2 nm2 when fully extended and this is comparable to the head group size for
-OH and -NH2, but smaller than -SO;. (b) Schematic arrangement of amphiphile (soap or
detergent) molecules at low concentration in water. (Note that counterions are not shown.) (c) The
situation above the critical micellization concentration (c.m.c.). Note that adsorption will also occur
on the walls of the vessel. The arrangement there is not shown because it is more problematical and
is, in any case, dependent on how hydrophilic the surface of the vessel is.
(hydrophilic) (Fig. 1.4.3(a)). The hydrophobic part is non-polar and usually consists
of aliphatic or aromatic hydrocarbon residues. The hydrophobic character is not much
affected by introducing halogens and similar groups (Laughlin 1981). The hydrophilic
part consists of polar groups which can interact strongly with water (especially
hydroxyl, carboxyl, and ionic groups).
The fatty acid soaps are typical examples:
CH3( C H Z ) O.SOzO-Na+
~
sodium dodecyl sulphate or SDS'
alkylbenzene sulphonate
(sodium salt)
sodium tetradecyl sulphate R is a (preferably linear) alkyl chain (NC12)
'Sometimes referred to as sodium lauryl sulphate. Lauryl alcohol, from which that
name is derived, is the C12 alcohol but lauric acid is the C12 carboxylic acid so it has a
C11 chain. Somewhat confusing.
14 I 1: N A T U R E OF COLLOIDAL DISPERSIONS
The most significant characteristic of this type of amphiphile is the tendency to adsorb
very strongly at the interface between air and water (Fig. 1.4.3(b)); in doing so, the
hydrophobic part of the molecule can escape from the aqueous environment whilst the
hydrophilic head group can remain immersed in the water. Such substances are said to
be strongly surface-active because they lower the surface (or interfacial) tension, y.
They therefore make the formation of new surface easier and are widely used as
foaming and dispersing agents. Commercial surface active agents (or surfactants) are used
for a variety of purposes: as cleaning agents (detergents), colloid chemical stabilizers,
and wetting agents. We will return to these matters in Chapter 2. For the moment we
are concerned with another of their properties.
At very low concentrations (< lop4 M, say) many surface active agents are soluble in
water to form simple solutions; if they are ionic, like the fatty acid soaps or the alkyl
sulphate detergents they will be dissociated as weak or strong electrolytes, respectively.
Some of the molecules will also be preferentially adsorbed at the surfaces of the
solution (i.e. the air-solution interface if there is one, and at the walls of the container)
(Fig. 1.4.3(b)). As the concentration rises this adsorption becomes stronger until
saturation is reached when the molecules are packed close together with strong lateral
interactions occurring between the hydrophobic chains, which tend to stick up out of
the water (Fig. 1.4.3(c)). [Note that some soap molecules will almost certainly be
adsorbed on the walls of the beaker. The arrangement there is not shown because it is
more problematic and depends to a considerable extent on how hydrophobic the
surface is.]
Another aggregation process, which often occurs at about the same surfactant
concentration, is the formation of micelles (Fig 1.4.3(c)). These are structures in
which the hydrophobic portions of the surfactant molecule associate together to
form regions from which the solvent, water, is excluded. T h e hydrophilic head
groups remain on the outer surface to maximize their interaction with the water
and the oppositely charged ions (called counterions). A significant fraction of the
counterions remains strongly bound to the head groups so that the lateral repulsive
force between those groups is greatly reduced. T h e precise structure of the micelle
depends upon the temperature and concentration but also on the details of the
molecular structure: size of head group, length and number of hydrocarbon chains,
presence of branches, double bonds or aromatic rings, etc. We will deal with those
matters in Chapter 9. For the moment we restrict attention to the simplest
amphiphiles: those with a polar (hydrophilic) head group at one end and one or two
straight hydrocarbon chains attached. This includes the simple soaps and
detergents and some natural lipids. These substances form micelles of colloidal
size and, as noted in Section 1.1, are called association colloids or, more rarely,
colloidal electrolytes.
The concentration at which micelles first form in the solution is called the critical
micellization concentration (c.m.c.). It is marked by quite sharp changes in slope when
various transport and equilibrium properties (like electrical conductivity and surface
tension) are plotted against concentration.
The initial suggestions on micellar structure by Hartley and by McBain have been
refined by the work of Stigter (1967) and many others. Suffice it to say at this stage that
the long chain fatty acid soaps and simple detergents like sodium dodecylsulphate
initially form micelles that are spherical in shape and have a fairly well defined
SOME TYPICAL COLLOIDAL DISPERSIONS I 15
Fig. 1.4.4 A sodium dodecyl sulphate micelle. The more detailed picture, which emerges from a
statistical mechanical calculation of the likely structure. (To be discussed in more detail in Chapter
9.) (Drawn by Dr J.N. Israelachvili from calculations by Dr D.W.R. Gruen, Australian National
University.)
aggregation number (- 50 molecules for sodium stearate) (Fig. 1.4.4). They are,
therefore, monodisperse.
As the surfactant concentration is increased above the c.m.c., the initially spherical
micelles become more distorted in shape, forming cylindrical rods or flattened discs
(Fig. 1.4.5). Ultimately, at high ratios of soap to water they form liquid crystals and
other so-called ‘mesomorphic phases’, a discussion of which would take us beyond the
scope of this book. An interesting recent development is the use of such surfactant
phases as templates for the formation of microporous solids of well-defined structure.
These are in great demand in molecular separation processes (Ciesla and Schuth 1999).
Under other circumstances amphiphilic substances can form two dimensional
membranes, or bilayers to separate two aqueous regions, very similar to a biological
membrane (Fig. 1.4.5(B)). If the bilayer is continuous and encloses an aqueous region
the result is a more or less spherical vesicle (Fig. 1.4.6(a)) or a microtubule in which a
double layer of surfactant encloses a cylindrical water region. This type of structure,
including the liquid crystal, provides a number of possible models for investigating
colloid chemical behaviour under controlled conditions.
In non-aqueous media, small amounts of water are able to act as nuclei for the
formation of inverse micelles in which the surfactant head groups point inwards to
stabilise the water droplets.
Fig. 1.4.5 (A) A cylindrical micelle and (B) a bilayer. (Reproduced from Evans and Wennerstrom
1999 with permission.)
1.4.4 Emulsions
Emulsions, like solid dispersions in liquids, can be formed by either condensation or
dispersion methods. The use of mechanical dispersion methods is more common in
this case because the energies involved are generally smaller. Apart from the
ultrasonication technique mentioned above (Section 1.4. I), high-speed stirring or
shaking of a two-phase liquid mixture can often induce emulsification, especially if a
dispersing agent is present. The resulting droplets are, of course, spherical, provided
that the interfacial tension (i.e. surface energy) is positive and sufficiently large.
Many two-phase systems are able to undergo spontaneous emulstJication especially if a
third component is present. This term applies strictly only to those systems in which
no mechanical energy at all is required, though it is sometimes applied to systems that
are simply easy to emulsify. In some cases, the spontaneous emulsification occurs
because of the presence of a surface active agent, which lowers the interfacial tension
essentially to zero. Negligible energy is then required for the formation of the
SOME TYPICAL COLLOIDAL DISPERSIONS I 17
Water
Water
Water
oil
Fig. 1.4.6 (a) Section through the centre of a bilayer vesicle. The wall thickness is about twice the
chain length and there is water on both sides of the surfactant. The conformation of the chains is
close to that of the liquid hydrocarbon. (b) A surfactant micelle swollen by the presence of some
solubilized oil. The molecules of oil are all in intimate contact with surfactant hydrocarbon chains.
Incorporation of more oil would lead to an oil in water microemulsion (c). In that case there would be
a separate pure oil region (containing no surfactant) in the interior. The microemulsion systems are
usually generated by using two surfactants -often an ionic detergent (filled circles) together with a
neutral dipolar compound of similar chain length (open circles) (called a co-surfuctunt).
emulsion. A special case of this sort is the formation of ‘microemulsions’, in which the
droplet size is very small (-10 nm). Such systems have attracted a good deal of
attention recently because of their technological significance. Although the surface
tension is essentially zero, the microemulsion droplets are spherical and almost
monodisperse (Overbeek 1980). The oily interior of a micelle can be used to take up
more oil (a process called sobbilkation, which is important in detergency). There is
then no sharp dividing line between an oil-swollen micelle (Fig. 1.4.6(b)) and a
18 I 1: N A T U R E OF COLLOIDAL DISPERSIONS
microemulsion of oil in water (Fig. 1.4.6(c)); nor, for that matter between a
microemulsion and a normal emulsion.
In some cases of spontaneous emulsification, the interfacial tension (or energy)
remains positive but the energy necessary for emulsification is supplied by the
redistribution of a solute between the two phases. Davies and Rideal (1963) give a
detailed account of such a phenomenon (with photographs), when a solution of toluene
in alcohol is mixed with water. The mechanism of emulsification is best described as
‘diffusion and stranding’. Toluene and alcohol diffuse simultaneously into the aqueous
phase and as the soluble alcohol diffuses ahead, the insoluble toluene is left stranded in
the aqueous phase as small droplets.
Provided the interfacial tension or energy, y, is sufficiently high, the emulsion
droplets must be spherical and this should make emulsion systems ideal candidates as
model systems for testing colloid theories. Unfortunately, they have some drawbacks:
the droplet size distribution is often rather wide, the kinetic behaviour of the droplets
is sometimes affected by the fact that the interior is fluid and therefore potentially
mobile and the possibility of coalescence leading to a change in droplet size is a further
complication. The microemulsions referred to earlier, being spherical and almost
monodisperse offer special advantages as model systems.
1.4.5Clay minerals
The inorganic fraction of soils and most natural sediments consists almost entirely of
silica and the various silicates. The term clay is used in soil science and agriculture to
mean any material of particle size less than 2 p m but the term clay mineral refers to
specific groups of silicate minerals. Some clay minerals have long been used in the
ceramic industry because their plate-like crystal habit and ability to bond to one
another when heated to high temperatures makes them suitable for making bricks,
earthenware, and pottery (including china). Clays are also used extensively as fillers in
making paper, paint, and rubber tyres, among other things. Indeed, so extensive is the
use of these materials in industry that clay minerals rank second only to oil in terms of
tonnages used.
We are concerned here with the special property of silicon, when bonded in a certain
way with oxygen, to form extensive flat plates or sheets?. When combined with similar
flat sheets of an aluminium oxide they can produce layered crystals, which in
favourable cases can be cleaved to yield surfaces that are believed to be atomically
smooth and flat over relatively large distances (of the order of a few square
millimetres). Such smooth solid surfaces have made it possible in recent years to make
measurements of the forces between solid particles with an accuracy and reliability not
previously possible (see Chapter 12).
Space does not permit a detailed discussion of the structure of the layer silicates.
The reader is referred to the texts of Grim (1953) or van Olphen(1977). We will
consider only the basic structures (talc, pyrophyllite, and kaolinite) and the ways in
tCrystallographers refer to these as layers rather than sheets. We will use ‘sheets’ to
avoid confusion with the electrical double layer formed when the particles are
dispersed in water (section 1.6).
SOME TYPICAL COLLOIDAL DISPERSIONS I 19
Fig. 1.4.7 Arrangement of silica tetrahedra in hexagonal rings to form a layer. Only the oxygens are
visible; a silicon sits at the centre of each tetrahedral arrangement of oxygens. The apical oxygens
(top layer) are shared with the adjoining alumina layer. Note the hole in the centre of the hexagonal
ring in the lower layer of oxygens. The counterions which balance the crystal charge can, in some
cases (e.g. K+) sit in those holes.
which the first two are modified to yield vermiculite, mica, and montmorillonite, since
these are the systems that have been most extensively used as colloid chemical models.
The basic silicon-oxygen unit is a tetrahedron with four oxygens surrounding the
central silicon. The bonds are approximately 50 per cent ionic and 50 per cent covalent
in character and in the clay minerals of interest to us the tetrahedra are linked to form
hexagonal rings (Fig. 1.4.7). This pattern can be repeated indefinitely in two
dimensions to form the sheet. Aluminium in combination with oxygen (and hydroxyls)
forms an octahedron with the aluminium at the centre and again these octahedra can be
linked to form a more closely packed two-dimensional sheet.
In the kaolinite crystal a sheet of alumina octahedra sits on top of a sheet of silica
tetrahedra with the apical oxygen atoms from the silica being shared with the
aluminium atoms of the upper layer (Fig. 1.4.8(a)). This ‘ideal’ structure would be
completely uncharged and the perfect kaolinite crystal would be built up by laying
these double sheets one on top of another. The bonding of one double sheet to the next
occurs partly through van der Waals forces and partly through hydrogen bonds from
the OHs of the octahedral sheet to the oxygens of the next silica sheet (Fig.1.4.8(b)).
These bonds are so numerous that only rather drastic treatments are able to prise open
the structure. The crystals are normally hexagonally shaped discs with an axial ratio of
20 I 1: NATURE OF COLLOIDAL DISPERSIONS
Fig. 1.4.8 (a) A diagrammatic sketch of the 'ideal' kaolin layer [Al(OH)2]2.O.(SiO2)2. One hydroxyl
ion is situated within the hexagonal ring of apical, tetrahedral oxygens and there are three others in
the uppermost plane of the octahedral sheet. The two sheets combined make up the kaolin layer.
(b) Simplified schematic diagram of the kaolinite structure. Note that the upper and lower cleavage
surfaces in the perfect crystal are quite different (but see text). A typical crystal would have about
100 or so layers.
(c) A typical kaolinite crystal of aspect ratio ( d u )about 0.1. Note the negative charges on the basal
planes (perpendicular to the c-axis of the crystal) and positive charges around the edges. The latter
are eliminated at pHs above about 7.
the order of 1O:l (Fig. 1.4.8(c)) and they seldom grow to sizes of more than a few
micrometres.
The real crystals are usually considered to carry a negative charge on the basal
surfaces (i.e. the larger flat surfaces). It cannot be attributed to dissociation of the few
Al-OH groups (since they would produce either positive or negative charges
depending on the pH). In any case, recent evidence (Ma and Eggleton 1999a) suggests
that in most kaolinite deposits the outer planes are modified so that only the silica
surfaces are exposed. This is also the probable result of the traditional preparation
procedure (Posner and Quirk 1964) which is designed to remove any contaminating
aluminium from the basal planes.
It is generally assumed that the crystal charge is due to substitution of aluminium for
silicon in the tetrahedral layer with a consequent imbalance of negative charge. The
edges of the crystal, where imperfections necessarily occur because of bond breakage,
carry a positive charge at low pH and this decreases to zero as the pH is raised to about
7 (Schofield and Samson 1954).The positive charge is generally considered to be due
to dissociation from the aluminium octahedra:
at low pH and is partly compensated by negative charges from the silica tetrahedra.
Recently Ma and Eggelton (19996) have argued that in the kaolinites which they have
studied the crystal charge is confined solely to the edges and there is no isomorphous
SOME TYPICAL COLLOIDAL DISPERSIONS 121
substitution. Whether this is always so is an open question; it has always been difficult
to verify the substitution because it does not influence the X-ray data and reverse
compensations (substituting silicon for aluminium) can also occur in the body of the
crystal.
Kaolinite is called a 1:l non-swelling dioctahedral clay because (i) it has one silica
layer to one alumina layer and (ii) the double sheets do not separate from one another
under any normal conditions. The term dioctahedral refers to the fact that only two out
of every three possible sites in the octahedral layer are occupied by aluminium ions.
The other clay minerals of interest to us are all of the 2:l type (i.e. two sheets of silica
to one of alumina or two of silica to one of magnesium oxide). The two parent materials
arepyrophyllite (with alumina in the central layer) and talc (with magnesia in the central
layer) (Fig. 1.4.9a). Pyrophyllite is again a dioctahedral whilst talc is a trioctahedral
mineral, since all three possible sites must be occupied by M$+ to obtain charge
balance.
One important difference between the 1:l and 2:l minerals is that in the 2: 1 minerals
there is no possibility of hydrogen bonding between successive triple sheets. The basal
oxygen planes can interact with each other only by way of van der Waals forces. They
are, therefore, very easily cleaved along this plane.
The parent minerals pyrophyllite and talc are not of much interest to us but if they
are modified by the substitution of some of the tetrahedral silicon by aluminium, they
develop very interesting new properties. Replacing one quarter of the silicons by
aluminium in pyrophyllite generates muscovite or white mica (Fig. 1.4.9(b)). This
confers a very large negative charge on each sheet and that charge must be balanced; in
muscovite it is balanced by the presence of potassium ions, which can fit snugly in the
hexagonal hole of the silica sheet shown by the unbroken lines in Fig. 1.4.9(b). This
greatly strengthens the bonding between each triple sheet so that mica does not tend to
expand in water. A similar replacement of silicon by aluminium in the talc structure
generates another form of mica called phlogopite. A further substitution of some of the
octahedral magnesium in phlogopite by other divalent metal (usually Fe2+) ions
produces another mica called biotite. Mica crystals can be very large (many centimetres
across) and can be readily cleaved in air or vacuum. Good specimens when carefully
cleaved can yield the macroscopic atomically smooth surfaces referred to earlier.
Montmorillonite can also be related to the pyrophyllite structure by the substitution
of approximately one in six of the aluminium ions in the octahedral layers by
magnesium or other divalent ions. Again this generates a negative charge throughout
each triple sheet and this must be compensated by the presence of cations in the inter-
layer region (Fig. 1.4.10). The material called Wyoming bentonite, which finds many
uses (as a filler, a catalytic support, and in drilling muds), consists of a mixture of
montmorillonite with a related material called beidellite in which aluminium ion is
isomorphously substituted for silicon in the tetrahedral layers.
When dry montmorillonite is placed in a moist atmosphere, it is able to take up
water vapour by adsorbing it between the triple-sheets (that is, in the interlayer region)
and the shape of the water vapour adsorption isotherm suggests that about four layers
of water can be taken up into this region. The same behaviour occurs when the clay is
immersed in concentrated salt solution (-1 M) (Norrish 1954). The spontaneous
hydration is presumably associated with the presence of the cations, since the parent
material (pyrophyllite), which has no such ions, is quite hydrophobic.
22 I 1: NATURE OF COLLOIDAL DISPERSIONS
Fig. 1.4.9 (a) A sketch of the ideal 2:l layer silicate. The trioctahedral mineral talc (MgO)z.Mg
(OH)z.(Si02)4] has all three octahedral sites occupied by Mg2+. In pyrophyllite [(AlO(OH))z.
(Sioz)~]only two of the three octahedral sites are occupied by the counterions required to balance
the crystal charge after isomorphous replacement. (b) Schematic diagram of the white mica
structure. The potassium ions are shown (large hatched circle in (a)) and they balance negative
charges in the silica layers caused by the substitution of about a quarter of the silicon ions by
aluminium. The ‘ideal’ formula is: K+[(AlO(OH))z .(AlSi308)]-.
In salt solutions, another important phenomenon occurs and one which has been
studied in clay mineral systems for over a century. This is the process of cation
exchange, which can occur between ions in solution and the ions in the interlayer region
(i.e. those on the basal cleavage planes). The process is made easier if the electrolyte
concentration is lowered because the triple-sheets (or platelets) can then separate from
one another to allow easier access to the adsorbed cations. The separation of the
platelets occurs more readily if the interlayer cations are monovalent and strongly
hydrated (Na+ or Li+ for example) because the inter-platelet repulsion is stronger in
that case, as we will show in Chapter 12. The number of cations adsorbed on the clay
(the cation exchange capacity) is an important characteristic of the material and is a
direct measure of the degree of isomorphous substitution that has occurred in the
Next Page
Fig. 1.4.10 Schematic diagram of the montmorillonite structure. It is very similar to that of mica
(Fig. 1.4.9(b)) but the negative charges are now in the central (octahedral) layer and are fewer in
number (about one in six aluminiums is replaced by Mg2+ to give: Na033[(A11.67Mgo.33) (0(OH))2
(Sioz)~]).The lower charge density, larger distance between the positive and negative charges, and
the poor fit of the counterion (when hydrated) make it possible for sodium montmorillonite to
expand readily when placed in water.
pyrophyllite structure. A typical value for montmorillonite would be about one mole of
univalent charge per kilogram of clay (see Exercise 1.4.7). All clay minerals that have
crystal structures with unbalanced charge exhibit ion exchange behaviour (e.g.
kaolinite, vermiculite), which is of particular importance in determining the retention
and availability of plant nutrients in the soil. Indeed, the strong affinity of the
potassium ion for certain clay minerals is, to a large extent, responsible for the relative
dominance of the sodium ion in sea water.
The final clay mineral we wish to discuss is vermiculite. Its structure is derived from
that of talc by the substitution of about one in six of the tetrahedral silicons by
aluminium. The balancing (exchangeable) cation is often magnesium but it may be
replaced by other divalent or monovalent ions. Vermiculite, like mica, can form large
sheets that show a pronounced cleavage parallel to the plane of the sheets.
Apart from the exchange of simple cations, there has been a very large amount of
work done on the adsorption of more complex ions and molecules onto
montmorillonite and vermiculite or their intercalation into the interlayer regions
(see, for example, Weiss 1963).
Before leaving the subject of clay minerals it should be emphasized that the chemical
compositions given above apply to the ideal (or idealized) crystals. The real materials,
as they occur in nature, seldom conform to these idealizations and that must be taken
into account in interpreting their behaviour. It should also be mentioned that there are
other geometries available in naturally occurring clay minerals (e.g. the thin cylindrical
rods of attapulgite; van Olphen 1977, p. 7), which have so far not been much exploited
for the testing of physical models of colloidal systems (though, see Buscall 1982).
r
Exercises
1.4.1 Prepare a paraffin wax sol in water by the method described above. Use about
5-10 ml ethanol saturated with paraffin wax and pour it into about 600 ml of
boiling distilled water in a 1 L beaker. (Careful: the alcohol boils very vigorously.)
Use this sol to examine the general statements made in the text. Note, for
example, the Tyndall effect when a light beam is passed through the sol. Put
a drop on a microscope slide and examine it in dark field illumination.
The rmody namics of Surfaces
2.1 Introduction
2.2 Surface energy and its consequences
2.2.1 Surface tension and surface free energy
2.2.2 Molecular origins o f surface tension
2.2.3 Pressure differences across curved surfaces - the Young-Laplace
equation
2.3 Thermodynamics of surfaces
2.3.1 Mechanical work done by a system with a surface
2.3.2 Surface excess quantities
2.3.3 Fundamental equations o f surface thermodynamics
2.4 The Gibbs adsorption equation
2.4.1 The relative adsorption
2.4.2 The general form o f the y-In c relation
2.4.3 Particular forms o f the Gibbs equation
2.4.4 Two liquid phases in contact
2.5 Thermodynamic behaviour of small particles
2.5.1 The Kelvin equation
2.5.2 Applications of the Kelvin equation
(a) Drops of liquid in a vapour
(b) Bubbles in a liquid
2.5.3 Effect of temperature on vapour pressure -the Thomson equation
(a) Liquid drop suspended in i t s vapour
(b) Bubble immersed in a liquid
2.5.4 Application of the Kelvin and Thomson equations t o solid particles
2.6 Equilibrium shape of a crystal
2.7 Behaviour of liquids in capillaries
2.7.1 Capillary pressure
2.7.2 Capillary condensation
2.7.3 Capillary rise in a powder
2.8 Homogeneous nucleation
2.9 Limits of applicability of the Kelvin and Young-Laplace equations
2.10 Contact angle and wetting behaviour
2.10.1 Adhesion, cohesion, and wetting
44
SURFACE ENERGY AND ITS CONSEQUENCES 145
2.1 Introduction
Lyophobic colloids are thermodynamically meta-stable (Section 1.6). Nevertheless,
the time-scale of many of the molecular exchanges occurring within a lyophobic
colloidal suspension is very short compared with the lifetime of the suspension.
Processes involving such molecular exchanges may, therefore, be treated by
equilibrium thermodynamics. In particular, adsorption equilibrium at the particle
surface is rapidly established, and so thermodynamics may be used to describe the
effects of surface active materials, including ions, on the properties of the suspension.
Such thermodynamic descriptions are an essential underpinning for the more
detailed molecular descriptions with which much of this book is concerned. The
analysis here follows the lines laid down in the introductory text by Aveyard and
Haydon (1973) and especially the more extensive analysis provided by Defay et al.
(1966), as translated by Everett.
We begin with a discussion of the important concepts of surface tension and surface
free energy and their origins at the molecular level. The consequent pressure
difference across a curved interface is then calculated (the Young-Laplace equation).
This gives rise to some important effects, which will be taken up after the basic
equations of surface thermodynamics have been introduced. The thermodynamic
treatment given here is intermediate between the rather oversimplified procedures,
which merely demonstrate the reasonableness of the key results, and the rigorous
procedures that are needed to cover all contingencies. In most cases we have chosen to
simplify the notation by sacrificing a little generality rather than by sacrificing rigour
on the grounds that the former is easier to recover than the latter.
The consequences of pressure differences across curved interfaces are particularly
relevant to colloidal particles (Sections 2.5 and 2.6) and are also important in capillary
phenomena (Section 2.7). A number of specific problems are then discussed, finishing
with some references to methods for measuring surface tension and contact angle
(Section 2.11).
where ZU,bulk is the number of molecules in the shell of nearest neighbours in the bulk
phase and q, is the average distance of these neighbours from the central molecule.
The energy EA,Sper molecule at the surface is, similarly:
so that
but
is positive, the interface will tend to shrink to its minimum possible area. However, if
SE < 0, the interface will tend to grow and the phases will tend to dissolve in each
other. [Strictly, we should deal with the free energy that would include an entropy
contribution favouring dissolution even if E were positive, provided it was not too
large.]
SURFACE ENERGY AND ITS CONSEQUENCES 147
Interface
Total foyce
f
Fig. 2.2.1 (a) Increasing the area of a surface of arbitrary shape. (b) The definition of surface
tension. 6A is an element of area in the surface bounded by a perimeter of length I (see text).
48 I 2: THERMODYNAMICS OF S U R F A C E S
Table 2.7
Surface and interfacial tensions of some liquids (in mN m-‘) at 293 K
(firom Aveyard and Haydon 1973, p . 70, with permission)
4y/d T
(liquid-vapour)
Liquid-vapour Water-liquid (mN m-l K-l)
6w = Fax (2.2.9)
y = F/l. (2.2.10)
That is, y acts like a force per unit length of the perimeter opposing any attempt to
increase the area, i.e. like a restoring force or tension (Section 3.4 below). More formally
(Fig 2.2.l(b), if we consider an area element in the surface, 6A bounded by a perimeter
of length 1 then y is conceived as a force per unit length exerted by the region outside
6A to keep it from shrinking. y is called the surface (or interfacial)+tension, and is
generally expressed as mN m-l (millinewton per metre, equivalent to dyne cm-’ in cgs
units). For a liquid-liquid or liquid-vapour interface, the equilibrium value of y is
independent of the orientation of the line element 61, so that the surface is in a state of
uniform tension in every direction if the surface is quiescent.
Typical values for the surface tensions of some pure liquids and interfacial tensions
for liquid-liquid systems are given in Table 2.1.
An important element in the above argument about ‘extending a surface’ or ‘creating
fresh surface’ is that we mean new surface with the same properties as the original
surface. In other words, the surface is created by adding new molecules from the bulk,
maintaining the same properties of the different molecules in a mixed system, and with
the molecules in their equilibrium configuration. If the increase in surface area were to
be achieved by increasing the average distances between the molecules (in the surface
+It is common usage to use the term ‘surface tension’ if one phase is a gas, but
otherwise to use ‘interfacial tension’.
SURFACE ENERGY AND ITS CONSEQUENCES 149
and bulk), then the extension of the surface would be accompanied by a change in the
bulk energy of the system. This would be an elastic deformation of the body; it would,
in general, require more energy input and the work done would almost certainly
depend non-linearly on the area increase, 6A. For a pure liquid in contact with its own
vapour or with another pure, immiscible liquid, the surface tension is numerically
equal to the excess surface free energy, as noted above. This is not the case if adsorbed
species are present at the interface, or if the surface is solid.
The surfaces of solids also possess a tension and confer an excess free energy on the
solid. Nevertheless, it is both experimentally and conceptually difficult to treat the
thermodynamics of solid surfaces in a similar way to that of liquid surfaces. The basic
problem is that a freshly cleaved, or indeed very aged, solid surface is not in
equilibrium, since the atoms are not, in general, free to move to positions of lower free
energy. In a liquid, on the other hand, dynamic equilibrium is established very quickly.
Thus, for example, a drop of spilt mercury will round up in a fraction of a second,
whereas the pyramids have retained their shapes (except for a minimal dynamic
interaction with tourists) over thousands of years.
pk = m k T (2.2.11)
where PN is the number density of the molecules and k and T have their usual
significance. This is the familiar pressure term in the kinetic theory of a perfect gas.
It is the same for a liquid and is positive.
50 I 2: THERMODYNAMICS OF S U R F A C E S
2. The cohesive contribution due to the time average of the net attractive and
repulsive forces between molecules on opposite sides of the imaginary test surface
in the body of the fluid. This second contribution, called the static pressure, p', is
normally negative (i.e. the attractive force dominates) and it is particularly
important for dense gases or liquids. The total pressure in a liquid is thus less than
the kinetic pressure, and must be equal to the applied pressure (i.e. the vapour
pressure for a one-component system) despite the fact that the kinetic pressure is
so much higher in the liquid (because p~ is higher).
T o simplify the argument we will neglect the repulsive forces when considering the
static pressure; the repulsive forces have a much shorter range than the attractive
forces and are important only at extremely high external pressures. The intermolecular
forces can be considered to have a sphere of influence beyond which they are negligible
(an idea introduced by Laplace in 1806). For molecules further from the surface than
the diameter of this sphere, the pressure must be isotropic because of the symmetry of
their surroundings. Near the surface between the two bulk phases, the tangential and
normal contributions to the static pressure are not the same because of the asymmetric
distribution of molecules within the sphere of influence.
When the sphere of influence cuts the surface (Fig. 2.2.2) there are fewer and fewer
pairs of molecules attracting across the test plane as the centre of the sphere
approaches the surface and hence the magnitude of the static pressure contribution
decreases for both the pressure across a plane parallel to the surface (Fig. 2.2.2(a)) and
for the pressure across a plane normal to the surface (Fig. 2.2.2(b)). It is apparent that
when the test plane is in the surface (Fig. 2.2.3), the decrease in magnitude of the
normal static pressure is greater than the decrease in magnitude of the tangential static
pressure because of the smaller number of interactions remaining in the former case.
Now consider the consequences of the fact that the forces on opposite sides of an
infinitesimally small cube must be equal and opposite for mechanical equilibrium. The
net pressure normal to the surface must be constant right through the surface. The
contribution from the kinetic and static pressures normal to the surface are shown
schematically in Fig. 2.2.4(a). The normal pressure is a constant equal to PO, the
pressure in the bulk phases. (We are neglecting here the hydrostatic pressure at
different depths in the liquid.)
Fig. 2.2.2 Sphere of influence around an imaginary test plane near a liquid surface. (a) With test
plane parallel to the surface; (b) with test plane normal to surface. In both cases the attraction
between molecules on opposite sides of the test plane must be less than in the bulk of the liquid
because of the deficiency of molecules in the region of vapour in the sphere of influence.
SURFACE ENERGY AND ITS CONSEQUENCESI 5 1
Fig. 2.2.3 Sphere of influence around imaginary planes in the liquid surface. (a) With test plane
parallel to the surface; (b) with test plane normal to the surface. Note the attraction between
+ +
molecules in quadrants (I 11) and (I11 IV) in (a) will be very small because there are very few
+
molecules in quadrants I11 and IV. Attraction between molecules in quadrants (I IV) and (I1 111) +
in (b) will be much greater as quadrants I and I1 are densely populated.
4 Pressure
k
P"
Liquid
Vapour
- - - - - - - - Pn'Po
(a> -
Surface region
A Pressure
P:
Liquid
Vapour
\ I
*Z
PS
(b) , ,
Surface
region
Fig. 2.2.4 (a) Variation of the kinetic (a:), the static (a:) and the net (p,) pressures normal to the
vapour-liquid interface. (b) Variation of kinetic (a:), static (as), and net (at), pressures tangential to
the vapour-liquid interface as a function of position.
52 I 2: THERMODYNAMICS OF S U R F A C E S
A similar graph for the pressure tangential to the surface cannot give a constant net
pressure because the kinetic contribution is identical to that for the normal pressure
k k
pt = p , = PNkT
while the static contribution is different (Ps # p i ) . The resultant pressure, p,, is less
than $0 on passing through the surface. The resulting net stress is the surface tension
and is equal to the integral of the deficit of the tangential pressure through the surface
layer:
(2.2.12)
-co
It is not, of course, an unbalanced force, but the balancing force must be applied
externally, for example by elastic deformation of the walls of the container. There may
appear to be no balancing force available for isolated liquid droplets, but we will find
that there is an increased hydrostatic pressure within the droplet, resulting from
surface tension, and this provides the necessary restoring force.
where y is again the surface tension, and rl and r2 are the radii of curvature of any two
normal sections of the surface perpendicular to one another.
This equation is now known as the Young-Laplace equation (sometimes, unfairly, the
Laplace equation (Pujado et al. 1972)) and is easily derived as follows (Defay et al. 1966).
Consider a small part ACBD of the surface of a static liquid drop (Fig. 2.2.5(a)). The
drop need not be spherical, and the circumference of the selected part of the surface is
simply defined as a line in the surface at a constant distance, d, from a chosen point X
on the surface. Through X draw any pair of orthogonal lines AB and CD in the surface.
' $1
\ \ '\
\
\
Fig. 2.2.5 (a) Equilibrium of a liquid cap. (b) A curved surface with no pressure drop across it
because 1/& = - (1/Rz). (After Adamson and Gast 1997, p. 9.) (See Exercise 2.10.6.)
For a sufficiently small value of d, AB, and CD can be considered as parts of circles
with radii ~1 and r2 respectively.
It is known from the differential geometry of surfaces (Weatherburn 1930) that the
directions of AB and CD can always be chosen so that ~1 is a maximum (R1) and r2 a
minimum (Rz).Furthermore,
-
1
+ 1 = 1 + 1 =J
- - - (a quantity independent of the orientation of the axes)
r1 r2 R1 R2
(2.2.14)
54 I 2: T H E R M O D Y N A M I C S OF SURFACES
(3is often referred to as the mean curvature of the surface.J-’is the harmonic mean
of R1 and R2.) Now consider the forces exerted by the cap of liquid surface ACBD.
The surface has a tension y , so that, for example, an element 61 of the boundary at A
exerts a resolved downward force Fvert parallel to XY and given by
2d 2d
ySl -+-
(Rl R2) -2dySl
-
-+- $,) .
(di
Since this expression is independent of the choice of AB and CD, it can be integrated
around the circumference (one quarter of a revolution, since there are four segments)
to give the total downward force due to surface tension of
But the drop is not in motion, so there must be an upward force to balance this
downward force. The upward force is provided by a pressure difference p” -p’
between the inside and outside of the drop. Equating the two forces:
(2.2.16)
or Ap(= p” - p’) = y
(dl + dJ
- - (2.2.17)
between two open cylindrical pipes. Similar considerations apply to a liquid drop
whose shape is unaffected by gravity, a situation that can be achieved experimentally
by floating the drop in an immiscible liquid of equal density.
The Young-Laplace equation (eqn (2.2.17)) has been introduced at this stage to
complete the argument of Section 2.2.2 and to provide us with the important relation
(see Exercise 2.2.2):
(d, d,)
dA= - + - d V (2.2.18)
r
Exercises
2.2.1 Calculate the number of molecules in 1 cm3 of water from the molar volume
(18 an3).If it is in the shape of a 1 cm cube show that the ratio of surface energy
to bulk energy is approximately lop7. How big a collection of molecules would be
needed so that the surface energy was the same as the bulk energy?
2.2.2 Consider the increase dA in area of a small surface element and the concomitant
volume, dvtraversed by the surface (Fig. 2.2.6). Show that dA = (1/R1 1/R2)dV +
where R1 and R2 are as defined in Section 2.2.3 above.
Volume dV
/
Fig. 2.2.6 Relation between increase in volume and surface area for a surface of arbitrary curvature.
Both radii increase by d R at constant angle 681, 60,.
2.2.3 Use the result derived in Exercise 2.2.2 to establish the Young-Laplace equation
using an energy minimization argument.
2.2.4. Calculate the excess pressure inside drops ofwater of radius lop5cm and lop6cm,
respectively. (Take y = 70 mN m-'.)
56 I 2: THERMODYNAMICS OF S U R F A C E S
dw = -pdV (2.3.1)
(i.e. the work done by the system is a negative contribution to the internal energy.)
If, however, the system contains two phases, we know from the Young-Laplace
equation (2.2.17) that, unless the interface between the phases has zero curvature, the
pressures p' and p" in the two phases will be different. Hence the total work done by
the system will be
Consider, for example the work done during evaporation of a liquid droplet having
some arbitrary shape of constant mean curvature (1/R1 +
1/&) (Fig. 2.3.1). If the
total volume of the system is V and the external pressure is p' then V = V' V" and+
+
d V = d V' d V". The work done on the system by its surrounding is:
P'
.dV
Vapour
P'
Liquid@'' V'y
V'
Fig. 2.3.1 Evaporation of a droplet whose surface has an arbitrary (but constant mean) curvature.
THERMODYNAMICS OF SURFACES 157
(2.3.3)
using eqn (2.2.18). One can thus identify a mechanical work term for each of the phases
and a ‘surface work’ term for the interface, without having to introduce a hypothetical
surface piston or the stretching effect depicted in Fig. 2.2.1.
It is useful at this stage to consider the droplet in Fig. 2.3.1 as a sphere of radius r.
Equation (2.3.4) then becomes:
There is, however, a gradient in density at the surface of the droplet, and for very small
droplets the distance over which the gradient extends may actually be a significant
fraction of the total radius of the droplet. It could reasonably be argued that the droplet
‘surface’ is anywhere within the region of changing density, and each arbitrary choice
would give a different value of r (and of d V’ and d V”). Fortunately, however, we know
that
and
It seems, then, that eqns (2.3.2) and (2.3.5) are equivalent, no matter where we choose
to place the surface. Equation (2.3.5), however, incorporates the Young-Laplace
equation, which defines the value of r to be used, sincep“-p’ would be different for any
other value of r. The imaginary, infinitesimally thin surface defined by this value of r is
called the surface of tension. What we have done is to replace the real droplet, with its
rather fuzzy radius, with an imaginary droplet having a sharp boundary and whose
mechanical properties are equivalent to those of the real droplet.
from the bulk phases must be u' Vf + u" V". The energy to be ascribed to the surface,
u" must then be given by:
(2.3.7)
Other surface quantities such as the surface entropy, S" surface Helmholtz free
energy, P and surface Gibbs free energy, G" may be similarly defined.
It must be clearly understood that these surface quantities can only be deJined in terms of a
particular model system and so their values will always depend on the model chosen.
The model described above (and indicated by superscript 0)is called the Gibbs
convention. Note that by dividing the total volume V into the precise volumes V'and
V" we have, in effect, constructed an imaginary system having the same
thermodynamic properties as the real system but with the two phases separated by
an infinitesimally thin dividing surface and having constant densities up to that surface.
This ideal system replaces the real system, with its finite dividing region where the
density is rapidly, but not discontinuously, changing. As with the droplet radius
(Section 2.3.1), the only reason for replacing the real system with a model is because it
is easier to think of such extensive quantities as energy and volume in discrete lumps
rather than as continuously varying quantities. It also means that intensive quantities
like pressure and density are given definite values in each phase even when the
interface is curved. The procedure was invented by Gibbs, and the imaginary dividing
surface is called the Gibbs dividing surface (Gibbs 1874-8). Other procedures are
possible (e.g. Guggenheim 1976) for planar surfaces, but generally become impossibly
cumbersome if the surface is curved. Melrose (1968) presents a very full and clear
discussion of the Gibbs treatment of curved surfaces and its integration with
hydrostatic treatments of the stress in the surface.
The presence of the interface also affects the molecular composition. In a two-
phase multicomponent system we can write L and c" as the concentrations of
component i in each phase and if the volumes of the phases are again Vf and V"(where
+
V' V" = V) then the numbers of moles of component i in each phase are:
Once again, the extra amount of i that can be accommodated in the system because of
the presence of the interface is evidently:
(2.3.9)
where ni is the total number of moles in the whole system. Note that all of the
quantities on the right-hand side of eqn (2.3.9) are unambiguously defined. A surface
concentration can be defined as:
n y / A = ri. (2.3.10)
The notation r; for the surface (excess) concentration of component i is used almost
universally and we will use it henceforth. Note that the value of ri may change
dramatically not only in magnitude but even in sign, as the chosen dividing surface is
THERMODYNAMICS OF SURFACES 159
Phase 11
------
Interfacial
region
---- - - -
Phase I
0 c: c: Ci
Fig. 2.3.2 The concentrations of a component i across an interface in a mixture (a) can be plotted
and will generally show rapid changes through the interfacial region (b). The Gibbs model (hatched
regions in (b)) ascribes constant compositions to both phases up to an arbitrarily defined interface.
The excess material (dotted region in (b) is ascribed to the infinitely thin Gibbs surface GG', (c). A
different arbitrary choice of Gibbs surface (d) can reduce the surface concentration to zero (note
equality of dotted and hatched areas in (d)) or even make it negative. In practice, it is useful to choose
GG' so that ny for the solvent is zero. The concentrations of all other components must then be
referred to this same surface. Note that for curved surfaces this choice of dividing surface is very
unlikely to coincide with the surface of tension (Melrose 1968), and so the model system would not
be mechanically equivalent to the real one.
varied even by a fraction of a nanometre (Fig. 2.3.2). For a curved surface the area, A,
also depends on the choice of dividing surface and this has a further effect on ri.
(2.3.12)
(2.3.13)
60 I 2: THERMODYNAMICS OF S U R F A C E S
The first term on the right is the heat absorbed by the system, the next three terms are
the mechanical work done on the system, while the last term is the chemical work done
on the system.
Equation (2.3.15) leads to some interesting results. First, consider a situation in which
the infinitesimal processes occur in a constant total volume, V, at fixed temperature, T.
Then the change, dF, in the Helmholtz free energy of the system is given by:
But the system does no mechanical work at constant volume, so the sum of the first
three terms on the right must be zero. Also, since dV’ = -dV” we have:
from which the Young-Laplace equation follows immediately. Note that it is here
derived on purely thermodynamic grounds.
Returning to eqn (2.3.16) we can introduce the condition
(2.3.20)
We thus reach the very reasonable conclusion that, at equilibrium, the chemical
potential of any component is the same in each bulk phase and at the surface.
THERMODYNAMICS OF SURFACES I61
We can now derive a series of formulae, of varying usefulness, relating to the other
thermodynamic parameters. From eqn (2.3.14):
(2.3.22)
This one is not very useful since there is no practical way to hold 3' and n: constant
whilst varying u" . The corresponding result from eqn (2.3.15):
is rather more useful but a better definition flows from the Helmholtz free energy.
Recall that for a system consisting of a single phase (eqn (A5.21)):
u = TS - PV + C pin;. (2.3.24)
The integration procedure? that leads to this equation (see Appendix AS) can be
applied to each of the phases (') and (") and to the surface to yield:
(2.3.29)
(2.3.3 1)
+The integration procedure corresponds to increasing the size of the system whilst
keeping all the intensive properties of the system constant (in particular keeping the
composition of the system constant). It is only in this case that the integrated form of the
equation makes sense.
62 I 2: THERMODYNAMICS OF S U R F A C E S
which is probably the most useful definition since the subscripted variables can readily
be held constant experimentally. Note also that (from eqn 2.3.27):
where ri is defined in eqn (2.3.10) and f " is the Helmholtz free energy per unit area of
the surface. This equation shows that f " is not equal to the surface tension, y, except
for a specific choice of the Gibbs dividing surface, namely that where C piri = 0; this
is a very easy and natural choice for a one-component system, but highly unusual for
multicomponent systems.
This is a convenient point to clear up a problem that often arises in the literature. It
is frequently stated that surface tension and 'specific surface free energy' (or some such
term) are equal only for a one-component system. Such statements can be properly
interpreted only when the terms are carefully defined (Morrow 1970). The surface
tension is defined by eqn (2.3.31) and if y is independent of area (which will be so if
there is an infinite reservoir of any adsorbing species) then this equation can be
integrated to give:
W ) T , V / , V y n ; ]= YAA (2.3.33)
or Y = (AF/AA)T,,,,V.,[ni]. (2.3.34)
Thus, the change in Helmholtz free energy of the system (model independent) per unit
change in surface area is numerically equal to the surface tension, y, and this is true for
a system with any number of components provided y is independent of area.
On the other hand, for the model system
[2.3.32]
and the specific Helmholtz free energy ascribed to the surface (and model dependent) is
only equal to y for a particular choice of dividing surface, such that C piri = 0. This
choice of dividing surface is a natural one only for a one component system.
Exercises
2.3.1 Establish eqns (2.3.27), (2.3.28), and (2.3.30).
2.3.2 The surface excess Gibbs free energy may be defined either as
G=U"-TS"+~V"
or G = U" - Ts" +pV" - yA.
G" = C/.&
i
which is analogous to the bulk phase equation (eqn (A5.25)). [Note that V" = 0
in the Gibbs model.]
G I B B S ADSORPTION E Q U A T I O N I63
(2.4.1)
Differentiating eqn (2.4.1) and comparing it with eqn (2.3.14) we have the Gibbs
adsorption equation:
(2.4.3)
This equation is one of the most widely used expressions in surface and colloid science
and we will explore its meaning at some length. It can be applied to systems where the
surface tension can be measured (e.g. those containing liquid-liquid or liquid-vapour
interfaces) in order to calculate the surface concentration of the adsorbed species
causing the surface tension change. Equally, if the surface concentration can be
measured directly but y cannot (as occurs in many solid-gas and solid-liquid systems),
the Gibbs adsorption equation can be used to calculate the lowering of y (i.e. the
spreading pressure, ll)from the measured adsorption.
Unfortunately, the absolute value of ri is extremely dependent on the choice of
dividing surface (see Fig. 2.3.2). We have already noted though (Fig. 2.3.2) that the
Gibbs dividing surface is normally chosen so that n: and, hence, I': for the solvent are
equal to zero so that all other components are measured with reference to that surface,
giving the relative surface concentrations. We will see shortly how to define the relative
surface concentration operationally so that its value is independent of the position of
the dividin plane. The resulting (experimentally useful) quantity, ri,1, (sometimes
cB
written ri ) is the surface excess of i relative to the solvent (1) and its value is
numerically equal to I'y with the convention I'y = 0. The superscript (T in this case
refers only to the use of a Gibbs dividing surface. We will also use the convention that
r? = 0 for the solvent, so our ri,1will always be equal to rjl).
2.4.1 The relative adsorption
The Gibbs-Duhem equation (eqn (A5.23)), when applied to the two bulk phases at
constant temperature, gives (remember pi is the same in both phases):
N N
V'dp' = nidpi and Vf'dpf' = nbdpi (2.4.4)
1 1
+Recallthat this integration process is done whilst keeping the intensive variables (and
in particular the composition) constant.
64 I 2: THERMODYNAMICS OF S U R F A C E S
If the interface is planar or if its curvature does not change so that (p" - p') is constant,
then dp" = dp' Equations (2.4.5) then give?, for the variation in p1:
(2.4.6)
We can now use this expression to eliminate the unknown p1 from the Gibbs
adsorption equation to obtain (Exercise 2.4.3):
(2.4.7)
The quantity inside the square brackets is defined for each component i as the relative
adsorption of that component and is written:
where Ac, = - 4 4.
Although it is not essential, it is customary in the case of a liquid-vapour interface to
take the solvent as component 1 and to compare the adsorption behaviour of all other
components to it, since this choice leads to the definition of quantities that are directly
accessible from experiments.
The definition of ri,1brings with it several advantages:
1. The value of r,,lis independent of the arbitrary choice of dividing surface. This is
obvious from our derivation of eqn (2.4.7) since it depends only on general
thermodynamic considerations. In a two-component system the quantity
where V’ and Y” are the volumes of phases (’) and (”) respectively and
+
Y = Y’ V” is the total volume. ny and ni are, as before, the number of moles
ascribed to the surface and total number of moles in the system respectively. For
component 1, eqn (2.4.9) gives:
ny = nl - Vci + V’’<c‘,- c‘,) (2.4.10)
and eliminating V’between the two eqns (2.4.9) and (2.4.10) (Exercise 2.4.3):
None of the quantities on the right-hand side of eqn (2.4.11) depends on the choice
of dividing surface, and if eqn (2.4.11) is divided by the surface area, A, we recover
the right-hand side of eqn (2.4.8), which defines rj,1.For a curved surface it may
be shown that A is the area of the surface of tension (Buff 1956).
2. r;,l has an intuitively appealing physical meaning. Since ri,1 is independent of the
choice of dividing surface, we are at liberty to pick any dividing surface that suits
us. In particular, we can choose the surface where rl = 0 (without needing to be
able to specify its position experimentally to a fraction of a nanometre). For that
surface, ri,l = r,.In other words, the relative adsorption of component i with respect
to component 1 isjust the surface excess concentration of component i at the surface where
the adsorption of component 1 is zero.
As a simple example, consider adsorption at the gas-liquid interface (e.g. in a
foam or aerosol). If phase (”) is the gas phase, then cy= c;= 0, and so
i.e. one can identify the relative adsorption or the surface excess simply by
determining the number of moles of solute that appear to have been removed from
the bulk solution and dividing by the surface area of the solid. Such direct
measurements of adsorption can be done using radio-isotopes or by measurement
of concentration changes in the bulk solution when the solid adsorbent is added.
3. r;,l can be introduced into the Gibbs adsorption equation, and hence related to
experimentally measurable quantities. We restrict ourselves to the simplest case, that
of a two- component system, for which, from eqn (2.4.3):
This equation describes how the surface tension of a solution of, say, component 2
in solvent 1 is changed as the activity of substance 2 (and hence p2) is altered, at
constant temperature.
66 I 2: THERMODYNAMICS OF S U R F A C E S
(2.4.17)
If the activity coefficient of component 2 is y2, so that a2 =~ 2 x 2we may write eqn
(2.4.17) as (Exercise 2.4.2):
(2.4.18)
(2.4.19)
and if the solution is fairly dilute so that x2 a 62 (the molar concentration) then
(2.4.20)
The Gibbs adsorption equation has been tested experimentally by McBain and
Humphreys (1932), who used a flying microtome to scoop thin (0.1 mm), hopefully
uniform, surface layers from a series of aqueous surfactant solutions. The relative
G I B B S ADSORPTION E Q U A T I O N I67
Fig. 2.4.1 Surface tension of aqueous solutions of n-aliphatic alcohols. A-E: G-Clo primary
alkanol. (After Defay et al. 1966, with permission).
adsorptions r 2 , 1 of the surfactant (relative to the water) were calculated from the mole
fractions of surfactant and water in the scooped-up and bulk phases, together with the
volumes of scooped-up liquid and the surface areas. Comparison of these experimental
values of r 2 , l with those calculated from eqn (2.4.20)showed good agreement (Defay
et ul. 1966). The remaining discrepancies are no doubt attributable to the experimental
difficulties rather than the thermodynamics.
Y I1
Pure
'"W solvent
-
1
I11
1
I
C .m.c.
Concentration of solute
Fig. 2.4.2 The three types of curves commonly observed when the surface tension of a solution is
plotted against concentration of solute. (After Davies and Rideall963.) The minimum in curve I11 is
due to impurities and disappears after rigorous purification.
later. At quite modest concentrations (often -lop3 - lO-'M) the decrease in surface
tension ceases with the formation of micelles. The critical micellization concentration
(c.m.c, Section 1.4.3) is indicated in the figure. The minimum shown in the figure is
caused by the presence of impurities? which are usually present in commercial
preparations. It disappears on rigorous purification.
NaA % Na+ + A-
where A- = R.S030-.
Assuming the salt is fully ionized, then
+Sodium dodecyl sulphate, for example, is commonly contaminated with the dodecyl
alcohol which is formed by hydrolysis.
G I B B S ADSORPTION E Q U A T I O N I69
Using the Gibbs convention ( r H z O = 0) and assuming that the interface as a whole is
electrically neutral we have r N a + , l = r A - , 1 and so:
or dy = - 2 r ~ - , l R T d l n a * (2.4.24)
a: = ayay- (2.4.25)
+
with v = v+ v-, the number of ions produced when the molecule dissociates into v+
cations and v- anions.
For a dilute solution in which the activity coefficienty (compare eqn 2.4.18) is close
to unity we have:
(2.4.26)
Note that the reduction in y for a given (molar) adsorption is doubled in this case: it is a
colligative property like osmotic pressure and freezing point depression and depends
on the number of solute particles in the interface.
If a large excess of a simple sodium salt (say NaCl) is added in addition to the
surfactant then the system can be held at constant ionic strength. The chloride ion can
be assumed (Fig. 2.4.2, curve 11) to have negligible surface activity (ra-M 0) and the
chemical potential of both the chloride and sodium ions can be held essentially
constant as the concentration of NaA is altered. In that case the counterpart to eqn
(2.4.23) is:
so that
(2.4.27)
Note that the factor 2 has disappeared in this case. This difference in behaviour of a
surfactant in the absence and in the presence of a swamping concentration of
electrolyte is quite significant.It can be investigated with the use of radiotracers such as
tritium (3H)that undergo p-decay. The p-radiation is rapidly absorbed in water so that
only radiation originating in the surface layers can escape. Using appropriately labelled
surfactants one can directly measure the amount of surfactant in the surface layers (see
Hiemenz 1977, p. 282).
f=Z+C-R-P-$+S (2.4.28)
f =2+c-P. (2.4.29)
The number of variables which must be specified in order to specify the state of the
phase (in the absence of external fields) is the usual 2 (temperature and pressure) plus
+
one each for the chemical potentials of each of the constituents (i.e. 2 C). There is,
however a Gibbs-Duhem equation (see Appendix A5.3) relating all the chemical
potentials of the constituents of each phase so this gives P restraints; the result is
equation (2.4.29).
We have, for two phases in contact:
f = C
Even this very simple form of the phase rule reveals a curious paradox. In a dispersion
of solid particles in a liquid medium we often refer to the surface area of the interface
between the two so we are inherently regarding the system as consisting of two phases.
Yet a similar dispersion of a protein in water would naturally be regarded as a single
phase system. How can it be possible to regard one dispersion as a two-phase and the
second as a one-phase system, and how does this affect the application of the Gibbs
phase rule? Kruyt (1952, p. 11) discusses this question and shows that one can choose
to treat a colloidal system as either one- or two-phase, whichever is most convenient.
The usual lyophobic colloid is best treated as a two-phase system since the disperse
phase has a negligible effect on the properties of the dispersion medium. If one chooses
to regard it as a single phase system one would have to acknowledge that the chemical
potential of the sol is not effectively a variable so that both C and P in equation (2.4.29)
are reduced by one and the value off remains unchanged.
The difference in behaviour of the two types of system can be illustrated by
considering the vapour pressure of the dispersion medium as a function of
temperature. For the lyophobic system, the vapour pressure is essentially unaffected
by the concentration of the colloid whereas the lyophilic colloid will lower the solvent
vapour pressure at any temperature, to an extent depending on its concentration. A
more complete resolution of this problem can be given using the theory of the
thermodynamics of small systems (Hill 1967). Hall and Pethica (1967) show how that
theory is applied to micellar systems and demonstrate the superiority of the one-phase
description in that case.
The simple case of a two-component solution in contact with its vapour is treated
extensively in most textbooks. The two independent variables in that case are usually
chosen to be the temperature and the concentration of solute. For an n-component
mixture, the temperature and the concentration of (n- 1) components are independent
G I B B S ADSORPTION E Q U A T I O N I71
and one can eliminate the solvent activity (PI) using the Gibbs convention. The
pressure in this case is a dependent variable because it is determined by the vapour
pressures of the n components at the temperature in question.
For two liquids in contact the more appropriate choice of independent variables is
T,p , and (n- 2) of the component concentrations because in this case the pressure and
temperature can be arbitrarily altered. What the phase rule tells us is that we can then
set the surface concentration of two components to zero. The most appropriate choice
is the two liquid solvents that form the bulk of the contacting phases. One then has two
Gibbs dividing surfaces and the surface concentration of the other (n - 2) components
is measured relative to both components 1 and 2. It can be shown that in that case:
r
Exercises
2.4.1 Imagine a sharply defined surface layer of adsorbed material at the liquid-vapour
interface. By placing the Gibbs dividing surface at the boundary between the
adsorbed layer and the liquid, use eqn (2.4.8) to show that the relative adsorption
ri.1is zero when
i.e. when the components i and 1 are present in the same molar ratio in the
surface layer as they are in the liquid. Show also that ri,l is positive if the surface
layer is relatively richer in i, negative if the surface layer is relatively poorer in i.
2.4.2 Establish eqn (2.4.18).
2.4.3 Establish eqn (2.4.7) and (2.4.11).
2.4.4 If two phases, (’) and (”) are in contact at constant T and p, the Gibbs-Duhem
equation requires that 1nidpi = 0 in each phase. Consider the situation where
solute 3 is distributed between the bulk solvents 1 and 2. Show that the surface
excess of solute 3 relative to 1 and 2 is given by:
(Hint: refer back to eqn (2.4.3) and use the Gibbs-Duhem equations to eliminate
dp1 and dpz. Note the similarity to eqn (2.4.30).)
72 I 2: THERMODYNAMICS OF S U R F A C E S
2.4.5 Show that if the two solvents 1 and 2 in Exercise 2.4.4 are mutually insoluble
then
2.4.6 Consider the system indicated below where the data for the surface region is
obtained from a sample which contains some bulk phase:
Calculate the surface excess of solute in this case, relative to both oil and water,
using the relation in Exercise 2.4.4. (Assume the area of the interface from which
the sample is taken is 250 cm’.)
2.4.7 What is the surface version of the Gibbs-Duhem relationship (cf. eqn. A5.23)?
2.4.8 Repeat the calculation in Exercise 2.4.6 using the Guggenheim (1976) concept of
a surface phase which has a finite thickness and is defined by boundaries placed
inside both the bulk phases. T o solve the problem, ensure first that you place the
boundaries in each phase so that the surface excess of both water and oil is zero.
Hence calculate the surface excess of the solute and verify that the result is the
same as in Exercise 2.4.6.
Fig. 2.5.1 Capillary rise of a wetting liquid in a cylindrical capillary tube contained in a closed
isothermal chamber. The space above the liquid column contains only the vapour.
An idea of the magnitude of the vapour pressure changes involved can be gained
from Kelvin's original, elegant physical dcrivation in which he considered a capillary
rise experiment in a closed, isothermal vessel (Fig. 2.5.1; Thomson 1870, 1871).
For thermodynamic equilibrium to hold, the liquid and the vapour must be in
equilibrium both at the flat liquid interface and the curved upper meniscus. Since the
hydrostatic pressure (in the vapour phase) is lower at the upper meniscus by p,gh than
it is at the plane liquid surface, it follows that the equilibrium vapour pressure at the
meniscus is lower than that at the plane liquid surface by this amount (p, is the density
of the vapour.)
The Young-Laplace equation (eqn (2.2.17)) gives, at the meniscus:
(2.5.3)
where p' is the equilibrium vapour pressure at the meniscus and po is the equilibrium
vapour pressure above the flat liquid surface. Equation (2.5.3) is the original
(approximate) form of the Kelvin equation.
A more rigorous analysis of the effect of surface curvature on p is given by Defay
et al. (1966, Chapter 15), from which a few results are summarized below. For a single
pure substance the fundamental equations are, for a sphere of radius r:
(2.5.5)
74 I 2: T H E R M O D Y N A M I C S OF SURFACES
where, by convention, (") and (') refer to the phase on the concave and convex sides of
the interface, respectively. Note that eqn (2.5.4) can be generalized to any curved
surface by replacing 2/r by (1/R1 +
1/Rz) where R1 and Rz are the principal radii of
curvature and may have the same or different signs. The superscript convention used
here allows the equations to be applied to both menisci and droplets. Note that r is
positive for droplets which, therefore, have an increased vapour pressure; it is negative
for a concave meniscus (Fig. 2.5.1).
For an infinitesimal process applied to a system initially at equilibrium, we have
and
dp" = d p f = dp . (2.5.7)
(2.5.8)
when applied to a pure substance (where S = S/n; and 7= V/ni, are the molar
entropy and volume respectively). Equations (2.5.6-8) are the fundamental relations
from which a large number of important results can be deduced.
For example, at constant temperature we have (from eqns (2.5.7) and (2.5.8)):
(2.5.10)
Equation (2.5.10) may be regarded as the most general form of the Kelvin equation as it
applies to spherical interfaces.
(2.5.1 1)
+ (PI
1
-p o ) (droplets) (2.5.12)
Equation (2.5.12) is the exact form of Kelvin’s equation. It shows that the equilibrium
vapour pressure of the drop is higher than that of the flat liquid surface. It follows,
therefore, that if a number of droplets of uniform radius are initially in equilibrium
with a surrounding vapour (of infinite volume) that equilibrium must be unstable
because if condensation occurs on a drop its radius will increase, its equilibrium vapour
pressure will decrease, and it will continue to grow. Conversely, if a little evaporation
occurs from a droplet, its radius decreases, its equilibrium vapour pressure increases,
and it continues to evaporate. If the droplets are initially of different size, then the large
ones will grow at the expense of the smaller ones.
This argument applies if there is a sufficient reservoir of vapour, so that droplets do
not change the partial vapour pressure as they evaporate or grow. The important point
to note is that, for a given vapour pressure there is a critical size, r,, equal to the size in
equilibrium with the vapour pressure, above which the droplets will grow but below
which they will evaporate. When a vapour is cooled so that it becomes supersaturated,
condensation to the liquid cannot occur until there are formed some nuclei of liquid of
a size that can continue to grow. We will return to the study of this nucleation the0y in
Section 2.8.
An expression analogous to eqn (2.5.12) can be written for the solubility of a solid
crystal in a surrounding liquid:
(2.5.13)
where C(r), C , are the solubilities of the particles and the bulk solid respectively and
y is the interfacial surface energy per unit area. The relation between the ‘effective’
radius, r, and the dimensions of the solid crystal will be examined shortly (Section 2.6).
Again, eqn (2.5.13) suggests that large crystals will grow at the expense of smaller ones
- a phenomenon known as Ostwald ripening (see Kahlweit 1975). It should be noted,
however, that this can only occur at an appreciable rate if the solubility of the substance
in the surrounding liquid is high enough. In many cases this is not so and a suspension
of particles of many different sizes can coexist in quasi-equilibrium for very long times
indeed.
(b) Bubbles in a liquid
In this case, the phase (”) is a gas and so we have (Exercise 2.5.2):
p” 2yVf
or ln-w
Po
-~
rR T
(ifV‘ << V”).
76 I 2: THERMODYNAMICS OF S U R F A C E S
Table 2.2
Influence of radius of curvature on the equilibrium vapour pressure above a spherical water surface,
calculated from eqns (2.5.12) and (2.5.14), assuming that y is independent of r
Hence, the vapour pressure inside a bubble is smaller than the value for a flat surface.
This explains the phenomenon of superheating of liquids above the normal boiling
point. Remember that there has to be an excess pressure inside a bubble and yet the
vapour pressure is actually lower than that over a flat surface at the same temperature.
Boiling can only occur if either (i) the liquid can vaporize into a pre-existing bubble of
reasonable size or (ii) the temperature is raised sufficiently high so that the equilibrium
vapour pressure even in a small bubble is large enough to allow it to continue to grow.
Case (i) is more usual, where the pre-existing bubble is, say, an air bubble previously
released from the liquid. Such bubbles form on the walls of the container as the liquid
is heated and dissolved gases become less soluble. Alternatively, they can be introduced
deliberately by adding porous solids ('boiling chips'); gas contained in the pores
expands as the temperature rises and produces seed bubbles into which vaporization
can occur.
Case (ii) is called homogeneous nucleation (Section 2.8) and it can only occur at
temperatures well above the normal boiling point. It turns out that a temperature of
almost 200 "C is required to induce boiling of water at atmospheric pressure if rigorous
efforts are made to exclude any extraneous nucleation bubbles and one relies entirely
on the formation of bubbles of the pure vapour. It should be noted that the changes in
vapour pressure due to surface curvature and predicted by the Kelvin equation are
negligible except for very highly curved surfaces (Table 2.2).
It should also be noted that the integrations in eqns (2.5.12-14) are carried out with
respect to the variable y / r . They remain valid, therefore, even if y changes with radius
and pressure, provided that the value substituted for is the appropriate one for the
value of r, and p' (or p") concerned.
+Not to be confused with William Thomson who became Lord Kelvin and made the
other important contribution in this area.
THERMODYNAMIC BEHAVIOUR OF SMALL PARTICLES 177
equation we will assume that the latent heat of vaporization AH, for the liquid is
independent of curvature. Although there is some slight dependence at very high
curvature it may generally be neglected. Defay et al. (1966) give a full discussion of this
question (pp. 23CL9) and point out that even at r = 1 nm the decrease in AHvapfor
water is only about 6 per cent.
With this approximation we can easily determine the effect of change in radius of
curvature (r) of a drop or bubble on the temperature required to establish equilibrium
across the interface, if the external pressure is kept constant.
(a) Liquid drop suspended in its vapour
From eqn (2.5.8), with (”) referring to the liquid, we have dp’ = 0 (since the external
pressure is constant) and so:
(2.5.17)
which is the Thomson equation (Thomson 1888). We have assumed here that 7”and
AH,, are unaffected by the temperature change. This is a reasonable assumption in
the present case because the temperature range involved in practice is fairly small (see
Table 2.3).
In this case, T < TO(since AH,, is always positive). Thus as Y decreases the
equilibrium temperature becomes lower and lower. In order to induce a vapour to
condense onto a small droplet of the liquid phase it is necessary to cool it down to a
temperature below the normal condensation temperature T, corresponding to the external
pressure p‘. This is the phenomenon calling supercooling. Supercooled (i.e. super-
saturated) vapours are used in the Wilson cloud chamber for the detection of
radioactive particles. The particles ionize the air in the chamber and this aids the
formation of nuclei of sufficient size to permit further deposition (i.e. drop growth) at
the temperature of the chamber.
It should be noted that the integration process leading to eqn (2.5.17) corresponds
to a gradual (reversible) bending of the interface, with a corresponding reduction in
temperature to maintain equilibrium between vapour and liquid at each stage. As the
bending progresses the pressure inside the droplet increases and in order to maintain
the chemical potential constant we make a corresponding reduction to the
temperature. The actual condensation process that occurs when nucleation is
induced is, of course, a highly irreversible process. The vapour deposits rapidly
because as the droplet grows the equilibrium temperature rises so the driving force
towards condensation tends to increase. This is offset, however, by the release of the
latent heat of condensation, which tends to raise the temperature of the whole
system.
78 I 2: THERMODYNAMICS OF S U R F A C E S
(2.5.18)
(In the integration process, AHvapis assumed constant with respect to temperature
but this assumption can easily be dispensed with (Exercise 2.5.3).) This form of
Thornson’s equation gives the temperature at which a bubble of vapour of radius Y can
exist in equilibrium inside a liquid. For small values of Y it follows that T >TOand,
indeed, the temperature can be much higher than the normal boiling point as we noted
above. Recall that there has to be an excess pressure inside the bubble so eqn (2.5.18)
calculates how high the temperature must be raised to achieve that higher vapour
pressure.
Table 2.3 gives values for the equilibrium temperature for drops and bubbles of
various sizes. Notice that the effect on the equilibrium of bubbles is much larger than
that of drops. That is not surprising when one recalls that the effect is due to the excess
pressure inside the sphere: the effect of pressure on the chemical potential of a gas is
much greater than its effect on a liquid. Note also that the more elaborate equation
derived in Exercise 2.5.3 has been used to estimate T for bubbles because significant
changes occur in AHvapover the temperature range involved.
“1. The vapour pressure of small crystals is greater than that of large crystals. In the
presence of vapour, large crystals will grow at the expense of small crystals.
2. Small crystals will melt at a temperature lower than the normal melting point. The
m.p. of a small crystal will be given by
(2.5.19)
where TOis the normal melting point at the same external pressure, y s ~is the
interfacial energy . . .7,
is the molar volume of the solid and AHf,, is the molar
heat of fusion . . . .
3. The melting point of a substance solidified in the pores of an inert material will
depend upon the size of the pores . . . .
4.Small crystals may have a heat of fusion and a heat of sublimation smaller than the
value for bulk solid.”
Table 2.3
Influence of curvature on the equilibrium temperature of droplets and bubbles. (Adaptedfrom D e f y et al. (1966) p . 242).
Droplet of water in water vapour at 1 atm Bubble of water vapour in water at 1 atm
U
CD
Next Page
80 I 2: THERMODYNAMICS OF S U R F A C E S
Point 1 has already been alluded to above. Point 2 is very important in the field of
ceramics where finely powdered materials are heated to high temperatures and sintering
occurs (i.e. fusion and partial inter-diffusion occurs at localized centres where the
radius of curvature is very small). Point 3 has important consequences in tertiary oil
recovery and the treatment of oil sands and shales to recover high molar mass
hydrocarbon fractions (i.e. heavy crude) since the melting point of such material is
affected by the capillary pressure to which it is subjected in the pores. (See Section
2.7.1 for further discussion of capillary pressure.)
Equation (2.5.13) for the effect of particle size on solubility can be made a little more
precise by using activities rather than concentrations:
(2.5.20)
a t = a7.a"- [2.4.251
where v is the number of ions produced when the salt dissociates. Equation (2.5.20)
then becomes (Defay et al. 1966, p. 272):
(2.5.21)
Fig. 2.5.2 Effect of particle size on the Ni-Ni(CO), equilibrium. (From Defay et al. 1966, with
permission.)
Response to External Fields
and Stresses
3.1 Response to gravitational and centrifugal fields
3.1.1 Settling under gravity
3.1.2 Sedimentation equilibrium under gravity
3.1.3 Settling in a centrifugal field
3.1.4 Sedimentation equilibrium in a centrifugal field
3.2 Response of a dielectric material to an electric field
3.2.1 Static electric fields
3.2.2 Response o f a bound electron t o an alternating field
3.2.3 The dielectric response function
3.2.4 The phase lag between D and E
3.2.5 The shape o f E ( W )
3.3 Response to electromagnetic (light) waves
3.3.1 Scattering by small particles (Rayleigh scattering)
3.3.2 Rayleigh-Cans-Debye (RGD) scattering
3.3.3 Mie scattering
3.4 Response to a mechanical stress
3.4.1 The rheology of colloidal materials
3.4.2 Ideal solids and liquids
3.4.3 The general response to a shearing stress
(a) Stress relaxation after a sudden strain
(b) Stress relaxation after cessation o f steady shear fl o w
(c) Creep after a sudden stress
3.4.4 Response t o an oscillating shear field
3.4.5 Viscous shear behaviour
115
116 I 3: R E S P O N S E TO E X T E R N A LFIELDS AND S T R E S S E S
simultaneously, or serially to the suspension. The ideas are not elementary and the
observed behaviour can be quite complex but we are interested here in giving an
overview so that the vocabulary connected with the basic features of colloidal systems
can be introduced.
The most pervasive effect is that of the Earth’s gravitational field and the extension
from gravitational to centrifugal fields is a natural one for colloidal systems, especially
when the particle size is less than about 1 pm.
The response of a dielectric material to an alternating electric field introduces some
important concepts which can be applied immediately to the subject of light scattering
by colloidal particles but that will be applied in a more definitive way in Chapter 1 1 to
develop the theory of the attractive forces between colloidal particles (long range van
der Waals forces). It is recognized that many students of colloid science have only
passing acquaintance with the mathematical tools used in this section so we have
included a number of practice problems to bring out the main features of the use of
complex functions to describe elastic and inelastic responses to an applied stress.
There follows an introduction to light scattering, which is a technique widely used
for investigating particle size (Chapter 5), especially in its more recent manifestation
(called photon correlation spectroscopy).
The elementary functions of a complex variable are again used to describe the
response of a system to a mechanical stress. An introduction to that theory is also
presented at this stage to bring home the similarities between this and the dielectric
response and to provide a sound basis for the following chapter on the viscous
behaviour of colloidal systems. Some of that material will not be needed again until
Chapter 15 when we take up the study of the rheology of suspensions.
so that
(3.1.3)
B = 6nva(l - ~ / 5 ) ;( E + 0) (3.1.4)
where E , the eccentricity, is (a - b)/b. Allen (1975, p. 169) gives a few similar relations
for regular shapes, but these are of limited value for colloid settling because they
require a knowledge of the initial orientation of the particles with respect to the
gravitational field. Unless one or other of the principal axes of symmetry (assuming
they exist) is aligned with the field, the particle’s motion is very irregular and certainly
not confined to a vertical direction. Fortunately, as the particle size decreases, these
considerations become rather less important because the particles become subject to
increasingly vigorous Brownian motion, which has two important effects. For
118 I 3: R E S P O N S E TO E X T E R N A LFIELDS AND S T R E S S E S
B = kT/V [1.5.18]
When that occurs, however, the settling rate is significantly affected by the thermal
diffusive motion of the particles. The r.m.s. displacement for a sphere is (from eqns
(1.5.18), (1.5.19), and (1.5.25)):
(3.1.6)
where ~1 and 72 are the viscosities of the drop (or bubble) and the fluid respectively.
Equation (3.1.6) assumes that the interface is perfectly fluid so that it cannot support
an applied stress. For air bubbles in water ( ~ <1< ~ 2 the
) drag is then reduced by one
third. In actual practice, this seldom occurs because the drop or bubble surface almost
always has an adsorbed layer of a surfactant (whether by accident or design) and this
tends to make the interface rigid so the correction is not applicable (Levich 1962) and
eqn (3.1.3) still holds.
at equilibrium. Here. 4 is the gravitational potential (= gh) and Mi is the molar mass.
Then since:
-M;gdh
d lnci = (3.1 .S)
RT .
~
120 I 3: R E S P O N S E TO E X T E R N A L F I E L D S AND S T R E S S E S
In colloidal dispersions, gravitationalforces are much more significant than they are for
molecular solutions or gases. In fact it is possible to observe the operation of gravity
directly over very small distances and this was the second method exploited by Perrin
(1909) (c.f. Section 1.5.4) to obtain an estimate of the Avogadro constant. Perrin and
his collaborators prepared dispersions of highly monodisperse spheres of a natural
colloid called gamboge. (This in itself was a very considerable feat in those days: in his
account of Perrin’s work, Overbeek (1982) points out that Perrin and his students
would start with 1 kg of gamboge and after several months of patient fractional
centrifugation might finish up with a few hundred milligrams of monodisperse
particles.) When this suspension was mounted on a vertical microscope slide
(Svedberg 1928) the number of colloidal particles could be directly counted at
different heights and from the resulting data a value of the Boltzmann constant, k,
could be calculated (Exercise 3.1.4). Then since k = R / N A , where R is the universal
gas constant, Avogadro’s constant NA could be obtained. Perrin obtained a value of
6-7 x mol-’ in his experiments and subsequent measurements of Westgren gave
6.05 x very close to the presently accepted value of 6.022 x (See Svedberg
1928, p. 101 for some discussion of this work.)
This procedure relies on the notion that the microscopically visible particles are
being continually bombarded by the surrounding molecules and so come to have the
same average translational energy (3k T/2). Their distribution in the gravitational field,
therefore, reflects the energy of the surrounding molecules. These two sets of
experiments by Perrin provided some of the very first direct evidence for the real
existence of atoms, which were regarded by many scientists at the time (ca. 1900) as no
more than convenient figments of the theoreticians’ imagination.
Fig. 3.1.2 Forces on a particle in a centrifuge tube. The outward force is an apparent (virtual) force
invoked to explain the motion of the particle with respect to a coordinate frame attached to the rotor
and moving with it.
RESPONSE TO GRAVITATIONAL AND CENTRIFUGAL FIELDS I 121
Consider a particle immersed in a liquid at a distance x from the axis of the
centrifuge head (or rotor), Fig. 3.1.2. If it were to stay at that distance as the head
revolved, it would have to be acted on by a (centripetal) force directed towards the
centre of the rotor and forcing the particle to travel in a circle. The magnitude of that
force is equal to (m- m’)w2x, where w is the angular velocity (radians s-l) of the rotor
and (m - m’) is the apparent mass (corrected for buoyancy) of the particle. In the
absence of that force, the particle moves away from the axis of rotation as if it were
acted on by a force of that magnitude acting outwards. Again it is retarded by a
frictional force that is proportional to its velocity and again it takes only a very short
time before these two forces are balanced (Exercise 3.1 .l):
where v2 is the (partial) molar volume of the polymer and VZ is the volume per unit
mass (= pi’); NAis the Avogadro constant. The sedimentation coefficient can then be
written (using eqn (1.5.18)):
MD
S = M(l - ZZ~ZP~)/NAB
= -(l - Vzpl) (3.1.11)
RT
and, hence, a knowledge of S and the diffusion coefficient, D, can be used to estimate
the molar mass (or molecular weight) of the polymer.
S is obtained by writing eqn (3.1.9) in the form:
dx/x = m[(l - pl/pS)w2/B]dt = S w2dt
and on integration:
In(x2/xl) = s w2(t2 - tl). (3.1.12)
(3.1.13)
122 I 3: R E S P O N S E TO E X T E R N A L F I E L D S AND S T R E S S E S
For non-spherical particles of regular shape, the corresponding equivalent sphere can
be calculated using the methods set out by Jennings and Parslow (1988).
Values of S range from about 1 ps for large colloidal particles down to values of the
order s for proteins. The time unit s is called 1 Svedberg, in recognition of
the man who developed the ultracentrifuge and the above procedure for establishing
protein molar masses.
The simplest application of eqn (3.1.12) is the two-layer sedimentation technique in
which the colloidal suspension is initially placed as a thin layer on top of the clear
suspension medium. As centrifugation proceeds the components of that layer travel
down the centrifuge tube at rates determined by eqn (3.1.12) and in favourable cases
their individual progress can be followed, usually by an optical procedure. This
method is particularly suited to the separation and characterization of mixtures of
distinct materials with well-defined mass and density characteristics, such as mixtures
of proteins. It does, however, have some problems (in particular the phenomenon of
‘streaming’) and more detailed treatments should be consulted (Allen 1975, pp. 266-
73) for methods of minimizing their effects.
(3.1.14)
where the potential energy of molecules in the Earth’s gravitational field (mgh) is
replaced by the particle potential energy in the centrifugal field. (See Exercise 3.1.7 for
an informal proof of this equation.) The distance x in eqn (3.1.14) is measured from
some arbitrarily chosen reference line where the concentration is Co. Note also that the
exponent is positive in this case since the centrifugal field acts to concentrate the
particles at larger values of x.
Equation (3.1.14) may be written:
(3.1.15)
which is the form used to calculate the molar mass (Mi) of a protein or polymer from
centrifugation measurements. Note that it requires no assumptions about the friction
factor B. Equation (3.1.15) does, however, assume ideal behaviour for the sedimenting
RESPONSE TO GRAVITATIONAL AND C EN T R IF U GA L F I E L D S I 123
material (Exercise 3.1.6). A more elaborate analysis, taking account of departures from
ideality leads to:
(3.1.16)
d(x2) - 2RT
where yi is the activity coefficient. This same activity coefficient correction can also be
applied to equation (3.1.1 1) in more concentrated polymer or protein systems. The
notion of activity coefficient is not really appropriate to particulate dispersions so we
will not discuss this procedure further. It is more relevant to the behaviour of lyophilic
than lyophobic colloids.
.
Exercises
3.1.I Consider a spherical particle settling under gravity according to eqn (3.1.1).
Show that during the period before it reaches terminal velocity its equation of
motion is:
where p = (ps- pl)/ps and G = 9~/2p,?. Hence show that u = (pg/G )[1- exp
( - Gt )] during this period. How long does it take for a particle of radius 1 p m
and density 3 x lo3 kg mP3 to reach 99 per cent of its terminal velocity in water
(density = lo3 kg mP3, viscosity = lop3NmP2 s)?Repeat the calculation for a
radius of 0.1 p m and 0.01 pm. (Allen 1975, p. 158).
3.1.2 A suspension of silica particles (ps= 2.8 g cmP3)in water is allowed to settle in a
cylinder at 20 O C. Calculate the time required for a particle of 2 p m radius to
settle a distance of 20 cm, assuming it is spherical. (Take v = lop2g cm-’ s-l =
1 centipoise.) Convert the data to SI units, and repeat the calculation.
3.1.3 Show that for colloidal particles dispersed in a liquid, the equilibrium number of
particles, N , at a height, k above a reference level, Izo is given by:
where NOis the number of particles at height ko and m’ is the mass of fluid
displaced by a particle of mass m (k is the Boltzmann constant).
3.1.4 Svedberg (1928, p. 101) gives the following table of Westgren’s date for the
number of particles at different heights in a gold sol at sedimentation equilibrium
under gravity.
Assume the particles have radius 21 nm and density 19.3 g cmP3 and the
temperature is 20 "C. Estimate k from the equation derived in Exercise 3.1.3 and
then calculate NA assuming R = 8.31 J K-' mol-'. Repeat the calculation with a
radius of 22 nm and note how sensitive the answer is to this variable.
3.1.5 The time taken for a particle to reach its terminal velocity under gravity is about
5 / G (Exercise 3.1.1). Show that the sedimentation coefficient, S,is also an
approximate measure of this time.
3.1.6 In a sedimentation equilibrium experiment, Svedberg found the following
concentration ((7-depth ( x ) profile for carboxyhaemoglobin:
The rotor speed was 8710 r.p.m. and the temperature was 20.3 "C. Take the
density of the water as 0.9988 g cm-3 and that of the dry protein as 1.338 g cmP3
and estimate the molar mass of the protein, assuming that the molecules
sediment as individual particles.
3.1.7 An informal demonstration of the reasonableness of eqn (3.1.14)can be given by
balancing the sedimentation force, (m - m')m2x against the driving force for the
diffusion process (eqn (1.5.11)).Derive eqn (3.1.14) by this method; then derive
eqn (3.1.15).
3.1.8 Calculate the centrifugal acceleration at a distance of 7 cm from the axis of an
ultracentrifuge rotor travelling at 20 000,40 000, and 60 000 r.p.m. and compare
this with the gravitational acceleration, g.
-a& tad
Fig. 3.2.1 Polarization of the molecules of a dielectric between two plates of area A on which is a
charge per unit area +(Tand d. (After Atkins 1978.)
where €0 is the dielectric permittivity of a vacuum (€0 = 8.85 x lo-'' F m-l or C V-'
m-'). With the dielectric material in place, the field drops to Q / E = a0/erc0,where E, is
a characteristic of the material (the relative dielectricpermittivity, E / E O , sometimes called
the dielectric constant). Note that E , > 1; the field is always reduced because the dielectric
material aligns itself as shown in Fig. 3.2.1 and this partly cancels the applied field.
The capacity of this parallel plate condenser is equal to the ratio of charge to
potential difference. If the space between the plates is empty, then:
C = Eoc,d/d (3.2.2)
where p is the local dipole moment of the molecules at r and p~ is their local number
density. P is again the dipole moment per unit volume.
The induced charge f P d near the plates reduces the effective charge to a value
(00 - P ) d so that the new field has a magnitude (from eqn (3.2.1)):
126 I 3: R E S P O N S E TO E X T E R N A LFIELDS AND S T R E S S E S
P = EOE(E,- 1) (3.2.4)
D = EOE+ P . (3.2.6)
(3.2.7)
where E(t) is the magnitude of the electric field which is assumed to be oscillating with
a frequency w (rad s-'). This could be represented by
+ -a
-4
+
I
L
I
I
q - 1
I
I
I
(a) exponentials are rather easier to handle in integration processes than are
trigonometric functions; and, more importantly,
(b) the use of the complex number procedure allows us to separate out the elastic
(storage) effects during the interaction from any dissipative (frictional or viscous
(lossy)) processes like absorption.
and direct differentiation and substitution in eqn (3.2.7) gives (for x = xo, t = 0):
The polarization of the system is, in this simple case, equal to the dipole moment: P(t)
= 9x6)
so that:
where:
Po = qxo. (3.2.13)
where
2 = ko/m
wo and aowi = 2/m.
Provided w is not too close to 9 we can write
(3.2.1 5 )
The constant of proportionality between PO and EO is, in this case, the (frequency
dependent) polarizability, a, which measures the ease with which electrons can be
displaced by an applied field. This is so because the polarization is here equal to the
induced dipole moment p and a is deJined by the relation:
p=aE (3.2.16)
128 I 3: R E S P O N S E TO E X T E R N A LFIELDS AND S T R E S S E S
It is obvious from eqn (3.2.15) that something rather dramatic occurs when w = wo since
Po would then increase without limit. This is the natural resonancefrequency of the system,
wo = ( & / m and at that frequency the dipole would be able to absorb energy from the
field during every oscillation and so gradually increase its amplitude until it was infinitely
large. There would be, at this frequency, an extremely sharp line in the absorption
spectrum. In practice, of course, apart from the quantum effect, the spring would not
remain harmonic in its action and in a real material other constraints would limit the
motion of the charge so that the absorption peak would be reduced in height and
broadened. The result is a Lorentzian function. (See Exercises 3.2.6 and 3.2.7)
In order to complete the solution of eqn (3.2.7) to find P(t) as a function of E(t) we
would have to introduce the notion of a complex polarizability (a=a’ ia”)and solve +
for a’ and a”(Richmond 1975). Rather than do that we will now examine the more
general relation between the vectors P and E, which applies when a collection of
interacting dipoles is involved. The vector P still measures the dipole moment per unit
volume but its relation to E requires further discussion.
3.2.3 The dielectric response function
The more general relation between P and E when the field varies with time has to take
account of the past influence of the field on the present state of polarization and that
involves the idea of a memory function. In place of eqn (3.2.3) we now write:
t> = PNP(I, t )
W - 9 (3.2.17)
Because the induced polarization takes a finite time to decay, the total polarization at
time t is dependent on the electric field at all previous times t - t (0 C t C 00). The
most general (linear) relation for P (I,t) is then (Landau and Lifshitz 1960):
s
M
..
where m ( t ) is the memory function, which tells us the contribution to P(r, t ) at time t
because of an electric field E(r, t-t), applied at a time t before t (Parsegian 1975). We
will encounter a similar memory function in Section 3.4 in connection with the
mechanical (rheological) properties of a colloidal material. [Such linear relations
between an imposed field and the resulting response can only be expected to hold for
low to moderate fields. Ultimately, at sufficiently high field strengths, the material will
respond non-linearly. This will introduce further complications that are beyond the
scope of the present treatment.]
The function m ( t ) describes the decay of an induced polarization with time and, of
physical necessity, must tend to zero for sufficiently large t:
lim(as t + 00) m ( t ) = 0 (3.2.20)
RESPONSE OF A DIELECTRIC MATERIAL TO AN ELECTRIC F I E L D I 129
Parsegian (1975) gives a useful discussion on m (t)in which he draws attention to the
fact that it must also be finite for all t (since otherwise one would have an infinite
polarization produced by a finite field. Recall the problem with the undamped
oscillating charge at the resonance frequency wo in Section 3.2.2.). Note also that t
must never be negative since that would imply that the field at some future time was
affecting the present polarization (i.e. the cause coming after the effect). (Such
violations of the causality principle may be quantum mechanically possible (Wheeler
and Feynman 1945) but will not concern us here.)
The function m ( t ) contains all the information about local electronic, vibrational,
rotational and translational relaxationst of the component molecules of the dielectric
material and for most purposes can be taken to be independent of the position Y within
the material.
From eqns (3.2.18) and (3.2.19):
~ ( rt ) ,= Eo(r)eiwf. (3.2.22)
We can then write, for the (complex) displacement vector:
m(t)Eo(r)exp(-iw(t - t)dt
1
or D(r, t) = ~ & ( rt), 1
where:
00
E,(w)= 1 + I
0
m ( t ) exp(iwt)dt (3.2.25)
and ~ ( w=
) EOE,(W)as before.
+Relaxationis the process whereby a system which has been perturbed in some way
returns to its equilibrium condition
130 I 3: RESPONSE TO E X T E R N A LFIELDS AND S T R E S S E S
s
00
The dielectric response function E ( W ) is simply the extension of the familiar static
relationship between D and E to take account of time-varying electric fields.
(3.2.28)
and
00
E”(w)= €0
J
0
m ( t ) sin ~t d t (3.2.29)
(Note that ~ ( 0 =
) ~ ’ ( 0 and
) ~ ” ( 0=
) 0.) (3.2.30)
T o explore the phase lag, E ( W ) can be written in the alternative form (Exercise 3.2.3):
where
(3.2.32)
and
(3.2.33)
Substituting eqn (3.2.31) in (3.2.24) and taking the real part of the resulting equation
we see that if the physical field E ( I , t) has the form of eqn (3.2.22) then the
displacement vector has the form (Exercise 3.2.4):
so that
2
- K 2 = EL and 2nlK = E:. (3.2.36)
So far we have said nothing about the explicit mathematical form of E ( W ) except what is
contained in eqn (3.2.25). It is possible to show from eqn (3.2.25), using Laplace
transforms (Parsegian 1975), that any physical material can be adequately represented
by a sum (or integral) of terms of the form
(3.2.37)
where the constants&, hj, gj, and WQ must be chosen to fit the experimental data. In
practice, one can limit the types of response to just two: (a) a Debye relaxation term to
characterize the response of a rotating dipole, and (b) a Sellmeier damped-oscillator
form to characterize infra-red, visible, and ultraviolet absorptions.
For the Debye relaxation there is no restoring force, since the dipole simply rotates
(restrained by friction) to follow the field. (This assumes that all of the dipoles can
behave independently which is obviously an approximation.) Since such a rotator has
no resonance frequency ( 9 j = 0) eqn (3.2.37) gives for such a system:
hi di
(3.2.38)
132 I 3: R E S P O N S E TO E X T E R N A LFIELDS AND S T R E S S E S
For the infra-red and ultraviolet absorptions a suitable form is (see Exercise 3.2.8):
.6 (3.2.39)
+ gj(-Zw) + (-Zw)2
and we could expect the following expression to represent a (non-conducting) polar
dielectric material over the entire frequency range:
Exercises
3.2.1 Establish eqn (3.2.14).
3.2.2 Draw an Argand diagram to represent the functions (3.2.31)- (3.2.33).
3.2.3 Establish eqns (3.2.31) - (3.2.33).
3.2.4 Establish eqn (3.2.34).
3.2.5 Consider a capacitor (for which C ( w ) = ~ ( w ) d / dsubjected
) to an alternating
voltage V = VOcos wt. The resulting current is the real part of I = V / Z where Z
is the capacitive impedance or reactance and Z = -l/ZwC. (This is the a.c.
analogue of Ohm’s Law.) Show that I = (od/d)VO[O[E”COS wt - &sin ot] where
d is the capacitor plate area.
By considering the integral 0 j2nI(wt).V(wt)d(wt) show that the power
dissipated in the capacitor is entirely determined by 6”:
Power dissipated per cycle = (nd/d)V;E”(O)
whereas E’ relates to the storage of electrical energy during the cycle.
3.2.6 Consider the extension of the problem treated in Section 3.2.2 to include a friction
(damping) term to the electron’s motion. The equation of motion will then be:
m -d2x
=-k OX - B -dx + qE(t).
dt2 dt
Show that A can be written in the form x’ + i x” where x’ and X” are real
functions and then:
incident beams, together with sensitive and stable photon-detection apparatus and
rapid data analysis by computer.
The theory of light scattering has been extensively described by van de Hulst (1957)
and by Kerker (1969). The following analysis owes much to recent reviews by
Sorensen (1997) and by Kralchevsky et al. (1997).
A proper treatment of the interaction between an electromagnetic wave of
wavelength h and a particle consisting of many thousands of atoms would require
the formalism of quantum field theory and would, in fact, be insoluble by present
methods. Fortunately, however, it turns out that the problem can be tackled from a
classical point of view, and was indeed solved in fairly general fashion before the
quantum theory gained much credibility (Mie 1908; Debye 1909). Indeed, an
approximate analysis, valid for very small particles was given by Lord Rayleigh in
1871.
where k is the wave vector which has a magnitude k = w/c = 2n/h where w is the
frequency. The relation between the incident wave and a wave scattered in any
direction is indicated in Fig 3.3.l(a). The incident wave is characterized by a wave
vector ki and the scattered wave by k,.In most practical applications of light scattering,
the observation of the scattered wave is confined to the horizontal plane (4 = 90°) and
we are concerned with elastic scattering so that the magnitude of the wave vector is
unaffected (ki = k, = 2n/h).
We must also consider the polarization of the incident and the scattered light (i.e.
the plane in which the electric field vector is considered to be oscillating). T h e state
of polarization can be described in terms of two independent polarization states:
horizontal (H) and vertical (V). For an initial wave travelling along the x axis as in
Fig. 3.3.l(b), we will take the vertical plane to be defined by they and z axes and the
horizontal plane by the x and y axes. Normal (incoherent) light from the sun or
from an incandescent lamp is said to be unpolarized which means that the plane of
polarization is randomly oriented. Such a source can be regarded as providing equal
intensities of vertical and horizontal polarization.
In principle it would be possible to use light which was initially polarized in the H or
V direction and to study the scattered light in either H or V polarization. We would
then have to consider the intensities:
R E S P O N S E TO ELECTROMAGNETIC (LIGHT) W A V E S I 135
I /r
Y
Fig. 3.3.1 (a) The incident wave is characterized by a wave vector ki and the scattered wave by wave
vector k,. The plane P is perpendicular to the scattered wave and the amplitude measured in that
plane will be proportional to sin 4 and to cos 8.
Fig. 3.3.1 (b) The usual configuration for light scattering in which the incident and scattered beam
lie in the xy plane which is horizontal. 8 is the scattering angle and in most modern systems the
incident beam is provided by a laser source which is vertically polarized (V).
136 I 3: R E S P O N S E TO E X T E R N A LFIELDS AND S T R E S S E S
In most early studies the light source was an incandescent lamp with unpolarized
( I H= Iv) light but in more recent apparatus the light is supplied by a laser, and is
almost invariably vertically polarized. Th e scattered light is also usually viewed in
vertical polarization so we will concentrate, unless it is stated otherwise, on IVV
which, for very small particles, is independent of the scattering angle 0 in the
horizontal plane.
The theory of light scattering has been developed along two distinct lines which
look somewhat different but which are in fact identical. When light passes through a
non-absorbing, perfectly homogeneous medium, the electric field vector causes the
molecules of the medium to be polarized. The fluctuating dipoles which result will
radiate energy in the form of light but this radiation is uniform in all directions and
cancels itself in the body of the medium; the light therefore appears to move through
the medium unchanged. In a real medium, however, there are always small
fluctuations occurring in the local molecular density and these local density
differences result in incomplete cancellation and, hence, some very weak scattering.
It is this weak scattering which gives rise to the colour of the oceans and the sky.
Theoretical descriptions can be given in terms of (spherical) solid particles of known
refractive index immersed in a medium of a different refractive index, or they can be
couched in terms of the density fluctuations occurring in a solution with a refractive
index which differs from that of the solvent; we will generally use the former
treatment.
Rayleigh’s initial (1871) analysis concerned the scattering from a collection of very
small, non-conducting, and non-absorbing particles acting independently of one
another. In that case we have
p = 2xa/h << 1 and n p << 1 (3.3.2)
where n = %,/no is the refractive index of the particles relative to that of the
surrounding medium.
The first condition ensures that the whole of the particle is subjected to the same
electric field strength at each instant in time so there is a negligible difference in the
phases of the scattered wave from different regions of the particle. The particles must
also be separated from one another so that the scattering from one does not interfere
with that from other particles.
Rayleigh argued that the amplitude of the scattered wave, relative to that of the
incident light would fall off inversely with distance, r from the scatterer. The electric
field of the incident light would produce a dipole of moment p which would be
proportional to the polarizability, a of the scattering material. The field of the scattered
wave produced by this oscillating dipole would, at any distance, also be proportional to
p and, hence, to a. Furthermore he knew that a was proportional to the volume, V,of
the scattering particle (Exercise 3.3.2). The relative amplitude of scattered and initial
electric fields is, of course, a dimensionless number and, apart from depending on the
ratio V/r could also depend on such parameters as the wavelength, the permittivity of
the medium, the density of the scatterer etc. But, apart from the wavelength, all of
these other parameters would involve dimensions of mass and/or time which are
apparently not involved. Only the wavelength can be involved and that as the inverse
square so that the relative amplitude of scattered and incident waves is proportional to
R E S P O N S E TO ELECTROMAGNETIC (LIGHT) W A V E S I 137
the dimensionless ratio V/rA2 or a/rA2. The intensity of the light beam is proportional
to the square of the amplitude and so the relative intensity should be given by:
I/Io c( a2/r2h4.
In terms of the magnitude of the wave vector, k, and using SI units, Rayleigh’s
equation becomes:
2
1 6n4
(3.3.3)
where the quantity in square brackets has the dimensions of volume and is the
polarizability of the scatterer, as that term was understood by Rayleigh. [The 4n€o
term is a consequence of the use of rationalized SI units. (Appendix A6)]
Rayleigh’s eqn (3.3.3) is particularly applicable to the scattering from molecules and
it is invoked to explain why the sky is blue: scattering from small fluctuations in the
density of molecules in the air is strongest for light of short wavelength so it is the blue
end of the visible spectrum that we see. At sunset it is the transmitted (red) end of the
spectrum that is most obvious, and the scattering is augmented by the presence of dust
particles and water droplets in the lower parts of the atmosphere.
A more familiar form of Rayleigh’s equation is obtained by expressing the
polarizabilities in terms of the refractive indices of the scatterers. It can be shown,
using the continuum theory of dielectrics (Section 3.2) that the polarizability of a
dielectric is given by the Clausius-Mossotti equation (see e.g. Atkins 1978, p. 756;
1982, p. 7 7 3 ) (Exercise 3.3.2):
(3.3.4)
(3.3.5)
(3.3.6)
(3.3.7)
where the function P(@)depends on the polarization of the incident and scattered light.
By considering the projections of the electric field vector onto the scattering plane
(Fig. 3.3.1) it can be seen that if 4 = 90":
Pw = 1; PHH = cos2 0; P
V H = PHv = 0. (3.3.8)
Few colloidal particles conform to the requirements of the Rayleigh model (j3 << 1)
but eqn (3.3.6) is important because it emphasizes the strong dependence of scattering
on particle size, and wavelength. Since the scattering is proportional to v2Np/h4at any
angle it is apparent that for any particle size (v constant) the scattering increases
directly with particle concentration and for a given mass concentration (vNp constant)
of the particles, the scattering will increase with particle size. This second property is
extensively used to follow the process of particle aggregation (called coagulation or
flocculation (Section 1.6)).
T o use eqn (3.3.4) for the excess polarizability of a suspension we require knowledge
of the ratio n (= np/no) of the refractive index of the particles to that of the
surrounding medium and must also assume that the particles are spherical. A slightly
different formulation (Kralchevsky et al. 1997) follows if the scatterers are treated as
forming a solution of refractive index n, compared with the refractive index of the
medium, no (Exercise 3.3.3):
This relation allows one to determine, in principle, the molar mass, M, from the scattering
behaviour. Note that this analysis applies only to very small particles (such as protein
molecules) for which the molar mass is a more appropriate size parameter than the radius.
As the particle size increases and a more elaborate theory is required, it is only the
detailed @-dependencethat changes; the dependence of I on h and the number and size
of the scatterers remains as indicated in eqn (3.3.7) and (3.3.10).
R E S P O N S E TO ELECTROMAGNETIC (LIGHT) W A V E S I 139
3.3.2 Rayleigh-Gans-Debye (RGD) scattering
When the particles are too large to satisfy the condition /3 << 1, Rayleigh (1910)
suggested a strategy which was developed by Debye (1915) and Gans (1921). It
depends on the assumption that the incident light beam, which generates the dipoles is
not affected (either in magnitude or in phase) by the presence of the particles. That will
be so for small particles but also for larger ones if the following conditions are met:
In - 11 << 1 and 2/3ln - 11 << 1. (3.3.1 1)
Note that the second condition allows the particles to be larger than for Rayleigh
scattering, provided the refractive index difference between particle and medium is
sufficiently small.
As the particle size increases the scattering pattern ceases to be symmetrical and the
amount of forward scattering (0" c 0 c 90") increases at the expense of back scattering
(90°C 0 C 180").The waves which are scattered in the strictly forward direction (0 =
0") remain in phase and the intensity in this direction is a maximum (P (0 = 0) = 1)
and so from eqn (3.3.10) we can set R(0) = CKM. Then we can define P(0) as
(Kralchevsky et al. (1997):
P(0) = R(O)/R(O). (3.3.12)
Expressions for P(0) can be calculated from purely geometrical considerations and this
has been done for a number of particle shapes. (See, for example, Kralchevsky et al.
(1997) Table 4.) The intensity of the scattered wave in any direction, 0, depends on the
phase relationships of the waves coming from different parts of the scattering particle
which will depend on the path differences involved. The results are therefore quoted
in terms of the scattering vector, Q, defined as shown in Fig. 3.3.2. As noted above, the
scattering is elastic so ki = ks = 2n/h and so we have:
(3.3.14)
and analogous expressions can be written for rods and Gaussian coils. Expansion of
eqn (3.3.14) in powers of Qa gives (Exercise 3.3.4):
and it can be shown that for particles of arbitrary shape, satisfying the RGD criteria,
Pusey (1982):
Fig. 3.3.2 The geometry of a scattering event. Path difference =a + b = (ki - ks).ri,= -Q.rij
+ O(Q4) ]
KC
- = [MP(0)]-' = (3.3.17)
R(0) M
so a plot of KC/R (0) against sin2(0/2) for small values of 0 will give values of both the
molecular weight and the radius of gyration of the scattering entities. This is called a
'Guinier plot'. Equation (3.3.16) can be used for particles in the range 20 < ac/(nm) <
100. The technique has become much more useful with the advent of lasers because
they can be sharply collimated so measurements of scattering intensity can be made at
very small angles to the propagation direction. The technique is known as low angle
laser light scattering (LALLS).
The values of P(0) used in eqns (3.3.12-17) are for vertically polarized initial and
scattered light but the relationships in eqn (3.3.8) still hold and can be used to
evaluate the other contributions once Pv(0) is known. Thus we have (Kralchevsky
1997):
The asymmetry in the scattering pattern for particles in the RGD region makes it
possible to obtain information about particle size and, in some cases, shape. Values of P
(0) as a function of the scattering vector are shown in Fig. 3.3.3 (a) for particles of
various shapes. Even the ratio of forward to back scattering, as measured by the
dissymetry ratio:
RD = 14511135 (3.3.19)
can give information on shape if the size is known and vice versa (Figure 3.3.3(b)).
R E S P O N S E TO ELECTROMAGNETIC (LIGHT) W A V E S I 141
3.3.3 Mie scattering
As the particle size increases, the scattering pattern becomes still more complicated as
the spherical waves from each scattering centre in the particle increasingly interfere
with one another so that the intensity shows pronounced maxima and minima at
particular angles, 13, determined by the size parameter /3 = 2 n d h and the particle
refractive index (or polarizability). If white light is used for illumination of
monodisperse sols, the result is the appearance of strong beams of light of particular
colours (chiefly green and red) at particular angles. These are called higher-order
Tyndall spectra (HOTS) and their analysis can lead to data on particle size.
Fig. 3.3.3 (a) Scattering function P(0) for macromoleculesof various shapes. (After Marshall 1978,
p. 477.) The abscissa represents Qd/2no where Q is the scattering vector and d is the diameter for
the sphere or disc, and the length for the rod. For a random coil, the abscissa represents Q d / J 6 n o
where d is the r.m.s. end-to-end distance.
Spheres
5-
4-
3-
2-
1
0.1 0.2 0.3 0.4
L I/?
Figure 3.3.3 (b) The dissymmetry ratio RD as a function of (relative) characteristic length, for
spheres, random coils, and rods. (From Kerker 1969, p. 432, with permission.)
142 I 3: R E S P O N S E TO E X T E R N A LFIELDS AND S T R E S S E S
-
In this region (d A) the complete theory developed by Mie (1908) must be used to
properly describe the scattering pattern. Mie solved the general problem for a sphere
of any size when the refractive indices of both the particle and the medium can be
complex (i.e. for the case where the light wave may be absorbed as well as scattered
(Section 3.2)). We will not describe that result in detail. Suffice it to say that as the
wave vector increases above the Guinier regime, the intensity falls off generally as
(Qu)-~ but with a pronounced pattern of maxima and minima as shown in Fig. 3.3.4.
The precise positions of the maxima and the depths of the minima depend on the
details of refractive index and size. A good general description of this region is
provided by Sorensen (1997).
When the system is moderately polydisperse, the maxima and minima from
different size spheres tend to smooth one another out but one still observes the general
fourth power fall off in the intensity with increase in the relative size (Qu). The region
where In I versus In (Qu) has slope -4 is called the Porod region.
T he success of the Mie theory may be judged from Fig. 3.3.5. That data was
obtained using an ingenious device designed by Gucker e t ul. (1973) and able to
determine the complete scattering pattern from a single spherical particle over
I I
I 0-5
0.1 1 10
Fig. 3.3.4 Mie scattering intensity for vertically polarized incident and detected light for spheres of
size parameter B = 0.4, 1,2,4, and 8 and refractive index n = 1.33 as a function of the dimensionless
factor Qa. For polydisperse systems the rapid variations in the region Qa > 2 tend to cancel one
another and the intensity falls smoothly with slope 4 in this log-log plot. (This is the Porod region.)
(From Sorensen 1997 with permission.)
R E S P O N S E TO ELECTROMAGNETIC (LIGHT) W A V E S I 143
-Experimental
--- Theoretical
0 360
Degree
Fig. 3.3.5 Comparison of experimental data and theoretical (Mie) curves for scattering from a
single polystyrene sphere. 0.771 is the least squares deviationbetween theory and experiment. (From
Marshall et al. 1976, with permission.)
360" in about 20 ms. Several scans of the same particle can be done consecutively
while it remains essentially stationary in the laser beam. Only symmetric patterns
(Fig. 3.3.5) are accepted and stored for analysis. (See also Davis and Ray 1980.) We
will examine the application of Mie theory to particle size determination in
Chapter 5.
Exercises
3.3.1 Describe briefly what occurs when visible light is absorbed by a particle. In
what way is the energy usually stored? Why does re-radiation usually occur at a
longer wavelength? When is this not the case? If re-radiation does occur at the
incident wavelength is the effect on the incident beam noticeable? Why?
3.3.2 Use the definitions of a and P from Section 3.2 to find a link between a and
E (w). In a region where there is no absorption, show that
a €0 (n; - 1)/N
for a collection of gas molecules of refractive index nl with N molecules per unit
volume. Show that this can be reconciled with eqn (3.3.4) for a condensed
medium (where N = (47ta3/3)-') if cr is related to the refractive index.
3.3.3 Establish eqns (3.3.9.) and (3.3.10).
3.3.4 Use the standard expansions:
3.3.5 Show that the radius of gyration of a spherical particle of uniform density is
related to its radius by: ac = &3/5) a. Note that this reconciles eqns (3.3.15)
and (3.3.16) for a sphere.
+The quantity F / d is usually refered to as the shearing stress and we will use that
expression henceforth. Strictly speaking, however, the traction is a vector quantity
whilst the stress at a point is a tensor that is the aggregate of all tractions acting on the
surface elements of different orientation that contain that point (Reiner 1960, p. 5). We
are assuming then that we can consider only one component of the stress on the upper
surface. (The stress tensor will be introduced in the next chapter.)
RESPONSE TO A M E C H A N I C A L STRESS I 145
0 XI
material between the plates will cause a deformation (or strain), y. When the force is
removed, we find that either:
These three behaviour patterns are characteristic of solids, liquids, and plastic
materials, respectively, but in practice, most materials can exhibit any or all of them,
depending on the time scale involved in the application of the stress and the
measurement of its effects. The time scale is measured by the Deborah number (Harris
1977, p. 21):
As DN + 0, materials tend to behave more like fluids and as DN + 00 they behave like
solids. Thus geologists can speak of the ‘flow’ of rocks over geological time, while a
person falling from a great height into water encounters its solid-like characteristics
when deformed over short time intervals.
where Y is called the Young’s modulus of the material. The strain, y, is in that case the
relative change in length. The corresponding behaviour under a shearing stress is
described thus:
146 I 3: R E S P O N S E TO E X T E R N A LFIELDS AND S T R E S S E S
F l d a y or F / d = S = G y (3.4.2)
where G is the shear modulus of the material and y is defined in Fig. 3.4.1. Many solids
conform to eqn (3.4.2) for small stresses and provided the stress remains below some
upper limit S L , they will recover their original shape completely when the stress is
removed. If S > SLthe material suffers permanent deformation, i.e. someflow or creep
occurs and the solid has begun to exhibit some of the characteristics of a plastic or a
liquid.
Ideal liquid-like behaviour is described as Newtonian behaviour and in that case the
applied shearing stress is directly proportional to the time rate of strain or rate of shear
( y = dy/dt):
S a y or S = q y (3.4.3)
where the proportionality constant, q, is the (first coefficient of shear) viscosity. This
equation was proposed by Newton to describe the flow behaviour of simple fluids
(gases and liquids like water) undergoing steady shear. Consider the system shown in
Fig. 3.4.2 where a simple liquid is confined between two plates. The lower plate is
stationary and the upper plate is being pulled at a velocity v by the application of a
force per unit area, S. It may be assumed that the liquid in contact with the lower plate
remains stationary whilst that in contact with the upper plate must move with the
velocity v. Between the two a gradual variation in velocity occurs. The rate of shear, 9,
for this simple shear regime is equal to the velocity gradient v/h and is normally
measured in s-'. In the more general case i. = dv/dxz.
Fig. 3.4.2 Deformation (flow) of a liquid under an applied shearing stress, S. If the velocity of the
upper plate is z, then i. = d(tan a)/dt = ( l / h ) dxl/dt) = v/h. (This coordinate system is used
throughout the next chapter.)
RESPONSE TO A M E C H A N I C A L STRESS I 147
for example, the clay minerals (Section 1.4.5)).Measurements at higher concentration
or on shorter time scales (higher Deborah number) may begin to reveal evidence of
solid-like (elastic) behaviour in what otherwise appears to be a (viscous) liquid. Such
materials are said to be vasco-elastic and they may be intrinsically solids or liquids,
depending on which of the two characteristics is dominant.
For materials of this sort, the entire deformation history may, to some extent,
influence its present structure. Once again we encounter the need for a memoy function
(as in Section 3.2.3) to properly describe the behaviour. An introduction to the theory
is given by Ferry (1980) from which the remainder of this section is taken.
Again we restrict ourselves to the linear theory (compare eqn (3.2.19))in which the
effects of sequential changes in strain are assumed to be additive:
S(t)=
i
--oo
G(t - t') p(t') dt'. (3.4.4)
Note that in this formulation of the memory process t - t' = t,so that t is the total
elapsed time and the integration is carried out over all past times up to the current time
t. G(t) is called the (shear) relaxation modulus of the material.
If the function G(t) approaches zero for very large t, then the substance is liquid-
like and an alternative formulation of eqn (3.4.4) can be given in terms of the strain,
rather than the rate of strain (Exercise 3.4.1):
s
I
y(t) = 1t
--oo
J(t - t ')i(t ')dt' (3.4.6)
‘I,
Fig. 3.4.3 Time profiles for a simple stress relaxation experiment following a sudden strain. The
behaviour of S for - 6 < t < to is not usually accessible but S(t) can be followed for t > to.
S(t) =
1
t0-c
G(t - t’)(y/$) dt’ (3.4.7)
Fig. 3.4.4The mean value theorem establishes that there is a horizontal line which can be drawn so
as to equalize the shaded areas. Then .[,?Gdt = area ABCD = tG(tl+et) for some e < 1.
RESPONSE TO A M E C H A N I C A L STRESS I 149
for times that are long compared to (. The ratio of a stress to the corresponding strain
is called a modulus (Section 3.4.2) and for a perfectly elastic body, the equilibrium shear
modulus G = S / y from eqn (3.4.2). G(t) is then the time-dependent analogue of G,
measured in an experiment with this sort of time pattern.
Provided the strain y is not too large, the value of G(t) should be independent of y
and it is only then that this simple procedure is of value. The concept can be applied to
liquid-like and solid-like materials.
(b) Stress relaxation after cessation of steady shear flow
As noted earlier, liquid-like materials can be grossly deformed and still retain some
structural characteristics. They can be deformed at a constant strain rate 9, under a
steady shearing stress S where S = Yqo. (qo is used for the viscosity to signify that it is
measured at sufficiently low shear rates for the behaviour to be linear). If the flow is
suddenly stopped, the shearing stress decays with time and it can be shown (Exercise
3.4.3) that, for to = 0 (Fig. 3.4.5):
00
(3.4.9)
I Stress ,
t0 Time
Fig. 3.4.5 A simple shear stress relaxation experiment, following cessation of flow at time t = to.
150 I 3: R E S P O N S E TO E X T E R N A LFIELDS AND S T R E S S E S
m
m Y
2
;
;
;I
I
t0 Time Time
Fig. 3.4.6 A creep compliance experiment in which a sudden stress is imposed and then maintained
constant while the resulting strain is measured.
ok I I I
I I I I
I I I I
I I I
Time
I I
Fig. 3.4.7 Stress-strain relationship for an ideal solid and an ideal liquid subjected to a sinusoidal
stress. Note that for the liquid it is the strain rate (p) that is in phase with the stress and this is shown
by the full curve (which also corresponds to the strain for an elastic solid). The stress in the liquid
leads the strain by n/2.
RESPONSE TO A M E C H A N I C A L STRESS I 151
find that once again the use of a complex variable enables us to keep track
simultaneously of the elastic (storage) and viscous (dissipative) characteristics of
the material (Fig. 3.4.7).
s
00
= yo w
[1 1
G(t)sin w t d t sin wt + yo w 1
00
0
G(t)cos w t d t COSWt
+
(since cos (A - B) = cos A cos B sin A sin B and t is a constant in the integration
process).
The integrals will converge if G + 0 as t + 00, and this will be so if the material is
liquid-like since, by definition, a liquid cannot permanently support a shearing stress.
The terms in square brackets are functions of w but not o f t and we can write
where G'(w) represents a modulus that measures the ratio of the in-phase stress to the
strain. This is the shear storage modulus. The quantity G"(w) likewise measures the
ratio of the stress to the strain which is 90" out of phase; it is the shear loss modulus.
The nomenclature here reflects the result obtained in Exercise 3.2.5: the lossy
(dissipative) part of the process is represented by G". For a purely elastic solid the
stress and strain remain in phase and so G" = 0 and G' = G.
For a purely viscous liquid it is the rate of strain y that remains in phase with the
applied stress, according to eqn (3.4.3), and the material has no elastic (storage)
character so G' = 0. Then
and then in eqn (3.4.15) q = q’ for purely viscous behaviour. For the viscoelastic
material
TI‘= G“/w and q” = G‘/w. (3.4.17)
An alternate method of describing a viscoelastic fluid (Fig. 3.4.8) is in terms of the
phase lag, 6, between stress and strain. Writing
SO SO
G’ = -cos6; G” = -sin6 (3.4.19)
YO Yo
and
G“/G‘ = tan 6. (3.4.20)
An oscillating measurement at frequency w corresponds to a transient measurement
over a time t = l/w and the result obtained gives two pieces of information: the ratio of
the amplitudes of stress to strain (S ‘ / y o ) and the phase lag, 6, or alternatively the
values of G’ and G” or of fand q’ (from eqn (3.4.17)).
The alternative representation in terms of a complex strain y* = yOexp(zwt)and the
corresponding complex stress S* = P e x p [z(wt+6)] yields a complex modulus G*=
S*/y* where G* =GI +
iG”and I G* I = (Gf2 G’”)~. +
Polymer solutions show very pronounced viscoelastic behaviour and much
experimental work has been done on the phenomenon (see Bird et al. 1977; Ferry
1980). The elastic component is much smaller in most colloidal dispersions but
coagulated colloidal sols do show some viscoelasticity (van de Ven and Hunter 1979)
and we will return to that behaviour in Chapter 15.
/ I
m
I
I
I
I !L
I
I I
I
I
I
I
I
Fig. 3.4.8 Stress-strain relationship for a visco-elastic fluid. S = 0 corresponds to elastic solid and
8 = n/2 corremonds to viscous liauid.
RESPONSE TO A M E C H A N I C A L STRESS I 153
3.4.5 Viscous shear behaviour
For many colloidal dispersions, the elastic effects play a rather secondary role in the
behaviour, especially if the system is being sheared very strongly. Thus, in the
pumping of a slurry or the high speed extrusion of magnetic iron oxide paste to make
recorder tape, we are more concerned with the viscous (dissipative) aspects of the flow
behaviour, even though the storage or elastic properties have to be recognized. The
fundamental assumptions of linearity, small strains, and small strain rates on which the
memory equations (eqns (3.4.4)-(3.4.5)) are based no longer apply. In one sense this
makes the analysis easier because the majority of the time effects may then have
disappeared.
In many such situations it is sufficient, at least as a first step, to investigate the
relationship between shear stress and shear rate, to obtain a more general form of eqn
(3.4.3)in which the viscosity, q, is no longer a constant. In the simplest case, when the
experimental measurement is conducted on a time scale which is long compared to the
relaxation time of the system, the viscosity becomes independent of time, though in
general it will depend on the shear rate. We can then define an apparent viscosity, qappas
or a dafferential viscosity:
rdiff ( Y ) = d S/ d j. (3.4.22)
Some of the behaviour patterns commonly exhibited by colloidal dispersions are
shown in Fig. 3.4.9. If rappand rdiff both decrease regularly with shear rate (curve 2)
the behaviour is called pseudoplastic and if they both increase (curve 3) it is dilatant. If
the material behaves like a solid until a certain value of stress is reached and then
deforms like a Newtonian liquid obeying:
S-SB=VPLY (3.4.23)
where T]PL is constant, this is called Bingham behaviour and it is the ‘ideal’ standard for
plastic behaviour. ~ P is L called the plastic viscosity and SBis called the Bingham yield
value. In this case rdiff is constant but qappdecreases continuously from its zero shear
value (TO + 00) to some limiting (infinite shear) value qoo. This type of behaviour is
observed in, among other things, concentrated dispersions (slurries) of coal in water at
fairly high volume fractions and low shear rates.
Newtonian behaviour is observed in dilute stable dispersions of spherical particles.
Pseudoplastic behaviour occurs even in dilute dispersions of anisometric particles
because the increasing shear rate tends to orient the particles along the fluid
streamlines and this decreases the viscosity. Dilatancy is common in concentrated
dispersions; in this case it is the flow of (lubricating) fluid between the particles that
dominates the behaviour and the shearing process tends to drive particles together
which constricts the flow channels.
Non-ideal plastic behaviour (Fig. 3.4.9, curve 5) is characteristicof fairly dilute (-10%)
coagulated colloidal dispersions in which every collision between two particles results in
the formation of a (temporary or permanent) link. The extrapolated value of stress
obtained from the linear high shear rate behaviour is again called a Bingham yield value and
154 I 3: R E S P O N S E TO EXTERNALFIELDS AND S T R E S S E S
Fig. 3.4.9 Common forms of flow behaviour for colloidal dispersions: (1) Newtonian; (2) shear
thinning (pseudoplastic); (3) shear thickening (dilatant); (4) (ideal) Bingham plastic; (5) non-ideal
plastic. So,the primary yield value, is a well-defined quantity for concentrated dispersions and pastes
(Nguyen and Boger 1983) but for dilute systems its value depends on the previous shear history of
the sample and erroneous values can easily be obtained as instrument artefacts.
the differential viscosity at high shear rate can be called a plastic viscosity. We will discuss
these forms of flow behaviour, and their interpretation in more detail in Chapter 15.
If the time scale on which the measurements are conducted is sufficiently short
(that is, comparable with or shorter than the characteristic relaxation time for the
structure of the flow units) then one may also observe changes in the apparent or
differential viscosity as a function o f time even at constant shear rate. The behaviour
shown in Fig. 3.4.9 can then be regarded as the steady-state behaviour, arrived at after
the system has had sufficient time to relax (i.e. to establish a (dynamic) structure in
response to the shearing stress). Some systems show a gradual decrease and some a
gradual increase in apparent or differential viscosity with time at a fixed rate of shear.
These behaviour patterns are referred to as thixotropy and rheopexy respectively.
Thixotropy occurs commonly in dispersions of very anisometric particles (e.g.
montmorillonite (bentonite), Section 1.43, especially at moderate and high shear rates.
The increasing shear rate requires an appreciable time to break down the particle linkages
to establish the steady-state structure for that shear rate. Rheopexy usually occurs under
conditions of gentle stirring where the slight degree of agitation apparently allows
particles to establish extensive structures, which resist the shearing action.
Exercises
3.4.1 Establish eqn (3.4.5) from eqn (3.4.4).
3.4.2 Establish eqn (3.4.8).
3.4.3 Show that in the stress relaxation experiment of Fig. 3.4.5
S(t) = 1; I
t-to
G ( t ) dt.
References
Allen, T. (1975). Particle size measurement. In the Powder technology series (ed.
J.C. Williams). Chapman and Hall, London.
Atkins, P.W. (1978). Physical chemistry. (2nd edn 1982), (3rd edn 1986). Oxford
University Press, Oxford.
Bird, R.B., Armstrong, R.C., and Hassager 0.(1977). Dynamics ofpolymerjuids,
Vol. 1, p. 277. Wiley, New York.
Davis, E.J. and Ray, A.K. (1980). 3. Colloid Interface Sci. 75, 566.
Debye, P. (1909). Ann. Phys. 30 (4), 57.
Debye, P. (1915). Ann. Phys., 46, 809.
Ferry, J.D. (1980). Viscoelasticproperties of polymers (3rd edn). Wiley, New York
Frumkin, A.N. and Levich, V.I. (1946). Acta physiochim. U R S S 21, 193.
Gans, R. (1921). Ann. Phys. 65, 97.
Gucker, F.T., T h n a , J., Lin, H.M., Huang, C.M., Ems, S.C., and Marshall, T.
R. (1973). Aerosol Sci. 4, 389.
Harris, J. (1977). Rheology and non-Newtonian j o w . Longmans, London.
Jennings, B.R. and Parslow, K. (1988). Proc. Roy. Soc. Lond. A419, 13749.
Kerker, M. (1969). The scattering of light and other electromagnetic radiation.
Academic Press, New York.
Kralchevsky, P.A., Danov, K.D., and Denkov, N.D. (1997). Chapter 11 in
Handbook of surface and colloid chemistry (ed. K.S. Birdi). CRC Press, New York.
Landau, L.D. and Lifshitz, E.M. (1960). Electrodynamics of continuous media.
Pergamon, New York.
Levich, V.I. (1962). Physicochemical hydrodynamics.Prentice Hall, Englewood Cliffs,
N.J.
Marshall, A.G. (1978). Biophysical chemistry -principles, techniques and applications.
Wiley. New York. [An excellent treatment giving valuable insights into physical
models.]
Marshall, T.R., Parmenter, C.S., and Seaver, M. (1976). 3.Colloid Interface Sci.
55, 624.
Mie, G. (1908). Ann. Phys (Leipzig) 25, 377.
Nguyen, Q D . and Boger, D.V. (1983). Acta Rheologica 27, 321; 29 (1985),
335.
Overbeek, J.T.G. (1982). Adv. Colloid Interface Sci. 15, 251-77.
Parsegian, V.A. (1975). Long range van der Waals interactions. In Physical
chemistry: enriching topicsfrom colloid and surface science (ed. H. van Olphen and K.
J. Mysels) Chapter 4.Theorex, La Jolla, California
Perrin, J. (1909). Ann. Chim. Phys. 18 (8), 5
Pusey, P.N. (1982). Light scattering. In Colloidal dispersions (ed. J.W. Goodwin)
Chapter 6. Royal Society of Chemistry, London.
Rayleigh, Lord (1871). Phil. Mag. 41, 107, 274, 447.
Rayleigh, Lord (1910). Proc. Roy. Soc. (London) A84, 25.
Reiner, M. (1960). Deformation, strain andjow. H.K. Lewis, London
Richmond, P. (1975). The theory and calculation of van de Waals forces. In Colloid
science (ed. D.H. Everett) Volume 2, Chapter 4. Specialist Periodical Report,
Chemical Society, London. [There are a number of typographical errors in the
treatment of the problem given there.]
Sorensen, C.M. (1997). Scattering and absorption of light by particles and aggregates.
In Handbook of surface and colloid chemistry (ed. K.S. Birdi). CRC Publishing,
New York.
156 I 3: R E S P O N S E TO E X T E R N A LFIELDS AND S T R E S S E S
4.1 Introduction
In this chapter we will describe the calculation of three of the transport properties
introduced in Chapters 1 and 3: the sedimentation coefficient, the Brownian diffusivity,
157
158 I 4: TRANSPORT PROPERTIES O F S U S P E N S I O N S
and the effective viscosity of a colloidal dispersion. For simplicity we will concentrate
initially on the case of a suspension of rigid, force-free particles; this is an idealization that
can be approximated in practice by a suspension in which the double-layer thickness is
much smaller than the particle radius, and the repulsive forces are large enough to prevent
coagulation. For most separations the colloidal forces are then unimportant.
To calculate colloidal transport properties it is necessary to determine the way in which
the solvent flows around the suspended particles. In preparation for this analysis we will
devote the first portion of the chapter to a discussion of the relevant aspects of fluid
mechanics. Although earlier colloidal texts were able to get by with a very elementary
treatment of this subject, research in the field of transport properties has now reached a
level where it is no longer possible, even in a qualitative sense, to understand the problem,
let alone the solution, without some basis in fluid mechanics. It is hoped that the first few
sections of this chapter will provide such a basis. In this description it will be assumed that
the reader is familiar with the notation and elementary methods of vector calculus; a few
revision notes are provided on this subject in Appendix 3.
Fig. 4.2.1 The small rectangular volume referred to in the derivation of the continuity eqn (4.2.3).
P has coordinates (XI, XZ, ~ 3 ) .
The total flux per unit volume out of the fluid element at point x is therefore
(4.2.1)
In this case the obvious measure of the flux into or out of a fluid element is the velocity
of the fluid and so eqn (4.2.1) can equally be written:
C,,=O
avi
or A A ~ = O (4.2.2)
t=l
which is called the continuity equation. The second form is read as div v (short for the
divergence of v). This relation is quite general; the flow lines do not need to be parallel
to the coordinate axes and the sample volume can have any shape.
The same argument can be used to establish that, if some component, i, of the
system (say a solute species) is moving with the fluid and diffusing in response to a
concentration gradient, say, it will have a flux in each direction through the volume
element and the net accumulation of the material in the volume element as a function
of time must be given by:
where ni is the number of molecules of species i per unit volume. The part of the flux
which is due to hydrodynamic flow is n p (compare Fig. 1.5.2)and the net contribution
of this component will be zero but there will be an additional component due to the
diffusion process which may well be non-zero. Equations (4.2.1-3) are all
consequences of the notion of mass conservation.
160 I 4: TRANSPORT PROPERTIES O F S U S P E N S I O N S
( Exercise I
I 4.2.1 Establish eqns 4.2.2 and 4.2.3 using the sort of argument developed in
Appendix 3.3.
Fig. 4.3.1 The small block of fluid used in the derivation of the formula (4.3.5) for the stress on a
plane BDC of arbitrary orientation.
STRESS I N A M O V I N G FLUID I161
where the summation is over the three orthogonal faces. This equation can be
simplified using the geometrical expression (Exercise 4.3.1):
AAj =AA2j ’ (4.3.2)
together with the relation S(x, 2j ) = -S(x, -2j), (4.3.3)
which follows from Newton’s third Law (action = reaction). Combining eqns (4.3.2
and 3) with (4.3.1) gives:
This formula enables us to calculate the stress on an arbitrary plane through x,given
the stress on the three planes with unit normals il, iz, and i3.
It is customary to write the components of the stresses S (x, 4) on these three
orthogonal planes as aij where:
S(x, 2j ) = Xi 0ij (x) 2i (4.3.5)
the zth component of the stress on the plane with unit normal 4. Equation (4.3.4)
aij is
can be written more compactly as
S(x, ii) = (x) . ii (4.3.6)
where is a 3 x 3 (Cartesian) tensor whose elements can be represented in the form of a
matrix (see Fig. 4.3.2):
[2 : 21.
011 012 013
(4.3.7)
Since students often encounter difficulties with tensors, we should emphasize that this
is simply a convenient piece of notation. The important thing to bear in mind about the
stress tensor is its physical significance; that is that the components aij are the
components of the stresses on three orthogonal planes through the point.
The net (turning) moment due to the stresses on any small block of fluid is
approximately zero+ so, (from Fig. 4.3.2), it is clear that:
g - 0..
0.. Jr. (4.3.8)
for all i andj. For this reason the stress tensor is said to be ‘symmetric’.
Finally, for future reference we note that the net force due to the stresses over a
macroscopic surface A is obtained by adding the contributions from the surface
elements. On taking the limit of very small area elements and using the formula (4.3.6)
we find that the net stress force is
f* *ii dA. (4.3.9)
t We are still assuming that the parcel of fluid is so small that inertial effects are
negligible.
162 I 4: TRANSPORT PROPERTIES O F S U S P E N S I O N S
Fig. 4.3.2 An illustration of the stresses acting on a small rectangular block with faces parallel to the
coordinate planes. The shear stress, S, depicted in Fig. 3.4.2 would correspond to 012. (Note that the
origin of the stresses labelled 0i3 lies in the plane of the front face of the block.)
Exercises
4.3.1 Derive the formula (4.3.2)for the area of the orthogonal faces of the block shown
in Fig. 4.3.1 using t%foll%g information. By the definition of the cross-
prod%, AA fi =(; BC x BD) where AA is the area of t h e 9 i n g face and BC
and BD are vectors along the e dge s3hat face. Show that BC = 1 3 4 - 1 2 4 and
+
BD = l l i l - 1 2 4 . Calculate BC x BD and use the result to verify eqn (4.3.2).
[k K 81
STRESS AND VELOCITY FIELD IN A FLUID I N THERMODYNAMIC EQUILIBRIUM I 163
If a fluid in equilibrium is in motion, it must move as a rigid body, for this is the motion
that maximizes the entropy (Landau and Lifshitz 1969, section 10). It can be shown
(Meriam 1966, section 30) that the velocity field for an arbitrary rigid body motion
must have the form
v(x) = v + x x (4.4.3)
where V and are independent of x, and are the translational and angular velocities
respectively.
Those readers who are unfamiliar with this formula may find it helpful to consider
the example of a glass of liquid placed on the centre of a steadily rotating turntable, as
shown in Fig. 4.4.1. In a frame of reference that moves with the turntable the fluid
appears to be at rest. In the laboratory frame the fluid particles will move steadily
around circles centred on the axis of rotation with a speed of 2nr/T, where T is the
time for each rotation and Y is the distance of the particle from the axis of rotation.
The components of the fluid velocity in the laboratory frame of reference are
therefore given by
-2nr . -2nx2
V] =-sin8 = ~
T T ’
2nr 2nx1
v2 = -cos8 = ~
T T ’
and v3 = 0.
These component expressions can be written in the vector form (Appendix A3):
v = ( 2 n & / T ) x x. (4.4.4)
Comparing this result with the general formula (4.4.3) for rigid body motion, we see
that the translational velocity is zero, while the angular velocity is given by:
Fig. 4.4.1 An illustration of a fluid that is in equilibrium and in motion: a glass of liquid on the
centre of a rotating turntable.
164 I 4: TRANSPORT PROPERTIES O F S U S P E N S I O N S
Exercise
4.4.1. Establish eqn (4.4.4).
=-jZ+ D
. (4.5.2)
The quantity is known as the ‘deviatoric stress tensor’, since it represents the
deviation of the stress tensor from the equilibrium form (eqn (4.4.2)). will depend on
the history of the motion of the fluid, being zero for a fluid which has been in steady
rigid body motion for a sufficiently long time.
Since the stresses arise from short-range forces in the liquid, the deviatoric stress
tensor at a point will only depend on the history of the motion of the fluid in the
neighbourhood of that point. Furthermore, since the fluid molecules jiggle around and
rearrange themselves very rapidly, the effect of past motions will soon fade. For most
liquids the time for this molecular rearrangement is much smaller than the time
required for the macroscopic velocity v (x, t) to change significantly. Thus, to a good
approximation, D(x, t ) will only depend on the instantaneous velocity field in the
neighbourhood of x,since this is the quantity that characterizes the recent deformation
history.
We do not expect to depend on the absolute velocity of the neighbouring fluid
particles, but only on their velocity relative to the particle at x . If the distance Ax
between neighbouring fluid particles is sufficiently small we may approximate this
relative velocity with the aid of the formula for a total differential, viz.
V(X + aX,t ) -V(X, t ) M Ej(a~/axj)Axj (4.5.3)
where the derivatives are evaluated at (x, t). Thus the local relative velocity field is
determined by the three derivatives &/ax,, and therefore D(x,t ) will be a function of
the nine scalar quantities avilaxj. In fact, because of the isotropic nature of simple
fluids, it turns out that we can express in terms of just two of these terms:
rl = q(avj/axj + avj/axi)
8 (4.5.4)
RELATIONSHIP B E T W E E N THE STRESS TENSOR AND THE VELOCITY F I E L D I 165
where q is the shear viscosity. For the simple shear flow shown in Fig. 3.4.2 we can
write:
v1 = yx2, v2 = 0, v3 = 0,
and then eqn (4.5.4) shows that 012 = qp, where is the shear rate.
where
1
e- -- (-hi+z). avj
(4.5.6)
-2 ax,
The nine quantities eij can be regarded as the components of a tensor called the ‘rate of
strain tensor’, denoted by e. In this notation eqn (4.5.5) becomes
= 2qe.
Combining this formula with the expression (4.5.2) for the total stress tensor, we get
Table 4.7
The viscosities of commonly occurring liquids at 15 “C and I atm. The unit in the SI system is
I N m-2 s = I Pascals = 1 Pa s = 10 poise. (I poise = 1 dyne cm-2 s = I g cm-’ s-‘)
Table 4.2
The viscosity of water over a range of temperatures
Temperature
(“C) 0 5 10 15 20 25
strain tensor e. With the aid of the formula (4.5.6)for the components of e we can write
the expression (4.5.3) for the relative velocity in the component form
where e i is evaluated at the point x. The first two terms on the right-hand side of this
formula are the components of the vector expression
v(x)+(x) x x (4.5.10)
where the components of the ‘local angular velocity’ (x) are related to the quantities:
On comparing eqn (4.5.9)with the general formula (eqn (4.4.3))for rigid body motion,
we see that the first two terms on the right-hand side of eqn (4.5.9) correspond to a
translation and a rigid body rotation of the fluid. Thus the local rate of deformation or
‘strain’ of the fluid in the neighbourhood o fx is entirely determined by the last term in
eqn (4.5.9), which in turn depends only on e(x). This is the reason why e is known as
the ‘rate of strain tensor’.
colloidal particles, and that this average therefore depends on the particle configuration.
For the case of a dilute suspension of rod-like particles for example, this configuration is
characterized by the particle orientation distribution.
In general the macroscopic flow tends to give the particles a ‘preferred’ configuration,
as for example in the case of a shear flow, where rod-like particles tend to be aligned with
the streamlines (Section 4.10.4). This ‘ordering’ tendency is opposed by the Brownian
motion, and the final particle configuration is determined by the balance between these
two opposing forces. In the limit of weak strain rates the Brownian motion dominates, and
the configuration becomes statistically isotropic. Thus in this limit the Newtonian form
(eqn (4.5.4))applies. At higher strain rates this is not generally the case, and as a result the
rheological behaviour is characterized not by one, but by a number of viscosity
coefficients, and these in turn usually depend on the strain rates.
For unsteady flows the situation can be even more complicated, because a change in
the imposed strain rates leads to a change in particle configuration, a change that may
take a significant time. In a suspension of rod-like particles this ‘relaxation time’ is
determined by the time required for Brownian motion to reorient the particles, which
for a 1 p m long particle is of the order of 1 s. If the imposed strain rates change
significantly over this period, the assumption that the stress depends on the
instantaneous rate of strain breaks down, and to determine the stress it is necessary to
look at the recent history of strain rates. Clearly the combination of rigid particles and
Newtonian liquid can lead to some formidable complications!
I
Exercises
4.5.1 By writing out the components of the expression (4.5.10)for the local velocity
field explicitly, and comparing the result with the first two terms of eqn (4.5.9),
show that the components of the angular velocity are of the form i ( h i / a x j
- &,,axi).
xz t
P is the point (x,,
x3
Evaluating the force on the bottom of the block and then subtracting, we find that the
net force per unit volume is
@Vl *
r12e1.
8x2
If v1 also depends on x3 there will be additional shear forces on the faces with unit
normals f&,leading to an extra force
@Vl*
r12e1.
8x3
Finally, in the general case, where v1 also depends on XI, and the components v2 and v3
are non-zero, there will be additional normal stresses on the f 21 faces, which lead to a
force per unit volume:
This force, which cannot be obtained from the simple Newtonian expression (eqn
(3.4.3)), arises not from shearing of the liquid but from stretching in the x1 direction.
Additional shear stresses in the x1 direction due to the velocity gradients &/ax1 and
av3/axl lead to a force per unit volume
Thus on summing these results, we find that the net force in the x1 direction is, per
unit volume:
with similar results for the x2 and x3 direction. The viscous forces on this parcel of
fluid can then be represented in general by the term qV2v. In addition there is the
force produced by the local gradient of the pressure so the total stress force per unit
volume is:
In addition to the pressure gradient and the viscous force there will also be a
contribution from the long-range forces. We let F denote the total long-range force per
unit volume. If gravity is the only such force then:
F = pg (4.6.2)
where g is the gravitational vector, directed vertically down. The total force per unit
volume on a Newtonian liquid is therefore given by the sum
F - Vp + qV2v. (4.6.3)
The required differential equation for v is obtained by equating this term to the mass
times the acceleration per unit volume of the block and writing the acceleration in
terms of the velocity field.
During a time At the velocity of a particle originally at x changes by
&/at +v . v v . (4.6.4)
Although there will be some slight variation in the acceleration over the block, we can
to first approximation take the acceleration to be uniform. By combining the above
result with the expression (4.6.3) for the force per unit volume on the block, we obtain
:(
p -+v.VV
) =F-Vp+qV2v. (4.6.5)
170 I 4: TRANSPORT PROPERTIES O F S U S P E N S I O N S
v-v=o (4.6.6)
are known as the Navier-Stokes equations. These are the equations that govern most
of the commonly occurring fluid flows, from the large scale flows in the oceans down to
the small scale flows of interest here.
In most applications the long-range force distribution F(x) is given+ and the
unknowns are the velocity and pressure fields. T o determine these quantities for any
situation it is necessary to solve the Navier-Stokes equations subject to conditions on
the boundary of the fluid. These conditions depend on the nature of the boundaries,
which in our case are the solid particle surfaces. Experimental observations indicate
that the layer of fluid next to the surface sticks to that surface; thus the particle and the
fluid adjacent to the particle move with the same local velocity. This is referred to as
the ‘no-slip’ boundary condition.
Exercises
4.6.1 Establish eqn (4.6.4) using the fact that Axi = viAt if At is small. I
4.7 Methods for measuring the viscosity
In this section we will study some simple devices for measuring viscosity, namely the
Couette, the Ostwald, and the cone/plate viscometers. We will not attempt to
present a detailed description of the experimental procedures, but will concentrate
instead on the theory behind these devices. The discussion is made simpler by the
fact that most viscometers are designed to operate with very simple (unidirectional)
flow regimes. We can therefore analyse the flow behaviour by simple force balance
arguments, which are more transparent than the formalism involving the general
stresses on a fluid element.
Torsion
wire
Bob
Drive
(a) shaft
Fig. 4.7.1 (a) A vertical cross sectional sketch of a Couette viscometer. The liquid occupies the
shaded area between the two cylinders. (b) Isometric and cross-sectional views of a Couette
viscometer. Ri and R, are the radii of the inner and outer cylinders respectively.
and the cylinder comes to rest when the torque imparted to it by the moving fluid is
equal to the restoring torque in the wire.
When the inner cylinder stops turning the flow velocity of the fluid varies from the
inner wall (v(r) = 0) to the outer wall (v(r) = QR,) where R, is the radius of the outer
cylinder (Fig. 4.7.l(b)). The shear rate in the gap is not simply dv/dr (compare
Fig. 3.4.2) for we must make allowance for the fact that some part of the fluid motion is
a rigid body rotation which does not shear the fluid. We then have
The resulting shear stress S over the area of each cylinder of fluid generates a torque,
T given by
where L is the length of the cylinder immersed in the fluid. Under steady flow
conditions, the torque on every cylinder is the same and equal to that on the inner
cylinder, T;:
(4.7.3)
(4.7.4)
172 I 4: TRANSPORT PROPERTIES O F S U S P E N S I O N S
where Ri is the radius of the inner cylinder. On the outer cylinder, v must be equal to
Ro!2 and substituting this in eqn (4.7.4) gives:
(4.7.5)
where C is an instrument constant. The torque on the inner cylinder causes it to turn
through an angle 0 and it can usually be assumed that the angle is linearly related to the
torque even for rotations approaching 0 = n.So T; = k0 where k is another instrument
constant. The plot of 0 against !2 for a Newtonian liquid should be linear with slope
Cy/k and given the data for a liquid of known viscosity (usually water) one can then
determine the viscosity of other fluids.
The shear rate at different points in the gap is given by the bracketed term in eqn
(4.7.3) and, hence, is equal to T;/2ngL? so for a viscometer in which R, = 2R; the
shear rate at the inner wall would be four times that at the outer wall. Such a device
would not be very suitable for studying the behaviour of a liquid for which the
viscosity varied with shear rate (Fig. 3.4.9). In that case the above analysis would be
invalid since g would vary across the gap. Although it is possible to modify the analysis
to take into account the shear rate dependence of the viscosity (Krieger and Maron
1952, 1954), the problem is usually overcome by making the gap width Ro-Ri much
smaller than the cylinder radius. In that case the shear (or strain) rate is approximately
uniform across the gap (Exercise 4.7.3) and so the Newtonian analysis can be applied.
The above results can, as we mentioned earlier, be derived directly from the Navier-
Stokes equations. Normally, a problem such as this would be solved by first writing the
Navier-Stokes equations in cylindrical coordinates in order to take advantage of the
symmetry of the problem. The appropriate forms of these equations in cylindrical and
other curvilinear coordinate systems are given in standard fluid texts (see for example
Appendix 2 of Batchelor 1967). The assumption that the flow is circumferential then
leads to three differential equations. The first equation is equivalent to eqn (4.7.3),
while the remaining two have the form (Landau and Lifshitz 1959, section 18)
(4.7.6)
The last equation simply represents the hydrostatic variation of pressure with depth x,
assuming the axis of the apparatus to be vertical, while the first expression represents
the balance between centrifugal forces and radial pressure gradients in the liquid. For
the case of a rotating inner cylinder, this balance breaks down at high speed and a
steady radial flow pattern develops. The speed at which this flow begins can be
predicted from stability analysis of the Navier-Stokes equation. For the fixed inner
cylinder device the circumferential flow pattern breaks down at much higher speeds; in
this case the breakdown leads to a turbulent flow. The speed at which the turbulence
begins cannot as yet be predicted theoretically. Since the measured torques can only be
interpreted theoretically for the circumferential flow regime it is important to test any
device to ensure that torque and speed are linearly related over the experimental range
of speeds; this will ensure that there is no radial flow. Clearly, Couette devices with a
rotating outer cylinder are preferable, since the circumferential flow assumption breaks
down at high speeds for these devices. It should be noted, however, that this problem
METHODS FOR MEA SU R IN G THE VISCOSITY I173
is not so serious for liquids of high viscosity and most commercial instruments actually
use the stationary outer cylinder configuration, presumably because it is easier to
manage mechanically.
One final point on the Couette viscometer. The flow regime in the bottom of the
cylinder is different from that in the annulus and an end correction may, therefore, be
necessary. It can be assessed (Alexander and Johnson 1949) by filling the cylinder to
two different depths, L1 and L2 and determining the difference in deflection (61 - Q2)
for a given rotational speed (Exercise 4.7.5). Hunter and Nicol (1968) in their
experiments shaped the bottom of the two cylinders in the form of cones calculated to
give approximately the same shear rate as that in the annulus. This seems to eliminate
the end effect quite satisfactorily. (See Van Wazer et al. 1963, pp. 68-72 for details.)
For most viscometers of this type, the inertia term in eqn (4.7.7) may be neglected, and
we may treat the flow as time independent (Exercise 4.7.12).
From eqn (4.7.8) it follows that the pressure is uniform over the cross-section of the
tube. Thus the eqn (4.7.7) contains a term, ap/axl, which is independent of x2 and xg,
while the remaining terms are independent of XI. It follows that
(4.7.9)
where G is a constant. Since the pressure gradient is uniform down the tube, we can
write:
G = ($2 - PlVL - Pg (4.7.10)
where pl and p2 are the pressures at the top and bottom of the capillary and L is the
tube length.
174 I 4:TRANSPORT PROPERTIES O F S U S P E N S I O N S
K
i
Fig. 4.7.2 (a) The Ostwald viscometer. The volume of liquid used is such that the height of the
right hand reservoir moves symmetrically about the centre of that reservoir as the height in the left
hand arm falls from the upper to the lower mark. The constriction makes timing more accurate.
Ii I
-+
Ar'
I
Fig. 4.7.2 (b) Forces on a cylinder of fluid moving through a cylindrical capillary.
METHODS FOR MEA SU R IN G THE VISCOSITY I175
Since the tube has a circular cross-section, v1 can only depend on distance r from
the centre line. Thus eqn (4.7.7) reduces to the ordinary differential equation (see
Exercise 4.7.6)
(4.7.11)
As in the previous section, this differential equation for the velocity can also be
obtained by combining the Newtonian expression for the stress tensor with a force
balance on a suitably chosen volume of fluid. In this case the appropriate volume is a
thin cylinder co-axial with the capillary tube (Fig. 4.7.2(b) and Exercise (4.7.13)).
The solution of eqn (4.7.11) is readily found to be (Exercise 4.7.7)
v1 = --(uG 2
--Y
2
). (4.7.12)
4rl
This parabolic flow profile is referred to as ‘Poiseuille flow’. It applies not just to
capillaries but to flow down any circular tube, provided the flow rate is not greater than
the value at which turbulence sets in; for capillaries that is not usually a problem.
Integration of the velocity over the tube cross-section leads to the Hagen-Poiseuille
expression for volume flow rate, Q, viz (Exercise 4.7.7):
a
Q = /2nrvldr = -~
nGa4 npga4
-
hl
l+-.
~
[ ;
hz]
(4.7.13)
8rl 811
0
The Ostwald viscometer is very useful for obtaining accurate measurements of the
viscosity of Newtonian fluids. Its major limitation in applications to colloidal systems is
that the shear rate varies so widely across the capillary. It is necessarily zero at the axis
and at the wall it can be as high as 2000 s-l. It would therefore seem to be impossible to
use such a device for measuring the viscosity of a non-Newtonian liquid, but correction
procedures are available and these will be discussed in Chapter 15. (See, for example,
Maron e t ul. 1954.)
Fig. 4.7.3 The cone and plate viscometer. The angle is typically very small (<3" and often 4").
The cone is usually rotated but this is not always so. Rotating the plate allows the cone to be more
easily suspended on a torsion wire. In most modern instruments the upper element is attached to a
piezo-electric crystal which permits the measurement of the normal stress generated by non-
Newtonian fluids. (See Chapter 15 for more details.)
closely spaced nearly parallel solid surfaces and forms the basis of most studies on
lubricant flow (Batchelor 1967, section 4.8). For this reason this parallel-plate
approximation is called the lubrication approximation.
In this case one of the parallel plates is fixed while the other moves tangentially. Thus
the flow is locally a simple shear flow with shear rate V / h where V is the local cone or
plate velocity, depending on which of the surfaces is fixed. By using the fact that
V=rQ and hmra! (4.7.14)
where r is the distance from the axis of rotation and 2
! is the angular velocity, we see
that the local shear rate is Q/a.Thus the shear rate is uniform in this device, and it is
therefore well suited to the study of non-Newtonian liquids.
The torque or moment M due to the uniform shear stresses on the cone or plate is
readily found to be
(4.7.15)
Exercises
4.7.1 Derive eqns (4.7.4) and (4.7.5). [Hint: Consider r(d(v/r)/dr).]
4.7.2 By solving eqn (4.7.5) show that the velocity field due to a rotating cylinder of
radius R1 in an infinite liquid is given by v = R : / r where Q is the angular
velocity of the cylinder. Calculate the torque per unit length required to rotate
the cylinder.
METHODS FOR MEA SU R IN G THE VISCOSITY I177
4.7.3 Derive an expression for the strain rate in the gap of a Couette viscometer in
terms of the inner and outer radii, R; and R, and the angular velocity, Q of the
outer cylinder. Show that for very small gap widths, d, this can be reduced to
strain rate = Q(R;/d).
4.7.4 If the radius of the cylinders in a Couette viscometer is large compared with the
gap width, d, it is possible to approximate the flow in the gap by assuming it
occurs between two parallel flat plates. The bottom plate, at R1, is stationary
and the top plate, at R2, moves with velocity QR2.
(a) Show that in that case the continuity equation gives (&l/axl) = 0, v2 = vg
= 0 and that the Navier-Stokes equation reduces to
where $1 and $2 are the corresponding angles of twist of the torsion wire, and
K' is an instrument constant.
4.7.6 Assuming that the velocity v1 in eqn (4.7.9) depends only on the distance Y from
the centreline of the capillary tube, show that this equation reduces to the form
(4.7.11). [Hint: first show that
ax; -
(
d2v -%)'+**
dY2 ax2 drax;
and ar/ax2 = x2/r]
4.7.7 Establish the results (4.7.12) and (4.7.13) for Poiseuille flow assuming that the
pressures, pi are purely hydrostatic.
4.7.8 A certain Ostwald viscometer has a capillary tube of length 5 cm and diameter
0.5 mm. Calculate the time required for the level of the water in the top
reservoir to drop by 1 cm, assuming both reservoirs have a circular cross-
section of diameter 1 cm. The temperature of the water is 20 "C. You may
assume that the term (hl - h2)/L << 1 in eqn (4.7.13) for the flow rate.
4.7.9 Calculate the shear rate at the wall of the capillary tube for the viscometer
described in the previous question.
4.7.10In applying the hydrostatic formula (eqn (4.7.13)) for the pressure in the
Ostwald viscometer, we have, in effect, neglected the vV2v term in the equation
of motion in the portion of the apparatus beyond the capillary tube. We aim to
estimate the resulting error in the formula (4.7.13) for the flow rate. The order
of magnitude of the neglected qV2v term in this region is V / ( U ' )where
~ Vand
178 I 4: TRANSPORT PROPERTIES O F S U S P E N S I O N S
a’ are the typical velocity and radius of the tube. These viscous forces must be
balanced by an extra pressure gradient. With the aid of this estimate, show that the
relative error in the final formula (4.7.13) for Q is of the order of (u/u’)~(L’/L)
where L’ is the length of the viscometer tube beyond the capillary.
4.7.1 1 Hiemenz (1977) gives the following expression for the torque on the cone of a
cone/plate viscometer
4
-nR3qS’2cos a
M= 3
t a n a +$[(I +sina)/(l - sina)]}cosa‘
Show that this reduces to eqn (4.7.15) for small a.(You will need the
+
approximation In (1 x ) x for small x . )
4.7.12 Use the kind of argument displayed in Exercise (4.7.10) to show that for an
Ostwald viscometer with a flow time of order 100 s, the inertial term (i.e. the L.
H.S.) in the Navier-Stokes equation may safely be ignored.
4.7.1 3 Establish eqn (4.7.11) using a force balance argument on a cylinder coaxial with
the capillary tube (Fig. 4.7.2(b)).
These are known as the Stokes equations. They are the equations which must be
solved for the calculation of the sedimentation coefficient and the viscosity of a
suspension of particles.
For a sedimenting particle, the boundary conditions are that v = 0 on the particle
surface r = a, and v + -V far from the particle. The sedimentation velocity V is
determined from the constraint that the net force on the particle is zero. The solution
to the problem, which is described in standard texts (see for example section 4.9 of
Batchelor 1970 or section 20 of Landau and Lifshitz 1959) is given by
3a la3)
-1 +-----
2 r 2r3 (
Vcos8r+ 1 ------
:ar : ar33 )
n
Vsin88 (4.8.4)
where 8 is the angle between the direction of particle motion and the radius vector to
the point in question. Note that the disturbance velocity v +V due to the presence of
the particle decays at large distances like r-l (see Exercise 4.8.2). As we shall see, this
relatively slow drop-off leads to significant particle interactions which limit the above
analysis to very dilute suspensions.
The force on the particle due to the fluid is obtained by integrating the fluid stresses
over the particle surface which produces two terms: the bouyancy force in the
direction of particle motion and the Stokes expression (6nqaV)for the viscous drag on
the particle. The final expression for the particle sedimentation velocity is then
(compare eqn 3.1.3):
(4.8.5)
where it is assumed that the body force on the particle is (p,/p)F, pp being the particle
density.
The Stokes equations have also been solved for ellipsoidal particles; in this case the
sedimentation velocity depends on the particle orientation. Perrin (1934) has
calculated the average sedimentation velocity for a dilute suspension of ellipsoids on
the assumption that the Brownian motion has given the particles a uniform orientation
distribution.
velocity due to an isolated particle drops off on a length-scale of the order of the
particle radius, the hydrodynamic interaction between a pair of particles will only be
significant if their separation is of the order of a or less. Thus at low particle volume
fractions 4, the fraction of interacting particles should be proportional to 4, and of
these the vast majority will be interacting with only one neighbour, since the
probability of finding two (or more) particles within one or two radii of a given
particle is proportional to 8.
The average sedimentation velocity (V) for such a suspension will therefore be given
by an expression of the form
where Vo is the sedimentation velocity of an isolated particle, and the coefficient a can
be obtained from the solution of Stokes equations for a pair of interacting particles,
averaged over all possible separations. The difficulties involved in the solution of this
Stokes flow problem are so great that at present it is only possible to calculate a for
spherical particles. For the important case of force-free spheres for which Brownian
motion has made all separations equally likely it has been found that a = -6.55
(Batchelor 1972).
In Fig. 4.8.1 the measured sedimentation velocities are shown for latex
suspensions over a range of volume fractions. The broken line represents the
approximate formula
obtained by truncating the series eqn (4.8.6) at the 4 term, and using Batchelor’s
value for a. From the figure it can be seen that the approximation is accurate to
about 5 per cent if 4 < 0.05. At higher volume fractions the O(4’) term in eqn (4.8.6)
becomes significant, reflecting interactions between groups of three or more
particles. The fact that these interactions are significant at such a low volume
fraction can be attributed to the slow l / r drop-off in the velocity field of a
sedimentary particle.
The unbroken line in Fig. 4.8.1 represents the formula (Ekdawi and Hunter 1985)
[
( V ) =vo 1 - -
T’* (4.8.8)
with the parameters k1 and p set equal to 5.4 and 0.585 respectively, in order to fit the
data. The origins of this type of representation are discussed briefly at the end of
Section 4.10.
T he above formulae for average sedimentation velocity are only valid over
regions in which the suspension is macroscopically homogeneous, because in an
inhomogeneous suspension the Brownian motion, which we have so far neglected,
can lead to a flux of particles in addition to that due to the sedimentation force. T h e
problem of calculating this diffusive flux in a dilute suspension has been discussed
in Section 1.5. In the following section we will look at the case of concentrated
suspensions.
BROWNIAN MOTION REVISITED I181
0.0 I I
0.0 0.1 0.2 0.3 0.4 0.5
4
Fig. 4.8.1 The average sedimentation velocity ( V ) is a function of particle volume fraction. The
points were obtained from measurements on latex suspensions and the curve represents eqn (4.8.8)
with k = 5.4 and p = 0.585. (From Buscall et al. 1982, with permission).
r
Exercises
4.8.1 Let v ~ , pand
~ ,v2,p2 be two solutions to the continuity equation and eqn (4.8.2).
+ +
Verify that the sum v1 v2, p l p 2 also satisfies the equations. (This is a
consequence of the linearity of the equations.) Show that this result is not true if
eqn (4.8.2) is replaced by the full Navier-Stokes equation.
4.8.2 Satisfy yourself that the disturbance velocity v V = v+ +
V (cos 6' i. - sin 6'6)
where ? and 6 are unit orthogonal vectors with i. directed radially from the
particle centre. Hence show that, for large distances:
v + V = (3~/2)[V+(V/2)sin6'6].(l/r)
centre. As a result of this field the equilibrium particle density n(x) will be non-
uniform. There will therefore be a flux due to Brownian motion, which will be
balanced by the sedimentation flux. As mentioned in the previous section, these two
fluxes can be calculated separately and superposed.
The sedimentation flux is obtained by multiplying the concentration by the velocity
(recall Fig. 1.5.3):
where n is the particle number density and K(4)/6nqa is the average sedimentation
velocity in a suspension in which the particles are acted on by a unit force. Comparing
with the formula (4.8.7) for the sedimentation velocity for rigid spheres in non-
concentrated suspensions, we find
(4.9.2)
For more concentrated suspensions (4 > 0.05), the empirical expression (4.8.8) can be
used for calculating K.
The Brownian flux will presumably depend on the local variations in particle
number density, variations that are characterized by the quantity Vn. By similar
arguments to those that lead to the Newtonian formula (eqn (4.5.7)) between stress
and rate of strain it can be shown that the flux density due to Brownian motion has
the form -2)Vn for a locally isotropic suspension. This is identical to Fick's first law
(eqn (1.5.17)), but now we must allow for the fact that 2) depends on the local
particle volume fraction.
In equilibrium, the net particle flux is zero, so
(4.9.3)
We also have the constraint that the net body force -n(x) f per unit volume of
suspension must be balanced by the surface forces on the volume.+ Since the
suspension is in equilibrium, the surface forces must take the form of an osmotic
pressure ll, acting normal to the surface. The force balance equation therefore takes
the form -n(x)f = Vll or, since the local osmotic pressure depends only on particle
density in this case:
dll
-n(xy = ~ Vn. (4.9.4)
dn
Using this equation to eliminate the nfterm in eqn (4.9.3) we find that the diffusivity is
given by
(4.9.5)
Thus the diffusivity can be calculated from sedimentation and osmotic pressure data
for the suspension. On replacing K(4)in the above expression by the formula (4.9.2)
and using the approximate result
where Do = kT/6nrla is the Einstein formula for the diffusivity in a dilute suspension,
and terms of order have been neglected. It can be seen that the effects of particle
interaction on sedimentation velocity and osmotic pressure nearly cancel out, leaving a
relatively weak diffusion-concentration dependence. The experimental verification of
eqn (4.9.7) is described in Russel’s (1981) review article.
In a non-equilibrium suspension, differences between the sedimentation and
diffusion fluxes lead to variations in the concentration n, variations that can be
calculated using the equation
where F is the net sedimentation force on the particle. The derivation of this equation
follows similar lines to that of eqn (1.5.20) for one-dimensional diffusion in a dilute
suspension.
For suspensions of non-spherical particles, and for suspensions of spheres at high
concentrations, the quantities K(4) and dll/dn must be determined experimentally.
Once these quantities are known, eqn (4.9.8) can be used for the prediction of
sedimentation behaviour in any situation. This approach has been successfully applied
to a number of colloids by Philip and Smiles (1982).
The formula for the tracer diffusion coefficient in a suspension of spheres is found to
be (Batchelor 1976)
2, = (1 - 1.83@)kT/6nva. (4.9.9)
Not surprisingly, the Brownian motion of a particle is hindered by interactions with its
neighbours (Russel 1981, p. 437).
for the velocity V of an isolated spherical particle. This is simply Newton’s second
law for a particle of mass m acted upon by a fluctuating Brownian force F(t),with the
liquid drag represented by the Stokes formula -6nqaV. Since this formula was
derived in Section (4.8.1) for a particle in steady motion, its application to the
present problem can only be justified if the inertia term in the equation of motion of
the liquid (eqn (4.8.2)) is negligible. After we have obtained our solution to the
Langevin equation, we must therefore check for consistency by estimating the size
of this neglected inertia term.
Since the particle is much more massive than the surrounding molecules, its
response time will be much greater than the time scale of the fluctuating force F(t),
which will presumably be of the order of the relaxation time for water molecules
(-10-13 s). Thus the velocity of a particle at any time is determined not just by the
instantaneous force on the particle, but by the history of that force over several particle
relaxation times. In order to estimate this relaxation time, we take the average of the
Langevin equation over all those particles that have a given velocity V at t = 0. By the
above argument, we expect that the particle velocity and the force will be statistically
independent. Hence the average of F(t) over this group of particles will be zero, and
the average of eqn (4.9.10) becomes
where t,, = m/6nqa. From eqn (4.9.12) it can be seen that <V> M 0 if t is large
compared with t,,. Since the average velocity of a random sample of particles is zero, t,,
provides a measure of the time required for a particle to ‘forget’ its initial velocity. For
a 0.1 p m radius particle of density 2 g cm-3 in water at 25 “C, t,, = 5 x lop9 s. For the
BROWNIAN MOTION REVISITED I185
random walk analogue of Chapter 1 to be valid, the step times t must be much greater
than this relaxation time.
We are now in a position to test the validity of the Langevin equation by estimating
the size of the neglected ‘inertia’ term p(av/at) in the equation of motion (eqn (4.8.2))
of the liquid. Since the particle velocity decays from its initial value VOover a time t,,,
this inertia term is of order pVo/t,,. The viscous term qV2v will be of order qVo/a2,
and so the ratio
(4.9.13)
From eqn (4.9.12) we see that this ratio is of order p/p,, where p,, is the particle
density. So for cases of practical interest, in which the particle density is not large
compared to the fluid density, the Langevin equation is invalid!
The correct analysis must be based on the equation of motion (eqn (4.8.2)), which
takes into account the inertial forces in the liquid, together with a fluctuating ‘body
force’ F,representing the effects of Brownian motion. As in the Langevin equation,
this fluctuating term can be removed by averaging. The solution of these equations for
the average velocity (V) yields a similar estimate for to,but the decay in (V) at large t is
found to have a t-3/2form, instead of the exponential form found earlier (Russel 198 1,
p. 428). Thus while the Langevin equation may not be strictly valid, it provides
qualitatively correct results, and serves as a simple model for illustrating the techniques
used in the analysis of the complete equations describing Brownian motion dynamics.
where 6 is the change in the particle orientation during time t, and Dr is termed the
rotational diffusion coefficient. This result is valid provided the change in orientation
is small (see Exercise 4.9.2).
The effect of Brownian motion on the particle orientation distribution in a dilute
suspension is described by the differential equation (McQuarrie 1976, p. 398).
aR
-
at
= Dr[ ~- (.
1 a sin0-a
sin 0 29
~ +--
) 1 a
sin20
2~] (4.9.15)
where the angles 0 and 4 used to specify the orientation of the particle axis are
illustrated in Fig. 4.9.1. By definition, the quantity R(0, 4, t) sin 0 A0 A 4 is equal
to the fraction of particles whose orientation lies in the range (0 f (A0/2),
4 f A4/2)).
186 I 4: TRANSPORT PROPERTIES O F S U S P E N S I O N S
Fig. 4.9.1 The angles 0 and 4 used to specify the orientationof a particle, relative to a fixed set of
Cartesian coordinates.
By multiplying eqn (4.9.15) by cos 8 and integrating over all orientations it can be
shown that
for a suspension in which all the particles are initially aligned in the 0 = 0 direction.
Equation (4.9.16) refers to what is called Debye relaxation, since it was used by Peter
Debye (1929) to describe the relaxation behaviour of molecular dipoles. From eqn
(4.9.16) it can be seen that the time required for the particle distribution to become
isotropic (i.e. Ccos 8> M 0) is of order l/D,..
The quantity Dr can be calculated using a variation of the Einstein argument
described in Section 4.9.1, that is by considering a suspension in equilibrium under the
action of an applied field which in this case tends to align the particles. The applied
field leads to an extra term in eqn (4.9.15) corresponding to the rotation of the particles
in the absence of Brownian motion. The solution of eqn (4.9.15) for this equilibrium
problem yields:
Dr = k T / B , (4.9.17)
where B, is the friction factor for particle rotation, equal in magnitude to the torque
required to rotate the particle with unit angular velocity, in the absence of Brownian
motion. From the solution of the Stokes equations around a rotating spheroid it is
found that (Perrin 1934):
for a cigar-shaped (prolate) spheroid, where q = b/u is the ratio of the length of the
minor axis to the major axis (the axis of rotation of the particle). The formula for disc-
shaped spheroids is given in Exercise 4.9.5.
BROWNIAN MOTION REVISITED I187
where BIIand BI are the friction coefficients for motion parallel and perpendicular to
the particle axis respectively. The formulae for these quantities are given in Perrin’s
(1934) paper.
With the aid of these formulae and the expression for the rotational diffusivity it is
possible to estimate the particle size and shape from combined translational and
rotational diffusivity measurements on dilute suspensions. The results relate of course
to the size of the particle in solution, and this may differ from that of the dry particle as
a result of solvent absorption or adsorption. This method has mainly been applied to
protein and virus particles, but some work has also been done on montmorillonite
suspensions (Shah 1963).
The extension of theoretical and experimental studies of translational and rotational
Brownian motion and diffusion to intermediate and higher concentrations has to date been
limited to spherical particles (Degiorgio and Piazza (1996) and Degiorgio et al. (1994)).
r
Exercises
4.9.1 Establish eqn (4.9.12).
4.9.2 By using a suitable approximation for small values of 8 in eqn (4.9.16),derive eqn
(4.9.14) for the mean squared orientation change of a particle during a small
period t.
4.9.3 Calculate the value of Vr, for values of u / b from 1 to 20 and a = lop7 m using
eqn (4.9.18). Show that for large a/b, this formula takes the approximate form
Compare values obtained with this formula with the exact value obtained earlier.
4.9.4 Alexander and Johnson (1949; p. 400) quote a value of Vr = 7 s-l obtained by
Edsall for the rotary diffusion constant of rabbit myosin at T = 276 K. Taking
the viscosity of water as 1.6 x lop3 Pa s at this temperature and a high value (say
u / b M 100) for the axial ratio, estimate the length of the myosin molecule.
(Independent measurements of the particle volume would permit a further
refinement of this result.)
4.9.5 For an oblate (disc-shaped) particle the rotational diffusivity Vr,is given by
RT
v - 16nyab2(1
- - 8)
[3&(2 - &)(q2 - 1)-f arctan(& - 1)f +3 1
Show that for q >>1 (i.e. a flat disc): V,.% 3kT/32yb3
188 I 4: TRANSPORT PROPERTIES O F S U S P E N S I O N S
where the sample volume Vis centred on the point x.In a flowing suspension this average
velocity will vary on a macroscopic length scale L determined by the apparatus; in flow
down a pipe for example, L will be of the order of the pipe radius. Thus in order to give
formula (4.10.1) an unambiguous meaning we must specify that the sample volume Vbe
much smaller than L3,while still being large enough to contain many particles.
Differential equations for the macroscopic velocity field v(x) can be derived in a
similar manner to the Navier-Stokes equations for the local velocity field; that is by
applying the Principle of Conservation of Mass and Newton’s Second Law to a
rectangular block of fluid. However, whereas in the earlier sections we took this block
to be much larger than the molecules and much smaller than the particles, the block in
this case is taken to have the dimensions of a sample volume, as defined above.
The most difficult step in the derivation of these differential equations for the
macroscopic velocity field is the determination of the relationship between the short-
range forces that act on the surface of the block and the macroscopic velocity field; that
is, the analogue of the Newtonian expression (eqn (4.5.8)) for the microscopic tensor in
a liquid. The remainder of this section is concerned with the derivation of this
relationship.
As in the case of the microscopic stress tensor, we can write (cf. eqn 4.5.2):
() = -PI + (D)
where P is an average pressure. CD> is the part of <> that depends on the motion. By
using the fact that = 2qe where e is the rate of strain tensor in the liquid, (c.f. eqn
(4.5.8)) we can write eqn (4.10.2) in the form
( ) = - P Z + 3V/ e d V + d / D d V (4.10.3)
VI VP
where V, is the volume occupied by the particles in V, and 6 is the liquid volume.
The rate of strain tensor e is zero inside the rigid particles (see Exercise 4.10.1), and
so the integral over 6 in eqn (4.10.3) can be formally extended over the particle
volume as well, resulting in
where
(e) = JV/ e d V (4.10.5)
V
is the average rate of strain tensor. As before, n denotes the particle number density,
and <S> denotes an average, over the particles in V, of the quantity
/ Dd V (4.10.6)
V’
where Vi is the volume of the ith particle in V. <S> is referred to as the average
particle dipole strength. The term ‘dipole strength’ arises from the application of this
volume averaging technique to the electrical transport properties of suspensions.
There each uncharged particle behaves, when viewed from a distance, like an electric
dipole with a dipole strength given by an integral form analogous to eqn (4.10.6).The
average dipole strength in (4.10.4)represents the particle contribution to the stress. T o
calculate this term we must solve the Stokes equations around the particles. In the
following section we will outline the solution to this problem for a dilute suspension of
spheres.
surface of the particle is the same as the particle velocity - the no-slip boundary
condition.) In this case it turns out that a spherical particle will remain at rest in the
ambient field and if it is rigid, the resulting value of the dipole strength is given by
(Landau and Lifshitz 1959):
S = (20/3)m3q(e). (4.10.7)
In the absence of particle interaction, the dipole strength is the same for each sphere in
the sample volume, and the formula (4.10.4) becomes
This has the same form as the expression (eqn (4.5.8)) for the stress in a Newtonian
liquid. Thus from the macroscopic point of view, a dilute suspension of spheres
behaves as a Newtonian liquid with viscosity q* where
This result was obtained by Einstein? from a calculation of the energy dissipation in
the suspension. This dissipation method suffers from the disadvantage that it yields
only a single number, namely the rate of energy dissipation. While this may be
adequate for isotropic suspensions, which can be characterized by a single viscosity, it
fails when applied to a non-isotropic suspension such as a suspension of rod-like
particles aligned by a flow. Such suspensions require a number of ‘viscosities’ to
describe their flow properties. For this reason, the recent papers on this subject have
adopted the volume average approach described here, an approach first devised by
Landau and Lifshitz (1959).
t Eistein initially calculated the coefficient of q5 as unity but one of Perrin’s students
measured the effect and found it fell between 2 and 3. Perrin alerted Einstein who
revised his analysis and obtained 2.5.
T H E FLOW P R O P E R T I E S OF S U S P E N S I O N S 1191
valid for those suspensions, and since the suspensions are dilute, we can again calculate
the dipole strength of each particle as if it were alone in an infinite liquid. This
calculation involves the solution of the Stokes equations subject to the same boundary
conditions as for the spherical particle. As before, the dipole strength due to the rigid
body component of the ambient field is zero and the particle simply translates and
rotates with the field. Unlike the spherical particle however, the dipole strength of the
spheroid in the ambient straining field is found to depend on its orientation relative to
the field, and furthermore this straining field induces an angular velocity in the particle
that depends on particle orientation. This rotation is superposed on the rigid body
rotation described earlier. Thus spheroidal particles do not simply rotate with a
uniform angular velocity like spheres, but move in a more complicated fashion that
depends on the type of flow. For example, in a shear flow it is found (Okagawa et al.
1973) that cigar- shaped particles rotate in a periodic fashion, with the minimum rate of
rotation occurring when the particle is most nearly aligned with the streamlines. In
other flows such as the extensional flow of a stream of liquid from a hole in the bottom
of a container, there is a preferred orientation (also along the streamlines in this case) at
which the particles have zero angular velocity. Such flows tend to align the particles in
the preferred direction. Since the average dipole strength depends on the orientation
distribution, R (0, $), the same suspension may behave quite differently for these
different flows.
T o calculate the distribution function R for any flow it is necessary to solve eqn
(4.9.15) with an extra term due to the rotation of the particles by the ambient flow.
Since the rotation rate is non-uniform, the ambient flow tends to make the distribution
function non-uniform. For example, in the shear flow case described above, the
particles move more slowly when they are aligned with the streamlines, and so R will be
larger for the aligned orientation. This aligning tendency is opposed by the Brownian
motion term in eqn (4.9.15), and the final form of the distribution function in a steady
flow is determined by the relative magnitude of these two opposing terms. In a shear
flow this quantity is characterized by a ‘Piclet number’ p/Dr. When the ratio is small,
the Brownian motion dominates and the orientation distribution is approximately
uniform. At large Piclet numbers the distribution is quite non-uniform, but rather
surprisingly the Brownian motion still plays an important part in the shear flow case,
for in the absence of Brownian motion the particles simply rotate in a periodic fashion
leading to an orientation distribution R that varies periodically. In such a situation the
effect of a small amount of Brownian motion over a long period of time results in a
steady-state orientation distribution.
The orientation distribution has been calculated for a number of limiting cases such
as nearly spherical particles at low or high PCclet numbers, for a range of flows (Leal
and Hinch 1973). Once the orientation distribution is known, the average dipole
strength can be calculated directly, but even at this stage there is an unexpected
complication, for in addition to the dipole strength obtained from the solution of the
Stokes equations in the ambient flow there is a direct ‘contribution’ from the Brownian
motion. The dipole strength of a particle undergoing Brownian motion will, like its
angular velocity, depend on the history of the fluctuating torque on the particle over
the previous few relaxation periods. In a suspension in which the orientation
distribution is non-uniform, the particles that instantaneously have a given orientation
represent a biased sample, since more will have come (via Brownian rotations) from
192 I 4: TRANSPORT PROPERTIES O F S U S P E N S I O N S
neighbouring orientations where R is large than from those where R is small. Thus on
average the particles will have experienced more torques in one direction that in
another, and as a result their average dipole strength due to these torques is non-zero.
Although this extra dipole strength is the consequence of fluctuating torques it can in
fact be calculated from the solution of Stokes’ equations around a spheroid rotating
under the action of a steady ‘thermal torque’ analogous to the thermal force introduced
in Section 1.5.2 (see Hinch and Leal 1972).
Since the orientation distribution depends on the type of flow, it is difficult to make
general statements about the flow properties of these suspensions; consequently we
will limit our attention to the results obtained for the important case of a steady shear
flow.
Even in this restricted case it is not possible to describe the flow properties in terms
of a single viscosity, for when a suspension is subjected to the idealized shear flow of
Fig. 3.4.2, the top plate would experience both a shear force and a normal force, both
depending on the shear rate. We will concentrate here on the calculation of the ‘shear
viscosity’, which represents the ratio of shear stress to shear strain rate.
It is found that the shear viscosity of the suspension, r]*, decreases with increasing
shear rate. Although this viscosity dependence cannot be calculated for arbitrary
particle shapes, the limiting viscosities at small and large Piclet numbers have been
calculated for arbitrary aspect ratios ~ ( = a / b (Hinch
) and Leal 1972). These results are
illustrated in Fig. 4.10.1, when [ r ] ] is the ‘intrinsic viscosity’ defined by
(4.10.10)
The broken lines represent approximate formulae valid for disc-shaped (la<< l), and
rod-like ( ~ > > l particles.
) With the aid of these results it should be possible to obtain
information about particle shape from measurements of the high and low shear rate
viscosities.
I I I I I I I I
I
1
300 -
/
100 - _._
32 1
I I I I I I I I 1
0.01 0.03 0.1 0.3 1 3 10 30
w=llq
Fig. 4.10.1. The high and low shear rate intrinsic viscosities of a suspension of spheroidal particles.
From Figs 5 and 8 of Hinch and Leal’s (1972) paper.
THE FLOW PROPERTIES OF S U S P E N S I O N SI193
where SOis the dipole strength of an isolated sphere, and S1 is related to the dipole
strength of one of a pair of particles in the ambient flow, averaged over all pair
configurations.
T o calculate this average it is necessary to determine the pair probability function
p(r), where p(r) is the probability of finding a particle with its centre in the small
volume A V around the point r, given that there is a sphere at the origin. p(r) is the
analogue of the orientation distribution R (@,$)for the dilute spheroid problem and,
like the orientation distribution, p(r) depends on the type of flow.
In Fig. 4.10.2 we show the relative trajectories of pairs of spheres in a shear flow
(Batchelor and Green 1972~).The lines represent the paths traced out by the centre
of one member of the pair relative to the other for the case of pairs lying in the plane
of the flow. Since the path lines are symmetrical about the xz and XI axes, only one
quadrant has been shown in the figure. If the ambient shear rate av?/axz is negative,
the second particle will move from right to left along the path lines in the diagram.
The quantity RZrepresents the xz coordinate of the particle far upstream, where the
path lines are parallel to the XI axis. From the diagram it can be seen that the flow
tends to bring sphere pairs together. For the case Rz/a C 1, the pairs are brought
into such proximity that even a slight attractive force may be enough to bring about
‘shear-induced’ coagulation.
2
32
2
I
R;la2=9
0 1 2 3 4 5
xlla
Fig. 4.10.2 The trajectories of one member of a pair of spheres in a shear flow, relative to the other
sphere, for pairs lying in the plane of the flow. The circle of radius 2 is the trajectory for touching
particles. (From Batchelor and Green 1972a, with permission.)
194 I 4: TRANSPORT PROPERTIES O F S U S P E N S I O N S
Those trajectories that lie below the R2 = 0 line represent pairs in closed orbits. In
the absence of Brownian motion the pair distribution function in that region will vary
periodically with time, like the orientation distribution for the spheroid problem. For
such particles Ri/a2 < 0 so there are no (real) values of R far upstream of the particle.
The difference between open and closed streamlines is shown more clearly in
Fig. 4.10.3 which again has the origin at the centre of particle 1. It shows the limiting
trajectory and the minimum separation distance which can be very small (< 1 nm).
This distance of closest approach can be used to estimate the roughness of the particle
surface (Arp and Mason 1977).
In extensional flows there is also a tendency for pairs to be brought together, but
there are no closed orbits. These flows lead to an increase in the probability p(r) at
small separations, a tendency that is of course opposed by the Brownian motion.
For a given flow the form of p(r)is determined by the relative magnitude of these
two opposing terms, characterized by the Piclet number j.a2/D, where is a
Fig. 4.10.3 Equatorial trajectories of two spheres in simple shear flow (schematic). The solid lines
are the relative trajectories of a sphere of radius a2 with respect to the reference sphere. The
minimum approach distance, &in, may be less than 1 nm. (After van de Ven 1989, p. 362 with
permission.)
THE FLOW PROPERTIES OF S U S P E N S I O N SI195
Fig. 4.10.4 The effect of particle crowding on viscosity. The broken line is drawn according to the
Einstein eqn (4.10.9) and indicates that these silica spheres show the theoretical intrinsic viscosity of
2.5. The points are experimental and the full curve is drawn using the polynomial expression
(4.10.12) with b = 5 and c = 53. (From Jones et al. 1991 with permission.)
and achieved best fit with b = 5 and c = 53. The fit (Fig. 4.10.4) what is more, is
excellent up to about 4 = 18%.
At high Piclet numbers the problem of determining p(r) in a shear flow is
complicated by the closed trajectories referred to earlier, for in the region of closed
trajectories a small amount of Brownian motion can have a significant effect on p(r),in
the same way that the Brownian motion has a significant effect in the spheroid case at
high shear rates. In the extensional flow, where there are no closed trajectories, the
effect of the Brownian motion is unimportant. The calculation of the bulk stress in this
flow is described by Batchelor and Green (19726).
For more concentrated suspensions there are a number of semi-empirical formulae
that can be used for calculating the shear viscosity. The data for the limiting shear
196 I 4: TRANSPORT PROPERTIES O F S U S P E N S I O N S
viscosities at high and low shear rates (that is, high and low Piclet numbers) can be
adequately represented by the Dougherty-Krieger formula
(4.10.13)
Sq* = 2.5q*.S@.
In order to take into account the fact that the suspension is rigid when the particles are
closely packed, S4.is replaced in the above formula by S$/( 1 - K4) where 1 - K+ is the
volume available for the added particles. Integration of the resulting formula yields
(Exercise 4.10.2):
(4.10.14)
Fig. 4.1 0.5 The low shear rate viscosities of concentrated suspensions of the same silica spheres as
shown in Fig. 4.10.4. The curve represents the Krieger-Dougherty eqn (4.10.13) with [qlp = 2 and
p = 0.631. Note that the relative viscosity is here shown on a log scale to accommodate values up to
100 000. The intrinsic viscosity used here is 2/0.63 = 3.17 which is a little higher than the measured
value (Fig. 4.10.4) as is indicated by the fact that the curve runs a little above the data points at the
lowest volume fractions. The fit is, however, remarkably good. (Data from Jones et al. 1991.)
THE FLOW PROPERTIES OF S U S P E N S I O N SI197
where p = 1/K is expected to be approximately equal to the volume fraction for close
packing. In practice, p is treated as an adjustable parameter, and the 2.5 is replaced by
the measured intrinsic viscosity [q], which can differ from the Einstein value due to
effects such as non-sphericity, particle charge, and the presence of small numbers of
permanent doublets. Equation (4.10.14) is an example of an effective medium
relationship. There are a number of these very useful formulae which can be used
to describe the phenomenology of, for example, flow and sedimentation, where the
more formal theoretical relations are very limited in their range.
From Fig. 4.10.5 it can be seen that the formula (4.10.13) provides an excellent fit
of the measured low shear rate viscosities of a suspension of silica spheres up to
concentrations near to the close-pack limit. In this case the value of [qbwas 2 and
p = 0.631. The same relation was used by Krieger to describe successfully his low and
high shear rate viscosity measurements for latex suspensions. In both cases [q] = 2.67,
while p = 0.57 for the low shear and 0.68 for the high shear viscosity. For such
suspensions, the viscosity at intermediate shear rates can be calculated using an
equation of the form (Krieger 1972)t:
rl* - rT
-.=(I+%) IS1 -I
(4.10.15)
r12 - rll
where r l l * and rl2 * are the high and low shear limiting viscosities respectively. S is the
shear stress, and Si is a characteristic shear stress related to the particle diffusivity
(Exercise 4.10.4):
- Predicted
10
,P
b 9
7
0.01 0.03 0.1 0.3 1.0 3.0 10 30
3.5
0.01 0.03 0.1 0.3 1.0 3.0 10 30
S,
Fig. 4.10.6 Measured shear viscosities of concentrated latex suspension as a function of non-
dimensional shear stress. Si = Sa3/kT.The volume fraction is 0.35 for (a) and 0.45 for (b). The lines
represent eqn (4.10.15)with a = 0.431. (From Krieger 1972, with permission.)
Exercises
4.1 0.1 In deriving the formula (4.10.4) for the macroscopic stress tensor in a
suspension of rigid particles, we used the fact that e = 0 inside the particles.
Establish this result by calculating the strain tensor for the rigid body velocity
field V + x x.
4.1 0.2 Establish eqn (4.10.14) using the suggested integration procedure.
4.10.3 Show that Si in eqn (4.10.15) is proportional to Q / a 2 where V is the particle
diffusion coefficient.
4.10.4 Calculate S, for the suspensions shown in Fig. 4.10.5 assuming the particle
radius is 100 nm and temperature is 25 "C.At what shear rate is the suspension
viscosity half way between ql * and q 2 * ? What do ~ l and * QZ * represent?
References
Alexander, A.E. and Johnson, P. (1949). Colloid science. Oxford University Press,
Oxford.
Arp, P.A. and Mason, S.G. (1977). J. Colloid Interface Sci. 61, 44.
Batchelor, G.K. (1967). An introduction to fluid dynamics. Cambridge University
Press, Cambridge.
REFERENCES 11%
200
GENERAL CONSIDERATIONS 1201
This chapter concerns itself first with the direct observation of the sizes and shapes of
particles commonly found in colloidal systems. Since the range of sizes in a suspension
is often very considerable we next consider how best to represent the distribution of
particle sizes by some convenient algebraic formula. Some of the more common
methods of particle size analysis are then described. The theoretical basis of each
method is given in sufficient detail to appreciate its scope and limitations but to obtain
practical experimental details of any particular method the reader should consult the
more specialized manuals (Kissa 1999, Allen 1990, Barth 1984) or reviews (Miller and
Lines 1988). This is an area of active instrument development and the range of
instruments and their characteristics are constantly changing. It is hoped that the
principles developed below will help the reader to safely navigate between the claims
and counterclaims for the various technologies.
I A X~S ofrotation
t
h
Fig. 5.1 .I (a) An oblate spheroid, obtained by rotating an ellipse around its short axis. The cross-
section in the plane where b is measured is circular. (b) A prolate spheroid, obtained by rotating an
ellipse around its long axis (like a football). Sections parallel to b are circular. (c) A disc may be
approximated as an oblate spheroid, a cylinder (d), or an ‘infinite’ flat plate (e).
Fig. 5.1.2 (a) Electron micrograph of calcium carbonate (calcite) crystals (normal habit). (b)
Crystals of the same calcium carbonate obtained from a solution containing an organophosphorus
crystal habit modifier. (Photographs obtained from the work of Leonard Dubin 1980, of Nalco
Chemical Company, Illinois, which markets these modifying compounds.)
+These are sets for which the concept of membership is fuzzy (i.e. one may not be
certain whether an object does or does not belong to the set). Nevertheless, they can be
handled logically with the help of fuzzy set theory.
204 I 5 : PARTICLE SIZE AND SHAPE
more conventional treatments of the subject are given in the standard works by Orr
and Dallavalle (1959) and Herdan (1960). A brief description of the more sophisticated
techniques for describing shape is given by Sutton (1976); the use of Fourier series is
described there and by Beddow et al. (1977) while Allen (1981) and Kissa (1999)
describe the commonly used shape coefficients and shape factors. For more regular
particles (discs, rods, and ellipsoids), Jennings and Parslow (1988) have given a general
account of the relations between the dimensions measured by different analytical
techniques, and the equivalent spherical diameter of the particle (Section 3.1.1). An
excellent introduction to the notion of fractals and their significance in chemistry
(including some reference to particle size and surface area measurement) is given by
Harrison (1995). Space here permits only a very brief outline of the methods of
treating irregular particles.
Finally, it must be emphasized that different methods of particle size analysis will
almost certainly yield different results but that may not mean that one method is
superior to the other. It may reflect the fact that different characteristics of the
particles are being measured. Combining the results from different measurements can
yield more significant information about the particles. A most striking example is
provided by Jennings (1993) who combined data from rotary diffusion and
sedimentation to estimate the axial ratio of disk-shaped particles of kaolin.
where no is the refractive index of the medium and 20 is the angle subtended by the
microscope objective at the focal plane (Fig. 5.2.l(a)).
Resolving power can evidently be improved (i.e. dp decreased) by reducing the
wavelength, or increasing no and 8. In practice only visible light is used (A-500 nm),
but no can be increased by filling the region between the lens and the sample with a
transparent oil (no % 1.5) instead of air. (A drop of immersion oil is placed on top of
the microscope cover slip and the ‘oil-immersion’ lens lowered into it.) Wide angle
lenses are also helpful (i.e. increased 0) but the angle is limited by other optical
problems like spherical and chromatic aberration. In effect the lower limit for dp is
about 0.2 pm, so optical microscopy is limited to the upper end of the colloidal size
range.
There are now available numerous commercial devices for automatically scanning
and analysing images obtained from light (and other forms of) microscopy. See Allen
(1997, pp. 13740) and also Kissa (1999) Chapter 10.
I
Microscope
- objective
- lens
I
I
+d+
a
C
b e
Fig. 5.2.1 (b) The slit ultramicroscope. An intense beam of light (from an arc, Xenon lamp, or laser)
a, emerges through a slit, b, and is focused by a lens, c, into a chamber, d, containingthe colloidal sol.
The light scattered from the particles can be viewed though the microscope, e, against a dark
background.
Fig. 5.2.2 (a) Schematic diagram of the electron microscope. (For description see text.)
208 I 5 : PARTICLE SIZE AND SHAPE
Fig. 5.2.2 (b). Shadowing (or shadow-casting) of a spherical or cylindrical particle produces a
characteristic intensity pattern in the transmitted beam, since the shadowing material (usually a
metal) is a stronger absorber - scatterer of electrons. I is the intensity (or number) of transmitted
electrons.
specimen in proportion to the local electron density, and the remainder are
transmitted.
An electromagnetic objective lens, L2, collects the transmitted electrons and
magnifies the image of the specimen 10 to 200 times, into the object plane of a magnetic
projector lens system (L3), which induces a further magnification of 50 to 400 times as
it projects the electrons onto a fluorescent screen, S. There the image may be viewed
directly or photographed with a fine-grained film to be enlarged a further 5 to 10 times.
The overall magnification factor ranges from about 100 to 1 000 000 times. Since the
human eye can discriminate between points separated by about 0.2 mm, the lower limit
of observation is then from 2 p m down to about 0.2 nm, so the instrument can resolve
individual atoms in favourable cases.
The image formed by this procedure is a two-dimensional representation of the
actual structure, and in some cases this is all that is required. In many cases, however, it
is helpful to have an idea of the surface topography and this is most readily achieved by
shadow-casting. A beam of metal atoms is fired, En vucuo, at an angle to the sample and
the deposit modifies its electron transmission characteristics. Figure 5.2.2(b) shows
schematically how the intensity might be expected to vary in the case shown. Although
the human eye will quickly interpret the resulting image in terms of a particle shape it
is important to realize that the pattern of lightness and darkness on the screen or
photograph is the result of a complex sequence of events and an exact shape analysis
may call for more detailed consideration of the influence of shadowing angle 0 and
direction on the apparent shape. Metals such as gold, platinum, palladium, nickel, and
chromium may be used for shadow casting.
D I R E C T MICROSCOPIC OBSERVATION I209
+PronouncedOh-zhay.
210 I 5 : PARTICLE SIZE A N D SHAPE
Fig. 5.2.3 Scanning electron micrograph of small (-0.1 pm) polystyrene (PS) particles
adsorbed on larger PS particles (-1 pm). (Photograph courtesy of Professor Brian Vincent,
University of Bristol.)
1. Martin’s diameter (&): the length of the line that bisects the image of the particle.
The lines may be drawn in any direction, but the direction must be maintained
constant for all the image measurements.
2. Feret’s diameter (df):the distance between two tangents on opposite sides of the
particle, parallel to some fixed direction.
3. The projected area diameter (d,): the diameter of a circle having the same area as
the particle, viewed normally to a plane surface on which the particle is at rest in a
stable position. (This is usually assumed to be the case for electron micrographs,
since the drying process would be expected to favour a stable particle orientation.)
There are differences of opinion over the usefulness of the Martin and Feret
diameters, although various relationships have been derived or experimentally
established between them and d,, for different materials. Thus, the ratio dJdf has
characteristic values (a little greater than one) for different ground solids and can be
used as an empirical measure of shape. For very irregular particles, these diameters will
depend strongly on the orientation of the particles with respect to the choice of
measurement direction.
Fig. 5.2.4 (a) Martin’s diameter shown (d,) Feret’s diameter shown (df). (b) The projected area
diameter, d,
212 I 5 : PARTICLE SIZE A N D SHAPE
The projected area diameter, d,, (Fig. 5.2.4) can be fairly readily measured, either using
a graticule or a semi-automatic procedure. The semi-automatic procedures, though still
tedious, are somewhat faster. The Zeiss-Endter particle sizer (Allen 1975, p.141), for
example, is best suited to analysing photomicrographs. It projects onto the photograph a
circle of light, the diameter of which is determined by an iris diaphragm controlled by the
operator. When the circle is of the correct size the operator depresses a switch, which
actuates one of a number of counters, each of which is associated with a pre-set size range.
The machine also marks the particle with a pin-hole to avoid double counting. The entire
image is projected onto a frosted screen for easier viewing. The main advantage claimed
for this instrument and the other similar ones described by Allen (1981) is that they
permit the operator to exercise some judgement, both in particle selection (in
heterogeneous systems) and in deciding what to do about overlapping particles.
Modern automatic imaging procedures measure d, and df and other length
characteristics in many directions; the average and the scatter then give much more
information about the particles. This is clearly an area where the h t h e r development of
microprocessor discrimination and control will prove increasingly effective. A related
procedure is that of chord measurement in which a laser beam traverses a sample of
particles (usually in air) and the length of the chord cut by the beam is calculated from the
back-scattered light. The mean chord length, C, is related uniquely to the area-to-
perimeter ratio for any particle outline or m a y of outlines (Scarlett 1997):
C = RA/P. (5.2.2)
The three dimensional equivalent of this equation relates the average chord length
(measured in random directions) to the volume to surface ratio:
c = 4 v/s. (5.2.3)
This relation applies to any array of particles with any shape and is a convenient way of
measuring parameters such as V / S . This ratio can also be estimated by combining a
sizing method, such as light scattering which depends on volume, with one which
depends on surface area (like gas adsorption).
Exercises
5.2.1 Show that the average radius, of spherical particles is given by:
?; = (3cV/4npN3
where p is the particle density, c is the concentration of the sol (by mass) and it is
found to contain N particles in a volume V.
5.2.2. The wavelength associated with an electron is given by the de Broglie relation, h
= h / p = h/(2meE)4, where h is Planck’s constant, p is the electron momentum,
meits mass, and E its kinetic energy. Estimate the wavelength of an electron that
has been accelerated through a voltage of 10 kV so that it has acquired an energy,
E, of 10 keV. [A typical acceleration voltage would be of this order.]
PARTICLE SIZE DISTRIBUTION I213
r-
I
I
!/
i 4
-7- -
II
Fig. 5.3.1 (a) The frequencyhistogram can be replaced by a smooth curve. Note that the modal size
is the most common one. (b) The continuous distribution function, F (di) and its relation to the
frequency histogram.
214 I 5 : PARTICLE SIZE AND SHAPE
Table 5.1
Class range Midpoint of Number of Fraction in Total Cumulative
(nm) class range particles this class number with percentage
di (nm) ni A d C di d C di
C n, = N = 2500. C J = 1.00
the trouble of plotting data each time, it is often sufficient to specify the main features
of the distribution using a few numbers.
(5.3.1)
whereA is the fraction in class i (Table 5.1). (To distinguish it from other mean
diameters we should strictly refer to this as the number length mean diameter.)
The standard deviation, (T,is defined as:
(5.3.2)
Note that just as in the random walk (eqn 1.5.9), we do not use (di - 2) as a measure of
the spread because it can be positive or negative, and for a symmetrical distribution
C ( d i - 2) = 0 even though the distribution might be quite broad.
PARTICLE SIZE DISTRIBUTION I215
The quantity inside the main brackets is called the variance (= a’) of the population.
We take the square root of the sum of the squares so that (T can be more readily
compared with the mean. It is easy to show (Exercise 5.3.2) that
which is often easier to compute than eqn (5.3.2). (d2 is the average value of the squares
of the sample diameters.)
(5.3.4)
i
From eqn (5.3.1) it is apparent that the first moment about the origin (do = 0) is the
mean whilst from eqn (5.3.2) the second moment about the mean is the variance (0’).
The third moment about the mean is a measure of the skewness of the distribution.
Since it is an odd function (f(x) = -f(-x)) it will be zero for a perfectly symmetric
distribution and its magnitude measures the departure from symmetry. The fourth
moment about the mean measures the kurtosis. It very heavily weights the points far
away from the mean and so is a measure of the length of the tail of the distribution.
These ‘moments’ may be interpreted in quite a different way in polydisperse
(heterodisperse) systems. When, for example, a measurement is made of the total
surface area, As, of a polydisperse system of spherical particles we can write:
As = C n j n d2j . (5.3.5)
(This total area can be determined by measuring the capacity of the solid, S, to adsorb a
gas.) If the total number of particles, N , is also known then the number area mean
diameter, ZNA is defined as the diameter of the sphere for which
so that
(5.3.6)
+This is often simply called the surface (or area) average diameter.
216 I 5 : PARTICLE SIZE AND SHAPE
Table 5.2
Some possible average dimensions for colloidal particles
This leads naturally to the concept of weighting a distribution (i.e. treating some
particles as more important than others). Table 5.2 lists only a few of the possible ways
of weighting a distribution. The first is weighted by number and length and the second
is weighted by number and area, since
Instead of weighting by volume and number as in distribution (iii) we can equally well
weight the distribution by mass and area as in (iv) or by mass and volume:
where mi is the mass of material in class i that is characterized by a volume Y,. The
constants &’ and K’’ in these relations are geometric factors that may be calculated for
simple geometries. (k’= n and k” = n/6 for spheres if we retain the definition of dab
suggested in Table 5.2; they are often ignored.)
where the function F(di) is called the (number) distribution function for dj. In
Fig. 5.3.l(a) it will be noticed that all of the classes have the same width, so that the
height of the rectangles is proportional to nj. In the more general case, we may choose
PARTICLE SIZE DISTRIBUTION I217
to vary the width of the classes and in that case it is preferable to draw rectangles whose
areas reflect the numbers of particles in the class. Then we can readily relate the
distribution function, F , to ni using eqn (5.3.7). It is apparent from Fig. 5.3.l(b) that
and the area under the F (di) curve will give the total number of particles. For the data
in Table 5.1 the value of F (di) is related to that o f 5 by a constant factor since
f
n, F(d$d,
--=-
“N N .
One could generate the distribution function (in nm-’), then by multiplying column 4
by (2500/50) = 50 and it would have the same shape as the broken line in Fig. 5.3.l(a).
Ind = C J l n d j (5.3.10)
i
218 I 5 : PARTICLE SIZE AND SHAPE
I l l I I 1500
I II
2000
I
2500
Lt
3000 3500
1
and this quantity is related, not to the arithmetic mean of the diameters, but to the
geometric mean dg,which is the Nth root of the product of all the diameters:
1IN
(5.3.11)
Table 5.3
Size range Interval Average Frequency Fraction lOOL d, ln(di)
100FIN =
(nm) 6di (nm) size di ni L ~
6di in nm
N =C ni =780 C L = 1.00
PARTICLE SIZE DISTRIBUTION I219
Fig. 5.3.3 The relative percentage frequency distributionfunction plotted against In (di/nm). (Data
from Fig. 5.3.2.)
(The symbol l7i here requires the arithmetic product to be taken of the i terms.)
Although in these definitions it is usual to use number averaging there is no reason, in
principle, why they cannot be extended to mass, area, or volume averages (Section 5.3.2)
1
so that
'-h
lnd--1n n(dF) =-xtzjlndi
( j
=CJIlndi=m.
N i ) (5.3.12)
i
(5.3.13)
220 I 5 : PARTICLE SIZE A N D SHAPE
Exercises
5.3.1 Calculate the mean, standard deviation, and the variance of the distribution
shown in Table 5.1. What is the difference between the mean and the mode in
this case?
5.3.2 Establish eqn (5.3.3) from (5.3.2).
5.3.3 Calculate the number area mean diameter, N A of the particles described in
Table 5.1 and compare it with the number length mean diameter. Show that &A
= (a2 + z2)e
and check this with the result obtained in Exercise 5.3.1 (note the
relevance to eqn (5.3.13)).
5.3.4 The accompanying Fig. 5.3.4 shows a simple distribution function which can be
approximated by a parabola:
Show that F= ~
3Ndj
2N2
(1 - 2)
where N is the number of particles. [Hint: First show that F = bu( 1 - u/2) where
u = dJd.1
5.3.5 Calculate the number (arithmetic) mean of dj (2 or 2m) and its standard
deviation and do the same for In di for the distribution in Table 5.3. What is the
(number length) geometric mean diameter of these particles?
5.3.6 Calculate the number volume mean diameter of the particles in Table 5.1 and
compare it with the number area mean diameter calculated in Exercise 5.3.3.
Why is it larger? Is this always true?
Show that the mass average diameter, zd, is the ratio of the fourth to the
third moment of the distribution.
[ 1.5.91
This is shown again in Fig. 5.4.l(a). Note that, in principle, the deviations from the
mean extend to infinity in both directions but in practice the bulk of the distribution
(68.2%) lies within f a of the mean and less than 0.3% lies outside the range X f 3 a .
The normal distribution function is closely related to the error function and lies at
the heart of the statistical treatment of errors. The justification for applying it to
particle size distributions is that the positive or negative dtflerences from the mean value
that occur can be assumed to be caused by the operation of a large number of
uncontrolled (and uncontrollable) influences. T o the extent that those influences
operate in a random fashion, eqn (1.5.9) ought to be applicable (Herdan 1960, p. 73). In
fact, we know that the distribution is often very different from eqn (1.5.9) but the
distribution of In d may then be close to normal (Fig. 5.3.3). We will, therefore,
concentrate attention on the normal and the log-normal distributions, and discuss
some others only briefly.
The significance of the variance (0’)is clear from eqn (1.5.9). It is the scaling
parameter that determines how rapidly the exponential function drops to zero. The
pre-exponential factor is introduced so that:
(5.4.1)
-co
222 I 5 : PARTICLE SIZE A N D SHAPE
t
3.08
2.33
1.28
0.84
0.53
0.25
n
-0.25
-0.53
-0.84
-1.28
-2.33
-3.08
Fig. 5.4.1 The normal (Gaussian) distribution and the cumulative distribution curves.
(This is the normalizing condition, and when used in this form fG gives the fraction of
the population in each interval. Then dni = NfG dx.)
The distribution can also be represented by the cumulative distribution curve, which
represents the fraction, 9, of material that is less than a given size (Fig. 5.4.l(b)).
Setting t = (x - ??))/(Twe have:
(5.4.2)
This function is closely related to the error function (erf t), which measures the area
under the error curve between the mean and some particular value oft:
(5.4.3)
THEORETICAL DISTRIBUTION FUNCTIONS I 223
In terms of the error function: 9 (t 5 ti) = k+ erf ti. (Note that the integral in eqn
i.)
(5.4.3) is negative for t i < 0 (i.e. x < X) so that 9(t 5 ti) is then less that Values of
the error function are given in standard texts (e.g. Herdan 1960, p. 77 or the CRC
Handbook). The fraction of material oversize is, of course, 9 ( t p ti) = 1 - 9 (t 5 ti)
i
= - erf ti.
The relation between the normal curve and the cumulative curve is illustrated in
Fig. 5.4.1. It is possible to purchase special graph paper ('probability paper') on which
the ordinate scale varies in such a way as to convert the sigmoidal cumulative
distribution function (Fig. 5.4.l(b)) into a straight line (5.4.1(c)). Such paper is useful
for testing whether a particular set of data does or does not approximate to a normal
distribution. If it does, then the standard deviation can be read from the graph by
identifying the size that is bigger than 15.9 per cent (i.e. (50- 68.2/2) per cent) of the
sample and subtracting this from the mean.
What, then, is the relation between the standard deviation as it appears in eqn (1.5.9)
and the formula given earlier for a discrete distribution (eqn (5.3.2))? We can replace
the right hand side of eqn (5.3.2) by its counterpart for a continuous variable:
[f (d Sdn
- z)2dn ]'=[k f(d -z)2dn]f
(5.4.4)
and substitute
dn = Nf dd = ~ N ex.[ -i(T)2]dd
d-d
0
dW
to show that the expression (5.4.4) is indeed equal to 0.This is exactly analogous to the
demonstration (using eqn (1.5.24)) that the r.m.s. displacement of a diffusing particle
is equal to 0 (Exercise 1.5.5). Any of the expressions given for discrete distributions
can be converted in the same way to treat the continuous distribution function (see
Exercise 5.4.2).
(5.4.5)
(5.4.6)
224 I 5: PARTICLE SIZE AND SHAPE
where d, is the geometric mean of the values of d, and a, is the geometric standard
deviation of the distribution of ratios around the geometric mean. Note that, by
analogy with eqn (5.3.12),the logarithm of a, is equal to the standard deviation of In d.
The number of particles between dl and d2 in a log-normal distribution is (Herdan
1960, p. 81):
n . - L
' -P & 4
1
d2lP
di If
exp[-$(
lnd - lnd,
)']d(lnd) (5.4.7)
where p = In a,.
(5.4.8)
Fig. 5.4.2 A bimodal distribution resulting from the overlap of two 'normal' distributions having
modes dM1 and d M 2 .
THEORETICAL DISTRIBUTION FUNCTIONS I 225
is a particularly simple one, derived by comparing the various sizes with that of the
largest particle. This produces a linear log-log plot with a slope, m, which is small for
widely spread populations, and larger for narrow spreads (Scarlett 1997). The Rosin-
Rammler distribution:
(5.4.9)
also compares all sizes with a characteristic size (dc) and is commonly used for
describing the results of a comminution (grinding) process. The function can be
plotted linearly on a lnln versus In plot and again a wider distribution has a smaller
value of n (Scarlett 1997).
The Schulz distribution:
~ ( a=
) (
z + 17 [?] a(z + 1)
exp)
z+l
(5.4.10)
2
r
Exercises
5.4.1 Show that the normal distribution curve has inflexions on either side of the mean
and the distance between them is 20.
5.4.2 (i) Show that, for the distribution described in Exercise 5.3.4, the standard
deviation is given by a = d / f i .
(ii) What fraction of the material lies between d f a in this case? Compare this
with the normal distribution.
226 I 5 : PARTICLE SIZE AND SHAPE
Method (i) depends on the fact that, as sedimentation proceeds, the particles which
fall out of the measuring zone are replaced by other particles coming from above, until
the time when the largest particles have fallen down through that level. Thus the
particle concentration remains constant for some little time and then falls gradually to
zero. For successful measurement the temperature must be maintained constant and
the particle concentration must be very accurately assessed. Using a pipette to
withdraw a sample creates some disturbance which limits accuracy. Using light
absorption requires that the particle concentration be very low so that there is no
multiple scattering and the obstruction caused by each particle is additive. Such low
concentrations limit the accuracy of the procedure. The use of X-ray absorption allows
higher particle concentrations to be used, at the expense of some particle interference
during settling. The same procedures can, of course, be applied to centrifugal settling
and that is more productive for most colloidal systems.
In method (ii) the density is directly related to particle concentration. The main
problem is that the large size of the normal hydrometer and its peculiar shape makes it
difficult to locate the height to which the measurement refers, and extensive
corrections are necessary (Orr and Dallavalle (1959). The diver method has the
advantage that a number of different divers with different densities can be used to
make measurements at different stages in the sedimentation process.
SEDIMENTATION METHODS OF DETERMINING PARTICLE S I Z E I227
0.7 --
0.6 ~~
0.5 --
0.4 ~~
0.3 --
0 1 2 3 4 5
In t
Fig. 5.5.1 (a) The mass accumulated on the balance pan as a function of time in a gravity
sedimentation analysis. (b) Alternative method of analysing the data from Fig. 5.5.l(a).
In method (iii) the pressure is generated by the entire column of suspension above
the measurement level so it gradually falls as sedimentation proceeds. See Allen (1981
p. 314) for details.
In method (iv) the material is allowed to accumulate on a balance pan suspended
in the dispersion. The method was introduced by Oden (1926) and improved on by
Bostock (1952). Figure 5.5.1 shows a plot of the accumulated mass, M , of sediment
on the pan as a function of time. This is made up of two contributions: (a) the mass,
m due to particles (for which d d l ) large enough to have sedimented through the
entire depth; and (b) a fraction of the smaller particles that have fallen a shorter
distance:
M(t) = m + 1
&I,
4
?M(d) dd =
h
&ax
4
1 M(d) d d +
&in
1
k
;M(d) dd (5.5.1)
Next Page
where v is the velocity of particles of diameter d and h is the total depth of the
suspension. Differentiating eqn (5.5.1) with respect to t gives:
dM
-=
dt
/ iM(d)dd (5.5.2)
(5.5.3)
A plot of m as a percentage of the total mass of material available for settling will have
the same shape as the cumulative oversize distribution function. T o see that the
second term (dM/dln t) measures the contribution of the smaller particles, consider
Fig. 5.5.l(a). The slope dM/dt at any time, t, measures the rate of incidence of
particles smaller than the cut-off size corresponding to that time, since all larger
particles have already deposited. But this process has been going on at the same rate
for the entire time, so it contributed a mass t dM/dt to the pan. Getting rn by the
method implied in Fig. 5.5.l(a) has some problems, however, because drawing
tangents to experimental data is never a very precise procedure. T h e use of the final
logarithmic form (Fig. 5.5.l(b)) is said to involve less uncertainty (Lloyd 1997) since
the tangents can be drawn more precisely at low values o f t . Only data of very high
accuracy can produce a reliable size distribution using this method, but it is usually
possible to determine the modal value with fair accuracy. T h e time axis must, of
course, be transformed into the value of d using eqn (3.1.3) (see Exercise 5.5.1):
(5.5.4)
One basic limitation of this technique is that the region immediately below the pan
becomes depleted of particles and so has a lower density than the solution immediately
above the pan. This results in a convective motion of the suspension medium, which
interferes with the sedimentation process to some extent. A possible correction
procedure is described by Allen (1975 p. 225).
Method (v) in which the suspension is initially placed in a thin layer on top of the
suspension medium (called the line start method) makes the analysis of results simpler.
It is commonly used in centrifugation and was discussed briefly in Section 3.1.3. For
more details see Chapter 9 of Kissa (1999) and for its application in the disc centrifuge
see Allen (1997 p. 320).
(b) Sedimentation equilibrium in a gravitational field
As noted in Section 3.1.2 this was the method used by Perrin to determine the
Avogadro number using a suspension in which the particle size was known. It does not
seem to have been developed for the reverse procedure but the corresponding
Adsorption onto Solid Surfaces
6.1 Vacuum characterization methods for solid surfaces
6.1.1 The field emission and field ion microscopes
6.1.2 LEED and RHEED
6.1-3 Auger electron spectroscopy (AES)
6.1.4 X-Ray photoelectron spectroscopy (XPS)
6.1.5 Secondary ion mass spectrometry (SIMS)
6.1.6 Scanning tunnelling microscopy (STM)
6.2 Some non-vacuum techniques
6.2.1 The surface force apparatus (SFA and MASIF)
6.2.2 The atomic (or scanning) force microscope (AFM or SFM)
6.2.3 Total internal reflectance microscopy (TIRM)
6.2.4 Attenuated total reflectance spectroscopy (ATR)
6.2.5 Total internal reflectance fluoroscopy (TIRF)
6.3 Adsorption and desorption at the solid-gas interface
6.3.1 The mechanics of adsorption
6.3.2 The Langmuir adsorption isotherm
6.3.3 The BET adsorption isotherm
6.3.4 Temperature programmed desorption (TPD) and thermal desorption
mass spectrometry (TDMS)
6.4 Adsorption at the solid-liquid interface
6.4.1 Adsorption from dilute solution
6.4.2 The Langmuir isotherm
6.4.3 The Freundlich isotherm
6.4.4 Thermodynamics of adsorption
6.5 Adsorption of neutral polymers
6.5.1 Polymer chains
6.5.2 Polymer chains in solution
6.5.3 Limitations of the simple free volume theory
6.5.4 General aspects of polymer adsorption
The surface structure of solids has been an important subject of scientific study for
much of the twentieth century, initially because of its significance in understanding the
259
260 I 6 : ADSORPTION ONTO SOLID SURFACES
behaviour of solid catalysts. The early work on the subject was, therefore, heavily
influenced by the adsorption behaviour of various gases on solids, a subject which was
given a dramatic impetus by the introduction of gas warfare in World War I. There
were, however, many limitations in these studies, due firstly to the very small numbers
of molecules involved in surface reactions and hence the high sensitivity needed for
detection. The other problem was contamination. It was not until the development of
ultra-high vacuum technology (as part of the space research program in the 1960s) that
it became possible to deal with truly clean surfaces on a more or less routine basis.
Using the well established techniques of the electron microscope (Section 5.2.3) and a
variety of special spectroscopic methods, the study of the details of surface atomic
structure expanded to become a highly specialized field from the late 1960s onwards.
The scientific, technological, and commercial driving forces for these developments
came from many sources: new catalyst requirements, harder wearing surfaces and
better lubrication (tribology), fracture resistant ceramics, chemical sensors, magnetic
tape and other memory devices, dry paper copying, and a host of others, including the
ever increasing demands of the computer micro-chip industry. The detailed study of
these solid surfaces can involve the application of a barrage of techniques to a carefully
prepared surface under the most stringent purity conditions.
Many colloidal solids would, however, change structure irreversibly under high
vacuum and our primary interest in hydro-colloids also makes much of the high vacuum
work of only marginal relevance, because of the overwhelming influence of water on the
structure of the solid-water interface. Water does not merely orient itself at the interface,
it often interacts chemically with the surface atoms and may even penetrate to some
extent into the surface layers. We will, therefore, concentrate on techniques which can
throw some light on those processes and leave the many other fields for the specialist
monographs (for example Somorjai 1994, Morrison 1990). Fortunately, there still
remain a large number of spectroscopic and microscopic techniques, as well as theoretical
modelling and computational methods which can throw light on the solid-liquid
interfaces of most interest to us. The following introduction is based on the specialist
works referred to above, the recent review by Soriaga et al. (1995) and Chapter 8 of the
recent revision of Adamson's well known text (Adamson and Gast 1997).
One of the more interesting developments of recent years was the construction of
complex instruments which enabled a surface to be examined in an aqueous and also a
vacuum environment. Figure 6.1.1 shows such a device. The details need not concern
us but note that the left hand side of the instrument has a number of manipulators
which can be used to subject the sample to electrochemical tests in an aqueous
environment. The sample can then be transferred through the gate to the right hand
side of the instrument which is the high vacuum chamber (at a pressure of order 5 x
lo-'' mbar) with a central platform for the sample manipulator and an array of
emitters and receivers on the extreme right. These can be used to fire electrons, ions,
or laser or X-ray beams at the sample and to collect the resulting emission from the
sample for analysis. The sample can then be returned to the aqueous environment for a
further check without being exposed to the normal laboratory atmosphere. Soriaga et
al. (1995) give an example in which 3-pyridylhydroquinone adsorbed on the (1 11) face
of a platinum crystal is examined with cyclic voltammetry. The sample is then
examined under high vacuum and returned to the electrochemical cell where it shows
almost the identical cyclic voltammagram.
TO ROUGHING PUMP
MANIPULATOR
r - 4
U
Fig. 6.1.1 Schematic drawing of an experimental arrangement which permits both aqueous electrochemical and ultra high vacuum study of the same surface
without exposure to the normal laboratory atmosphere. (Reproduced from Hanson and Yeager 1988.) 0American Chemical Society.
262 I 6 : ADSORPTION ONTO SOLID SURFACES
.-I
x 1200
I
400 300 200 100 0
ENERGY LOSS (mev)
1( 200
ENERGY (eV)
Fig. 6.1.2 Intensity versus energy of scattered electrons from a (rhenium) metal surface covered
with a chemisorbed monolayer. (a) Auger electron spectrum and (b) a high resolution electron
energy loss spectrum. (c) The LEED pattern. (Reproduced from Somorjai and Bent 1985 with
permission.)
264 I 6 : ADSORPTION ONTO SOLID SURFACES
Fig. 6.1.3 The field emission microscope. A is an anode, connected to the screen, S. Electrons
emitted from the tip, T are attracted to the screen. The (glass) envelope, E, encloses a high vacuum,
V. (From Gomer 1955. 0Academic Press.)
,4 k9 voltage
Screen
Fig. 6.1.4 Schematic arrangement of the apparatus for Low Energy Electron Diffraction (LEED).
(From Sanchez 1992.)
It is common practice to combine both LEED and AES in the same apparatus. LEED
uses normal incidence and AES uses grazing incidence to maximize the effect of
surface atoms and adsorbed molecules. The main problem in AES is to discriminate
accurately between electrons of slightly different energies. One procedure (called the
retarding field analyser) involves arranging the voltages on the supressor grids
(Fig. 6.1.4) to make that discrimination. A more recent method called the cylindrical
metal analyser (Fig. 6.1.5) uses two coaxial metal cylinders to achieve the same result.
The voltages on the two cylinders are adjusted so that only those electrons with exactly
the right kinetic energy are able to pass through both slits and land on the detector.
Sample
Fig. 6.1.5 Schematic arrangement of the cylindrical mirror analyser for Auger electron
spectroscopy. The voltage on the two coaxial cylinders is arranged so that the curved path of the
chosen electrons allows them to pass through both slits and land on the detector. (After Soriaga
et al. 1995.)
directly. It is much more precise than Auger and can be used to detect changes in valence
states of adsorbed species. Adamson and Gast (1997) give an example of the obvious
change in the XP spectrum when a clean aluminium surface is allowed to oxidize; a
distinct increase occurs in the ratio of the height of the A13+ peak to the Al0 peak.
The kinetic energy of the emitted photo-electron is determined in the spectrometer
and it is given by:
K.E. = hu - BE - ecjsp
where u is the frequency of the initial X-ray, BE is the binding energy of the electron,
and cjsp is the work function of the spectrometer (in volts) which must be known. The
binding energy is influenced principally by the nature of the atom from which it came,
but is affected also by the valence state of the atom and, to some extent, by the
disposition and polarity of the bonds from adjacent atoms. These ‘chemical shifts’ can
also be used to identify more subtle effects in adsorbate molecules. The kinetic
energies of the emitted photo-electrons are normally very low so only electrons from
the upper 1-5 nm are able to escape. The method is therefore very sensitive to surface
268 I 6 : ADSORPTION ONTO SOLID SURFACES
structure. Most elements (except for hydrogen and helium) have practical detection
limits of about 1-10 Yo of a monolayer.
Fig. 6.1.6 STM image of a pyridine substituted porphyrin derivative on iodine modified Au (111) in
0.1 M HClO4. Obtained at 0.82 V w.r.t. the reversible H electrode. (14nm image obtained by K. Itaya,
Tohuku University, taken from the Digital Instruments brochure on ScanningTunnelling Microscopy.)
Exercises
6.1 .I The number of collisions which gas molecules make with unit area of the
container walls is given by Atkins (1982, p. 874): Z, = p/(2~rmk7)i where p is
the gas pressure and m is the mass of the gas molecule. Calculate the number of
collisions per cm2 on any surface in contact with oxygen gas at 25 "C and a
pressure of lop6mbar. (Take 1bar = 1 atm and check the unit system before you
begin.) The surface can be assumed to contain some 1015 atoms per cm2. To
what pressure must the oxygen gas be reduced in order for the collision rate
per surface atom to be of the order of once per hour?
?Any dispersion of a solid, liquid or gas in water is a hydrocolloid, though the term is
most commonly applied to dispersions of solid in water, or in predominantly aqueous
dispersion medium.
270 I 6 : ADSORPTION ONTO SOLID SURFACES
large surface areas were relatively straightforward, because the problems of contamination
were then easily manageable. Such systems were, however, only capable of providing
information on average surface properties and there were usually formidable theoretical
problems in dealing with the multiple interactions which occur in concentrated systems.
As the demand for more precise information grew, it became necessary to concentrate on
very small areas of well-defined surface (much as was done in the high vacuum studies).
For many years it proved impossible to achieve the necessary levels of cleanliness to do
such studies with any certainty. Even Deryaguin and his colleagues, who had shown the
superiority of their methods of measuring the forces between macroscopic objects in the
1950s were to fall foul of impurities in the study of minute amounts of ‘polywater’ in
the 1960s. By the 1970s, however, the problems had been ironed out to the point where
it became possible to undertake the detailed study of small areas of the solid-water
interface on a more or less routine basis, with a gradual development of commercial
instruments for the purpose. Here we consider a few techniques which are related to
the vacuum techniques but which can function effectively in aqueous media.
light to
spectrometer
micrometer
microscope objective rod
piezoelectric tube
thermistor
conductivit
-moin SIJpport
-.
.... __ . .
_- -
cantilever spring
spring
0
- cm 5
white u ',
water inlet
water outlet via pH cell
Fig. 6.2.1 Schematic drawing of apparatus to measure long-range forces between two crossed
-
cylindrical sheets of mica (of thickness 1 p m and radius of curvature -1 cm) immersed in liquid.
(From Israelachvih and Adams 1978, with permission.)
lower disc only about 10 nm. A final positional adjustment can be made by applying a
varying voltage to the piezoelectric tube on which the upper disc in mounted. The
separation can be accurately monitored from the corresponding shift in fringe
wavelength. The forces were calculated from the observed deformation of the spring
and the known spring constant.
Measurement of the interaction forces under a wide range of solution conditions can
give a very good test of double-layer theory. Since forces can be measured down to
small separations, both the attractive van der Waals force and any solvent structural
forces can also be measured. We will discuss the results of such measurements in detail
in Chapter 12. The application of the SFA to the study of the adsorption of surfactants
has recently been reviewed by Claesson and Kjellin (1999).
A variation of this instrument, called the MASIF (for Measurement and Analysis of
Surface Interaction Forces) is described by Parker (1994). It uses a bimorph strip (i.e.
two slices of piezo-electric material sandwiched together) to sense the pressure
between the approaching surfaces The distance between the surfaces is controlled by a
translation stage fitted to a piezo-electric displacement transducer which allows fine
adjustment to about 0.1 nm. The apparatus is very simple in principle, has a small
volume (about 10 mL) and can be used with non-transparent specimens, whereas the
original SFA requires transparency for the development of the interference fringes. It
is not without its difficulties, however, the most important of which are discussed by
Claesson et al. (1995).
272 I 6 : ADSORPTION ONTO SOLID SURFACES
Split
Photodiode
Laser
Voltage -
deflection of
cantilever
To computer
for force-distance Cantilever
display
SiN tip
Fig. 6.2.2 Schematic diagram of the Atomic (Scanning) Force Microscope (AFM or SFM) when
used in non-contact (colloid probe) mode. In contact mode the colloidal particles would be absent
and the Si3N4 tip would make direct contact with the specimen on the upper surface of the piezo
crystal. (After Claesson et al. 1995.)
SOME NON-VACUUM TECHNIQUES I273
For fundamental colloid science studies where one is seeking the relation between
force and distance over a range of separations (the 'non-contact' mode), the interaction
between the very fine Si3N4 tip and the underlying surface would change with
separation. T o overcome this effect Ducker et al. (1991, 1992) showed that it was
possible to attach a small spherical colloidal particle (of order 5 p m in diameter) to the
tip and to study the interaction between that sphere and another surface attached to
the bottom piezo-electric plate. That procedure has greatly expanded the range of
surfaces for which fundamental forcedistance information can now be obtained. For a
description of how the measurements are transformed into data on force versus
distance, see Claesson et al. (1995).
The more recent devices of this type are similar in principle to the SFA and the
AFM and are referred to as electrochemical scanning probe microscope (ECSPM)
devices. They can be used directly in aqueous solution or in the high vacuum mode
discussed in Section 6.1. They also incorporate advanced techniques for studying the
electrochemical reaction properties of the adsorbed molecules.
Incident light
Fig. 6.2.3 The evanescent wave formed by total internal reflection from a glass-water interface.
(After Lassen and Malmsten 1996.) dp is the characteristic penetration depth of the wave.
274 I 6 : ADSORPTION ONTO SOLID SURFACES
decreases with distance from the interface and the scattering from the sphere follows a
similar exponential relation:
with g-’ - 100 nm. The scattered intensity can be measured to about 1% so the
precision of the position measurement is about 1 nm. The particle undergoes
Brownian motion so it moves erratically about its most favoured position but is
restricted in lateral movement by a light trap.+ At low electrolyte concentrations
(< 1 mM) the repulsive double layer force dominates and so the separation distance is
about 7-10 times the Debye length. The method is extraordinarily sensitive (being able
to weigh a single particle) and gives results in close accord with the expectations of
electrical double layer theory (Chapter 12) when the particle is well separated from the
surface. There remain, however, some discrepancies at higher electrolyte concentra-
tions, when the particle sits closer to the surface. Estimates of the attractive (van der
Waals) force do not agree with those obtained with the surface force apparatus (Section
6.2.1). Prieve (1999) suggested that this discrepancy was due to ‘surface roughness’
since neither the sphere nor the glass plate could be guaranteed to be smooth to the
same degree as the mica sheets in the SFA apparatus. That proposal has since been
confirmed by Bevan and Prieve (1999).
The TIRM has also been used to show how strongly the diffusion coefficient is
altered by proximity of the sphere to the surface (N 25-fold reduction) and also to
investigate the interaction between protein covered surfaces (Liebert and Prieve 1995)
and vesicles.
The most powerful aspect of the technique stems from the fact that the particle is
essentially unconfined (at least in the vertical direction). That could well prove to be
vital for the study of the interaction between surfaces covered with protein or polymer.
Whereas in the SFA and ATM apparatus the surfaces are forced together, the TIRM
allows the particle to sense the presence of the polymer chain as it approaches; its own
motion will then reflect the relaxation of the polymer chain.
Y’I
122; 1095 I
Fig. 6.2.4 ATR-CIR (cylindrical internal reflectance) spectra showing the influence of phosphate
adsorption on goethite in aqueous suspension. (a) pH 4, lo-’ M NaC1, no phosphate. (b) pH 4,
lo-’ M NaCl, 48 pmol of phophate per g of goethite; (c) pH 4, lo-’ M NaCI, 100 pmol of
phosphate per g of goethite; (d) spectrum of aqueous solution of NazHF’O4 at pH 4 and I = lo-’ M
in NaC1. (From Tejedor-Tejedor and Anderson 1986 with permission.)
et al. 1996) but that is not its only advantage. Since water has such a very strong
absorption in the IR region, the IR spectra of aqueous systems are normally studied in
extremely short path-length cells. The ATR method is a way of circumventing that
problem, especially for suspensions, since such cells cannot be easily filled or cleaned
and would be easily plugged with the solid (Tejedor-Tejedor and Anderson 1986).
The paper by Baty et al. (1996), referred to above, is a good example of the way in
which application of XPS (Section 6.1.4), ATR-FTIR, and AFM (Section 6.2.2) can
be applied to the same system to reveal details of the adsorption behaviour of a protein
at an interface.
4\
c
JI \
1
\
Fig. 6.2.5 Schematic illustration of the TIRF apparatus. The excitation side consists of the laser
with a neutral density filter (OD), a shutter, S, a chopper& and two mirrors, MI and Mz. On the
emission side are two lenses, the monochromator, the photomultipliertube (PMT), a fast amplifier, a
photon counter (PhC), and personal computer (PC) for data analysis. (After Lassen and Malmsten
1996.)
ADSORPTION AND DESORPTIONAT THE SOLID-GAS INTERFACE I277
atoms. There is thus a tendency for these more highly active sites on the surface to be
filled up first. The extent of adsorption is measured by the coverage, $, defined by:
The rate at which the first adsorbed layer is built up decreases as the layer nears
completion but the time required for equilbrium to be reached is expected to be less
than one second, unless the gas is able to penetrate into the interior of the solid.
Experimental data on gas adsorption is collected by placing a solid surface (often in
the form of a powder) in an evacuated chamber and heating it for some time, whilst
continuing to pump out any evolved gases. When the surface is considered to be
‘clean’, the sample is taken to the (usually lower) temperature of the experiment.
Known amounts of the adsorbing gas are then allowed into the sample chamber and
the residual gas pressure is measured after equilibrium has been established. This
allows one to estimate the amount adsorbed as a function of the equilibrium pressure at
the temperature of the experiment. The resulting plot is called an adsorption isotherm.
Theoretical descriptions of the adsorption process are usually given in terms of
adsorption isotherms but, as Adamson (1990 Chapter 16) points out in his very
detailed discussion of this area, mere agreement with an isotherm does not tell the
whole story. T o obtain a proper understanding of adsorption one must also examine
the energy involved in the adsorption and desorption process. Here we confine
ourselves to a few of the basic isotherms and only briefly describe the technique used to
estimate the desorption energy. For more details see Adamson (1990) or Adamson and
Gast (1997).
and the rate coefficients for the adsorption and desorption process are k., and k d
respectively. The rate of adsorption is proportional to the pressure of A, and the
number of sites available on the surface, N(1- 6) where N is the total number of sites.
ADSORPTION AND DESORPTIONAT THE SOLID-GAS INTERFACE I279
1.0r ...
-
. _ . . . . . .- .
0 2 4 6 8 10 12 14 16
Pressure (am)
Fig. 6.3.1 The Langmuir adsorption isotherm (eqn 6.3.8) for various values of the constant K.
Therefore:
Rate of adsorption = kalpaN(l - 6). (6.3.3)
At equilibrium the rates are equal and by equating eqns (6.3.3 and 4 ) and solving for 6
we arrive at the Langmuar Isotherm:
where K = k,/kd.
The form of the isotherm for various values of K is shown in Fig. 6.3.1. Higher
values of K mean a higher afinity of the gas for the solid. Note that the coverage
increases with pressure and eventually reaches a maximum, corresponding to a
monolayer, if K or the pressure is large enough. The quantity K is the equilibrium
constant corresponding to the reaction (6.3.2) and it depends on the temperature. For
physical adsorption, lower temperatures favour the adsorption process and so K is
expected to increase as the temperature falls. Since K is an equilibrium constant we can
use the usual equations of chemical thermodynamics to relate the temperature
dependence of K to the enthalpy of adsorption. The appropriate equation is usually
referred to as the van't Hoff isochore:
280 I 6 : ADSORPTION ONTO SOLID SURFACES
The subscript .$ means that the equilibrium constant at each temperature is measured
at constant coverage. Under those conditions, K = const. x (l/pa) and so:
(6.3.7)
(Note the change in sign.) This value of A g d , is called the isosteric heat of adsorption
referring to the fact that it applies to a certain value of the coverage.+ The Langmuir
model implies that A g , , should be constant but it is more likely to be a function of
coverage: at low .$ because of the preferential filling of highly active sites and at high 6
because of lateral interactions between the adsorbed molecules as they approach close
pack on the surface. Allen (1997, Vol. 2 p. 46) gives examples of related isotherms
based on alternative assumptions about the dependence of A H on coverage. A
logarithmic relation between A H and ,$,for example, gives rise to the Freundlich
isotherm (6 = kpl’q) which we will discuss in Section 6.4.3 and Chapter 10.
The Langmuir isotherm can be tested by rearranging it into a linear form, e.g.:
This would be most suitable if we had a direct method of determining the coverage.
More usually we have measurements of the volume, v of gas which is taken up by a
given mass of solid, m, at different pressures. Then, if V, is the maximum uptake,
assumed to correspond to total coverage, we would have .$ = v / V, and so, v / ( V, - v )
= Kpa which can be rearranged to give (Exercise 6.3.1):
Fig. 6.3.2 Test of the Langmuir isotherm for CO adsorbed on charcoal. (From Atkins 1982 with
permission.)
leading ultimately to condensation of the vapour as a liquid onto the solid, a better
description is provided by the isotherm developed by Brunauer, Emmett, and Teller
(1938) and universally known by their initials. The lirst layer in that case may sometimes
(though rarely) involve chemisorption but all subsequent layers will be physisorbed. A
derivation of this equation is given by Atkins (1982 p. 1026) and we will not repeat it
here. It is an extension of Langmuir’s argument with successive layers being formed
independently. An arriving molecule may land on a bare spot or one on which there is
already one or more molecular layers stacked up. The final result, for the volume of gas
adsorbed, Y compared to the amount, Vmonrequired to form a monolayer is:
(6.3.10)
where p* is the equilibrium vapour pressure of the adsorbate and c is a constant related
to the adsorption and desorption rate constants.
The coverage as a function of relative pressure is shown in Fig. 6.3.3. As the value of
c increases one can see the clear development of a knee in the curve corresponding to
the formation of a monolayer. As the vapour pressure approaches the value of the
equilibrium vapour pressure for the condensed liquid, the adsorption increases
exponentially and ultimately a macroscopic film of liquid is formed on the surface.
Equation (6.3.10) can be written (Exercise 6.3.4):
1)x P
(1
2
- z)Y
-- 1
- CVmon
+-(c-
CVmm
where x = -.
P*
(6.3.11)
Knowing the vapour pressure of the adsorbate allows one to plot the function
x/( 1 - x) V as a function of x to establish the validity of the isotherm, and to obtain the
282 I 6 : ADSORPTION ONTO SOLID SURFACES
values of c and V,,, from the intercept and slope. Most systems give a linear plot up to at
least p/p* = x = 0.3 (Fig. 6.3.4). A suitable gas for determining the area of a solid will
have a high value of c and in that case the intercept is zero and the slope is l/Vmon.This
allows one to obtain an estimate of the area from a single point measurement (accurate to
better than 5% if c > 100) in the relative pressure region 0.14.3 (Exercise 6.3.3). The
usual gas used is nitrogen at a temperature near to the boiling point (-196 "C)but krypton
is a common alternative, particularly useful for samples of small surface area.
Although for many solids, the shape of the adsorption isotherm is very much like
that shown in Fig. 6.3.3 for some value of c, the agreement is by no means exact, as is
evident from the fact that the data begin to depart from the linear plot in Fig. 6.3.4 for
values of x (= p/p*) greater than about 0.3. There is a variety of reasons for this sort of
'non-ideal' behaviour.
So far we have treated the solid as a smooth uniform surface to which the gas all has
ready access. That is not always the case, especially if the solid is finely divided and/or
porous. In that case the shape of the isotherm may become very different from that
shown in Fig. 6.3.3, especially for high values of z where the gas is approaching the
point at which it can condense into a liquid. It may also show some hysteresis effects,
that is, the curve may be different depending on whether the pressure is being
increased or decreased (Fig. 2.7.4). One reason for this has already been discussed in
Section 2.7.2. As the gas approaches its saturation vapour pressure it becomes possible
for condensation to occur in fine capillaries in the solid even before its saturation value
is reached, if the resulting radius of curvature of the liquid surface is small enough.
Allen (1997 Vol. 2) discusses other possible reasons for observed departures from
the BET isotherm. The most obvious is the assumption that the heat of adsorption is
independent of coverage and is the same for all layers except the first. There are two
errors here: the heat of adsorption of molecules in the first layer depends on coverage
because the first adsorbate molecules occupy the more energetic sites. For the second
Fig. 6.3.3 The BET adsorption isotherm for various values of c (= 0.1,0.5, 1,2,5, 10, 100, 1000.)
ADSORPTION AND DESORPTIONAT THE SOLID-GAS INTERFACE I283
h
0.06
0.05 -
///
I I
Fig. 6.3.4 The BET adsorption isotherm for two kaolinite samples, plotted according to eqn (6.3.11).
and subsequent layers this is not expected to be significant since they are occurring on
a layer of the adsorbate molecules; the energy of adsorption of these layers is generally
assumed to be closely related to the liquefaction energy. The effect of lateral
interaction will, however, be important for both the first and the subsequent layers.
When the layer is approaching saturation the number of near neighbour interactions
increases rapidly with (. Equation 6.3.10 is, then, unlikely to be very satisfactory for
very low (C0.05) and very high (-1) values of x.
There is some confusion concerning the actual model used in deriving the BET
equation. Some texts suggest that the derivation implies that atoms or molecules can be
stacked on top of one another in columns of arbitrary height and still yield the normal
adsorption energy. There is, however, no reason to assume such unlikely stacking
arrangements. It is sufficient if one permits the formation of islands of the first layer to be
used as substrates for the formation of the second layer and for islands of the second layer,
before that layer is completed, to function as substrates for the third layer and so on. This
is the picture suggested by Atkins (1982 Fig. 29.21) and supported by the calculations of
Lowell (1975) who showed that, when the total amount of adsorption corresponded to a
monolayer, the fraction of the surface covered by molecules i layers deep was:
(6.3.12)
2 0 1 2 3 4 5
(Ci;.)m 0.0909 0.8264 0.0751 0.00683 0.00062 0.00006
284 I 6 : ADSORPTION ONTO SOLID SURFACES
So for this value of c, corresponding to a modest degree of affinity, the surface is about
9% bare at monolayer coverage, less than a tenth of the first layer is covered with a
second layer of molecules, and likewise for the third and subsequent layers.
When a molecule hits the surface and sticks, whether on the bare surface or on a
previously adsorbed film, it may diffuse around the surface and, in that case, will
eventually meet an island of adsorbed molecules if one exists. Whether it stays
associated with that group or remains independent will depend on the
temperature, and the balance between the energy associated with the lateral
interactions and the entropy gain of remaining free to move over the entire surface.
This picture seems straightforward and one would expect to be able to distinguish
relatively easily between a process in which the adsorbate is attached firmly to fixed
sites and one which is mobile. In multilayer adsorption it would seem to be
intuitively obvious that the molecule would be fairly mobile, but for the first layer
Adamson (1990 Chapter XVI Section 3B) examines the question at some length
using a statistical mechanical argument. He concludes that the distinction is not
at all easy to make, even from measurements of the entropy of the adsorption
process.
Whether the first layer of adsorbate is fixed or mobile, the most important
prediction of the BET equation is the monolayer volume, V,,. For the high values of
c for which the method is most satisfactory, Vmo, corresponds to a low value of the
relative pressure, x (C0.4). The values given by the equation for values of x greater
than about 0.4 should be taken as only a rough guide to the anticipated adsorption
behaviour. A critical review of the theory is given by Dollimore et al. (1976) and Allen
(1997 Vol. 2) gives a discussion of the BET equation and its modifications, including a
description of the n-layer BET model which can be used to treat adsorption in
restricted spaces, such as occur in microporous solids (Exercise 6.3.6). Adamson
(1990) also gives a derivation of the BET equation and describes briefly some of the
important modifications. He also introduces Polanyi’s theory of the adsorption
potential and describes the isotherms which result from that approach, together with
the general thermodynamic theory of adsorption.
Most commercial adsorbents are porous and this is an important aspect of their
nature. They are characterized by their specific surface area (area per unit mass of
solid) and classified as macroporous (pores > -50 nm), mesoporous (pores between 2 and
50 nm) or microporous (pores < 2 nm). These limits are somewhat flexible since the
detailed behaviour depends on pore shape and the size of the adsorbing molecule.
Macroporous solids behave much like simple surfaces. Mesoporous solids show the
phenomenon of capillary condensation referred to above. Microporous systems show
capillary condensation at very low pressures so the isotherm may look like a high
affinity Langmuir, but the plateau often slopes upwards as the external surfaces
become covered with more than one layer at higher gas pressures. It is important to
recognize the microporous behaviour because the plateau in that case corresponds not
to a monolayer but to the micropore volume. The effective micropore volume should
be quoted for the particular adsorbate, say nitrogen at 77 K, since it will be influenced
by the accessibility of the pores and the molar volume of the adsorbate, which may
differ from its bulk value. For a description of the procedures for determining pore
volume and pore size, see Kissa (1999 Chapter 3) or Allen (1997 Vol. 2) who also
includes a listing of commercial devices for the measurement.
ADSORPTION AND DESORPTIONAT THE SOLID-GAS INTERFACE I285
4 0.5 L
0 200 400
Temperature ("C)
Fig. 6.3.5 TPD spectra for hydrogen chemisorbed on flat (1 ll), stepped (557), and kinked (12,9,8)
single crystal surface of platinum. Curves correspond to exposure to different volumes of the gas [in
litres (L)]. (From Somorjai 1994, p. 350 with permission.)
286 I 6 : ADSORPTION ONTO SOLID SURFACES
(6.3.13)
where T, is the temperature of the maximum desorption rate, /3 is the rate of temperature
rise, and A is the frequency factor for desorption (usually taken as 1013 s-').
If the gas is able to react with the surface (chemisorption) it is more usual to use a
rapid heating process (flash desorption) and then to analyse the products with a mass
spectrometer. (This is TDMS.)
Exercises
6.3.1 Establish eqns (6.3.5) and (6.3.9).
6.3.2When the Langmuir equation is used to measure the surface area of a solid, S,
(m2/g) it is assumed that the area occupied by a molecule of the adsorbing gas
(nitrogen) at its normal boiling point (-196 "C) is 0.162 nm2. Show that S, =
4.35 V, where V, is the volume corresponding to a monolayer of gas. [ V, is
usually corrected to standard temperature and pressure (i.e. 1 atm and 0 "C)
(Allen (1997) Vol. 2 p. 44).]
6.3.3Atkins (1982) gives the following data for the pressures of CO required to cause
adsorption of 10.0 cm3 of gas to be adsorbed onto a sample (3.022 g) of charcoal.
(All volumes corrected to 1 atm. pressure and 273 K.) Show from eqn (6.3.7) that
the enthalpy of adsorption can be obtained from a plot of In p as a function of
1/T. What is the value of AH,, at this coverage? This is the isosteric heat of
adsorption.
Assuming that a CO molecule occupies an area of 0.2 nm2 and that the coverage
in this case is 0.4, estimate the area of the solid (m2/g).
6.3.4(a) Derive eqn (6.3.11) from (6.3.10).
(b) The following table gives the volume (Vin cm3)of nitrogen gas (corrected to
1 atm pressure and 273 K) adsorbed on a carbon black sample (mass 1 g) as a
function of pressure (p (mm of Hg)). Plot the volume as a function of
pressure. The lowest point of the linear part of the curve is usually identified
with monolayer formation. Replot the data in terms of eqn (6.3.11) assuming
that p* is 760 mm (i.e. the measuring temperature is the temperature of
boiling liquid nitrogen). Do the estimates of monolayer coverage agree? Take
the area of a nitrogen molecule as 0.16 nm2 and estimate the surface area of
the sample of carbon black (m2/g).
ADSORPTION AT THE SOLID-LIQUID INTERFACE I287
V 17 22 25 28 30 37 47 54 75
P 15 20 50 80 100 210 320 440 525
6.3.5Estimate the surface areas of the kaolinite samples in Fig. 6.3.4 assuming that the
mass of solid is 1 g in each case and the volume is measured in cm3 corrected to
1 atm pressure and 273 K.
6.3.6 The n-layer BET equation is derived on the assumption that adsorption is
limited to a maximum of n layers. It can be written (Allen (1997) Vol. 2 p. 56):
V
-
cz + +
1 - (n 1)zn nz”+l
v, +
(1 - 2) 1 (c - 1)x - czn+l
Show that this reduces to the Langmuir equation for n = 1. How is c related to
the value of K in the Langmuir equation?
6.3.7The rate of desorption, D may be written: D = -df/dt =Af exp (-E/RT) where
E is the desorption energy and A is the frequency factor. Assuming the
+
temperature is rising linearly with time: T = TO Pt, establish eqn (6.3.13) by
determining the maximum desorption rate.
Use the data from Fig. 6.3.5 for the desorption of hydrogen from the Pt (111)
face to estimate the energy of desorption from that face assuming that the heating
rate was 8 K/s (Adamson 1990 p. 692.)
Unfortunately, for the solid-liquid systems with which we are principally concerned,
eqn (2.4.3) is of rather limited value, due to the problems associated with the
measurement of the surface tension (or surface free energy) of solids (Section 2.2.1).
Indeed, in this case the equation is used in the opposite manner, to obtain estimates of
changes in the surface energy from independent measurements of the amount of
adsorption. Because of the large surface areas which commonly occur in colloidal
systems, the amount of adsorption can often be determined directly from the depletion
of the adjoining liquid phase, especially if the adsorbate is present in fairly dilute solution
(Section 7.7). Furthermore, it often turns out that we are most interested in the
adsorption properties of relatively minor components of the liquid phase, especially if
they are surface active (i.e. strongly adsorbed). Such substances are widely used to modify
and control the equilibrium and kinetic (transport) properties of colloidal dispersions. In
this section, therefore, we will confine attention almost entirely to the adsorption of
components which are relatively dilute (on a mole fraction basis).
The adsorption behaviour of colloidal dispersions in mixed solvent systems is
discussed by Everett (1981) and by Lyklema (1995 Chapter 2) and that ground will not
be traversed here. There is, of course, much literature on the general theory of
adsorption from mixed solvents onto solid surfaces. Fortunately, that has also been
very ably reviewed by Everett et al. in a number of papers in the series Colloid Science
(Everett 1973: Brown and Everett 1975; Everett and Podoll 1979; Davis and Everett
1983). Lyklema (1995) also provides a comprehensive coverage of this area.
r
/
Linear
C
r
r
f L-Langmuir
C
c F-Freundlich
7
H-High affinity
c
i S- Sigmoidal
c C
the solute is behaving according to the postulates of Langmuir’s theory (Section 6.3.2).
Freundlich’s isotherm is also commonly observed and, as noted in Section (6.3), can be
explained on the basis of a logarithmic variation in the energy of surface sites. The high
affinity isotherm is normally regarded as a variation of the Languir in which the initial
region shows such strong adsorption that none of the adsorbate remains in solution until
the surface nears saturation. The sigmoid suggests that some nucleation process must
first occur either in solution or, more likely, on the surface, before adsorption can
proceed. The step isotherm is sometimes interpreted in terms of bilayer formation but
may also be due to the reorientation of the surface layer to accommodate more material
(e.g. the change from horizontal to vertical orientation of a long chain molecule).
Note that none of the isotherms shows a maximum. Lyklema (1995 p. 2.66) points
out that such behaviour is the result of complications such as the appearance of new
phases, scavenging of the adsorbate by micelles, competition from spurious
contaminants (including homologues of the adsorbate), or simply analytical artefacts.
A true maximum in the surface excess, r, with increase in solution concentration is
impossible since it would indicate a decrease in chemical potential with increasing
concentration and that is thermodynamically impossible.
We study isotherms in order to understand the mechanisms and the forces involved
in the adsorption process and to gain an understanding of the structure of the adsorbed
species. Control of the extent of adsorption and the orientation of the adsorbate can be
useful for controlling the transport and equilibrium properties of a suspension. In
addition to the study of the isotherm at various temperatures, it is also valuable to have
supporting calorimetric and spectroscopic information.
290 I 6 : ADSORPTION ONTO SOLID SURFACES
and if the molecules are of similar size, the thermodynamic equilibrium constant, K, is
given by (Everett 1965):
(6.4.2)
where fA and f s are (rational) activity coefficients of the adsorbing solute (A) and the
solvent (S) in the liquid phase (1) and at the surface (a)respectively; x is the mole
fraction of A in the liquid phase and $ is the coverage (i.e. the fraction of the surface
sites occupied by A ) . The standard free energy of adsorption per molecule is given by:
(6.4.4)
where n i is the number of molecules adsorbed per unit area and N, is the total number
of sites available per unit area. Equation (6.4.4) is a form of the Langmuir adsorption
isotherm suitable for solutions. Note its similarity to eqn (6.3.5) for gases. Despite its
simple form, eqn (6.4.4) has been successfully used in a variety of different situations.
Kitchener (1965) has given a good introductory account of the conditions under
which we can expect to draw meaningful conclusions from the shape of the adsorption
isotherm. The simple Langmuir form is expected to hold for clean, smooth, non-
porous surfaces, showing reversible, physical adsorption of a pure solute, if the
adsorption occurs uniformly over the surface and the adsorbed molecules interact
laterally merely as hard spheres (Fig. 6.4.1). The amount of adsorption (in moles of
adsorbate per unit mass of adsorbent) is
(6.4.5)
where A is the specific surface area (i.e. the area per unit mass) and N A is the Avogadro
number. If mi is the maximum value of mA (corresponding to monolayer coverage)
then mi = N,A/NA and
mA KX - KC; K'c;
.$-=--- - - (6.4.6)
+
mi - 1 Kx 55.5 Kcj 1 K'cj + +
where c; is the molar concentration and eqn (6.4.6) is written for aqueous solution. It
follows from eqn (6.4.6) that a plot of c;/mA against cj should be linear for a system
ADSORPTION AT THE SOLID-LIQUID INTERFACE I291
following a Langmuir isotherm (Exercise 6.4.2 and 3) and this is the way the data is
usually tested. The plot enables one to evaluate both the maximum adsorption,
mi, and the equilibrium constant, K’. At low concentrations (when K’ci << l),
eqn (6.4.6) would suggest a linear relation between mA and ci (or x)
and this describes the first adsorption isotherm, which corresponds to Henry’s law
for dilute solutions. Th e concentration at which the surface is half covered ( x 1 / 2 ) is
inversely related to the equilibrium constant (Exercise 6.4.1) and K (or x1/2) may,
therefore, be taken as the measure of the affinity of the surface for the adsorbate.
1.6 I I
1
C
.-
*
0
0.8
0
cn
0.6
.-e
- 0.4
d 0.2
0 0.2 0.4 0.6 0.8 1
0 4
X
0 0.2 0.4 0.6 0.8 1
Solution Concentration
Fig. 6.4.2 Surface heterogeneity and the Freundlich isotherm. Curve (a) shows the result of
combining two Langmuir isotherms with very different affinities. In the extreme case of sites with
many different affinities, the result is the Freundlich isotherm (b).
292 I 6 : ADSORPTION ONTO SOLID SURFACES
where q is an empirical constant (usually lying between 2 and 10). [Davies and Rideal
(1963) point out that this isotherm was actually proposed by Kuster and initially
rejected by Freundlich, presumably because it does not have the correct limiting
behaviour at low and high concentration. Freundlich did, however, demonstrate its
applicability to a number of systems.] As noted above, eqn (6.4.7) can be shown (see
Adamson 1990, p. 425) to result from a modification to the Langmuir derivation in
which the affinity of the surface sites (as measured by the heat of adsorption, Q) varies
continuously, in accordance with an expression of the form:
where A€#' = -Q As before, (compare eqn (6.3.7)) we can obtain a measure of A€#' by
measuring the concentration C at which the coverage has a certain value, as a function
of temperature:
( -AHO,,/RT~
(aln c / ~ T ) = (6.4.10)
6 I- t
0 1 2 3 4 5 6 7 8 1 2 3 4 5
x2/10-3 A ad,H(ref.)/Jg1
Fig. 6.4.3 Adsorption of long chain alkanes from n-heptane on various grades of graphite. Left:
Enthalpy isotherms for n-docosane (n-C22H%);right: correlation with the enthalpy of adsorption for
a reference sample. (From Lyklema 1995, with permission. Redrawn from Kern and Findenegg
1980.)
ADSORPTION OF NEUTRAL POLYMERS I293
r
Exercises
6.4.1Establish eqn (6.4.4)and (6.4.6).Show that the concentration when the surface is
half covered is given by x1/2 = K-'.
6.4.2Derive an appropriate form of eqn (6.4.4) to test it by a linear plot. How do you
evaluate the constants appearing in the equation?
6.4.3Check the following data for conformity to the Langmuir isotherm and estimate
the area of the solid assuming that the cross-sectional area of the adsorbate
molecules is 0.20 nm2.
because they affect the flexibility of the chain but for the usual polymers with which
we are concerned the restrictions are not very important and the polymer can be
treated as highly flexible. Synthetic polymers are sometimes made from a single
monomer and sometimes with a mixture of two or three monomers. The latter may be
randomly mixed in the final chain or, more commonly, consist of blocks of one
monomer unit interspersed with blocks of the other monomer. The block copolymers
have some useful properties which are exploited to provide colloid stabilization
(Section 12.9).
The adsorption behaviour of polymers at interfaces is not unlike that of the non-
ionic surfactants treated in Section (6.4.5), though the greater size of the polymer
molecules makes for even more varied behaviour. Mention was made in Section 1.7.2
of the notion that polymers are adsorbed at the solidsolution interface in the form of
trains, loops, and tails (Fig. 1.7.1) (Jenkel and Rumbach 1951). That arrangement is a
compromise forced by the usual balance between energy and entropy considerations.
In this case the relative interaction energy between the polymer segments and (i) the
solvent and (ii) the sites on the surface must be balanced with the very large entropy
effects which result from adsorption: the reduction in the entropy of the polymer and
the increase in the entropy of the solvent as many molecules of solvent are released for
each molecule of polymer adsorbed. The treatment given here is drawn from the much
more extensive discussion given by Lyklema (1995, Sections 5.3-5.9). We will discuss
briefly the theory of polymer solutions and then extend the ideas in a more qualitative
manner to the adsorbed state.
Fig. 6.5.1 The polymer chain (in this case in two dimensions) can be modelled as a random walk
with a common step length, 1, which results in a fluctuatingend-to-end length r after N steps. (The
atoms in this case are assumed to have zero volume.)
ADSORPTION OF NEUTRAL POLYMERS I295
The end-to-end distance r is a characteristic feature of the chain and the mean square
value of r is given by Exercise (6.5.1):
(2)= N l2 (6.5.1)
where N is the number of bonds in the chain and 1is the length of the bond. The mean
end-to-end distance (r2)ican be taken as a measure of the mean coil diameter. A rather
better measure is the radius of gyration, aG (Exercise 3.3.5) which is given by a& =
N 12/6 for a long, perfectly flexible chain and measures the r.m.s. distance of the
segments from the centre of mass of the molecule.
Real polymers are not quite so flexible and so the chain is not quite so tightly coiled.
The effect is accounted for by a persistence parameter, p so that:
(2)= 6 p N 1 2 or 2 = p N l 2.
aG (6.5.2)
p is then 1 / 6 for a highly flexible chain and typical values for real polymers fall in the
range 0 . 5 4 .
It is possible to treat any long chain, no matter what its p value is as though it were
flexible by grouping a sufficient number of segments together (Kuhn 1934) into what
are called Kuhn segments. The orientation of these segments with respect to one
another will be random, provided the group is large enough. The number, n of bonds
in such a statistical chain element (s.c.e) will be larger the more rigid the chain is. The
real chain of N bonds is then modelled as an ideal chain of Nk (= N / k ) beads, each of
length lk (= bl) where k and b are greater than unity. Then (?) = Nklt = ( b 2 / k )N12 =
6pN12 = 6ak since b and k must satisfy the condition b 2 / k = 6p so that p remains the
parameter describing the stiffness. T o specify the system completely we require
another condition on b and k and that can be obtained by requiring that the contour
length of the Kuhn chain (= NkZk) is the same as the contour length (=N I) of the real
chain. The result is k = b = 6p. The typical Kuhn segment contains 5 to 20 backbone
atoms, depending on the flexibility of the chain (Lyklema 1995 p. 5.5).
An important consequence,of this analysis is that the ‘size’ of the polymer molecule
is proportional to id or to MT where M is the molar mass of the polymer.
6.5.2 Polymer chains in solution
The above discussion takes no account of the finite size of the atoms in the polymer
chain. Including that volume has the effect of expanding the size of the polymer. The
interaction between solvent and polymer also has an effect: strong attraction will
increase the amount of bound solvent and so increase the effective volume of the
polymer. Poor solvent/polymer interaction will lead to a tighter polymer chain and
lower polymer volume. Lyklema (1995 p. 5.6) gives a brief discussion of how these
effects can be incorporated into the simple statistical models discussed in Section 6.5.1.
Here we will use the alternative approach of describing the interaction initially in terms
of the volume contributions of polymer and solvent. The statistical calculation of those
volumes can be left for the time being.
The thermodynamics of polymer solutions was developed independently by Flory
and by Huggins in the early 1940s (Flory 1941; Huggins 1941). The theory estimates
296 I 6 : ADSORPTION ONTO SOLID SURFACES
the free energy of mixing of pure amorphous polymer molecules with pure solvent by
separately calculating the entropy of mixing (which is a combinatorial term) and the
energy of mixing (which measures the energy of interaction between polymer and
solvent when they are in contact). These two terms are then combined in the usual
manner:
~ auM- T A P
A F = (6.5.3)
where v1 is the volume fraction of solvent in the polymer solution. Similarly, the
entropy change of the amorphous polymer molecules on mixing with the solvent is -
n2k In v2. Hence the combinatorial entropy of mixing of the polymer and solvent is:
immersed in pure polymer, the energy change would be nlXlkT. In a polymer solution,
the probability of a solvent molecule being in contact with a polymer segment is simply
vz. It follows that the change in the contact dissimilarity energy, is given by:
The total free energy of mixing according to the Flory-Huggins theory is then (from
eqns 6.5.3, 4, and 6):
Since x1 is often positive for non-aqueous solvents (i.e. mixing is endothermic), the
contact dissimilarity term often opposes mixing. In contrast, both the combinatorial
entropy terms are negative and promote mixing. When the polymer is of high
molecular weight, nz is comparatively small and the dominant term promoting mixing
is nlkT In V I . The primary reason why polymer and solvent molecules mix to form
polymer solutions is now apparent: the entropy of the solvent molecules is increased as
a result of the additional space available when the domains of the polymer molecules
become accessible to the solvent. The detailed structure of the polymer is irrelevant
according to the precepts of the Flory-Huggins theory; rods are just as effective as
coils in providing space for the solvent molecules (Flory 1970). This point is stressed
because it is sometimes erroneously asserted that polymer molecules dissolve primarily
because of an increase in their configurational entropy.
Note that although x1k T was originally introduced as a change in internal energy,
the arguments presented would be essentially unchanged if XlkT were a free energy
change. In this way X I ,as determined experimentally, may incorporate both energy
and non-combinatorial entropy contributions.
Finally it may be noted that relation (6.5.7) for the free energy of mixing is very
similar to the Bragg-Williams equation for the mixing of small molecules:
(6.5.8)
The only difference is in the use of volume fraction statistics for the polymer solution
instead of the mole fraction statistics used for molecules of comparable size.
by entropic considerations. The corresponding energy terms are relatively small and of
variable sign. The experimental results imply that there is a non-combinatorial entropy
change that opposes the mixing of polymer and solvent and is not accounted for by the
Flory-Huggins theory.
Second, x1 is found experimentally to depend on the polymer concentration.
Usually XI,which has values lying in the range 0.1- 0.5, becomes more positive as the
polymer concentration increases, (e.g. poly(iso-butylene) in benzene) but some
exceptions are known (e.g. polystyrene in toluene).
Third, the Flory-Huggins theory predicts that, as mixing of polymer and solvent is
an entropically driven process, it should be favoured as the temperature is increased.
Yet experimentally it is found that most, if not all, polymer solutions can be induced to
undergo phase separation as the temperature is raised to near the critical point of the
solvent. In this regard, polymer solutions differ significantly from solutions of small
molecules.
The phase separation which occurs near the critical point gives an important clue to
the origin of the problem. This is the failure of the theory to account properly for the
difference in the free volume of the solvent compared to the polymer. This difference
is dramatically magnified near the critical point of the solvent when the solvent
molecules become gas-like, with a large free volume. The polymer molecules, in
contrast, undergo a relatively small change in free volume since they are constrained by
their connectivity. In effect, the polymer segments cause the gas-like solvent molecules
to undergo ‘condensation’ (Patterson 1969). The resulting decrease in entropy of the
solvent on contact with the polymer can outweigh the combinatorial entropy of mixing.
It is important to note that this difference in free volume can persist down to room
temperature.
Aqueous polymer solutions often show phase separations on heating but at
temperatures well below the critical temperature of water. The explanation for this
phenomenon must be sought elsewhere. It appears to be associated with the
directionality of the hydrogen bonds formed between water and the polymer but that is
far from certain. For further discussion of this point and Flory’s (1970) more elaborate
analysis of the free volume problem, the reader is referred to Napper (1983). Lyklema
(1995) also takes the discussion much further, extending the dilute solution theory into
the region of strong polymer overlap.
i z
concentration of segments which arise from polymer molecules which are not
adsorbed.
At low polymer concentration, the adsorption profile tends to be steeper, with most
of the segments close to the adsorbing surface. As the concentration rises, the loops
and tails become more extended so the profile decreases more gently with distance.
The higher the polymer molecular weight the more closely it tends to sit to the surface
but this is also influenced by the relative size of polymer-surface, solvent-surface, and
polymer-solvent interactions. An alternate measure of the profile is the bound fraction
or train fraction which measures the amount of polymer within one segment length, 1,
of the surface, divided by the total adsorbed amount. It can sometimes be estimated by
spectroscopic methods, if the polymer has some appropriate markers, but results are
often difficult to interpret. The thickness of the polymer layer can be comparable to
the radius of gyration, aG, of the polymer and this can be detected in hydrodynamic
studies (see, for example, Firth et al. 1974). The result obtained is again a little difficult
to interpret because the apparent thickness of the adsorbed layer depends on the
degree to which the solvent can drain through the polymer layer.
The amount of adsorption can, in principle, be either positive or negative
(depletion) but we will leave the latter possibility until Chapter 12. [It is important at
rather high particle concentrations when the spatial constraints prevent the polymer
from entering the regions between two particles.] The most important parameters of
300 I 6 : ADSORPTION ONTO SOLID SURFACES
positive polymer adsorption are illustrated in Fig. 6.5.2. The polymer excess
concentration rexis represented by the upper shaded area in the figure, corresponding
to the amount which is in excess of the bulk concentration. For the dilute solutions
with which we are concerned, it can be assumed that the bulk concentration is near
enough to zero and then rexcan be equated to the total adsorption, r. The bound
fraction discussed above will be given by p = 1 c(l)/I'.
Despite the complexity of the adsorption behaviour of polymers, they commonly
exhibit a Langmuir isotherm of the high affinity type, with a plateau at about 1 mg
mP2.The solution concentration at which this occurs depends on the surface area to be
covered but equilibrium solution concentrations of order 0.1- 1 mg/L are typical. The
free energy of adsorption depends on the competition between the polymer segments
and the solvent for sites on the surface. With solid substrates, the van der Waals forces
may be augmented by hydrogen bonding or more specific bonding interactions. It may
also be affected by hydrophobic bonding, in which case the polymer adsorption may
have more to do with rejection of polymer by the solvent than positive interaction with
the substrate.
The other notable feature of polymer adsorption is that it may require quite some
time to come to equilibrium, or to reach some sort of steady state. This is in contrast to
the behaviour of small molecules on non-porous surfaces for which adsorption is near
enough to instantaneous (equilibrium in times of order milliseconds). The polymer
molecule will diffuse more slowly to the region of the interface and once there it will
begin a process of segmental adsorption and desorption. Segment attachment points
might even diffuse slowly over the surface, if the attachment is purely physical. Thus it
may take some time before it has achieved a true free energy minimum. A further
problem is caused by the fact that the polymer is usually polydisperse so that some
partitioning of the molecules occurs in the adsorption process. The lower molecular
weight members can diffuse to the surface faster but the final equilibrium adsorption
usually favours the higher molecular weight species. The establishment of that
equilibrium can be very slow indeed.
The energetic considerations discussed above can be made a little more quantitative
by introducing a dimensionless adsorption energy parameter xs (Silberberg 1968)
defined by:
where A UI is the adsorption energy for a solvent molecule and subscript 2 refers to a
polymer segment. xs will be positive if there is net segment-surface interaction, since
both A U terms are negative.
For negative values of xs there will be no adsorption because, apart from the
unfavourable energetics, the entropy effect is also unfavourable. When the polymer
molecule gets close to the surface, half of the configurational space becomes
unavailable to it so its configurational entropy decreases. For long polymer chains,
accumulation occurs at an interface only if the energy effect ( f ) exceeds a critical
value, of the order of a few tenths of KT. Once this value is exceeded, however,
adsorption is of the high affinity type described above, because of the possibility of
making many contacts.
REFERENCES I301
Exercise
6.5.1 Use the equations of Section 1.5 to derive eqn (6.5.1).
6.5.2 Calculate the relative contributions of each of the terms in eqn (6.5.7) for the
mixing of poly(methy1 methacrylate) (5.0 g) of molecular weight 500 000 with
toluene (200 g) at 20 'C, assuming that the interaction parameter is 0.45 under
these conditions. Take the densities of the polymer and solvent as 1.190 and
0.866 g ~ m - respectively,
~, and assume ideal mixing. Specify which terms favour
and which disfavour mixing.
References
Adamson, A.W. (1990). Physical chemistry of surfaces. Interscience, New York.
Adamson, A.W. and Gast, A. (1997). Physical chemistry of surfaces (8th edn).
Interscience, New York.
Allen, T. (1990). Particle size measurement. 4th edn. [Powder Technology series
(ed. J.C. Williams)]. 5th edn. (1997). In two vols. Chapman and Hall, London.
Atkins, P.W. (1982). Physical chemistry 2nd edn. Oxford University Press, Oxford.
Baty, A.M., Suci, P.A., Tyler, B.J., and Geesey, G.G. (1996).J. Colloid Interface
Sci. 177, 307-15.
Bevan, M. and Prieve, D.C. (1999). Langmuir 15,7925-36.
Brown, C.E. and Everett, D.H. (1975). Adsorption at the solid/liquid interface. In
Colloid science (ed. D.H. Everett) Vol. 2, Chapter 2. Chemical Society, London.
Brunauer, S., Emmett, P.H., and Teller, E. (1938).J. Amer.Chem. SOL.60,309.
Claesson, P.M., Ederth, T., Bergeron, V., and Rutland, M.W. (1995). Surface
force measurements. I n Trends in Physical Chemistry 5, 161-94.
Claesson, P.M. and Kjellin, U.R.M. (1999). Studies of interactions between
interfaces across surfactant solutions employing various surface force techniques. In
Modern characterization methods of surfactant systems (ed. B.P. Binks). Chapter 8.
Marcel Dekker, New York.
Davis, J. and Everett D.H. (1983). Adsorption from solution. In Colloid science
(ed. D.H. Everett) Vol. 4, Chapter 3. Chemical Society, London.
Davies, J.T.and Rideal, E.K. (1963). Interfacial phenomena, 2nd edn. Academic
Press, New York and London.
Derham, K.W., Goldsbrough,J., and Gordon, M. (1974). Pure appl. Chem. 38,97.
DiNardo, J. (1994). Nanoscale characterization of surfaces and interfaces. VCH,
Weinheim.
Dollimore, D., Spooner, P., and Turner, A. (1976). Surface Technol. 4[2],
121-60.
302 I 6 : ADSORPTION ONTO SOLID SURFACES
Ducker, W.A, Senden, T.J., and Pashley, R.M. (1991). Nature, 353,239;
Langmuir 8, 1831 (1992).
Ertl, G. and Kuppers, J. (1985). Low energy electrons and surface chemistry. VCH
Publishers, New York.
Everett, D.H. (1965). Trans. Faraday. SOL.61, 2478.
Everett, D.H. (1973). Adsorption at the solid/liquid interface: non-aqueous systems. In
Colloid science (ed. D.H. Everett) Vol. 1, Chapter 2. Chemical Society, London.
Everett, D.H. (1981). In Colloid dispersions (ed. J.W.Goodwin) Chapter 4, pp. 71-97.
Royal Society of Chemistry, London.
Everett, D.H. and Podoll, R.T. (1979). Adsorption at the solid/liquid interface.
In Colloid science (ed. D.H.Everett) Vol. 3, Chapter 2. Chemical Society,
London.
Firth, B.A., Neville, P.C., and Hunter, R.J. (1974). J. Colloid Interface Sci., 49,
214-20.
Flory, P.J. (1941). J. Chem. Phys. 9, 660.
Flory, P.J. (1970). Disc. Faraday SOL.49, 7.
Gomer, R. (1955). Adv. Catalysis 7, 93.
Hanson, M.E. and Yeager, E. (1988). Electrochemical surface science. Molecular
phenomena at electrode surfaces (ed. Manuel P. Soriaga). ACS symposium series,
No. 378 American Chemical Society, Washington, DC.
Huggins, M.L. (1941).J Chem. Phys. 9,440.
Israelachvili, J.N. and Adams, G.E. (1978). 3’.Chem. SOL.Faraday I 74, 975
Jenkel, R. and Rumbach, B. (1951). 2. Electrochem. 55, 612.
Kern, H.E. and Findenegg, G.H. (1980).J.Colloid Interface Sci. 75, 346.
Kissa, E. (1999). Dispersions. Characterization, testing and measurement. Surfactant
Science Series 84, (ed. A.T. Hubbard) Marcel Dekker, New York.
Kitchener, J.A. (1965). J. Photograph. Sci. 13 152-9.
Kuhn, W. (1934). Kolloid 2. 68, 2.
Lassen B. and Malmsten, M. (1996). J.Colloid Interface Sci. 179, 470-7.
Liebert, R.B. and Prieve, D.C. (1995). Biophys. J. 69,66.
Lowell, S. (1975). Powder Technol. 2, 291-3.
Lyklema, J. (1995). Fundamentals of interface and colloid science. Vol 11: Solid-liquid
interfaces. Academic Press, New York and London.
Muller, E.W. (1943). Z.Phys3. 120, 266 & 270; 131 (1951), 136.
Morrison, S.R. (1990). The chemicalphysicsof surfaces, 2nd edn. Plenum Press, New
York & London.
Napper, D.H. (1983). Polymeric stabilization of colloidal dispersions, pp. 428.
Academic Press, London and New York.
Parker, J.L. (1994). Prog. Surface Sci. 47, 205.
Patterson, D. (1969). Macromolecules 2, 672.
Prieve, D.C. (1999). Adv. Colloid interface Sci. (Feb) 82, 93-125.
Rabe, J.P. (1999). Scanning tunnelling microscopy at solid-liquid interfaces. In Modern
characterization methods of surfactant systems (ed. B.P. Binks) Chapter 2. Marcel
Dekker, New York.
Sanchez, I.C. (1992). Physics of polymer surfaces and interfaces, Butterworth-
Heinemann.
Scales, P.J. (1999). Atomic force microscopy. In Modern characterization methods of
surfactant systems (ed. B.P. Binks) Chapter 3. Marcel Dekker, New York.
Silberberg, A. (1968). J Chem. Phys. 48,2835.
Somorjai, G.A. (1990). Introduction to surface chemistry and catalysis 2nd edn.
(1994). John Wiley, New York.
REFERENCES I303
Somorjai, G.A. and Bent, B.E. (1985). Prog. Colloid Polymer Scz. 70, 38.
Soriaga M.P., Harrington, D.A., Stickney, J.L., and Wieckowski, A. (1995). In
Modern Aspects of Electrochem. [28] (ed. B.E. Conway) Plenum Press, New York.
Tabor, D. and Winterton, R.H.S. (1969). Proc. Roy. Soc. (London) A312,435-50.
Tejedor-Tejedor, M.I. and Anderson, M.A. (1986). Langmuir 2, 203-10.
Electrified Interfaces:
The Electrical Double Layer
7.1 The electrostatic potential of a phase
7.1 .I The outer (Volta) and inner (Galvani) potential of a phase
7.1.2 The potential difference between two phases
7.2 The mercury-solution interface
7.3 Potential distribution at a flat surface -the Gouy-Chapman model
7.3.1 The Debye-Huckel approximation
7.3.2 Solution of the complete Poisson-Boltzmann equation
7.3.3 The diffuse layer charge
7.3.4 The inner (compact) double layer
(a) Charge-free inner region
(b) Adsorbed charge in the inner region
7.4 Comparison with experiment
7.4.1 Presence or absence of specific adsorption
7.4.2 No specific adsorption
7.4.3 Interpretation of specific adsorption
7.4.4 The discreteness of charge (or adsorbed ion self-atmosphere)effect
7.5 Adsorption of (uncharged) molecules at the mercury-solution interface
7.6 Limitations of the Poisson-Boltzmann equation
7.7 The silver iodide-solution interface
7.7.1 Potential-determiningions
7.7.2 The completely reversible electrode
7.7.3 Determining the point of zero charge
7.7.4 Determination of surface area
(a) Positive adsorption
(b) Negative adsorption
7.7.5 The capacitance of the Agl interface
7.7.6 The suspension effect
7.8 Other Nernstian surfaces
7.9 Mechanisms of surface charge generation
7.9.1 Dissociation of a single site
7.9.2 Two-site dissociation models
304
THE ELECTROSTATICPOTENTIAL OF A PH A SE I305
Y= Q (7.1.1)
4n€o(a + Y)
Fig. 7.1 .I (a) The original sphere with its surface layer; (b) A sphere of homogeneous composition;
(c) Spherical shell with charge and dipole layers.
THE ELECTROSTATICPOTENTIAL OF A PH A SE I307
same arrangement of charge and dipole material and (ii) a sphere of uniform
composition with no charge or dipole layers (Fig. 7.1.1). The total work done in
transferring a particle i of charge zie from infinity to the point B, in the interior of the
sphere, can be broken down into two contributions ( W = W' W" = + x). The first
part ( W ') is the work involved in bringing the charge to the point B' in the interior of
the homogeneous sphere (Fig. 7.1.1(b)). This measures the way in which the charge
interacts with the bulk of phase a and we will equate that with the chemicalpotential of
the particle i in phase a, py. It should be made clear, however, that this will depend on
the size and charge of the particle i. If i is a small sphere of radius r, this term will
contain a term of the form
due to the polarization of the medium by the charged sphere (the Born effect) (Smith
1973).
The other term, W" measures an electrical potential in the interior of a and this is
the inner (or Galvani) potential, 4". It is in this sense that we write: +
= py zie4". It
is convenient to break 4" down further into a part due to the overall net charge on the
phase (Y")and a part called the chi- orjump-potential, xffdue to the arrangement of
dipoles and any charge separation which may occur at the surface:
-B
pi - = ( pBi + zie4B) - (p~.g+ zje4") = zie(qV - 4") (7.1.3)
III
Fig. 7.1.2 Measurement of the contact or Volta potential difference between two metals.
the two phases are not the same we cannot assume that the ‘chemical’ part of is the
same in both phases. As Guggenheim (1929) points out, the problem stems from the
very notion of an electrostatic potential at a point. It measures the work down in
bringing an injinitesimal charge (not an electron) from infinity up to the point in
question.
Equation (7.1.2) implies that the difference in the inner potential “As4 can be
written:
and it is then tempting to think of the first term as representing the effect of charge
which has moved entirely from one phase to the other. The second term would then be
due solely to the disposition of dipoles at the interface. Unfortunately that is not a
distinction which can be made with any certainty.
In the case of a macroscopic metallic electrode in contact with an aqueous solution, it
is still useful to break the term A 4 into two parts: a contribution due to the separation
of the free charges (ions and electrons) (A$) and a contribution (Axdipole)due to the
orientation of dipolar molecules at the interface. (Note, however, that in general A$ #
A’€’ and A x # Axdipolein eqn (7.1.4).) The potential due to the ion distribution in the
aqueous phase can then be assumed to change from zero, in the bulk, to some value $0
on the electrode surface with $0 M A$ and:
‘4 = $0 + A Xdipole. (7.1.5)
are so small that it makes no sense to speak of a macroscopic charge separation that is
measurable from outside the solid phase. In that case it is again preferable to treat the
entire double layer region as an arrangement of charges and dipoles that is uncharged
overall. It is, however, still profitable to separate out the contribution due to the
disposition of free charges from that due to dipole orientation (eqn (7.1.5)).
f Exercise I
I 7.1 . I . Sketch the curve of Y/Yo against loglo Y from eqn (7.1.1) for lo-’ < r(cm) <
100 and a = 1 cm. I
7.2 The mercury-solution interface
We noted in Section 1.6 that many important properties of colloidal systems are
influenced by the electric charges on the particle surface. When immersed in an
electrolyte solution a charged colloidal particle will be surrounded by ions of opposite
sign so that, from a distance, it appears to be electrically neutral. The surrounding ions
are, however, able to move under the influence of thermal diffusion so that the region of
charge imbalance, due to the presence of the particle, can be quite significant, relative to
the size of the particle itself. Indeed, for very small particles (-50 nm) the disturbance it
creates can stretch out to several particle diameters. The arrangement of electric charge
on the particle, together with the balancing charge in the solution, is called an electrical
double layer, and it has been studied on various surfaces for about 200 years.
There are many recent reviews of the structure of the double layer, since it is the
basis of the entire field of electrochemistry: the same double layer forms at the surface
of an electrode and determines its behaviour in an electrochemical cell, an electrolysis
apparatus or an electroanalytical chemical device. We undertake a review here to
provide a coherent description on which to base the remainder of our work in colloid
chemistry. More detailed treatments are given in the texts by Bockris and Reddy
(1970), Delahay (1966), Sparnaay (1972), Hunter (1981), Bockris et al. (1980) and
volume I1 of the treatise by Lyklema (1995).
The most reliable information on the components of charge at an electrified
interface comes from the study of the mercury-aqueous electrolyte solution
interface. A small drop of mercury issuing from a glass capillary under the surface of
an electrolyte solution is probably as close as we can get to an ideal system for study
(Fig. 7.2.1). The mercury is contained in a reservoir, M, and as it drops from the
lower end of capillary %2 a new (and clean) surface is continually created. T h e surface
is also molecularly smooth, which makes interpretation of results easier.
If the solution contains no ions that can readily undergo oxidation or reduction then
there is no mechanism for transport of electric charge across the mercury interface. The
mercury is then said to be perfectly polarized.? It is then possible to adjust the electrical
t The word polarization is used by electrochemists to mean any process that leads to a
limitation in current flow.
310 I 7: ELECTRIFIED INTERFACES: THE ELECTRICAL DOUBLE LAYER
Fig. 7.2.1 Schematic arrangement for determining the electrical capacitance across the surface of a
mercury drop D (see text). G is an electrode made of, say Pt gauze and R is a H2/Pt reference
electrode.
potential difference across the mercury interface by altering the setting on the
potentiometer, P. The potential drop between the interior of the reference electrode, R,
and the solution is determined by the activity of H+ ions in solution (and the pressure of
the Hz gas). Any change dE in the setting of the potentiometer, P is, therefore,
immediately transmitted to the surface of the drop. Since no current can flow through
the drop interface, the only effect is a gradual build up of charge on the drop, with a
counterbalancing charge in the surrounding solution. This double layer of charge
(compare Fig. 1.6.1) behaves like an electrical capacitor and the magnitude of the
capacitance can be measured with a suitable instrument at terminals AB in Fig. 7.2.1.
The system shown in Fig. 7.2.1 could be used to study the behaviour of, say, HC104 or
HNO3 solution, since it turns out that the H+ ion is not very easily reduced on the
mercurysolution interface (i.e. negligible current flows through the interface so long as
the mercury is not made too negative with respect to the reference electrode). Typical
THE MERCURY-SOLUTION INTERFACE I311
experiments of this simple kind take two forms. Either the bulk activity of HC104 is kept
constant and the setting, E, on the potentiometer is varied, or E is kept fixed and the
activity (or concentration) of HC104 in the solution is changed. In both cases we want to
know how the surface tension of the mercury-solution interface is affected. We will find
that such studies provide a wealth of information about adsorption in charged systems
and they can, in some important cases, be extended to the solid-solution interfaces of
interest in colloid chemistry.
The thermodynamics of such a charged interface can be analysed using the form of
the Gibbs adsorption equation appropriate to charged species (Exercise 7.2.1):
The electrochemical potential pi, is defined in eqn (A5.19) where it is pointed out
(Appendix 5) that in systems involving electric charges, the equilibrium condition
between two phases at constant temperature requires equality of the electrochemical
potential for any species (including electrons) that has access to both phases. In
particular it should be noted that the potentiometer P measures the difference in the
electrochemical potential of the electrons in the wires Cu( 1) and Cu(2):
where F is the Faraday of charge. Equation (7.2.3) involves the electron equilibrium
with the metal electrodes. We now seek to apply eqn (7.2.1) to the simple system of
HC104 in aqueous solution in contact with the mercury surface. The procedure is that
of Parsons (1975~).
The components in the mercury may be taken to be Hg2+, e-, and Hg. They are not
independent, of course, and equilibrium requires that
and now we need to introduce the equilibrium conditions in both phases to eliminate
quantities which are not accessible to independent measurement. We note that:
1
djZ - -dpHz - djZe(Pt) = -djZe(Pt) (7.2.9)
+-2
and dp- = dpsdt - dP+. (7.2.10)
(Equations (7.2.8) and (7.2.9) assume that in the experiments referred to above the
activity of the mercury and the Hz gas are kept constant. The subscript salt in this case
refers to HC104 .)
Substituting eqns (7.2.8-10) in eqn (7.2.7) and rearranging terms (Exercise 7.2.2)
gives:
The coefficient of djZe(M) measures the excess of electrons in the interface above that
required to neutralize the Hg2+ in the interface. It corresponds, therefore, to -ao/F
where (TO is the charge per unit area on the mercury surface. Likewise (r+- r - ) F is
the balancing charge in the aqueous solution (where F = 96 485 C mol-'). Substituting
these quantities in eqn (7.2.11) and using eqn (7.2.3) we have:
where the subscript + on E refers to the fact that E is measured in a system with a
reference electrode reversible to the cation. If the reference electrode is made reversible
to the anion, a similar argument gives:
(The superscript (T on r refers to the use of the Gibbs convention with r H z O set equal
to zero to get relative surface excesses of cations and anions (Section 2.4.1). We will
drop it in future to avoid confusion with (TO.)
Equation (7.2.12) leads directly to the Lippmann equation:
(7.2.14)
which was put forward over a century ago and forms the basis of the study of
electrocapillarity. The interfacial tension can be measured in a variety of ways
(Section 2.11) of which the most common for the dropping mercury electrode
(DME) is the drop weight method. The high surface tension of mercury makes the
drop almost exactly spherical in shape (Perram et al. 1973a) and by weighing a
THE MERCURY-SOLUTION INTERFACE I313
known number of drops the (relative) value of y can be estimated. T o obtain exact
data, however, a numerical solution of the curvature equations is required and that
is beyond the scope of the present discussion (see Perram et al. 1 9 7 3 ~ )In . the
capillary electrometer (Fig. 7.2.2), the mercury-electrolyte interface is formed in a
tube of varying radius. As y varies with E the hydrostatic pressure required to
return the interface to the same position in the capillary is measured. Since that is
determined only by the diameter at that point (and y ) one can calculate y (or a
relative y ) from the Young-Laplace equation (Section 2.2.3). For details of more
recent apparatus see, for example, Mohilner and Kakiuchi (1981), which references
earlier work on computer-controlled devices.
Note that when there is no charge on the mercury surface (ay/aE), = 0 and at this
value of E the interfacial tension is a maximum. Figure 7.2.3 shows some typical plots
of y against the applied potential difference. When E is made negative (cathodic
polarization) the mercury surface is negative with respect to the solution. The
predominant counterions on the solution side are then the cations and Fig. 7.2.3 shows
that Na+ and K+ behave essentially identically with respect to y. On the other hand,
under anodic polarization, the counterions are anions and it is clear from the figure that
even halide ions behave very differently from one another. It is perhaps not surprising
that anions show some of their chemical character against a metal surface whereas
cations tend to behave more like simple positive charges with no special affinity for the
surface. We will discuss this behaviour in more detail below (Section 7.3).
The most striking feature of the curves in Fig. 7.2.3 is the maximum in y, called the
electrocapillay maximum (e.c.m.). The maximum value identifies the point at which a0
= 0 (from eqn (7.2.14))and so is also referred to as the point of zero charge (P.z.c.).
Putting a charge on the mercury surface, whether positive or negative, has the same
effect as does a surfactant at other liquid surfaces (Section 2.4.2); it makes it easier to
expand the surface because that reduces the lateral repulsion between the charges.
The charge at any other value of E can be obtained by differentiation of the y
versus E curve. At the dropping mercury electrode, however, it is more usual to
measure the electrical capacitance of the drop surface by surrounding it with a
Mercury
Microscope
Aqueous solution
I I I I I I I I I 1 I l l 1 1 I f I I I
420 - -
400 -
-
380 -
- -
- 360 - -
-
$ 340 - -
- -
320 -
- -
300 -
- -
280 -
- -
260 - -
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ~ 1 I 1~ 1 I .
0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1.0 -1.2 -1.4
Relative potentiometer setting ( E (calomel) +0.48)V
Fig. 7.2.3 Electrocapillary curves (after Grahame 1947). The potentiometer reading E, has been
adjusted so that E = 0 at the electrocapillary maximum for sodium fluoride (see text). (Copyright
American Chemical Society.)
C= (2) iu.
(7.2.15)
which, from eqn (7.2.14) is equal to -d2 y/dE2. The importance of the capacitance lies
in the fact that, being a differential quantity, it contains a great deal more detailed
information than does the surface tension y.
T o see this more clearly we note that the curves in Fig. 7.2.3 are very nearly
parabolic. If they are represented by an equation of the form
(where the subscript ecm refers to the electrocapillary maximum) we would have (from
eqn 7.2.14):
-(dy/dE) = 2b'(E - EKm)= DO
and C = dao/dE = 2b' (i.e. a constant). The actual experimental values of C are far
from constant as is shown in Fig. 7.2.4. This data is for sodium fluoride, which is now
recognized to be about the simplest possible electrolyte behaviour to interpret. We will
attempt some interpretation in Section 7.3.
The data in Fig. 7.2.4 can be used to determine the charge on the electrode at any
value of E since (from eqn (7.2.15)):
00 = f CdE. (7.2.17)
E=E,,
y-yem=-/crodE=-/ /
n
C(dE)' (7.2.18)
and this value can be compared with the direct measurements as a check on the
reliability of the data.
I I I I I I
Fig. 7.2.4 Differential capacitance at the DME in contact with NaF solutions at 25" C (After
Grahame 1947.) (Copyright American Chemical Society.)
316 I 7: ELECTRIFIED INTERFACES: THE ELECTRICAL DOUBLE LAYER
If measurements are done at a variety of salt concentrations (as shown in Fig. 7.2.4)
one can determine the values of r+and r- from eqn (7.2.12) or (7.2.13):
(7.2.19)
Only one of these is needed since the electroneutrality condition requires that
F ( r - - r+)= 00. In this way a huge body of data has been built up on the relation
between applied e.m.f. and interfacial adsorption for a large number of systems.
The (perfectly polarized) mercury-solution interface has been examined in some
detail here because it is important to establish the sound thermodynamic basis of the
experimental work in that area: it is the underpinning for the models of the electrical
double layer to be discussed below (Section 7.3). In colloid systems the presence of the
solid-solution interface introduces all the problems concerning the quantity y s ~
(Section 2.5.4). They can be circumvented almost entirely, however, by appeal to basic
thermodynamic relationships like eqns (7.2.12-19) with the interpretation of y s ~ as a
surface free energy (Section 2.2.1). The surface charge density, 00, is a relatively easy
quantity to measure in a solid-liquid system.
This does not mean that our descriptions of the solid-liquid interface rest solely on
the certainties of classical thermodynamics. An understanding of the microscopic
structure at the interface can only come from the introduction of extra-thermodynamic
assumptions. The thermodynamics assists in manipulation of experimental data
without introducing new and possibly erroneous assumptions. In that respect, modern
electrochemistry may be regarded, in the words of one of its most respected modern
exponents as 'the most remarkable of the applications of classical thermodynamics'
(Parsons 1980).
In the derivation of the important expressions (7.2.12) and (7.2.13) it was tacitly
assumed that each of the phases in contact could be characterized by their inner
(Galvani) potential, 4 so that the electrochemical potential of any species in that phase
was fixed by eqn (A5.19). As we noted in Section 7.1, the difference A 4 in Galvani
potentials between two phases can only be directly measured if the phases have the
same chemical composition. The system shown in Fig. 7.2.1 can be represented by:
When a measurement is made of the difference in voltage between the two copper
wires (using either some electronic measuring device or a potentiometer) the quantity
which is measured is the difference in the Galvani potential in the two wires:
E = 41 - 45 (7.2.20)
and this is possible because the two copper wires have the same chemical composition.
The quantity which is being measured is the difference in the electrochemicalpotential
of the electron in the two wires:
E = F-'{jZe(Cu(2)) - jZe(Cu(l))} = FP1{pe(Cu(2))
(7.2.21)
POTENTIAL DISTRIBUTIONAT A FLAT SURFACE - T H E GOUY-CHAPMAN M O D E L I317
where pe is the chemical part of the electrochemical potential of the electron in the
copper. Since the chemical composition is the same in the two copper wires this term
cancels and eqn (7.2.20) is reconciled with eqn (7.2.2).
Now writing:
we see that the first and last terms are in the nature of metal-metal contact potentials
that are characteristic of the metals and are, therefore, constant. Likewise (43 -44) is
determined by the activity of H+ in the solution and if it is kept constant while the
potentiometer is adjusted we have:
where A 4 is the potential difference between the interior of the mercury (or its
surface) and the interior of the aqueous solution. It is this potential A 4 which we will
wish to investigate at the microscopic level in the next section.
Exercises.
7.2.1 The Helmholtz free energy function for an electrified surface phase can be
written (compare eqn (2.3.28)):
where 4 C d q = 4 ziF Cdni". Use this, together with the usual f integration
procedure (Appendix AS) to obtain eqn (7.2.1). [Note: 4 is the electrostatic
potential characterizing the surface and the 4 C d y term represents the work done
in placing charged species in the interface.] (Don't confuse F' with the Faraday, F.)
7.2.2 Establish eqn (7.2.11).
assume that the charge on it is confined to the surface and that that surface can be
regarded as a surface of constant potential. In the solution, on the other hand, ions of
opposite sign predominate over ions of the same sign but the latter are not completely
excluded from the surface region (Fig. 1.6.1). That surface region is also of significant
thickness. We will anticipate the final result and state (as we did in Section 1.6) that
this region of varying charge density stretches over distances of order 100 nm in dilute
electrolyte solution and rather less at higher concentrations. The curvature of the
mercury drop is of the order of 100 p m so it can be reasonably approximated as flat so
far as the electrical double layer is concerned.
We must now determine the electrical charge and potential distribution in this
diffuse charge region by solving the relevant electrical and statistical thermodynamic
equations. The problem was first tackled by Gouy (in 1910) and, independently, by
Chapman (1913) and the result is referred to as the Gouy-Chapman model. Solutions of
that model are available in the standard texts (e.g. Adamson 1967; Overbeek 1952) and
reviews (Grahame 1947). They depend on the solution of what is called the Poisson-
Boltzmann equation, one of the most important equations of statisticalphysics. Although
some criticism can be levelled at this equation on strictly statistical mechanical grounds
it has been shown to be remarkably accurate in its representation of the diffuse double
layer so we will save the criticism for later (Section 7.6). Rather than repeat the
treatment so readily available elsewhere we will discuss a rather better model than the
simple one proposed by Gouy and Chapman. We will recognize, from the outset, that
the ions in the solution (whether bare or hydrated) have a finite size and so are not able
to get closer than a certain distance from the metal surface. There is, therefore, a
charge-free region near the surface, which must be treated differently from the rest of
the double layer. The thickness of the charge-free region varies from about one bare ion
radius (say 0.1 nm) up to or a little beyond one hydrated ion radius (- 0.5 nm).
The transition from $2 to $3 across the interface occurs over a finite distance and it
may be assumed that $2 remains constant almost right up to the interface (Fig. 7.3.1).
I
O d X
Fig. 7.3.1 A possible potential distribution across the metalkelectrolyteinterface. The region between
x = 0 and x = d is assumed to be free of any charges and to have a permittivity~i which differs from the
bulk value E , and may be a function of position. Outside that layer, E takes its bulk values in each phase.
(The potential distribution in the inner region (0 < x < d) will be considered later.)
POTENTIAL DISTRIBUTIONAT A FLAT SURFACE - T H E GOUY-CHAPMAN M O D E L I319
This is a reasonable assumption for a metal. The potential at the plane x = 0 is q5M
(= 42) and for x > 0 it is determined by the Poisson equation (see Appendix A3 for a
discussion of the meaning of this equation):
div D = div EE = p (7.3.1)
where D is the dielectric displacement vector (eqn (3.2.6)) and p is the local volume
density of charge (i.e. the number of charges per unit volume). Now
E = -grad 4 (7.3.2)
and so from eqns (7.3.1) and (7.3.2):
div (E grad 4) = -p. (7.3.3)
For the region x > d, where E (= E,) is constant:
where wi represents the work done in bringing an ion i up from the bulk solution
(where 4 = 4 3 ) to a point in the double layer where the potential is, say, 4 and np is the
bulk concentration of ions of type i.
As a first approximation, we assume that
v + = --1E
2 njO x;e exp ( -zie+
F). (7.3.9)
EOEr i
320 I 7: ELECTRIFIED INTERFACES: THE ELECTRICAL DOUBLE L A Y E R
xien! -
i
xte2ny$/kT
1
.
The first summation term must be zero to preserve electroneutrality in the bulk
(7.3.10)
solution, so:
(7.3.11)
(7.3.13)
= 3.2881/I(nm-l) (7.3.14)
where F is Faraday's constant and I is the ionic strength (= C ci xi2 where ci is the
ionic concentration in mol L-'). In M 1:l aqueous electrolyte solution 1 / = ~ 9.6
nm and for the systems of most interest in colloid science 1 / ranges
~ from a fraction of
a nanometre to about 100 nm.
1" ( 3 ) 2 d x = /-sinh-d$.
dx dx
4n0ze
E
ze$
kT
(7.3.17)
Integrating from some point out in the bulk solution where $ = 0 and d$/dx = 0 (see
Fig. 7.3.1) up to some point in the double layer (x>d), we have (Exercise 7.3.4):
(g)2= 4nokT [ c o s h g - 11
(7.3.18)
or (Exercise 7.3.5)
-
d$=
dx
-~
8nokT
[ E
'] . ze$
sinh-
2kT
2 ~ k T ze$
= --
ze
sinh -
2kT
(7.3.19)
from eqn (7.3.12). Note that the negative sign is chosen so that d$/dx is always
negative for $ > 0 and positive for I) < 0. This ensures that I $1 always decreases
going towards the bulk solution and becomes zero far from the surface.
Equation (7.3.19) can be integrated from a point in the double layer up to the plane
x = d to give (Exercise 7.3.6).
For very low potentials the substitution tanh p = p can be made and eqn (7.3.20)
reduces to
322 I 7: ELECTRIFIED INTERFACES: THE ELECTRICAL DOUBLE LAYER
which is the solution to the linear equation (7.3.11) (see Exercise 7.3.3). A comparison
between eqns (7.3.20) and (7.3.21) is shown in Fig. 7.3.2. Note that for X$d = Ze$d/
kT < 2 (i.e. z$rd < 51.4 mV at room temperature) the approximate expression is
reasonably accurate. Note also that far out in the double layer, where the potential is
low, we can substitute tanh p M p and then:
4k T
$= ~ Z exp[-lc(x - d)] (7.3.22)
ze
where
(7.3.23)
Since Z approaches unity for high values of z$d one can expect the potential far from
the wall to resemble that for a wall of potential $d = 4kT/ze irrespective of the actual
potential provided it is sufficiently high (compare eqn (7.3.22) with (7.3.21)). In colloid
situations, any measurement of a highly charged system at ordinary temperatures will
suggest that $rd M (lOO/z) mV if the measuring method only samples the outer region
of the double layer. Likewise, if one wants to predict the behaviour of a highly charged
system in a situation in which only the outer part of the diffuse layer is important, the
approximation $rd M (lOO/z) mV should be a good one.
Fig. 7.3.2 Electrical potential in the diffuse double layer according to the Gouy-Chapman mpdel.
Full curves are from eqn (7.3.20) and broken lines from eqn (7.3.21) for z$,j = 2 and 4. ($ is a
dimensionless quantity called the reduced potential.) (After Overbeek 1952.)
POTENTIAL DISTRIBUTIONAT A FLAT SURFACE - T H E GOUY-CHAPMAN M O D E L I323
= 1
00
d
p dx (7.3.24)
(7.3.25)
ad = -11.74~sinh(19.46)x$d) (7.3.28)
ad = -6K$d (7.3.29)
(7.3.30)
The quantity Kd is called the integral capacitance of the (diffuse) double layer and eqn
(7.3.30) shows that, for low potentials, the diffuse layer behaves like a parallel plate
capacitor with a spacing of 1 / between
~ the plates, a charge of + a d and - a d on them,
and a potential difference of $d. For low potentials then, the Helmholtz model is quite
satisfactory for many purposes. [It may be noted in passing that in this case also the
integral capacitance is equal to the differential capacitance (-dad /d$d).]
One further quantity of importance is the differential capacitance of the diffuse
double layer, Cd, defined by (from eqn (7.3.27)):
(7.3.31)
324 I 7: ELECTRIFIED INTERFACES: THE ELECTRICAL DOUBLE LAYER
0.1M -
0.01M -
0.001M -
0.0001M -
60
2ol
,
10
0
-1.0 -0.5 0 +0.5 +1.0
Vdv)
Fig. 7.3.3 Differential capacity of the diffuse double layer (from eqn (7.3.31)).
or
d d$ (7.3.32)
-(c;grad$)=O so ~(x)-==Ql
dx dx
(7.3.33)
(The symbol d- means that d $/dx is evaluated on the left-hand side of the line x = d
whilst d+ is evaluated on the right-hand side.) At points where the permittivity
changes value there is a discontinuity in the derivative of $. The quantity on the left of
eqn (7.3.33) is fixed throughout the region 0 < x < d and so:
11.0 0
1
11.d
d+=
d (7.3.34)
(7.3.35)
(7.3.36)
where 00 is the charge on the metal (which balances the charge in the diffuse layer).
Note that when ai = 0, both the integral and differential capacitance of the inner
region can be treated as parallel plate capacitors in series with the corresponding
diffuse layer capacitances to form the total capacitance.
(b) Adsorbed charge in the inner region
The simplest model to account for the possibility of additional charge adsorbed in the
inner region is that proposed initially by Stern (1924) and refined by Grahame (1947).
All of the ions are assumed to be confined to a layer at a distance x = b from the metal
surface and again they are treated as point charges, in a first approximation (Fig. 7.3.4).
Since the regions 0 C x < b and b < x C d are again free of any charge, the same
procedure can be used to arrive at expressions for the potential drop across each
region:
$0 - h = aob/G (7.3.37)
326 I 7: ELECTRIFIED INTERFACES: THE ELECTRICAL DOUBLE LAYER
(7.3.38)
with average values of permittivity defined in each region. Charge balance also requires
that
It is common practice to assume that €1 and €2 are constant so that eqns (7.3.37) and
(7.3.38) predict a linear change in potential in each region but the above derivation
shows that this is not necessary. If E;(x) is allowed to vary smoothly from x = 0 to x = d,
then $ will also vary smoothly, rather than discontinuously. Such models have been
examined by Buff and Goel (1969), Levine (1971), and Robinson and Levine (1973).
A complete solution of the potential and charge distribution in the double layer
would give values for $0, $i, $d, ao,ai, and a d . s o far we have four equations (7.3.27),
(7.3.37), (7.3.38), and (7.3.39) - and six unknowns. In the case where there is no
charge in the plane x = 6, there are four unknowns (@o, $d, 00 and a d ) and three
equations (eqns (7.3.27), (7.3.34 or 36), and (7.3.39) with ai = 0). One further equation
comes from the imposed external e.m.f. E. Using eqn (7.1.5) we have
I IHP
I /
OHP
I
I Lous solution
I
I
I
I
I
I
b d
Fig. 7.3.4 The Gouy-Chapman-Grahame model of the electrical double layer. The plane where
the diffuse layer begins (at x = d) is called the outer Helmholtz plane (OHP) and additional adsorbed
charge is assumed to be confined to another plane called the inner Helmholtz plane (IHP). At the
mercuryaqueous solution interface the IHP is assumed to be the locus of the centres of adsorbed
(dehydrated) anions whilst the OHF' is the plane of closest approach of (hydrated) cations.
POTENTIAL DISTRIBUTIONAT A FLAT SURFACE - T H E GOUY-CHAPMAN M O D E L I327
and for the moment we will ignore the last term. Then, using eqn (7.2.23):
The assumption that (Axdipole)is constant would appear at first sight to be a difficult
one to justify. We will use eqn (7.3.41) to develop eqn (7.4.9), relating the measured
capacitance of the mercury drop to the capacitances Ci and c d of the model.
Fortunately, the error introduced by this procedure is small because the contribution
of the dipoles to the capacity of the interface is so large that, in a series arrangement, as
we assume, the effect on the total capacitance is negligible. (We will return to this point
in Section 7.4.)
The four equations would appear to be sufficient to completely describe the system
if q = 0. But what if a; # O? That means that some ions are sitting very close to the
mercury surface. It is usually assumed (Grahame 1947) that the plane x = d is the plane
of closest approach of hydrated ions, so any ion in the plane x = b must be dehydrated
(at least on the side next to the mercury surface). It can be in that position only if the
free energy of adsorption more than compensates for the work done in dehydrating it.
T o describe such adsorbed ions requires an isotherm of the form:
where ny is the number of ions adsorbed per unit area, which is expected to be a
function of (a) the number of adsorption sites on the surface ( N J ,(b) the activity of ion
i in the solution (ai),(c) the local electrostatic potential, and (d) 8i, an extra free energy
term that takes into account all other special effects that the ion experiences when it is
in the plane x = b. This will involve its special interaction with the metal, which will
include purely physical interactions (like the image force)+ but may also involve
'chemical' effects (i.e. effects that are only described in terms of molecular orbital
overlap and bond formation).
Exercises
7.3.1.Establish eqns (7.3.13) and (7.3.14) using the fact that F = 96 485 C mol-l,
R = 8.31 J K-' mol-', €0 = 8.85 x F m-' (i.e. CV-' m-l) and cr = 78.5.
~ each of the following solutions: (i) lop2 M KC1; (ii)
7.3.2 Calculate the value of 1 / in
lop4M KCl; (iii) lop6M KCl; (iv) lop3M NaCl + M Na2S04; (v) lop3M
K2SO4; (vi) 5 x lop3 M MgS04.
7.3.3 Show that the solution to eqn (7.3.11) for a flat surface can be written:
328 I 7: ELECTRIFIED INTERFACES: THE ELECTRICAL DOUBLE LAYER
and so
where (sgn @d)is the algebraic sign of the potential (i.e. sgn @ = @/ I @I ).
7.3.8 Establish the value of the conversion factors in eqns (7.3.28) and (7.3.31).
7.3.9 The diffuse layer charge a d can be broken up into a contribution :a from the
cations and ad from the anions. Use the fact that, for a symmetric electrolyte,
metal. They are then said to be speczjically adsorbed. Not all anions are able to do
this, however; the evidence given below suggests that the fluoride ion, at least at
modest concentrations, is not able to penetrate into the inner region.
The question is: can we identify when specific adsorption is occurring? It would be
preferable if we could do this in a completely model-independent way. Although that is
not quite possible it is certainly possible to offer an experimental criterion that can
distinguish those situations in which there is no need to invoke specific adsorption. If
this criterion is obeyed it means that the adsorption behaviour can be quite adequately
accounted for by appealing to diffuse double layer theory alone.
The test for the absence of specific adsorption relies on an examination of the
dependence of the electrocapillary maximum (e.c.m.) or point of zero charge (P.z.c.) on
the concentration (or activity) of the background electrolyte solution.The influence of
electrolyte on the e.c.m. was studied by Esin and Markov (1939) and the coefficient, B,
defined by
B= F ( E ) (7.4.1)
00
is called the Esin and Markov coefficient. When it is evaluated at the e.c.m. (i.e.
where 00 = 0), it is found to have a value of zero in some cases, and for such
electrolytes we can assert that there is no need to invoke specific adsorption to
account for their behaviour in the neighbourhood of the e.c.m. The behaviour can
be quite adequately described in terms of the ion concentrations in the dzfluse part of
the double layer.
Of course, we are interested in the behaviour for all values of 00 and the complete
analysis of the problem is not restricted to the immediate neighbourhood of the e.c.m.
Delahay (1966) sets out the development, as it was applied by Parsons (1957), and we
will follow that approach here.
T he e.m.f., E, appearing in the expression (7.4.1) for /3 is normally measured in a
cell with a liquidjunction. (That means that the test solution is separated from the
reference electrode solution by a membrane or porous plug which permits ion
transport but prevents physical mixing). Th e Galvani potential of the reference
electrolyte is then different from that of the test solution and the difference is
called the liquid junction potential. Considerable effort goes into minimizing (or at
least maintaining constant) this potential difference. The exact thermodynamic
analysis of the Esin and Markov effect is best done using a slightly modified
coefficient, PO,involving a reference electrode without a liquidjunction (like the one
treated in Section 7.2).
It is not difficult to show from eqn (7.2.12) that (Exercise 7.4.1):
The coefficient thus measures how the surface charge is balanced by an excess of
anions or a deficit of cations. In the absence of specific adsorption the balance is
determined solely by the Poisson-Boltzmann equation and it can be shown (Parsons
1957) that, in that case (Exercise 7.4.3):
Here ad is the amount of negative charge per unit area in the diffuse layer and a+ is the
mean ionic activity of the electrolyte. B is a constant (=2zen0/~from Exercise 7.3.9).
This equation can be applied to data over the entire range of p and E+ values for a
particular salt to determine whether its behaviour can be explained without invoking
specific adsorption. (See Delahay 1966, p. 55). In particular, it can be seen from eqn
(7.4.3) that as the surface is polarized more positively or more negatively, the limiting
values of Po are (Exercise 7.4.2):
Parsons (1957) shows that for sodium fluoride on mercury, eqn (7.4.3) is valid for all of
the data obtained by Grahame.
In colloid systems it is rare to apply such a strong test. More usually we examine the
behaviour near the point of zero charge (or the e.c.m.). From eqn (7.4.3), the Po
i
coefficient is then - and it is independent of electrolyte concentration. In simple physical
terms this means that the diffuse layer charge is made up of a certain quantity of anions
and an equal deficit of cations (or vice versa). This is a direct consequence of the linear
form of the Poisson-Boltzmann equation for low potentials.
Now suppose that instead of using a reference electrode which is reversible to the
cation, as we have done so far, we use a reference electrode of fixed electrolyte activity
(like a calomel electrode) with a liquid junction leading to the aqueous solution. The
e.m.f. imposed is then say E,, where (Exercise 7.4.7):
and the constant includes the liquid junction potential difference between the
reference electrode and the aqueous solution. Equation (7.4.2) now becomes
That is
($I,= & [PO.I
Thus, at the e.c.m. where Po = 4,the quantity F(aEr/ap),=o = P' = 0.
COMPARISON WITH EXPER IMEN T I331
That is to say, the point of zero charge is not affected by the electrolyte
concentration i f it is measured with respect t o a reference electrode with a liquid
junction.
This is the usual test that is used in colloid chemistry. If the point of zero charge can
be determined at a number of electrolyte concentrations and it turns out to be
independent of the electrolyte concentration, this serves to establish that the
electrolyte being used is an indifferent one (i.e. is not specifically adsorbed).
How is this definition related to our ideas about the nature of specific adsorption?
If an ion is specifically adsorbed it has a special relation with the surface and we
expect some adsorption to occur even when the charge on the metal is zero (at the
e.c.m. or P.z.c.). Suppose it is the anion that is specifically adsorbed (as is often the
case on mercury). Then the presence of anions near the surface tends to drive
electrons back into the bulk mercury and in order to restore the condition 00 = 0 it
is necessary to give the mercury a more negative polarization. In other words, the
position of the e.c.m. shifts to more negative values. As the activity of that ion in the
aqueous solution is increased, the tendency to adsorb is increased and the e.c.m.
moves further to the cathodic side. If there is no specific adsorption, this effect is
absent and so:
(7.4.8)
The difference in the surface tension behaviour is illustrated schematically in Fig. 7.4.1.
Note that if there is no specific adsorption, ye.c.m.
is unaffected by the salt concentration
whereas when 0; # 0 the maximum is lowered and shlfted as p increases and so specific
adsorption is readily identified.
I
I
I
I
I
I q=o
I
I
Fig. 7.4.1 Illustrating the effect of specific adsorption of an anion on the electrocapillary curves at
different salt concentrations. (a) No specific adsorption. (b) With specific adsorption.
332 I 7: ELECTRIFIED INTERFACES: THE ELECTRICAL DOUBLE LAYER
(7.4.9)
where CT is the total differential capacity of the double layer. The system behaves as a
pair of capacitors in series. (This is obvious when 0; = 0 but is not strictly true for o;# 0
(see Exercise 7.4.4).)
It is apparent from eqn (7.4.9)that the smaller of the two capacitances on the right is
the one that determines the observed value CT. Incorporating into the model an inner
double layer region of limited capacitance has the effect of limiting CT so that the very
large values calculated for Cd (Fig. 7.3.3) are not observed. In fact, Cd only influences
the observed CT in the neighbourhood of E, and then only at low electrolyte
concentrations (Fig. 7.4.2). A more detailed examination of the data in Fig. 7.2.4 shows
that C; is not constant, as is suggested in Fig. 7.4.2, but depends upon the charge on
the metal surface. Grahame (1947) made the important suggestion that Ci could be
estimated from the total capacitance at high electrolyte concentration (-1 M) when c d
is so large that CT M C;. He then calculated the expected capacitance at lower
concentrations, from eqn (7.4.9),assuming only that C; was the same when the charge on
the metal was the same.
I I
I I
I I
I I
I I d‘
I I
I I
I I
Fig. 7.4.2 Schematic diagram of the effect of adding a capacitances in series: at low concentrations
and near the e.c.m the differentialcapacitance determines the behaviour but in other situations Ci is
more important.
COMPARISON WITH EXPERIMENT 1333
Figures 7.4.3 and 7.4.4 show that these calculations go a long way towards
accounting for the capacitance data for the mercurysodium fluoride solution
interface. T he value of C; can be calculated as a function of a0 at various
temperatures (Fig. 7.4.5). T o interpret these in terms of eqn (7.3.37) we need some
idea of the distance d which we have earlier suggested is of the order of the radius
of a hydrated cation. If d = 0.5 nm then a Ci value of 32 pF cmP2 corresponds to
~i =EOE, = 32 x lop6 x lo4 F m-2 x 5 x lo-'' m = 1.6 x lo-'' F m-l so that E , = 18
compared with the normal value for bulk water of about 80. Is it reasonableto expect such a
low value for the relative permittivity in this region? The high value of E, in bulk water is
due to the ability of the water molecules to orient themselves in an applied field (Section
3.2) and to reorient themselves to follow the field if it is changing. The measurement of Ci
is done with an alternating applied field (Section 7.2) but the water molecules near the
mercury surface are not able to follow that field as easily as those in bulk water because they
are already oriented to a considerable extent by the very high electric field near the surface.
The potential drop across the inner layer is, from eqn (7.3.36), equal to ao/Ci; for C; =
32 pF and a0 = 16 pC mP2 this has a value of 0.5 volt. If that potential drop occurs
across a distance of 0.5 nm the field strength is lo9 Vm-'. The energy of a dipole in an
electric field is given byp.E wherep is the dipole moment which, for water, is 6 x lop3'
C m. Its potential energy in the field is, therefore, of order 6 x lop2' J which is about
1.5 kT. Although this effect alone might not be expected to lower E, from 80 to 18 there
are other effects (including the local field of the ions in the OHP and the image force in
the mercury) which further restrict the orientational motion of the water molecules in
response to the field. In the extreme case of a completely oriented layer, the anticipated
value of E, is about 6 for water, so a mean value of 18 is not unreasonable. At least in
this respect the model is self-consistent.
It should also be noted that since q = 0, and 00 = -ad the potential in the double
layer can be calculated from eqn (7.3.27). It turns out that, even at extreme
polarizations, I $rd I is never more than about 0.2 V, and then only at very low
electrolyte concentrations (Exercise 7.4.6). As the electrolyte concentration increases
-
-
O.01MNaF -
-Observed -
..... Calculated -
-
- - -
-
14 -
t i 8 1 1 1 l 1 1 1 I 1 1 1 1 1 1 1 1 1 1 I GI
0 -0.4-0.8 -1.2 -1.6 Volt
Potential relative to the calomel electrode (with 1M KCl)
Fig. 7.4.3 Comparison of calculated and experimental differential capacitances at 25 "C in water
using Grahame's method. (After Payne 1972, with permission.)
334 I 7: E L E C T R I F I E D I N T E R F A C E S : T H E E L E C T R I C A L D O U B L E L A Y E R
25
23
T 21
8
19
’
v
17
15
13
12
I
8 4
I
0 -4
I I
-8 -12
I I
-16
I I
i
1
Fig. 7.4.4 Differential capacitance as a function of surface charge calculated by Grahame’s method
at two different temperatures. (After Payne 1972, with permission.)
I $d I falls to less than 50 mV in 1 M solution. This is why the diffuse layer capacitance
is well described by the simple Poisson-Boltzmann equation, at least at modest
electrolyte concentrations.
Returning now to the neglect of the Axdipoleterm in eqn (7.3.41) we can see how
Grahame’s procedure (Figs 7.4.3-5) largely circumvents the problem. Using the more
exact expression for dE in eqn (7.4.9) we have:
1 dE -
-- - - d[$O
+ Axdipole]
CT do0 do0 (7.4.10)
- d(’h - $d)
do0
+-d$d
do0
and the first term can still be identified with l/Ci. Grahame’s procedure amounts to
assuming that the charge on the metal is much more important than, say the electrolyte
concentration, in determining the detailed structure (including dipole orientation) of
the inner layer. It does not assume that Axdipoleis constant under all conditions but
only that at any particular value of no, Axdipole is unaffected by the electrolyte
concentration. Figure 7.4.5 shows clearly that the inner layer capacitance varies
significantly with the charge on the metal and if the variation is attributed largely to
dipole orientation (so that the thickness parameters, b and d, are constant) then
Axdipolewill also vary significantly with charge.
Even at the e.c.m. in the absence of specific adsorption when $0 = 0 (from eqns
(7.3.27) and (7.3.36)), the value of ( 4 2 +)ecm = Axdipolemay be quite large (of the
COMPARISON WITH EXPERIMENT 1335
I I I I I I I I I I
Fig. 7.4.5 Capacity of the inner region of the double layer on mercury in the presence of NaF
order of tens or hundreds of millivolts). Further discussion of this point can be found
in Bockris and Reddy (1970) and Sparnaay (1972). It is of particular importance in the
study of the adsorption of uncharged (organic) molecules because they usually act as
dipoles, which compete with water molecules for sites at the surface and hence
profoundly affect the x (‘chi’) potential.
A complete molecular model of the interfacial region would require a description, in
molecular terms, of the permittivity E; and distance d as functions of the polarization
(00) and temperature. Quite a lot of work has been done in this area. The early models
are reviewed by MacDonald and Barlow (1964), by Levine et al. (1967), and by Bockris
and Reddy (1970), while Sparnaay (1972, pp. 92-104) gives a very good description of
the models current up to that time. More recently, Parsons (19756) and Oldham and
Parsons (1977) have developed a four-state model for the water molecules, based on an
earlier suggestion of Damaskin and Frumkin (1974). Salem (1976) and Damaskin
(1977) also offer simple models of the same system. All are attempting, with varying
degrees of success, to describe the dependence of E; on 00 and temperature. A more
general description would need to take account of the metal surface (Trasatti 1971;
Gardiner 1975). The subtlety of the behaviour even when q = 0 makes it obvious that
a complete description of the inner region when q # 0 would be a very difficult task
indeed. A good model description should also allow one to transfer the calculation to
other solvents (on which there is also a great deal of data) with a realistic adjustment of
the distance parameter d and the use of independently determinable properties like
dipole moment.
336 I 7: ELECTRIFIED INTERFACES: THE ELECTRICAL DOUBLE LAYER
In colloid chemical systems it is rare to find solid surfaces that are atomically
smooth. There is, therefore, seldom much point in attempting a detailed description of
ci and di n the inner region. It often suffices to postulate a value for the capacitance C;
and to use that as a fundamental parameter of the system. The main point we will take
from this analysis is that a good deal of information about the nature of a charged
interface can be obtained from the study of the adsorption of non-specifically adsorbed
ions. These are called indafferent electrolyte ions and a systematic study of any colloidal
system always begins with a study of its behaviour towards such electrolytes
(commonly the alkali nitrates on silver iodide or alkali halides on other systems).
Indifferent ions are assumed to interact with the surface only in response to the ‘long-
range’ forces that operate beyond the Outer Helmholtz Plane (OHP) (Fig. 7.3.4). Any
ion that can penetrate into the inner region becomes subject to other short range and
much more highly specific interactions. Because they are more strongly hydrated,
cations tend to be indifferent on the mercury surface whereas anions are often
specifically adsorbed.
15
h
Tg
u 10
3
G
I
5
0
0 -0.2 -0.4 -0.6 -0.8 -1.0 -1.2 Volt
E (relativeto normal calomel electrode,i.e. Calomel electrodewith 1M KCl)
Fig. 7.4.6 Amount of specifically adsorbed anions in various electrolytes (0.1 M) in contact with
mercury at 25 C. Curves computed by Parsons from Grahame’s data. Vertical lines indicate P.Z.C.
O
shown in Fig. 7.4.6. Note that, except for Br- and I-, the amounts of anions adsorbed
at the e.c.m. are quite small ( 5 2.5 pC cm-’).
A great deal of work has been done on the development of adsorption isotherms to
describe curves of this sort using equations of the form of (7.3.42). Since our primary
concern is with colloidal systems we will not examine this material in detail because it is
adequately treated elsewhere. (See, for example, Delahay 1966.) There are, however,
some general ideas that come out of this work and that have a bearing on the treatment
of colloidal systems.
Most, if not all, treatments assume that the adsorbed ion is, in effect, in a separate
phase and that its electrochemical potential is equal to that in the bulk. They differ
only in the degree of sophistication that is brought to bear in calculating for the
specifically adsorbed ions. The original Stern isotherm can be derived as an extension
of the Langmuir isotherm (eqn (6.4.6)) by incorporating a suitable expression for the
adsorption equilibrium constant, K (per ion) = exp (-AG :d,/kT). We can then
write:
x;eN,x; exp(- A G:,/k T)
a;= x;eN,( = (7.4.11)
+
1 xi exp(-AG:&/kT)
where .$ is the coverage, xi is the mole fraction of i in the bulk solution, and N, is the
number of available adsorption sites for the ion, per unit area. The term in the
denominator accounts for the effect of ions already present in the layer and is
important when the sites are almost fully occupied. This is seldom the case for
adsorbed charges (because of lateral repulsion) so Grahame (1947) neglects that term
and uses a simpler expression:
where r is the radius of the adsorbed ion. This amounts to estimating N, on the surface
as equal to 2rN, where N, is the number of water molecules per unit volume.
Both of these isotherms when expressed in terms of the coverage, can be put in the
form (Exercise 6.4.4):
where h(6) = -ln(l - () for the Stern (Langmuir) isotherm and a; (the activity of ion i
in the bulk) is usually set equal to xi. Improvements in the description can then involve
either the function A((), which is an entropic correction for ion size (put equal to zero
in eqn (7.4.12)) or in the calculation of Actd,. The simplest analysis would begin with:
(7.4.14)
where $i is the electrostatic potential in the Inner Helmholtz Plane (IHP) (called the
macropotential) and 8; incorporates all interactions other than the ‘macroscopic’
electrical one.
When this is done it is found that 0; depends upon the state of charge or polarization
of the surface (but see Section 7.4.4 below). When testing a particular isotherm (i.e. the
Electrokinetics and the Zeta Potential
8.1 Introduction
8.2 Equilibrium double layer theory of electrokinetics
8.2.1 Electro-osmosis
8.2.2Streaming potential
8.2.3 Electrophoresis: the Smoluchowski and Huckel formulae
8.2.4The Henry formula
8.3 Reciprocity relations
8.4 The surface of shear
8.5 Measuring electrokinetic properties
8.5.1 Streaming potential and streaming current measurement
8.5.2 Electrophoresis
8.6 Limitations of the elementary theory
8.7 The standard double layer model
8.7.1 The electrokinetic equations
8.7.2 Boundary conditions
8.8 Double layer dynamics
8.8.1 Development of a double layer on a conductor
8.8.2 Double layer due to ion sources a t a dielectric surface
8.8.3 Application to a colloidal problem
8.8.4Check of the linearization approximation
8.9 Electrokinetic effects in thin double layer systems
8.9.1 Limitations of Smoluchowski's formula
8.9.2 Dukhin's analysis
8.9.3 Solution for an isolated spherical particle
8.9.4 Extension to other electrokinetic calculations
8.10 Numerical solutions of the linearized electrokinetic equations
8.11 Electrokinetics in alternating fields
8.11 .I Low frequency behaviour
8.11.2 High frequency conductance (or dielectric dispersion)
8.11.3 Electroacoustics
8.12 Validity of the electrokinetic equations
373
374 I 8: E L E C T R O K I N E T I C S A N D T H E Z E T A P O T E N T I A L
8.1 Introduction
In the previous chapter we saw how some of the main features of the electrical double
layer could be examined experimentally. Our models of the interface are attempts to
provide a detailed picture of the charge and potential distribution in the
neighbourhood of the electrically charged surface. Since it is not possible to probe
the electrostatic potential directly, we must subject the proposed models to as rigorous
a testing program as is possible.
One of the most fruitful methods of obtaining further information on the
structure of the electrical double layer is the use of electrokinetic procedures.
Electrokinetics refers to all those processes in which the boundary layer between
one charged phase and another is forced to undergo some sort of shearing process.
The charge attached to one phase (say the solid) will then move in one direction
and that associated with the adjoining phase will move (more or less tangentially) in
the opposite direction. Th e relative motion can be analysed and from this it is
possible to infer something of the way the double layer reacts to the shearing
regime. In favourable cases it is then possible to calculate how the moving charge is
distributed between the two phases. If our models of double layer structure are
correct we should expect the electrokinetic data to be reconcilable with the
equilibrium data referred to above. That expectation is met in some important
cases but in others there remain significant discrepancies which are the object of
current research.
The shearing process referred to above occurs even when a particle is undergoing its
normal Brownian motion or when a colloidal suspension is made to flow. A complete
treatment of those processes for charged particles would therefore require a proper
consideration of the electrokinetic effects. We will find, however, that these effects are
more clearly defined when the particles are studied in a particular way. When, for
example, a suspension of positively charged particles is subjected to an electric field,
the particles will move towards the negatively charged cathode whilst the surrounding
double layer ions will be drawn towards the anode. That process is called
electrophoresis. It was one of the first of the electrokinetic effects to be studied (Reuss
1809) and remains one of the most important manifestations. In other cases, the solid
may remain stationary but the charges in the adjoining fluid may move when the
electric field is applied. This is what happens when a capillary, or a porous medium,
containing an electrolyte solution, is subjected to an electric field; the process is called
electro-osmosis.
If, instead of applying an electric field, we force the electrolyte solution through a
porous medium, (or a capillary) under a hydrostatic pressure then there is generated an
electrical potential between the ends of the capillary (or porous medium). This is called
the streaming potential. Finally, if a suspension of charged particles is allowed to settle,
the resulting particle motion causes the development of a potential difference between
the upper and lower parts of the suspension. The process is called the Dorn effect and
it gives rise to the sedimentation potential. There are various other effects which come
under the general purview of electrokinetics but these are its most important
manifestations.
EQUILIBRIUM DOUBLE LAYER THEORY OF ELECTROKINETICS 1375
and
v.v=o (8.2.2)
where pe is the volume density of charge. [We use the subscript e to avoid confusion
with the normal density of the fluid.] The treatment assumes that inertia forces are
negligible.
In general the ion density pe is unknown, and eqn (8.2.1) must be supplemented by a
set of ion-conservation equations, one for each species of ion. In certain circumstances,
however, the disturbing influence (e.g. an applied electric field) does not affect the ion
density and thus pe can be approximated in eqn (8.2.1) by its equilibrium value,
obtained from the solution of the Poisson-Boltzmann equation (Section 7.3). In this
initial analysis we will concentrate on problems of this type.
8.2.1 Electro-osmosis
This term refers to the motion of liquid induced by an applied electric field. Such a
motion occurs when an electric field is applied across a porous plug, but we begin with
the much simpler case of the flow induced in a capillary tube by an electric field E
parallel to the tube axis.
The flow results from the presence of a double layer at the tube wall. For a glass
capillary containing a simple aqueous electrolyte solution, the charge on the tube wall
arises from the dissociation of surface silanol groups (-SOH) or the preferential
adsorption of OH- ions and is almost invariably negative. It will be balanced by an
equal and opposite charge in the electrolyte. Application of the electric field causes the
ions in the double layer to move towards one electrode or the other. Since the ions are
predominantly of one sign their motion gives rise to a body force on the liquid in the
double layer, and it is this body force which sets the liquid in motion. In most cases, the
tube radius is much larger than the double-layer thickness, and we can analyse the flow
in the double layer on the assumption that the surface is locally flat.
Since the applied field is parallel to the tube surface, the resulting ion migration will
not affect the charge density pe. (As ions move towards the electrode they are replaced
by ions of the same sign from further along the tube.) Thus the body force on the
liquid may be written as
I
+ve
Fig. 8.2.1 The local Cartesian coordinate system used in the analysis of electro-osmotic flow near a
solid boundary with the velocity, P, and potential, $, superimposed.
where we have used Poisson’s equation (eqn (7.3.4)) relating pe to the equilibrium
potential @e. Note that it is only the external field, E which drives the flow. In
analysing the flow we use the Cartesian coordinates shown in Fig. 8.2.1, where the x1
axis is parallel to the applied field, and the x2 axis is normal to the local tube surface. In
this coordinate system the forcebalance equation (8.2.1) becomes
d2vl ap
q---=-peE (8.2.4)
dx; axl
and
- (8.2.5)
In setting up these equations we have assumed that the fluid flows parallel to the tube.
From the continuity equation (8.2.2) it then follows that 711 is independent of XI.It is
further assumed that none of the quantities will depend on xg, and for this reason we
have only shown the x1 and x2 components of the force balance equation.
In the absence of an applied pressure gradient and assuming that gravity is
unimportant, p will be independent of x1 and eqn (8.2.4) reduces to:
(8.2.6)
EQUILIBRIUM DOUBLE LAYER THEORY OF ELECTROKINETICS 1377
where we have replaced the electrical force by the formula (8.2.3).The first integration
of this equation yields:
A second integration with respect to x2 and using the no-slip boundary condition,
gives (Exercise 8.2.1):
From the macroscopic point of view the fluid appears to slip past the surface with this
velocity, hence the subscripts (Fig. 8.2.1 and Fig. 8.5.2).
Since only a small fraction of the total fluid volume lies in the double layer, the total
flow rate is approximately given by v,A, where A is the cross-sectional area of the tube.
Thus with the aid of the formula (8.2.8) it is possible to calculate the zeta potential (5)
from measurements of the electro-osmotic flow rate. Such measurements provide
useful information about the charging process at the glass-solution interface, and they
will be discussed further in Chapter 10.
Equation (8.2.8) was first obtained by Smoluchowski in 1903. Since that time the
analysis has been extended to the case of capillaries in which the radius is not large
compared with K - ~ (Hunter 1981, section 3.5) and, more importantly it has also been
extended to the case of porous plugs with thin double-layers. In the latter case the local
analysis of the velocity field given here is still valid, but the local applied field E is
distorted by the presence of the particles. Overbeek (1952) has shown that the velocity
is given by eqn (8.2.8) everywhere in the pores beyond the double layer, provided E
represents the local electric field.
378 I 8 : ELECTROKINETICS AND THE ZETA POTENTIAL
v = -kC/rllE (8.2.9)
in the Stokes equations, and using the fact that, V . E = 0 we see that the continuity
equation (8.2.2) is automatically satisfied while the force balance equation (8.2.1) is
satisfied beyond the double layer with zero pressure gradient. Thus the formula
(8.2.9) is the required solution to Stokes’ equations with zero applied pressure
gradient.
The macroscopic electro-osmotic velocity in a porous plug is therefore given by
(8.2.10)
where the integral extends over the liquid in the sample volume Vand the contribution
from the double layer is assumed to be negligible. The integral in this expression can
be evaluated by measuring the macroscopic electric current density, given by
(8.2.1 1)
V
v,
(8.2.12)
an equation which can be used for the calculation of zeta potentials in porous plugs,
provided that the contribution to the conductivity from the double layer is negligible.
(8.2.13)
EQUILIBRIUM DOUBLE LAYER THEORY OF ELECTROKINETICS 1379
where Ap is the pressure difference across the tube, L is the tube length, and a is the
radius. The electric current due to convection of the charge with the flow is:
I1 = ]
0
~J~Y~,(Y)v~(Y)~Y. (8.2.14)
This integral is dominated by the contribution from the double layer where pe is non-
zero. In this region the formula (8.2.13) for the velocity can be approximated by
(Exercise 8.2.2):
(8.2.16)
This current is balanced by the current due to the induced field E,, a current that is
approximately given by
I2 = K, na2E, (8.2.17)
in the case when the double-layer contribution can be neglected. This assumption is
valid if
exp(xe{/2kT)
Ka
<< 1 (8.2.18)
for a symmetrical electrolyte. Assuming that this constraint is satisfied, we find that the
condition of zero net current yields:
(8.2.19)
380 I 8 : ELECTROKINETICS AND THE ZETA POTENTIAL
Overbeek (1952, p. 204) has shown that this equation can be extended to porous plugs,
provided the constraint (8.2.18) is satisfied, where a is the particle radius.
Equation (8.2.19) is most likely to fail when the electrolyte concentration is low
(< 5 x lop3 M) since K is then small and the zeta potential is likely to be large. The
problem lies mainly in the assumption that the back conduction occurs only through
the bulk electrolyte. It can be improved significantly by including a term for the
conductivity through the double layer or surface regions. This is accounted for by the
surface conduction, Ks which commonly has values of order lop9 !X1. K, is measured
per unit length of the perimeter of the tube so the total conduction is given by na2Ke +
2na K,. Applying this correction to eqn (8.2.19) gives:
(8.2.20)
Rutgers (1940) showed that the true zeta potential could be estimated using capillaries of
different radii to eliminate Ks from eqn (8.2.20) assuming it to be the same at the same
electrolyte concentration. An alternative, and much simpler, procedure is to measure the
conductivity in the actual cell (or in the porous medium) rather than relying on the use of
the bulk conductivity of the electrolyte (Briggs 1928). One measures the actual resistance
of the electrolyte in the capillary or porous medium (Rexp) and compares this with the
value expected from measurements at high electrolyte concentration, when the effects of
surface conduction should be negligible. Equation (8.2.19) then becomes:
(8.2.21)
Very accurate estimates of 5 in vitreous silica capillaries have been obtained by Wood
and Robinson (1946) using this relation.
general case. Smoluchowski concluded that the relationship between particle velocity and
electric field should have the form
= WVlE
and hence that the electrophoretic mobility is given by
pE = €</TI. (8.2.23)
The justification for this step, which was given by Overbeek in (1952) (p. 207) follows
very similar lines to the analysis of the electro-osmotic flow described in Section 8.2.1.
The formula (8.2.23) can be applied to a particle of arbitrary shape provided the
particle dimensions are much greater than the double-layer thickness and that the
constraint (eqn (8.2.18)) is satisfied (O’Brien 1983). Since the double layer is thin this
constraint can only be violated at high 5 potentials (Exercise 8.2.3).
Huckel ( 1924) solved the electrophoresis problem for the opposite extreme
condition of very thick double layers (KU << 1). In that case, the field lines are almost
unaffected by the particle (Fig. 8.2.2) and the electrical force on the particle (Q . E) is
balanced by the viscous drag of the fluid (eqn 1.5.19) and so (Exercise 8.2.4):
p~ = v / E = Q/bnqa = [ 2 ~ < / 3(1 +
~ ] ~ a M) 2 ~ < / 3 ~ (8.2.24)
The discrepancy between these two relations (23 and 24) was resolved by Henry (1931)
by taking proper account of the way the particle influences the electric field lines in its
vicinity.
Fig. 8.2.2 Effect of a non-conducting particle on the applied field. (a) K a << 1; (b) K a >> 1. The
broken line is at a distance of 1 / from
~ the particle surface.
382 I 8 : ELECTROKINETICS AND THE ZETA POTENTIAL
(8.2.26)
He has also given a corresponding relation for a cylindrical particle (Ohshima 1996).
Figure 8.2.4 gives an idea of the range of validity of Henry's formula. His results are
compared with some curves of mobility versus (-potential obtained from a computer
Fig. 8.2.3 The value of Henry's (1931) functionh ( K U ) for the effect of the particle size and double
layer thickness on the electrophoretic mobility. Recalculated function using the formula of Ohshima
(1994) with permission.
EQUILIBRIUM DOUBLE L A Y E R THEORY OF ELECTROKINETICS I383
solution of the exact equations for the flow and the ion densities around a sphere
(O'Brien and White 1978). From these curves it can be seen that the Henry formula,
which represents the tangent at the origin, is valid for a range of {-potentials that
increases with KU, from a value of about 50 mV at KU = 1 to the value indicated by the
constraint (8.2.18) at large KU values.
Exercises
8.2.1Establish eqn (8.2.7). Calculate the electro-osmotic velocity for a capillary tube
using eqn (8.2.8), for a (-potential of 45 mV, and an applied field of 1000 V m-l.
The water temperature is 20 "C. [Check the unit system.]
8.2.2Establish the approximate relation (8.2.15) for vl and then establish eqn (8.2.16).
8.2.3For KU = 50, determine the value of < at which [exp(e{/2kg/~u = 1; the
Smoluchowski formula is invalid for such < potentials (see Section 8.2.4).
8.2.4Establish eqn (8.2.24).
e[lkT=r
0 -[=e[lkT
5 10
Fig. 8.2.4 Computed electrophoretic mobilities for a spherical particle in a KCl electrolyte. The
ordinate is a non-dimensional mobility, given by [ 3 q e / 2 ~ k T ] p(a)
~ . Small KU. (b) Large Ka. The
broken lines are discussed in Section 8.9. [(a) reproduced from O'Brien and White, 1978 with
permission and (b) from O'Brien and Hunter 1981.1
384 I 8 : ELECTROKINETICS AND THE ZETA POTENTIAL
(8.3.1)
where A p is the applied pressure difference, VTis the total volume flow per unit time,
and i is the current carried. This is known as Saxen's relation and it has been
recognized for a long time (Saxen 1892). It is one of a general class of relations that can
be derived from Onsager's reciprocity relations, using the thermodynamics of
irreversible processes. Similar relations can be established between, for example,
electrophoretic mobility and the sedimentation potential. Mazur and Overbeek (195 1)
showed that, for porous plugs as well as single capillaries, eqn (8.3.1) can be
supplemented by:
(8.3.2)
surface then the charge density should be represented by the more general expression
(compare eqn (7.3.3)):
The corresponding form of eqn (8.2.8) then becomes (Overbeek 1952, p. 199):
(8.4.2)
(8.4.3)
The electric field in the double layer is expected to reduce the value of E and increase the
value of r] so the overall effect is to limit the mobility. When that mobility is then
substituted in the usual Smoluchowskiequation (8.2.23), the magnitude of the (-potential
is also limited, no matter how large the difise double layer potential, I $d I may be.
Significant effects are expected to appear at field strengths of order lo7 V/m and such
values are expected to o m at distances within 1 nm of the surface (Fig. 8.4.1).
Davies and %deal (1963) made some estimates of the ratio E / V and concluded that the
shear plane would be of order 0.3 nm from the plane of the head groups in their model for
a surfactant stabilized oil-in-water emulsion system. A better comparison would,
however, be with the diffuse layer potential and this was done analytically by Lyklema and
Overbeek (1961). They considered the case where only the viscosity was assumed to vary
according to an expression suggested by Andrade and Dodd (1946, 1951):
(8.4.4)
wherefis called the viscoelectric coeficient. Values off are not available for water (since
it breaks down under high electric fields) sofwas estimated from model considerations
and given a value of mP2 V2 which is about 3 to 5 times larger than the
measured values for typical organic solvents. The field strengths, E, in the double layer
were calculated from eqn (7.3.19) and substituting this in the integral in eqn (8.4.3)
gives (Exercise 8.4.1):
(8.4.5)
x2 -
x2 - x.2 -
Fig. 8.4.1 Viscosity (top) and tangential velocity profile (bottom) for a real double layer (left) and a
layer containing a discrete slip plane at a distance 6 from the surface. The surface is moving upward
in the x1 direction with respect to the stationary liquid. (Redrawn from Lyklema 1995, p. 4.40.)
for the <-potential decreases as the electrolyte concentration increases, since the
electric field in the double layer then increases. Lyklema (1995) shows that such
limiting behaviour is commonly observed, at least on solid surfaces (Fig. 8.4.3). Note
that the comparison is made in terms of the electrokinetic charge and the surface
charge (determined by titration) since one cannot unequivocally estimate the surface
potential. Although there are other possible explanations for the limiting of <-potential
this is the most plausible (Lyklema 1995 p. 4.43).
The data on some oil-water systems suggest (Hunter 1966) that the estimates off used
by Lyklema and Overbeek may be too high for those systems and when reduced to values
more consistent with the zeta data would require both E and q to be varied. These two
effects are more than additive and the end result is qualitativelymuch the same as Fig. 8.4.3
but the plateaux occur at higher concentrationson the oil-water interface. The problem is
discussed at some length in Hunter 1981 (section 5.3). Israelachvili (1991) presents a
variety of evidence for the proposition that the ordering and increased viscosity of water is
stronger near a solid surface than near a deformable one and that appears to be borne out by
the electrokinetic data. We will therefore take the Lyklema and Overbeek model, with their
estimate of the viscoelectric coefficient, as a good basis for interpreting the <-potential for
the hydrophilic oxidesolution interface. For some oil-in-water emulsion systems (Hunter
MEASURING ELECTROKINETIC PROPERTIES I 387
180
160
140
120
100
c = lCFz(A< 1)
80 _--
60 c = 3.5 x lCFz (A = 1)
40
c = lCF1 (A > 1)
20
1 1 1 1 1 1 I l l I I I I
0 40 80 120 160 200 240 mV
“kd
Fig. 8.4.2 Apparent electrokinetic potential as a function of the OHP potential, @d when the slip
process is determined by the viscoelectric effect. G is the molar concentration. (From Lyklema 1995,
Fig. 4.12 with permission.) [A=A,]
and O’Brien 1997) and some latex systems (Midmore et al. 1996) values of I 5 I in excess of
150 mV at 0.001 M suggest that the viscosity effect is much smaller in those cases. In any
case, the region over which the viscosity changes abruptly to high values is usually small
compared to the diameter of a water molecule (Exercise 8.4.1) so that the concept of the
plane of shear remains a reasonable one.
Exercise
8.4.1 The viscosity of water in the double layer may be affected by the high electric fields
there. The magnitude of the effect may be estimated from the relation (8.4.4):
q = qo[l +f ( d @ / d ~ ) ~Use
] . the value forfestimated by Lyklema and Overbeek
(1O-l’ VP2m2).Devise a spreadsheet program to estimate the values of q for various
@d values from 200 mV down to O mV in solutions of concentration
lop2, and 10-1 M (1:l electrolyte) near a flat charged electrode. Calculate the
corresponding distances from the plane @d and hence plot q/qo as a function of
distance for each concentration out to about 5 nm. Estimate the distance between the
point where the viscosity is twice its bulk value to that where it is ten times its bulk
value, as a function of electrolyte concentration. Comment on the result.
-3 I *ek = *O
N
I
TiO,, 1 6 , KNO,
$
,.
U
3
3
-2
TiO,, 1 6 , KNO,
-1
10-2
6 4 2 F -2 4 4
FeOOH, l e 3
-8 -10 -12
2-47
FeOOH, l e 3
\ TiO,, 10-3 or 1e2
KNO,
a o / p Ccm-*
- FeOOH, 1 t 2
Fig. 8.4.3 Observed electrokinetic charge as a function of the surface charge for a number of
systems. (From Lyklema 1995 with permission.)
were reviewed by Hunter (1981 Chapter 4) and Lyklema (1995 section 4.5) gives a
brief update. Rather than repeat that material we will concentrate attention on the
important new developments in the measurement of electrophoretic mobility,
streaming potential, and current. The dielectric dispersion and ultrasonic or
electroacoustic effects will be discussed after we have dealt with the dynamics of the
double layer. Phenomena like sedimentation potential and electro-osmosis have
important theoretical implications but they have not so far found much use as methods
for the measurement of zeta potential except in a few research situations.
The theory of streaming potential and current was developed first for capillaries of
regular shape and many measurements have been done on capillary tubes made from
suitable materials like glass, silica, and quartz. Measurements between flat plates of
materials like mica (Section 1.4.5) are also possible (Lyons e t al. 1981), but the most
flexible arrangement is the packed bed which allows almost any particulate material to
be studied. The analysis of the results in terms of zeta potential is, however, made very
difficult if the pores between the particles are not sufficiently large to permit full
development of the double layers. For more or less spherical particles the pore size ( r )
MEASURING ELECTROKINETIC PROPERTIES I 389
is of order 15% of the particle size (a). Since the double layer extends for about 3 / ~
from the surface (Section 7.3),in order to avoid overlap one needs KT >3 or Ka > -20.
This thin double layer region is of special interest and we will examine it in some detail
in Section 8.9.
8.5.2 Electrophoresis
The standard methods for determining p~ are (i) micro-electrophoresis and
(ii) moving boundary electrophoresis. In the micro-electrophoresis procedure, the
particles, in very dilute suspension, are placed in a closed capillary tube of circular
or rectangular cross-section and viewed under ultra-microscope conditions
(Fig. 5.2.l(b)). An electric field is applied at the ends of the tube and the velocity of the
particles determined by following their progress against a grid in the ocular lens system
of the microscope. The traditional process was tedious, time consuming and not very
accurate since only a limited number of particles (of order 20) would normally be
timed. The main problem, however, was that the charge on the walls of the capillary
390 I 8 : ELECTROKINETICS AND THE ZETA POTENTIAL
103-104 Q
YzZl- Amplifier
.u.
Recorder
Fig. 8.5.1 Measurement of streaming current. The electrodes are normally platinized platinum
gauze to avoid polarization and impedance to liquid flow. The resistor should have a resistance small
compared with that of the cell and the microvolt amplifier must have an impedance approaching
1 Ma.
caused the fluid in the cell to flow by electro-osmosis as soon as the field was applied
(Fig. 8.5.2(a)). In a closed cell, that flow produces a build up of pressure which causes
the fluid to flow back down the centre of the tube (Fig. 8.5.2(b)). This flow and
counter flow causes the particles to move with different velocities at different depths
in the tube and it is necessary to concentrate attention on particles at a particular
point (called the stationary level), where the fluid velocity is zero, in order to
determine the true electrophoretic mobility. Since the apparent velocity of the
particles varies significantly across the cell it is necessary to locate the stationary level
with some precision.
The initial commercial devices concentrated on (i) improving the illumination and the
precision of location of the stationary layer and (ii) devising methods to sample the motion
of a large number of particles. In the Malvern instrument (Fig. 8.5.3) two coherent beams
of red light, derived by splitting the output from a low powered H e N e laser, are made to
cross at the stationary level in the capillary cell. The resulting interference process causes
bands of high and low intensity illumination and the particles are drawn across this pattern
by the field. The scattered light shows a similar fluctuation and the frequency of the
fluctuations is related to the speed of the particles. The scattered light is collected by a
photomultiplier and analysed by a digital correlator which is able to extract the frequency
component due to the particle mobility, and hence to construct a distribution function of
particle mobilities. T o determine the sign of the particle charge, one of the mirrors is set
into oscillatory motion. When the particles are moving in the opposite direction to the
mirror they appear to move faster and this can be correlated with the field direction. In an
alternative procedure (Sephy, France) the motion of the particles in the field is subjected to
MEASURING ELECTROKINETIC PROPERTIES I 391
-31~ -31~
+iY- -cI P-
I Counter I
Fig. 8.5.2 Velocity profile in a capillary during (a) electro-osmosis and (b) electro-osmotic counter
pressure measurement or closed tube electro-osmosis. The thickness of the layer of varying velocity
at the wall has been exaggerated. It is not visible with an ordinary microscopeso the liquid right up to
the wall appears to move with velocity v, = v, (the slip velocity). 1 / is~ the double layer thickness.
standard optical image analysis to determine the distribution of particle mobilities and,
hence, the distribution of zeta potentials.
We should note in passing that the phenomenon of electrophoresis is widely used in
biochemical analysis for the separation and (partial) identificationof proteins. Differences
in the electrophoretic mobility of proteins, either through a gel or over a wet paper surface
are due in part to the <-potential.They are, however, also influenced by, for example, the
differential adsorbability of the different proteins onto the paper surface and how this is
affected by pH. These methods are therefore not suitable for determining absolute
mobilities of protein molecules and we will not discuss them further.
The moving boundary method of determining electrophoretic mobility can, in
principle, be used to determine the absolute mobility of proteins and was once the only
way of doing so. It involved creating a boundary between the suspension and the clear
suspension medium, usually in a U-tube. An electric field was then applied and the
velocity of the boundary was measured as a function of the applied field strength. It has
now been almost entirely superseded by the laser Doppler light scattering procedure.
This is an extension of the dynamic light scattering (DLS, QELS, or PCS) procedure
discussed in Section 5.7 for determining particle size.
If an electric field is applied to the particles in a DLS experiment, the particles will
all move in a particular direction in response to the field. This drift velocity is
superposed on that of the normal Brownian diffusion so that the spectrum of scattered
light is not only broadened by diffusion (Fig. 5.7.4), but shifted in frequency by a
certain amount. The autocorrelation function in this case is given by:
392 I 8 : ELECTROKINETICS AND THE ZETA POTENTIAL
I
Moving mirror dtionary level
where q = p ~E .is the particle velocity in response to the field and Q is the
magnitude of the scattering vector [= (4nno/ho) sinf3/2]. [The g function was
introduced in Section 5.7 and will be treated more formally in Chapters 13 and 14.1
The intensity spectrum is the same shape as in Fig. 5.7.4 but shifted along the
frequency axis by an amount (Ware 1974):
Q . Y E = l Q q cos (8/2) = Q ~ E cos
E (8/2) (8.5.2)
where f3 is the scattering angle (Section 3.3). The correlation function, instead of
simply decreasing exponentially to zero with the parameter t (Fig. 5.7.6), oscillates
(like the cosine function) with a decreasing amplitude until it reaches zero. The cosine
function shows that the shift becomes smaller at larger scattering angles. The
resolution of the method, R is defined as the ratio of the Doppler shift (Q . V E ) to the
Doppler broadening ( D @ ) and so:
for small f3 values (tan f3 M 0). T o obtain sufficient resolution (R M 20 say) for small
particles (high D values) requires rather high field strengths (>lo0 V/cm). Such field
strengths can cause polarization and heating problems even at moderate salt
concentrations and it is usually necessary to use a pulsed or alternating rather than a
steady field. Again we will find it useful to examine the dynamics of the double layer
(Section 8.8) before exploring this matter further. Suffice it to say at this stage that
small colloidal particles are able to move in phase with an applied field, and with their
d.c. mobility values, at frequencies well into the kilohertz range.
LIMITATIONS OF THE ELEMENTARY THEORY 1393
complexities of the more general electrokinetic processes. Only in the case of very large
particles with very thin double layers (KU > 250) can one confidently apply the
Smoluchowski equation (8.2.23)and then only for I { I 5 100 mV. For lower values of
KU and modest I {I potentials (-75 mV) the error in that procedure can be 30% or
more.
The first effect to take account of is the size of the particle, since the field lines must
bend around it. That effect was discussed in Section 8.2.3. Provided that the 5-
potential is not too high it is still possible to assume that the charge density in the
double layer is unaffected by the applied field and by the resulting fluid flows around
the particle. The electrophoretic mobility for a given zeta potential then becomes a
complicated function of the particle size (eqn (8.2.25 and 26)). This behaviour is
described by the limiting slope of the curves in Fig. 8.2.4 near the origin.
T o lift the restriction on the zeta potential and permit the examination of the general
electrokinetic problem, even for the simplest geometry (an isolated sphere), requires us
to examine the physical processes involved in some depth, because the imposition of
the field will then alter the charge distribution in the double layer. The principles thus
established will provide the groundwork for a better study of the traditional areas of
electro-osmosis and streaming potential, as well as for electrophoresis and the
sedimentation potential. They will also provide insight into the effect of electric charge
on the viscosity of suspensions (the primary and secondary electroviscous effects)
(Hunter 1981, Chapter 5). In addition, they provide an understanding of the electrical
conduction and dielectric impedance behaviour of colloidal suspensions and the
complex interactions which occur when an ultrasonic pressure wave passes through a
colloidal suspension (O’Brien 1988).
The calculation of these effects involves the solution of a set of partial differential
equations, known as the ‘electrokinetic equations’. The unknowns in these equations
are the ion number densities, the electrical potential, and the pressure and velocity
fields in the electrolyte surrounding the individual particles. These equations involve a
number of local transport properties, such as the ionic mobilities and fluid viscosity in
the double layer, properties which cannot be directly measured at present. Hence the
theoreticians have had to adopt a number of ad hoc assumptions in setting up the
electrokinetic equations. In the following section, we will outline these assumptions,
and describe a procedure for testing their validity. The test involves the calculation,
and measurement of a number of electrokinetic effects on a well characterized
suspension.
The methods for carrying out such calculations are outlined in Sections 8.9-8.10. In
Section 8.7 we set out the arguments used in the derivation of the electrokinetic
equations. The calculation of any electrokinetic effect involves the solution of these
equations subject to certain boundary conditions (Section 8.7.2), the form of which
depends to some extent on the transport property of interest.
In order to understand the physical processes responsible for the various
electrokinetic effects, we need to have some grasp of the time scales on which
processes occur in double layer systems. There has been a considerable volume of work
in this area (relating particularly to the electrical impedance of colloidal dispersions as a
function of frequency) but the insights gained from those studies have not previously
been applied in a coherent fashion to the electrokinetic effects. Here (Section 8.8) we
present two solutions of the (time-dependent) electrokinetic equations for a flat double
THE STANDARD DOUBLE LAYER MODEL 1395
layer and then show how those results shed light on the electrokinetic problems.
In Section 8.9 we describe the theoretical techniques that have recently been developed
for the solution of the electrokinetic equations in thin double layer systems, and we
describe applications of these techniques to transport property calculations. If the double
layer is not thin in comparison with the particle radius, it is usually necessary to resort to
computer solutions of the electrokinetic equations in order to calculate transport
properties. In Section 8.10 we outline the computer solutions that have so far been
obtained. In Section 8.11 we look at some low and high frequency electrokinetic
processes. Finally, in Section 8.12, results are described of the few tests which have so far
been carried out to determine the validity of the electrokinetic equations.
Some of the material in the remainder of this chapter calls for more than the normal
level of mathematical analysis. My attempts to make it accessible will not always
succeed so it should be emphasized that it is possible to gain a quite effective
conception of the application and utility of the zeta potential without having a full
understanding of the mathematical theory behind its calculation. For many
applications, all one needs is the basic concept of the slipping plane and the
electrostatic potential characterizing that plane. That is so if one is concerned only with
the use of zeta in exploring equilibrium adsorption behaviour at charged interfaces.
Even in many kinetic processes, like the collisions occurring during coagulation, the
‘equilibrium’ zeta potential seems to govern the behaviour. The moral is: ‘don’t give up
too easily, but don’t despair if it all looks like more than you wanted to know. Zeta can
still prove useful’.
conduction process is certainly fairly common, and the mobility of ions in that layer is
often close to that of the diffuse layer ions (Lyklema and Minor 1998).
Nevertheless, the fact remains that in the classical analyses of the electrophoresis
problem by Overbeek (1943), Booth (1950), Wiersema et al. (1966), and O’Brien and
White (1978) the assumption was made that surface conduction was confined to the ions
above the shear plane. The initial theoretical analysis of surface conduction itself, which
was elaborated by Bikerman (1935, 1940), was also confined to conduction in the diffuse
layer. We will,therefore, continue to regard the conduction process in the Stern layer as
anomalous surface conduction and look at its effects after we have examined the primary
problem. Lyklema refers to this classical treatment as referring only to rigid particles but
that would seem to imply that Stern layer conductance is in some way connected with
Stern layer fluidity or distortability which is certainly not the intention.
Although it is not possible to directly test all the features of this classical model,
there is an obvious indirect test: by comparing an electrokinetic measurement on a
well-characterized suspension with the corresponding theoretical value obtained with
the standard model it is possible to determine the equilibrium particle charge or {-
potential on the shear plane. If the standard model is correct, such comparisons should
yield the same potential for every type of electrokinetic measurement carried out on
that suspension. It will be clear from the results of Section 8.3 that a true test of
consistency must involve correlations other than those which can be predicted from
the Onsager reciprocity relations, since such correlations (e.g. between streaming
potential and electro-osmosis (eqn (8.3.2)) remain true even when the simple model
breaks down. One valuable insight provided by the argument over surface conduction
is the realization that when anomalous surface conduction is occurring it leads to an
overestimate of 5 from conduction and dielectric dispersion type meansurements and
an underestimate from electrophoresis measurements. If, as seems to be the case, that
discrepancy can be removed by the introduction of Stern layer conduction, then that
would seem to be a vindication of the approach.
In any case, in order to apply a consistency test it is necessary to have theories for a
number of different electrokinetic effects. In the following two sections we will set out
the mathematical problem that must be solved in the development of such theories.
an,lat = -v . f j (8.7.1)
where nj is the number of ions of type j per unit volume, and& is the flux density (i.e.
the number of ions per unit area of type j passing through the surface of a volume
element). The term on the left represents the rate of change of the ion number in a unit
volume while the term on the right is the nett rate at which ions enter that volume (see
Fig. 1.5.3 and Appendix A3).
The local flux density& is related to the local fluid velocity and gradients in ion
density and electrical potential. It is in the formulation of this relationship that
THE STANDARD DOUBLE LAYER MODEL 1397
assumptions must be made about the local transport properties of the double layer. In
the standard double layer model, the flux density is assumed to be given by
+
where D, and zje are the diffusivity and charge of the ion, is the electrical potential,
and v the fluid velocity. The terms on the right of this expression represent the
contributions from Brownian motion, the local electric field, and convection with the
surrounding fluid respectively. The combination of eqns (8.7.1) and (8.7.2) yields a set
of differential equations, one for each species of ion, in which the unknowns are nj, +,
and v.
+
The quantities nj and are also related by Poisson’s equation (Section 7.3):
..N
v2+= - C zieni/r (8.7.3)
i= 1
where the permittivity of the electrolyte E is the same everywhere. The sum in this
expression represents the local free charge density, and the summation is carried out
over the N ionic species in the electrolyte. In most applications the electrolyte can be
treated as incompressible, in which case the mass conservation equation takes the form:
v . v = 0. [8.2.2]
The set of equations is completed by Newton’s Second Law, applied to unit volume of
fluid, viz (compare with eqn (4.6.5)):
N
pfavlat = qv2v - VP - C njzjev+. (8.7.4)
j= 1
Here p is the pressure, and pf and q the fluid density and viscosity respectively. In
accordance with the standard double layer model (Section 8.6) q is assumed to be
uniform across the double layer. The terms on the right hand side of the equation
represent the nett frictional force, the pressure force, and the electrical force per unit
volume respectively. The inertia term on the left hand side has been linearized on the
grounds that the Reynolds number of the flow (eqn (4.8.1)) is small for colloidal
systems.
At equilibrium the fluid velocity and ionic fluxes are identically zero. The
combination of eqns (8.7.2) and (8.7.3) under these conditions yields the familiar
Poisson-Boltzmann equation (7.3.9) while eqn (8.7.4) gives the osmotic pressure
distribution in the electrolyte (Exercise 8.7.2).
The solution of the electrokinetic equations in a non-equilibrium situation is a
formidable problem, thanks to the presence of the non-lineart terms nj v and njV+. T o
overcome this problem we consider systems which are only slightly disturbed from the
equilibrium state; these terms can then be approximated by linear expressions. This
approximation, which greatly simplifies the calculation is common to nearly every
theoretical study in this area.
We use the symbol to denote an equilibrium quantity and a prefix 6 to denote the
+
small disturbance. Hence the ion density is written as nj = nj 6nj where 6njln; is
assumed to be much less than unity. On substituting this expression and a similar form
+
for in eqns (8.7.1), and (8.7.2) and combining, we get (Exercise 8.7.1)
(8.7.6)
Since this set of equations is linear, a sum of solutions is also a solution. This property
greatly simplifies the calculation of electrokinetic effects and we will concentrate on
such linearized equations in this chapter.
v=o (8.7.7)
on the particle surface. Since the surface is impervious to ions the component of the
ion flux normal to the surface is also zero, that is
fj.i=o (8.7.8)
where & is the unit normal directed outwards from the surface. The final boundary
condition at the particle surface arises from electrostatics, viz (compare with eqn
(7.3.33)):
€(Z),-
€.(Z) P
= 6a (8.7.9)
THE STANDARD DOUBLE LAYER MODEL 1399
where a/& is the direction normal to the surface and 6a is the change in surface charge
density due to the external disturbance. The subscripts p and f denote the values on
the particle side and the fluid side of the interface respectively. [Note that we will also
be using the symbol n for number of ions per unit volume but it will always be
subscripted or superscripted.]
The remaining boundary conditions vary from one problem to the next. In the
electrophoresis problem, for example, we can say that with increasing distance from
the particle:
where E is the applied electric field. In addition, the fluid motion caused by the particle
movement approaches zero at large distances. Thus in a frame of reference moving
with the particle, the liquid velocity far from the particle?, ~(oo),is opposite to that of
the particle:
where U is the sedimentation velocity, while Vy9 tends to the uniform value required to
balance the current flow caused by the sedimentation process (Saville 1982). It is this
field which gives rise to the ‘sedimentation potential’, the voltage difference between
the top and bottom of a sedimenting charged suspension.
For concentrated suspensions and porous media it is not possible to treat each
particle as being alone in an infinite liquid. In that case the above outer boundary
conditions are replaced by the requirements that the volume averages of VGnj,V6$,
and v(00) or Vp take the prescribed macroscopic values. For example, in the case of a
porous medium subjected to a steady macroscopic field E, and a zero applied pressure
gradient, the appropriate boundary conditions are
where ( ) denotes a volume average over a representative portion of the plug away from
the boundaries (where some electrolyte inhomogeneities may have built up).
+Weassume, to simplify the exposition, that the particle moves in the direction of the
applied field without rotating.
400 I 8 : ELECTROKINETICS AND THE ZETA POTENTIAL
I
Exercises
+ +
8.7.1 Establish eqns (8.7.5) and (8.7.6). [Note that V(# @) = V# V@ and
+
v . (#v) = v -V# #V. v.]
8.7.2 Show that when a flat plate is immersed in an electrolyte, the equilibrium ion
distribution (from eqn (8.7.2)) is a Boltzmann distribution.
(8.8.1)
(8.8.2)
DOUBLE LAYER D Y N A M I C S I401
1 1
2
h = -(an1 + 6nz) 2
and g = -(an1 - 6nz). (8.8.3)
Note that the charge density is 2gze where z is the valency of ion type 1. The function g
thus monitors the local changes in charge density whilst h monitors the local changes in
salt concentration.
In terms of h and g the eqns (8.8.2) for both ion species become (Exercise 8.8.2):
and
aglat = v(v2g - K’g) (8.8.5)
where
K’ = 2 n0z2e2/EkT
a0 being the equilibrium density of either ion species. [In this case a0 = noo, the ion
concentration in the bulk electrolyte, so K has its usual significance (Section 7.3. l).]
By symmetry, h and g must be the same for all points at a given distance from the
plate, so h and g depend only on t and x and they must both decay to zero far from the
plate. The remaining boundary conditions are
(8.8.6)
The solution for h is simply h = 0. That is to say, the charging of the plate produces
symmetric and opposite changes in the concentrations of the two ions at each point in
the solution near the plate. This symmetry was noted earlier (Section 7.4.1). It is a
consequence of the linearization of the Poisson-Boltzmann equation for small
perturbations. This amounts to the introduction of the Debye-Huckel approximation.
The solution to the problem for g is found, using Laplace transform techniques (see,
Stephenson 1977 Chap 7, 8), to be [this is a correction to the earlier text provided by
Dr M. Minor]:
K(T
g(x, t) = -{exp(-Kx)erfc[X - T] - exp(lcx)erfc[X
4ze
+ q) (8.8.7)
Next Page
where the error function, erfc, is defined by [compare with eqn (5.4.3)]:
The time required for the double layer to reach this equilibrium form can be seen from
Fig. 8.8.1, where g(x, t) is plotted as a function of KX for three different values of K2Dt.
The equilibrium form is indicated by the broken line. Clearly the double layer is
approximately in equilibrium if t is larger than a few multiples of ~ / K ~This D . quantity
therefore provides a measure of the double layer relaxation time and for lop3 M KC1
the time is approximately lo-’ seconds. In most electrokinetic studies the time scale of
the macroscopic disturbance is much greater than 1//c2D,and hence the double layers
are at each instant in local equilibrium with the surrounding electrolyte. This
observation forms the basis of much of the recent work on thin double layer
electrokinetics (Section 8.9). We will find (Section 8.8.3) that when a double layer
system (such as a colloidal suspension) is subjected to a rapidly varying electric field
1
0.9
+k2Dt=O.l
0.8
--t k2Dt=0.5
0.7
0.6
.-
d
\
0.5
aJ
fi 0.4
cu
0.3
0.2
0.1
0
0 0.5 1 1.5 2
kx
Fig. 8.8.1 The form of the charge density as a function of distance from a flat conducting surface at
a number of instants after the appearance of a uniform charge on the plate. [Figure kindly provided
by Dr M. Minor (2003 priv. comm.).] The abscissa is KX and the ordinate is 2zeg/~awhilst the
parameter is T2 = 2 D t .
Association Colloids
9.1 The critical micellization concentration (c.m.c)
9.2 Factors affecting the c.m.c.
9.2.1 Effect of head group and chain length
9.2.2 Effect of counterion
9.2.3 Effect of temperature and pressure
9.2.4 Effect of added salt
9.2.5 Effect of organic molecules
9.3 Equilibrium constant treatment of micelle formation
9.3.1 The closed association model
9.3.2 Multiple equilibrium models
(a) Dimerization
(b) K,, values of similar magnitude
(c) Strong dependence of K,, on tz
9.4 Thermodynamics of micelle formation
9.4.1 Estimation of ae
9.4.2 Estimation of z ( h g )
9.4.3 Choice of standard state and concentration units
9.4.4 Enthalpy and entropy of micelle formation
9.5 Spectroscopic techniques for investigating micelle structure
9.5.1 Methods for determining the c.m.c.
(a) Solubilization of additives
(b) Spectral change of additives
9.5.2 Fluorescence methods for determining aggregation numbers
(a) Steady-state emission quenching method
(b) Time-resolvedemission quenching method
9.5.3 Interfacial electrostatic potential of micelles using solubilized pH
indicators
9.5.4 Polarity of the micelle-water interface
9.6 Micellar dynamics
9.6.1 Kinetics of micelle formation
9.6.2 Residence times of probe molecules in micelles
434
THE CRITICAL MlCELLlZATlON CONCENTRATION (C.M.C.) I 435
9.6.3 Determination of microfluidity using fluorescence probes
9.6.4 Other spectroscopic techniques
9.7 Molecular packing and its effect on aggregate formation
9.8 Statistical thermodynamics of chain packing in micelles
c.m.c.
Concentration of surfactant
Fig. 9.1.I Schematic representation of the concentration dependence of some physical properties
for solutions of micelleforming amphiphiles. (After Lindman and Wennerstrom 1980, with
uermission.)
436 I 9: ASSOCIATIONC O L L O I D S
derived from the same conductance data can vary significantly, depending on whether a
plot is made of the molar conductance against C’/’ or the specific conductivity
(conductance) against log C.) Despite these limitations the concept of critical
micellization concentration remains an important one. It can be defined in terms of one
or other of the properties suggested by Fig. 9.1.1 but a more general definition is
(Phillips 1955):
(9.1.1)
where q5 is any one of the properties (Exercise 9.1.1) and CT is the total concentration
of the amphiphile or surfactant.
Micelle formation is only one of a number of characteristic aggregation phenomena
which amphiphilic molecules undergo and one might well ask what causes this
behaviour pattern, intermediate as it is between true solution and a separation of the
components into two distinct phases. After all, it does not occur with long chain
alcohols, amides, or amines which also have a polar head group and a lipophilic or
hydrophobic chain attached. It seems to be necessary to have either a charged head
group (carboxylate, sulphate, or quaternary ammonium), a zwitterionic group, or a
rather bulky oxygen-containing hydrophilic group (polyoxyethylene, phosphine oxide,
amine oxide, or a sugar residue) able to undergo significant hydrogen bonding as well
as dipolar interaction with water. (We concentrate mainly on the formation of micelles
in aqueous solution because the characteristics of micelles formed in non-aqueous
solutions are less well established. For strongly hydrogen bonded solvents (e.g.
formamide and hydrazine), however, the behaviour is similar in many respects to water
(Kresheck 1975). Evidently, the balance of forces which leads to micelle formation is a
subtle one. If the hydrophilic effect is sufficiently strong the molecule can enjoy
complete solution; if it is too weak the substance is merely insoluble (e.g. octadecanol).
The structure of micelles has been a subject of controversy for many years, since the
pioneering works of Hartley (1936) and McBain (1950) arguing the cases for spherical
and lamellar structures respectively. The spherical (or near-spherical) form was
generally accepted as the dominant species in dilute aqueous solutions until recent
times. Tanford (1980) in his extensive review of the mechanism of micelle formation
prefers the description ‘disc-like’ since the weight of evidence from transport and
equilibrium properties suggests that the structures are better described as oblate
spheroids (Fig. 5.1.1). This view is not, however, universally accepted and the most
elaborate theoretical calculations (to be discussed in Section 9.8) are based on a
spherical shape, at least for the dodecyl sulphate micelle near its c.m.c. Throughout the
following discussion it would be wise to bear in mind the fact that the forces
controlling micellar structure are delicately balanced so that distortions of the shape
(either due to the shearing processes or fluctuations) may occur fairly easily (but see
Section 9.7).
Furthermore, any attempt to represent the structure can at best be a statement of
the average shape over some discrete time interval. The exchange that occurs between
monomer molecules and those in the micelle occupies a time-scale of the order of 1-10
ps (Aniansson 1978) and the entry or exit of a monomer presumably occurs by a series
THE CRITICAL MlCELLlZATlON CONCENTRATION (C.M.C.) I 437
of diffusive steps, one methylene group at a time. The surface must, therefore, be
somewhat rough on the nanosecond time scale (Section 9.6).
More detailed thermodynamic analysis (Section 9.4) reveals some, at first sight,
quite surprising results. For one thing, the aggregation process in water at room
temperature is accompanied by a significant increase in entropy, which is the main
contribution to the negative AGO value for micellization. The A@ value is usually
small, at least for systems with small degrees of aggregation (50-loo), and often slightly
positive. The traditional view of the mechanism of micelle formation has been based on
the study of the (very slight) solubility of hydrocarbons in water, and what has come to
be known as the hydrojhobic effect (Tanford 1980).It seems that, at room temperature,
the presence of a hydrocarbon molecule in water causes a significant decrease in the
(partial molar) entropy of the water, suggesting that it induces an increase in the degree
of structuring of the water molecules. The isolated hydrocarbon molecule forms a
cavity in the water structure and the walls of that cavity are lined with water molecules
with a bonding pattern that differs, on average, from the bulk pattern and that,
furthermore, varies in a complex and subtle way with change in temperature. The
predominant effect of a hydrocarbon molecule is to increase the degree of structure in
the immediately surrounding water and this is one of the main features of the
hydrophobic effect. The other major effect is to disrupt the extensive hydrogen
bonding pattern in the water. Evidently the entropy increase associated with this latter
process is outweighed by the energy increase involved and its contribution to AGO is
again positive (Frank and Evans 1945).
When hydrocarbon residues aggregate in aqueous solution to produce a micelle, the
reverse process occurs: the hydrogen bonding structure in the water is, to a large
extent, restored. For this process, both enthalpy and entropy changes are negative.
The 'melting' of the cavities that surrounded the hydrocarbon chains gives rise to an
entropy increase in the water that more than compensates for the decreased
randomness of the hydrocarbon chains as they enter the micelles. This view of the
micellization process, in which the principal driving force is the partial molar entropy
increase of the water, has been strongly challenged by some studies of aqueous systems
at high temperatures (up to 166 "C) and micellization in pure hydrazine solutions
(Ramadan et al. 1983) and we will return to this discussion when we have established
the basis for a thermodynamic analysis (Section 9.4).
Since the head group remains surrounded by water its contribution to the energetics
of micellization is much less but its role is essential. It is the nature of the head group
and the interactions that occur between head groups that, in principle, determine the
size and shape of the aggregate structure (Israelachvili et al. 1976). We will however,
defer more detailed discussion of these matters until later (Section 9.7). We will first
look at the phenomenology of micelle formation - what effect factors like
temperature, pressure, and salt content have on the c.m.c. We will then examine
some of the simple mass-action models of micelle formation and discuss in more detail
the thermodynamics of the process.
We then consider some of the methods (especially spectroscopic procedures) that
have been used to study other features of micellar structure and dynamics. T h e
chapter concludes with a brief description of the current ideas about the geometric
factors that determine aggregate shapes and the statistical mechanical description of
a typical micelle. More detailed discussion of these matters will be deferred to
438 I 9: ASSOCIATIONC O L L O I D S
Chapter 14, when the theory of liquid structure and scattering behaviour can be
brought to bear on the problem.
Binks (1999) provides a survey of the current microscopic, optical and non-optical
techniques being applied to the characterization of surfactant systems. These include,
among others, S T M (Section 6.1.6) and AFM (Section 6.2.2), surface light scattering
(Section 3.2 and 5.7) and reflection and non-linear optical techniques as well as X-ray
and neutron reflectivity (Section 14.5), electroacoustics and ultrasonics (Section 5.9).
Exercises
9.1 . I Show, using appropriate sketches, that eqn (9.1.1) is a general statement of the
fact that the plot of the property ~ ( C Tagainst
) CT undergoes a sharp change
in slope at CT = c.m.c.
where bo and bl are constants. Figure 9.2.1 illustrates this point for a number of typical
ionic and non-ionic surfactants. Some of the reasons for the differences in bo and bl
values revealed by this figure and Table 9.1 will be discussed below (Section 9.4.2). It is
hardly surprising that the nature of the head group should effect the value of bo, but it
is also apparent that it profoundly effects bl as well.
It should also be noted that the non-ionics usually have much lower c.m.c.s than the
ionics despite their generally larger bo values. Most ionics of given chain length show
very similar values for c.m.c. Lindman and Wennerstrom (1980) quote, for the straight
chain Clz (dodecyl) surfactants the following values (with the counterion in brackets):
.z -7
VJ
h
5
F
.Ei
* -6
0 -5
-
4
i2 -4
?
c!
-3
v
M
2 -2
6 8 10 12 14 16 18
Fig. 9.2.1 Plots ofloglo c.m.c. (in mole fraction units) versus m,, the number of carbon atoms in the
alkyl chain. (Temperature in general 25 "C.) (a) Alkyl hexaoxyethylene glycol monoethers. (b) Alkyl
trimethylammonium bromides in 0.5 M NaBr. (c) N-alkyl betaines. (d) Sodium alkyl sulphates
(40 "C). (e) Sodium alkylcarboxylates. (From Lindman and Wennerstrom 1980, with permission.)
Table 9.1
Values of bo and bl in eqn (9.2.1) for various surfactants. (Selectedfrom Kresheck 1975.)
+These values are likely to be p H dependent because this amphiphile becomes cationic at low pH.
c.m.c., but the introduction of a benzene ring is equivalent to adding about 3.5
methylene groups to the chain length. Introducing fluorine in place of hydrogen has a
quite marked effect, at first increasing and then decreasing the c.m.c. (ultimately to less
than 10 per cent of the hydrocarbon value), as the proportion of fluorine is increased
towards saturation (Mukerjee and Mysels 1975).We will return to this point in Section
9.6.4.
the explanation of smaller variations. [The lyotropic series is an ordering of ions of the
same valency according to their size. We will have a little more to say about it in
connection with the coagulation behaviour of salts (Chapter 12).] The values for bo and
bl quoted in Table 9.1 for the sodium and potassium salts of carboxylic acids are,
however, rather more difficult to rationalize. Small differences in bo can be attributed
to differences in degree of hydration and cation binding to the head group, but it seems
surprising that the bl value should vary so much in this case.
and
log(c.m.c) = b4 + b5 log C (ionic). (9.2.3)
The constants, bi, depend upon the nature of the electrolyte (Fig. 9.2.2). For the
ionics, the principal effect of the salt is to partially screen the electrostatic repulsion
between the head groups and so lower the c.m.c. Values of b5 of from -0.6 to -1.2 are
given (Kresheck 1975) for the influence of sodium salts on sodium carboxylates. The
more subtle influences of salts of the same valence type are again usually discussed in
terms of the lyotropic series.
FACTORS AFFECTING THE C.M.C. 1441
c.m.c+CNaCl
(mol L-’1
Fig. 9.2.2 The effect of added salt on the c.m.c. of SDS and dodecylaminehydrochloride(DHC).
(From Stigter 1975a, with permission.)
For the non-ionics the concentrations of salts required to produce significant effects
are much higher and the discussion of such behaviour introduces the notion of ‘salting
in’ and ‘salting out’ of non-electrolytes by the electrolyte (Ray and Nemethy 1971).
This amounts to a description of the competition between the surfactant (chiefly the
head group) and the electrolyte for the opportunity to associate with the water. T o put
it more precisely, the activity coefficient of the monomer surfactant changes as the
electrolyte concentration and type alters. If the monomer is salted out by electrolyte
then micellization is thermodynamically favoured and the c.m.c. is reduced. The
reverse situation applies if the monomer is salted in.
The short-chain alcohols like ethanol and methyl propanol-2 can act as both group A
and B materials.
EQUILIBRIUM CONSTANT TREATMENT OF MICELLE F OR M A T ION 1443
Exercises
9.2.1 Estimate the c.m.c.s for the sodium salts of the CS, C10, C12, and C14 straight-
chain alkyl sulphates from the data in Table 9.1. The experimental values at
25 "C are:
130.3 33.0 8.08 and 2.05 mM
and at 40 "C are:
136 33.5 8.7 and 2.21 mM, respectively.
Comment on the relation between the results.
[micelles] --C,
K= - (9.3.2)
monomer^]^ C; .
The inherent assumption here is that the activities may be replaced by concentrations.
For the monomer this amounts to assuming that the only departure from ideal
behaviour is the aggregation process. It could, in principle, be removed by estimating
other activity corrections from solution theory. Assuming ideal behaviour for the
444 I 9: ASSOCIATION COLLOIDS
micelles is more problematical because of the large size difference between monomers
and micelles. The micelles will also interact strongly. For ionics the interaction will
become very significant as soon as the mean separation is less than about (8-10)/~,
where K is the Debye-Huckel parameter (Section 7.3), and that occurs at surfactant
concentrations not far above the c.m.c. (Exercise 9.3.1). It must also be noted that
when ionic micelles are formed there is a strong tendency for the counterions to be
associated closely with the head groups, because of the high electrostatic potential in
that region. This is a further source of non-ideality, which is discussed in Section 9.5
(Evans and Ninham 1983).
From eqn (9.3.2) we have (Exercise 9.3.4):
Mukerjee (1975) shows how eqn ( 9 . 3 4 , with a K value of unity and n = 100 gives rise
to a sharp transition from a system in which all of the surfactant is present as monomer
Fig. 9.3.1 Variation of dC,,,/dCT with total surfactant concentration for different values of the
aggregation number, n. COis the critical micellization concentration and C, the concentration of
micelles.
EQUILIBRIUM CONSTANT TREATMENT OF MICELLE F OR M A T ION 1445
to one in which the monomer concentration remains essentially constant above the
c.m.c. and all additional surfactant goes into micelle formation (Exercise 9.3.2). As an
alternative approach (Exercise 9.3.3), eqn (9.3.5) can be differentiated to obtain:
(9.3.6)
The plot of dC,,,/dCT against concentration for K = 1 and various values of n is shown
in Fig. 9.3.1.
When it is realised that for most of the commonly used surfactants the value of n is at
least 50 it becomes clear that the concept of a critical micellization concentration (being
the concentration in the neighbourhood of which, micelle formation begins) is a
reasonable one. As n+ 00 the transition becomes sharper and ultimately approaches
the behaviour expected of a first-order phase transition. In this extreme case it is
possible to treat the micelle formation as a phase separation. For smaller values of n
there are problems involved in the phase separation model (Hall and Pethica 1967) and
these can only be properly resolved by resorting to the formalism of small system
thermodynamics (Hill 1963).
(9.3.7)
The overall association constant, then, for the formation of x, from xi (i.e. n q $ x,) is:
(i.e. "K, is the product of all stepwise association constants K2, K3,. . ... up to K,.).
Again in defining "K, we replace activities with concentrations with all of the
uncertainties previously referred to.
The concentration of all solute species, S, (in mol dmP3) is:
(9.3.9)
(9.3.10)
Then the number average degree of association of all species (including the monomer)
is (see Section 5.3.2):
(9.3.11)
(9.3.12)
We normally concern ourselves only with associated species and exclude the monomer.
In these terms, the corresponding values of number and mass average are:
(9.3.13)
(9.3.14)
It is evident from eqn (9.3.8) that, as [XI] increases, the concentration of each
associated species increases. The value of n influences each associated species and the
percentage increase of each increases with n. From eqns (9.3.11)-(9.3.14) both number
average and mass average increases. On the other hand, with increasing dilution A41 +
0, N,+ 1, xw
+ 1 and M I M [XI].
The individual K, values define the concentrations at which particular n-mers
become most important and, hence, they control the values of Enand ww at any
concentration. The products of association and their size distribution depend on
concentration and on K,(n), i.e. how K, depends upon n, which is a reflection of the
molecular architecture of the associated species.
Three main types of association behaviour may be identified:
(a) simple dimerization (K2 dominant);
(b) formation of micelles with a wide range of aggregation numbers (all values of K,
approximately equal); and
(c) formation of micelles with a narrow size range (strong dependence of K, on n).
EQUILIBRIUM CONSTANT TREATMENT OF MICELLE F OR M A T ION 1447
The first two cases are rather trivial, while the third covers most of the interesting
micelle-forming compounds and will, therefore, be discussed in more detail.
(a) Dimerization
The formation of dimers will of course take place in all self-associating systems.
Whether or not the process is limited to dimer formation or continues predominantly
on to multimers will be decided by the value of KZin comparison with other K, values.
Dimerization appears to be restricted to dilute aqueous solutions of some flexible chain
surfactants, such as carboxylic acids (Oakenfull and Fenwick 1974) and to solutions of
some bile salts, notably sodium cholate (Small 1968; Oakenfull and Fisher 1977). Both
of these are rather special cases, in which the structure of the interacting molecules
favours formation of a ‘closed’ dimer. The carboxylic acids for example, show
cooperative hydrogen bonding to form a cyclic structure:
Bile salts may show a similar interaction, with three cooperative hydrogen bonds,
although the evidence for this has been disputed (Zana 1978; Oakenfull and Fisher 1978).
(b) K, values of similar magnitude
For the equilibrium between monomers and micelles (eqn. (9.3.l)), let
K2=K3= . . .K,=K. (9.3.15 )
The total concentration, S, can now be directly related to the equilibrium monomer
concentration [XI]. Defining X = K [XI] then
+ + + . . . . . . .Kn-’[~1IM-’)
= [ ~ l ] ( l K[x~] K2[x1I2
= [X1](1 + x + x2+ x3+ . . . . x n - l ) (9.3.16)
-
-- IX1l for large n,
1-x
if X < 1, which it must be in real systems that conform to the association scheme
outlined above. MI, the total monomer concentration, now becomes (Exercise 9.3.5):
(9.3. 7)
and thus
(g) 112
= 1 - K[xl] (9.3.
448 I 9: ASSOCIATIONC O L L O I D S
from which the modified number and mass average degrees of association (eqns
(9.3.13) and (9.3.14)) follow (Exercise 9.3.6):
(9.3.22)
which approaches 1 for small X and 2 as X approaches 1. The breadth of the size
distribution thus increases with the degree of association. It is not always broad as is
suggested by Mukerjee (1975).
This model often gives a good description of the behaviour of molecules that are
rigid and flat, with faces of approximately equal hydrophobicity. Such molecules
associate by a simple stacking arrangement as had been demonstrated with the cationic
dye, methylene blue (Mukerjee and Ghosh 1970). As the size of the stack increases,
charge repulsion builds up and K, gradually decreases at higher concentrations. This
association model is also a fairly good representation of the behaviour of many
nucleosides and of the stability of double stranded DNA (Ts’o 1968). In both cases,
stacking interactions occur between the organic bases.
(c) Strong dependence of K, on n. This is the situation that was crudely examined
in the closed association model. The delicate force balance involved in micelle
formation leads, not to an increase or decrease in K, with n, but rather to a value of n
for which K, is a maximum. The values of K,-1 and K,+1 will be of comparable
magnitude but one will expect to find the resulting size distribution to be fairly narrow.
EQUILIBRIUM CONSTANT TREATMENT OF MICELLE F OR M A T ION 1449
As noted earlier (Section 9.1), the existence of a preferred micelle size is easily
explained on the basis of the competing requirements for bringing the hydrocarbon
chains into intimate contact, away from the aqueous environment, while maintaining
the head groups as far apart as possible. There is very little energy change in separating
the hydrocarbon chains from the water (Exercise 9.1.2) and the negative free energy
associated with that process stems largely from the concomitant entropy increase in
the water. In the case of an ionic surfactant, as each additional monomer is added to the
micelle, the contribution to the free energy change becomes less negative because the
developing micellar charge causes an increasing (positive) free energy change,
reflecting the repulsion between the head groups.
Even when n is large, a relatively broad maximum in the values of K, may produce a
narrow size distribution. The free energy change involved in the formation of the n-
mer from the monomer is, from eqns (9.3.3) and (9.3.8):
and only a slight minimum in the plot of AGt/n against n is required to produce
narrow size distributions of micelles, as Stigter and Overbeek (1957) have shown.
Mukerjee (1975) gives a striking illustration of this point. Using the following
empirical expression for In *K,:
0 25 50 75 100
n
Fig. 9.3.2 Variation in the concentration of monomers existing in the form of micelles (n[x,]), as a
function of the number of monomers per micelle, n, for an assumed free energy profile, AG:/n.
(From Mukerjee 1975, with permission.)
450 I 9: ASSOCIATIONC O L L O I D S
which exhibits a very broad maximum (Fig. 9.3.2) Mukerjee calculated the values of
n[x,] from eqn (9.3.8) (i.e. the concentration of n-mers expressed in terms of monomer
concentration), for a monomer concentration of 4.11 x mol L-' (i.e. ln[xl] =
-10). The result is a narrow size distribution (Fig. 9.3.2) peaking at n = 97 and with a
half-width of less than 10, even though A e / n changes by less than 2 per cent over the
whole range from 70-120 for n.
This figure illustrates clearly why the calculation of the anticipated size distribution
in a particular case requires accurately measured "K, values near the maximum or
good estimates of A G . We will examine the extent to which A G can be estimated in
the next section.
When the monomer chains are very long, very large polydisperse aggregates
apparently form, even in dilute solution (Debye and Anacker 1951). These micelles are
thought to be flexible cylinders and may be described by a self-association model
similar to that discussed above (Mukerjee 1974). The molecular requirements of
micelles of various geometries will be discussed in Section 9.7. Our immediate aim is to
estimate the value of AG: as accurately as possible and, from this, to determine the
micelle size distribution and other properties.
Exercises
9.3.1 Estimate the mean centre-to-centre distance between the micelles of sodium
dodecyl sulphate (SDS) when the surfactant concentration is 5 x lo-' M.
Assume that the c.m.c. is 8 mM. Show that this corresponds to about 4 / when~ K
is calculated on the basis of the residual monomer concentration. (Note that the
minimum micelle radius is about 2.5 nm, corresponding to a stretched C12 alkyl
chain.)
9.3.2Take K = 1 and n = 100 in eqn (9.3.5) and discuss the variation of C, (=CT - nC,)
and nC, with C,/Co in the range 0 5 CT/CO5 3. (Assume that in this formulation
the concentration of micelles is expressed in moles of micelles per litre.)
9.3.3Establish eqn (9.3.6).
9.3.4Integrate eqn (A5.11) to establish eqn (9.3.3) for the case where only P,Vwork is
involved. (Use the fact that AG = 0 at equilibrium if P and Tare constant.)
9.3.5Establish eqn (9.3.17) using the series expansion for (1-a-'. Establish eqns
(9.3.19) and (9.3.20) by showing that Z = [xl](l a3.
+ a/(]-
9.3.6Establish the expressions (9.3.21) and use them to derive (9.3.22).
This would give the contribution per mole of micelles of size n, assuming ideal
behaviour. (The use of mole fractions is connected with the most appropriate
choice of standard state and will be discussed in more detail in Section 9.4.3.)
Again (cf. Fig. 9.3.2) it is more convenient to express this contribution in terms of
the concentration of the monomeric surfactant:
RT In (X,/n) where X , is the mole fraction of monomer in micelles of size n. The
cratic contribution per mole of monomeric amphiphile is l / n of this and SO:
(9.4.1)
[Do not confuse X , with [x,] as used in Section 9.3; the latter is the concentration
expressed in terms of moles of n-mers.] Note that in this formulation, each of the
micellar sizes is treated as a separate component, with its own standard state chemical
potential. Equating pmic,,with the value for the free surfactant gives:
and Tanford (1980) points out that if the size distribution is reasonably narrow, the
value of n" is experimentally indistinguishable from the number average or mass
average micellar size. All can then be set approximately equal to a mean size, n. At this
level of approximation, all micelles are treated as the same and lumped together with a
standard chemical potential, pLic:
(9.4.5)
The activity of the free surfactant a1 is, of course, given by ylX1 where y1 is the
activity coefficient and it is tempting to assume that y1 M 1, since X I is usually fairly
452 I 9: ASSOCIATIONC O L L O I D S
small. (Even above the c.m.c. the concentration of the free surfactant remains close to
the c.m.c. value.) The problem with this procedure is that even at the concentrations
normally encountered, ionic surfactants exhibit activity coefficient effects due to
interactions other than micelle formation. Some account could be taken of such effects
using, say, the extended DebyeHiickel theory but the experimental procedure of
extrapolating to infinite dilution breaks down in this case because the presence of the
micelles may itself influence the interactions between the free surfactant molecules.
This effect may not be as serious as Tanford (1980) suggests, however, since the free
surfactant ions will tend to be excluded from the double layer regions surrounding the
micelles. For non-ionic surfactants the activity coefficient correction can probably be
dispensed with altogether with negligible error (Desnoyers et al. 1983).
The relation between the standard chemical potential change and the c.m.c. (Xo)can
be obtained from eqn (9.4.5) by introducing the ratio cr = Xmic/Xoand recognizing
that, at the c.m.c., XI = XO- Xmic.We then have (Exercise 9.4.2):
(piic- &/RT = [(E - l)/?i]lnXo + lnyl + ln(1 - o)+ (1/E) In @/a). (9.4.6)
T o estimate values of (pLic- p!) thus requires a knowledge of the c.m.c. (XO)and the
mean aggregation number, E. The value of cr can be taken to be anywhere from 0.01 to
0.1 with negligible effect on the result but, of course, one has to either assumeyl = 1 or
make some correction for it. This latter correction should be relatively unimportant if
one wishes only to evaluate the change in @Lie
- p;) as the chain length changes.
Indeed, for many purposes, the following approximation, valid for large ki and small cr
(Exercise 9.4.3) is sufficiently accurate:
0 -
pIL;, - p1 = AGO= R T In x,. (9.4.7)
Note that this is identical to eqn (9.3.3a) if n is very large and the surfactant
concentration is expressed in mole fraction terms. It cannot be applied in this form to
ionic surfactants, for which a further correction is essential. (See Section 9.5 below.)
Further discussion of the thermodynamic determinants of micellar size and shape is
given by Missel et al. (1980). See also Ekwall et al. (1971).
~
(9.4.8)
+ AGo(hg)
~~ ~ ~
9.4.2Estimation of AGO(hg)
~
The contribution of the hydrophilic head group AGo(hg) to the free energy of
micellization is invariably p o s i t i v e d s o opposes the process. Very little progress has
been made on the calculation of AGo(hg) for non-ionic molecules. It is assumed to
arise from steric interactions as the large head groups crowd the surface but beyond
that there is little that can be said from a theoretical point of view. The contributions of
different non-ionic head groups can, however, be evaluated from experimental data.
By contrast, more progress has been made on the calculation of AGo(hg) for ionic
surfactants, assuming that it is dominated by the electrostatic effects predictable from
the Gouy-Chapman theory of the double layer. We consider the energy change
involved in establishing an n-mer (of charge q = ne). The electrical contribution per
mole of monomer is calculated by imagining small increments of charge dq to be
transferred to the initially uncharged spherical n-mer:
where q' and $0' refer to the charge and potential on the micelle surface during the
charging process. The problem is that, for spherical particles, there is no explicit
relation between the surface charge, q, and the surface potential. Only at low potentials
(strictly $0 < 25 mV) can the DebyeHuckel relation (eqn (7.11.6)) be invoked:
where e is the proton charge. Unfortunately, this approximation is rarely, if ever, valid
in micellar systems. Stigter (19754 has shown, however, that a more exact calculation,
using the computer-calculated relation between $0 and q for spherical particles, can
satisfactorily account for the variation of c.m.c. with salt concentration in SDS solutions
at 25 "C (Table 9.2) over the range from 0 to 0.2 M sodium chloride. In this region, the
micellar aggregation number changes over less than a factor of two (from 65-1 10) and
the concentration of free surfactant remains approximately equal to the c.m.c.
454 I 9: ASSOCIATION COLLOIDS
S In Xo = 6 In Co = S[AGO(hg)/RT] (9.4.13)
where 6 measures the change compared to the value in the absence of salt. Equation
(9.4.13) assumes that CO (mol L-l) c( XO at low concentrations. The agreement
between columns 5 and 6 is excellent, suggesting that double-layer theory gives a good
account of the effect of added salt, at least up to 0.2 M. The remaining discrepancy is
of the order of the activity coefficient correction for unassociated ions and this has not
been included. The DebyeHuckel procedure would obviously give a very poor
representation, being some 5&65 per cent higher at all concentrations. Note that in ~
this approach, no attempt was made to relate the absolute magnitude of AGO(e1ec) to
~
the experimental estimates of A@(hg). That would require an assumption about the
degree of ion binding and a choice of the appropriate value of 11.0 (or 11.d) to characterize
the surface. More elaborate calculations of this nature are given by Stigter (19756). In
Table 9.2 it is assumed that ion b i n d i n d n o t affected over the concentration range
involved. A more recent calculation of AGO(elec)along similar lines is given by Evans
and Ninham (1983).
At somewhat higher salt concentrations (above about 0.45 M) the aggregation
number of SDS increases dramatically to over 1000 and a more elaborate model is
required (see, for example, Mazer et al. 1977 and Gunnarson et al. 1980).
The ionic strength effect (Table 9.2) has another important consequence that was
noted in connection with Table 9.1. The constant 61, which measures the effect of each
additional CHZ- group on the c.m.c., is much smaller for most ionics than it is for non-
ionics. This may be partly due to the much better shielding of the hydrocarbon chains
from the water in the case of non-ionics. In the ionics, the repulsion forces prevent the
head groups from packing close together so that there is always a significant
hydrocarbon-water interface (Fig. 1.4.4). Tanford (1980), however, claims that the
Table 9.2
Calculation of the electrical contribution to the free energy of sodium dodecyl sulphate (SDS) micelles
at 25 "C. (From Stigter 1975a, with permission.)
Column 3 uses the DebyeHuckel approximation (eqn (9.4.13)) while column 4 uses the 'exact'
$0' - q' relation from Gouy-Chapman (G-C) theory. A@ values in units of RT.
T H E R M O D Y N A M I C S OF MICELLE F O R M A T I O N 1455
difference in bl values can be almost entirely accounted for by the ionic strength effect.
As the number of carbon atoms decreases, the c.m.c. tends to increase but for ionics
this is partly offset by the increasing ionic strength due to the surfactant itself. The
resulting tendency to lower c.m.c. (Table 9.2) thus opposes the effect of shortening the
chain. This explanation is supported by the behaviour of the n-alkytrimethylammo-
nium bromides in 0.5 M salt solution. (Exercise 9.4.6).
Solid
l *
Monomeric
surfactant
hydrate
5 * Micellar
surfactant
hydrate
surfactant (H~btion) (unit mole Micellization (unit mole
fraction) under standard fraction)
state conditions
(Formation
of micelle
hydrate)
Micellar
Surfactant
solution 3
4 surfactant
- - - - - - - - -
solution
(at c.m.c.)
(Equilibrium (at c.m.c.)
between monomers,
counterions, and micelles.)
AG5 may be identified with the standard free energy change in terms of unit mole
fraction, so that AG5 becomes A@ and may be calculated through eqn (9.4.2) (with
-
AGO = pii, - py). The condition under which micelles are formed spontaneously is
given by:
AG1 + AG2 5 0. (9.4.14)
For compounds such as alcohols, amides, and like substances, the free energy of
+
solution, AG1, is positive to the extent that AGl G2 > 0 and no micelles are formed.
AG1 is reduced by the presence of an ionic head group, or a strongly polar head group
of the ethylene oxide variety and micelles will begin to form when AG1 AG2 = 0. As +
the temperature is lowered, the positive entropy of micelle formation means that at
some point (called the Krafft temperature) the overall free energy change for micelle
formation is no longer negative and the solubility of the surfactant decreases
dramatically. (See Section 9.2.3.)
456 I 9: ASSOCIATIONC O L L O I D S
RT
= p:, - py = R T In a1 - -lnX,i,
n
RT
M RTln(c.m.c) - ~ lnX,i, (9.4.15)
n
neglecting the (n-' In n) term. The temperature and pressure derivatives of eqn
(9.4.15) give the standard enthalpy change A H o and volume change A V o ,per mole of
monomer (Kresheck 1975; Exercise 9.4.5):
aT ) p + TR( T T) ~a l n ~ , ~ ,
~ = - R T z (aln(c.m.c)
(9.4.16)
P
-40 I I I I
-20 0 20 40
TAZ(H
rno1-l)
Fig. 9.4.1 Compensation plot of data for a variety of ionic and non-ionic surfactants in various
liquids. Open circles are for water, with or without additives. Filled circles are for benzene and filled
triangles for formamide. (Modified from Kresheck 1975, Fig. 9 with permission.)
T H E R M O D Y N A M I C S OF MICELLE F O R M A T I O N 1457
In the neighbourhood of the c.m.c. the value of Xmicis small, and since n is not
always known, it is common practice to n&ct the second term on the right of each
of these expressions and to estimate A@ from the approximate eqn (9.4.7).
Although Kresheck (1975) defends this procedure, chiefly on the grounds that the
~
AHo data obtained agree with the calorimetric estimates, Muller (1977) has called it
into question, on account of the large changes in aggregation number that can occur
with temperature. ~ ~
Despite some reservations then, we will examine the resulting A H o and ASo
data because they have been used to develop a deeper insight into the micelle ~
-
AHo TZF
kJ mol-’ kJ mol-’
Although these data were derived at different temperatures (299 K and 313 K,
respectively) it is obvious that no nave interpretation will ‘explain’ such a difference. This
could be a situation in which the changes in aggregation number over the temperature
range are impossible to ignore, as Muller (1977) has argued. Certainly it is important to
recognize that these molecules, when dissolved in benzene at low concentrations (before
they aggregate), are present as ion pairs, so the starting point for the aggregation processes
is very different from that in aqueous solution. Indeed, Kertes (1977) has called into
question the whole concept of micelle formation as applied to solutions in aprotic non-
aqueous media (such as benzene). It should also be noted that the presence of trace
amounts of water has a profound effect on the micelliiation process.
One thing is clear from the minimal amount of data considered here: the very simple
behaviour of the free energy function often masks a very complex shift in A S and AH
values as temperature changes. Sometimes the study of these detailed shifts provides
deeper insights into the process. In the present case it may well have led to an
overemphasis of the role of ‘structured water’ in the phenomenon referred to as ‘the
hydrophobic bond’. For a further discussion of the problems involved in the
interpretation of AG in terms of A H and A S in this case see Evans and Wennerstrom
(1999) section 5.6.
Tables 9.3 and 9.4 provide some illustrative data on micellization for the common
ionic surfactants. The degree of ionization is the ratio of the apparent charge on the
micelle (p) to the aggregation number.
Exercises
+
9.4.1 Note that as % + 00 eqn (9.4.5) reduces to pLi, = py RT In a l . How would
this be interpreted in terms of a phase separation model?
9.4.2 Establish eqn (9.4.6). ~
9.4.3 Show that a crude estimate of AGO for the micellization process can be obtained
from:
~
AGO M RT In Xo
+
~
I 9.4.6 Discuss the relation between the bl values obtained for alkyltrimethyl-
ammonium bromides in the presence of salt (Fig. 9.2.1) and those for the
corresponding chlorides in the absence of salt (Table 9.1).
Table 9.3
Critical micelle concentrations, aggregation numbers, effective degree of ionization of micelles (p/n),
and free energies of micelleformation for various ionic surfactants. (Philldps 1955; Ford et al. 1966).
(Bracketed values (fir SDS) are more recent values from Kratohvill980.)
~
AQ
Material Solvent c.m.c(mM) n p/n ~
RT
+These data from Ford et al. (1966) seem to be on the low side.
460 I 9: ASSOCIATIONC O L L O I D S
Table 9.4
~~ ~
AGO, ASo, and AHo contributions to micellization (Kishimoto and Sumida 1974; Tori and
Nakagawa 1963; Corkill et al. 1964.)
~ ~ ~
0 5 10 15 20x 10-~
(b) C, (mol L-')
Fig. 9.5.1 (a) Pyrene monomer fluorescence in aqueous sodium dodecyl sulphate (SDS) solutions,
at SDS concentrationsbelow and above the c.m.c. (After Kalyanasundaramand Thomas 1977, with
permission. Copyright American Chemical Society 1977.) (b) Variation of the ratio of intensity of
peaks I11 and I from Fig. 9.5.1 (a) as a function of the SDS concentration.
(i) both the probe and quencher are completely associated with the micelles;
(ii) both the probe and quencher remain attached to the micelle for times much
longer than the unquenched lifetime of the luminescent probe;
(iii) the quenching in or on a micelle containing both a probe and a quencher
molecule is much faster than the emission lifetime of the probe so that emission is
observed only from micelles without quenchers;
(iv) the distribution of the probe and quencher among the micelles is known; in
practice, a Poisson distribution is assumed (Exercise 9.6.1).
If the above assumptions are valid, the relative intensity of the fluorescence of a
probe is related to the quencher concentration and micelle concentration by the
relationship.
where I0 is the emission intensity in the absence of quencher (C, = 0) and I the
emission in the presence of quencher. The micelle concentration, C,, is given by:
where CT is the total surfactant concentration. Rearranging eqns (9.5.1) and (9.5.2)
yields (Exercise 9.6.2):
ln(Io/I) =
c, .?z (9.5.3)
CT - c.m.c'
(9.5.4)
where k, (s-l) is the quenching rate constant in the micelle, and to the emission
lifetime of the solubilized excited probe in the absence of a quencher.
SPECTROSCOPIC TECHNIQUESFOR INVESTIGATING MICELLE STRUCTURE I463
It is instructive to consider the form of eqn (9.5.4) at long and short times. For t +
00 we have
(9.5.5)
(9.5.6)
Thus, the logarithmic emission decay curves for different Cq/Cm (changing C,
rather than Cm),will be a family of curves that have the same slope at long times
equal to - l / q . However, at short times the slopes depend on C, according to eqn
(9.5.6). An example of such a set of emission-time quenching curves is presented in
Fig. 9.5.2.
Computer fitting routines are generally used to determine k, and Cq/Cm (and,
subsequently, n ) from the emission decay curves. A simple method of determining n is
to extrapolate the long-time slopes to t = 0 and, if the family of curves has been
standardized, the intercept gives Cq/C, from eqn (9.5.5) and % follows from eqn
(9.5.2) at the known C,. The advantage of the time-resolved method over the steady-
state method for determining ? isithat it can be used over a much wider range of
micelle sizes.
I I I I
0 200 400 600 800
Time (ns)
Fig. 9.5.2 Semilog plots of the fluorescence quenching of excited pyrene in 0.05 M SDS as a
function of Cu2+ concentration. (From Grieser and Tausch-Treml 1980, with permission.
Copyright American Chemical Society 1980.)
464 I 9: ASSOCIATIONC O L L O I D S
where $r is the interfacial potential and the other constants have their usual
meaning.
However, the shift in the pK of the indicator in micellar solutions may be caused not
only by the electrostatic potential but also by a different local environment at the
micellar surface, e.g. by a lower dielectric constant as compared to bulk water.
Mukerjee and Banerjee (1964) pointed out that to measure the electrostatic
contribution to the pK shift, the intrinsic interfacial pK, pK', must be known. The
'apparent' pK (pKa) is related to pK' by:
1.0
0
.-*
.+
s
0.5
u
O
J
B
0
3 4 5 6 7 8 9 1 0 1 1 1 2 1 3
Bulk pH
Fig. 9.5.3 Degree of dissociation of acid pH indicator versus bulk pH. The figure compares the
titration of hydroxycoumarin chromophore, hydrophobically bound to positively (CTAB) and
negatively (SDS) charged micelles in 24 mM surfactant concentrationin water. (From Fernandez
and Fromherz 1977, with permission. Copyright American Chemical Society 1977.)
SPECTROSCOPIC T E C H N I Q U E SF O R INVESTIGATING MICELLE STRUCTURE I 465
titration curves obtained. T o calculate an electrostatic potential for the charged system,
the apparent pK in the neutral micelles was taken as the intrinsic interfacial pK'. Their
assumption that the intrinsic pK is similar in the charged and uncharged interface was
supported by other experimental data. The extent of the pH shift due to the zero-
charge surface environment can be seen in Fig. 9.5.3, where the titration curve of the
coumarins in water is shown alongside the titration curve in neutral Triton X-100
micelles.
The surface potentials calculated using eqn (9.5.8) for CTAB and SDS were
+148 mV and -134 mV respectively. The potentials are sensitive to the total ionic
strength of the solutions and the background electrolyte, but these values are very
reasonable estimates of the surface potential, $0, expected for spherical micelles of the
size suggested in Table 9.3 (Exercise 9.6.3). It appears, therefore, that the probe
molecule is in this case sampling the potential in the plane of the head groups and is not
merely being affected by the diffuse layer potential, which would be much lower in this
case.
Table 9.5
Effective permittivity at the surface of various micelles.
+In some cases three relaxation processes have been observed for ionic surfactant
solution -a very fast process (ca.50 ns) has been attributed to counter-ion relaxation
in the system (see Section 8.8).
MICELLAR D Y N A M I C S I467
(9.6.5)
It is interesting to note that although a non-equilibrated system can relax by all of the
above processes, only two well-defined relaxation times are consistently observed. This
is partly a consequence of the widely varying rates involved in the above reaction steps.
As already mentioned, process A is very rapid, and generally outside the time regime
studied. Process B is the only other process that can occur rapidly and in a single step,
and hence it is assigned to the fast relaxation time. (An independent pulse radiolysis
study (Almgren et al. 1979b) has confirmed that the fast reaction is due to monomer
exchange, at least in the case of SDS.)
Process C is actually a shorthand representation of a mechanism with ( n - 1) steps,
1.e.
This sequence has been identified with the slower relaxation event. It has been argued
(Muller 1977) that although reactions (9.6.4) and (9.6.6) may also be slow, because they
must also proceed through a sequence such as (9.6.7), the relative concentration of
species involved is small and therefore not easily recognized.
where k- is the reverse rate constant for reaction (9.6.2), and C, is the monomer
concentration at equilibrium following the perturbation. This result is in accord with
468 I 9: ASSOCIATIONC O L L O I D S
A n-
Fig. 9.6.1 Micelle size distribution.M,, is the number of aggregates of size a. The aggregates on the
left side of the minimum (L) are called submicellar, those on the right-hand side (proper) micelles
with mean size of ?i and the width of their size distribution is given as 0.
the observed linear rise of l/tl with CT, the total surfactant concentration, noted in
most experiments. A brief discussion of the derivation of this kind of relaxation is given
by Vold and Vold (1983,pp. 612-14).
The derivation of the slow relaxation process involves a number of assumptions and
approximations. The reason for this can be readily appreciated when it is remembered
that n steps are required, the rate constants of the individual steps are not all equal, and
the concentrations of the intermediate sized aggregates need to be specified. Based on a
model of mass flow, the simplifications made by Aniansson and Wall allowed a single
relaxation time to be derived and it is given by,
1
(9.6.9)
Table 9.6
Relaxation times tl (ps) and t2 (ms) for some sodium alkyl sulphates.
1 760 350
2.1 320 41
2.1 245 19
2.1 155 7
3 125 34
10 - 1.8
50 - 50
The relaxation times are, of course, related to the rate constants for association ( K + )
and dissociation ( K - ) as represented by reaction (9.6.2) earlier. The rate constants for a
series of sodium alkyl sulphates are given in Table 9.7. The obvious pattern shown in
Table 9.7 is that, as the alkyl chain becomes larger, the exit rate of monomer becomes
appreciably slower, as one would expect with increasing hydrophobic character of the
monomer. The association rate constant also decreases although only slightly,
reflecting basically a diffusion controlled rate step with some electrostatic repulsion
involved between the charged micelle and the anionic monomer (Aniansson e t al.
1976).
In summary it may be said that although some uncertainties remain in
understanding the results of the relaxation processes in micellar systems, a reasonably
good picture of the equilibria has been established.
Table 9.7
Kinetic parameters of association and dissociation of alkyl sulphatesfrom their micelles (from Aniansson
et al. 1976).
Note: The equilibrium constant K = k+ii/k- % ii/(c.m.c); ii is the average aggregation number.
470 I 9: ASSOCIATIONC O L L O I D S
Table 9.8
Residence times in micelles (ps).
(9.6.10)
+Thisis the time taken for a fraction (1 - e-'), i.e. 63 per cent, of the probe molecules
to escape.
MICELLAR D Y N A M I C S I471
The Perrin (1932) equation relates the degree of polarization d to other parameters of
the system by:
(9.6.11)
r
Exercises
9.6.1.
The probability of finding a micelle with i solute species is given by the Poisson
relationship,
-.
have made little allowance for the existence of rod-like micelles, vesicles, or bilayers
but Tanford (1972), Mitchell and Ninham (1981), and Israelachvili et al. (1976, 1977,
1980) have revived an idea originally proposed by Hartley (1941), viz. that molecular
packing plays a crucial role in determining allowed structures, at least in the case of
dilute surfactant solutions.
Aggregation is, as outlined previously, described by either the mass action or phase
models. Typically, chemical potentials of the monomer in the aggregate and in solution
at equilibrium are equated and the normal relationships are derived. In this section we
wish to examine the contributions to the standard chemical potential or molar free
energy per surfactant molecule in the aggregate, &:
These terms may be further described, referring to Fig. 9.7.l(a) for clarification. The
bulk term, p:, is a constant and is a measure of the free energy change involved in
removing hydrophobic tails from an aqueous environment into the micelle interior.
The latter is assumed to be liquid-like which, as we have already seen, is a sound
assumption (Vikingstad and Hoiland 1978).The surface term, pk,includes a quantity
yA to account for the fact that the hydrophobic tails have some residual contact with
the aqueous phase. A is the area per surfactant head group and y is the hydrocarbon-
water interfacial tension. Head group interactions, which may be due to steric,
hydration, electrostatic, and other forces contribute a repulsion energy. Their
quantitative description is still elusive; however, for electrostatic repulsion, an energy
contribution varying inversely with A would be expected. pk then takes the form:
(9.7.2)
reaching a maximum at 2yAo where A0 is a limiting or optimal area per head group.
The curvature term, pk,accounts for reductions in the effective surface tension and
alterations in the electrostatic energy when a spherical surface is formed, rather than a
planar one. Generally speaking these corrections may be ignored, except for very
special cases (Israelachvili et al. 1976).
Lastly the packing term needs to be considered. One needs to account for the fact
that the interior of the aggregate is fluid-like and incompressible. The aggregate radius
or radii, hydrophobic chain volume, and surface area per head group are of paramount
importance here, as discussed below.
In more concentrated systems, where other interactions take place, eqn (9.7.2) will
contain additional contributions. For the present, however, Israelachvili and Ninham
(1977) have shown that when &, is decomposed into the first three terms, the correct
dependence of c.m.c. on hydrocarbon chain length is predicted as are the effects of
temperature and ionic strength. However the existence of rod-like micelles or bilayers
suggests certain geometric constraints and it is with these that we shall now concern
ourselves, i.e. the nature of the molecular packing term.
For simplicity, consider firstly a spherical micelle (Fig. 9.7.l(a)). The radius, R",
surface area per head group A and hydrophobic chain volume V are linked by
V / A = R"/3. (9.7.3)
474 I 9:ASSOCIATION COLLOIDS
The radius of a spherical micelle cannot be greater than a specific critical length I,
slightly less than the fully extended length of the hydrocarbon chain, if it is assumed
that the head group never moves into the core. Clearly, then, when V/A& > 1/3
/Area A
Volume
Fig. 9.7.1 (a) Schematic representation of a model spherical micelle of core radius R”. (After
Mitchell and Ninham 1981, with permission.) (b) Geometric packing of a a hydrocarbon region of
volume Vand surface area A, at a surface with two radii of curvature, RI and Rz. Equation (9.7.7)
gives the relation between V ,A, R1, and Rz and the length, 1, of the hydrocarbon. For both R1, Rz > 1
a void region is formed behind the hydrocarbon region. (After Israelachvili et al. 1976, with
permission.)
MOLECULAR PACKING A N D ITS EFFECT O N AGGREGATE F OR M A T ION 1475
spherical micelles will not form unless A > Ao. The critical condition for the formation
of spheres is
1
VIA01 --. (9.7.4)
“-3
Similarly, for cylindrical micelles, it may be shown that
1
VIA01 - - (9.7.5)
“-2
and for planar bilayers
V/AOlC= 1. (9.7.6)
Any aggregated structure must satisfy two basic criteria: (a) no point within the
aggregate can be farther from the surface of tension than I,; (b) the total hydrocarbon
core volume, V ,and the total surface area, d,
must approximatelysatisfy V / Z = &/A0
= n, the aggregation number. Here V is the volume occupied by a single hydrocarbon
molecule of the appropriate chain length in the liquid state. This criterion is only an
approximate one since the average surface area per surfactant head group is assumed to
be equal to Ao.
Between a sphere and a cylinder, one might reasonably expect to find a variety of
transition shapes. Thus if we consider a surfactant (Fig. 9.7.l(b))with a head group
of surface area A and liquid hydrocarbon core volume V in a micelle or bilayer
vesicle where the local radii of curvature are R1 and R2, the following equation
results:
(9.7.7)
where 1 is the length of the hydrocarbon region of the amphiphile. Equation (9.7.7)is
exact for spheres ( R I = R2), cylinders (R2 = co),and planar surfaces (R1 = R2 = 00)
and is accurate to within one per cent for other cases (Israelachvili et al. 1976). It may
then be predicted that bilayers or vesicles exist when
1
- < V/AOl, < 1 (9.7.8)
2
and inverted structures occur when
This deceptively simple packing model allows many physical properties of micelles and
vesicles such as size, shape, polydispersity, etc. to be predicted. The interested reader
should consult the original references for more detailed discussions.
476 I 9: ASSOCIATIONC O L L O I D S
the length of a fully extended chain (1.67 nm), it will contain approximately 56 chains.
The model predicts that the free energy cost of packing the chains into this structure is
less than 0.5 kT per chain. By comparison, the free energy gained when a C12 chain is
-
transferred from water to bulk n-alkane is 20kT (recall Section 9.4 above).
In the micelle one or two chains must be completely straight (in their all-trans state)
in order to fill the volume at the centre of the micelle, but no more than one or two. On
average, the model predicts a loss of only 0.2 gauche bonds per chain on transfer from a
bulk n-alkane environment to the micelle.
The fact that half the volume of the hydrophobic core occurs within 0.34 nm of its
surface has important consequences. The mean position of each segment in the chain
is nearer to the surface of the aggregate than to the micelle centre. The terminal CH3
group, although on average closer to the centre than any other group, sits a mean
distance of 1.04 nm from the centre (and only 0.63 nm from the core surface).
Because of the liquid-like micelle interior, and the flexibility of the chains, all
segments sample the surface of the micelle. Even the terminal CH3 group is in
contact with the surface (and hence in contact with a partly hydrophilic
environment) approximately 20 per cent of the time. (This observation explains
why probes attached to any part of the chain behave as though they were in a partly
hydrophilic environment.)
The model suggests that the hydrocarbon-water interface is fairly sharp. 0.2 nm
beyond the core surface, the average hydrocarbon volume fraction is 0.02 and falling
fast. It may seem extraordinary that a structure as dynamic as a micelle (with
monomers being associated with the micelle for only lop6- lo-’ s) could have such a
smooth surface. It occurs because the hydrophobic effect is very strong and so the vast
majority of time that any monomer is associated with the micelle it sits with almost all
of its chain inside the hydrophobic core.
It is this picture of the SDS micelle that was used in Fig. 1.4.4 for the spherical
micelle structure. Each of the five spherical shells contains approximately the correct
number of chain segments to ensure an even packing density throughout. Note
particularly the large fraction that is in the outermost shell. This model is capable of
reconciling a considerable body of experimental evidence but there remain some
disagreements over details. Hayter and Penfold (1981), for example, argue that the
neutron scattering results suggest a slightly rougher hydrocarbon-water interface,
though the difference is not great.
References
Almgren, M., Grieser, F., and Thomas, J.K. (1979a).J.Am. Chem. Soc. 101,279.
Almgren, M., Grieser, F., and Thomas, J.K. (19796).J. Chem. Soc. Faraday
Trans. I 75, 1674.
Almgren, M. and Lofroth, J.E. (1980).J Colloid Interface Sci. 81,486.
Anacker, E.W. (1970).Micelle formation of cationic surfactants in aqueous media.
In Cationic surfactants (ed. E. Jungermann). Marcel Dekker, New York.
Aniansson, E.G. (1978).J Phys. Chem. 82,2805.
Aniansson, E.G. and Wall, S.N. (1974).J.Phys. Chern.78, 1024.
Aniansson, E.G., Wall, S.N., Almgren, M., Hoffmann, H., Kielmann, I.,
Ulbricht, W., Zana, R, Lang, J., and Tondre, C. (1976).J Phys. Chem. 80,905.
478 I 9: ASSOCIATIONC O L L O I D S
Bendedouch, D., Chen, S.-H., and Koehler, W.C. (1983).3. Phys. Chem. 87,153.
Bids, B.P. (ed.) (1999). Modern characterization methods ofsurfactant system, pp. 601.
No. 83 in Surfactant science series. Marcel Dekker, New York.
Cabane, B., Duplessix, R, and Zemb, T. (1983). In Surfactants in solution (ed. K.
L. Mittal and B. Lindman). Plenum Press, New York.
Chan, S.-K. and Kahlweit, M. (1977). Ber. Buns. Phys. Chem. 81, 1294.
Chan, S.-K., Herrman, U., Ostner, W., and Kahlweit, M. (1977). Ber. Buns.
Phys. Chem. 81,60 and 396.
Chan, S.-K., Herrman, U., Ostner, W., and Kahlweit, M. (1978). Ber. Buns.
Phys. Chem. 82, 380.
+
Clint, J.H. (1992). Surfactant aggregation, pp. xi 283. Chapman and Hall New
York.
Corkill, J.M., Goodmann, J.F., and Harrold, S.P. (1964). Trans. Faraday SOL.
60, 202.
Debye, P. and Anacker, E.W. (1951). 3. Phys. Colloid Chem. 55, 644.
Desnoyers, J.E., Caron, G., DeLisi, R., Roberts, D., Roux, A., and Perron, G.
(1983). 3. Phys. Chem. 87, 1397406.
Ekwall, P., Mandell, L., and Solyom, P. (1971). J. Colloid Interface Sci. 36,
519-28.
Evans, D.F. and Ninham, B.W. (1983). J. Phys. Chem. 87, 5025-32.
Evans, D.F. and Ninham, B.W. (1986). J. Phys. Chem. 90,226-34.
Evans, D.F., Allen, M., Ninham, B.W., and Fouda, A. (1984).3. Solution Chem.
13, 87-101.
Evans, D.F. and Wennerstrom, H. (1999). The Colloidal Domain (2nd edn),
pp. 632. Wiley-VCH, New York.
Fernandez, M.C. and Fromherz, P. (1977). 3. Phys. Chem. 81, 1755.
Fisher, L.R. and Oakenfull, D.G. (1977). Chem. Soc. Rev. 6,25.
Ford, W.P.J., Ottewill, R.H., and Parriera, H.C. (1966) 3. Colloid Interface Sci.
21, 522.
Frank, H.S. and Evans, M.W. (1945). 3. Chem. Phys. 13, 507.
Franks, F. (1983). Water. Royal Society of Chemistry, London.
Fuerstenau, M.C. (1976). Flotation, Vols 1 and 2. Am. Inst. MMPE, New York.
Gaines, G.L. (1966). Insoluble monolayers at liquid-gas interfaces. Interscience,
New York.
Grieser, F. and Tausch-Treml, R. (1980). 3. Am. Chem. SOL.102, 7258.
Gruen, D.W.R. (1981).J. Colloid Interface Sci. 84. 281.
Gunnarson, G., Jonsson, B., and Wennerstrom, H. (1980). J. Phys. Chem. 84,
3 114-21.
Hall, D.G. and Pethica, B.A. (1967). Thermodynamics of micelle formation. In
Nonionic surfactants (ed. M. Schick), Chapter 16. Marcel Dekker, New York.
Halle, B. and Carlstrom, G. (1981). 3. Phys. Chem. 85, 2142.
Hartley, G.S. (1936). Aqueous solutions ofparaf3n chain salts. Herman et Cie, Paris.
Hartley, G.S. (1941). Trans. Faraday Soc. 37, 130.
Hartley, G.S. and Roe, J.W. (1940). Trans. Faraday SOL.36, 101.
Hayter, J.B. and Penfold, J. (1981).3. Chem. SOL.Faraday Trans. 177, 185143.
Hill, T.L. (1963). Thermodynamics of small systems, Vol. 1. Benjamin, New York.
Israelachvili, J.N. and Ninham, B.W. (1977). 3. Colloid Interface Sci. 58, 14-25.
Israelachvili, J.N., Mitchell, D.J., and Ninham, B.W. (1976). 3. Chem SOL.
Faraday Trans. 2 72, 1525.
Israelachvili, J.N., Mitchell, D.J., and Ninham, B.W. (1977). Biochem. Biophys.
Acta 470, 185.
STATISTICAL T H E R M O D Y N A M I C S OF C H A I N PACKING I N MICELLES 1479
Israelachvili, J.N., Marcelja, S., and Horn, R.G. (1980).Q Rev. Biophys. 13,121.
Kahlweit, M. (1981). Pure Appl. Chem. 53,2069.
Kalyanasundaram, K. and Thomas, J.K. (1977).J. Am. Chem. Soc. 99,203944.
Kertes, A.S. (1977). Aggregation of surfactants in hydrocarbons. In Micellization,
solubilization and microemulsions (ed. K.L. Mittal) Vol. 1, pp. 445-54. Plenum
Press, New York.
Kishimoto, J. and Sumida, K. (1974). Chem. Pharm. Bull. (Japan) 22, 1108.
Kitahara, A. (1967). In Nonionic surfactants (ed. M.J. Schick), p. 289. Marcel
Dekker, New York.
Kitchener, J.A. (1964). In Recent progress in surface science (ed. J.F. Danielli,
K. Pankhurst, and A.C. Riddiford) Vol. 1. Academic Press, New York.
Kratohvil, J. (1980). 3. Colloid Interface Sci. 75, 271-5.
Kresheck, G.C. (1975). Surfactants. In Water - a comprehensive treatise (ed.
F. Franks) Chapter 2, pp. 95-167. Plenum Press, New York.
Leja, J. (1982). Surface chemistry offroth flotation, pp. 284-6. Plenum Press,
New York.
Lindman, B., Lindblom, G. Wennerstrom, H., and Gustavsson, H. (1977).
Ionic interactions in amphiphilic systems studied by n.m.r. In Micellization,
solubilization and microemulsions (ed. K.L. Mittal) pp. 195-227. Plenum Press,
New York.
Lindman, B. and Wennerstrom, H. (1980). Topics in current chemistry 87,
pp. 1-83. Springer, Berlin.
Loeb, A.L., Wiersema, P.H., and Overbeek, J. Th. G. (1961). The electrical double
layer around a spherical colloid particle, p. 37. M I T Press, Cambridge, Mass.
Mazer, N.A., Carey, M.C., and Benedek, G.B. (1977). The size, shape and
thermodynamics of sodium dodecyl sulphate (SDS) micelles using quasielastic
light-scattering spectroscopy. In Micellization, solubilization and microemulsions
(ed. K.L. Mittal) Vol. 1, pp. 359-81. Plenum Press, New York.
McBain, J.W. (1950). Colloid science. D.C. Heath, Boston.
Menger, F.M., Jerkumica, J.M., and Johnson, J.C. (1978).J. Am. Chem. Soc.
100,4676.
Missel, P.J., Mazer, N.A., Benedek, G.B., Young, C.Y., and Carey, M.C.
(1980). J. Phys. Chem. 84, 1044-57.
Mitchell, D.J. and Ninham, B.W. (1981). 3’. Chem. Soc. Faraday Trans. 2 77,
601.
Mukerjee, P. (1974). J Pharm. Sci. 63,972.
Mukerjee, P. (1975). Differing patterns of self-association and micelle formation. In
Physical chemistry, enriching topics from colloid and surface science (ed. H. van
Olphen and K.J. Mysels) Chapter 9; IUPAC Commission 1.6. Theorex, La Jolla,
California.
Mukerjee, P. and Banerjee, K. (1964). 3’.Phys. Chem. 68, 3567.
Mukerjee, P. and Ghosh, A.-K. (1970).3. Am. Chem. Soc. 92, 6419
Mukerjee, P. and Mysels, K.J. (1971). Critical micelle concentrations of aqueous
surfactant systems. NSRDS-NBS 36, National Bureau of Standards. US
Government Printing Ofice, Washington, D. C.
Mukerjee, P. and Mysels, K.J. (1975). Anomalies of partially fluorinated surfactant
micelles. A.C.S. symposia (ed. K.L. Mittal) Series 9, p. 239. American Chemical
Society, Washington, D.C.
Mukerjee, P., Cardinal, J.R., and Desai, N.R. (1977). The nature of the local
micro-environments in aqueous micellar systems. In Micellization, solubilization
and micro- emulsions (ed. K.L. Mittal) pp. 24141. Plenum Press, New York.
480 I 9: ASSOCIATIONC O L L O I D S
Vold, R.D. and Vold, M.J. (1983). Colloid and interface chemistry. Addison-Wesley,
Reading Mass.
Wennerstrom, H. and Lindman, B. (1979). Phys. Rep. 1, 1.
Wennerstrom, H., Lindman, B., Soderman, O., Drakenberg, T., and
Rosenholm, J.B. (1979). 3’.Am. Chem. Soc. 101, 6860.
Zachariasse, K.A. (1978). Chem. Phys. Lett. 57,429.
Zachariasse, K.A., Nguyen van Phuc, and Kozankiewicz, B. (1981).J. Phys.
Chem. 85, 267&83.
Zana, R. (1978). 3’.Phys. Chem. 82,2440-3.
Adsorption a t Charged Interfaces
10.1 Introduction
10.1 .I The Gouy-Chapman-Stern-Grahame (GCSG) model revisited
10.2 Adsorption of potential determining ions
10.2.1 Surface charge density and surface potential
10.2.2 Surface complexation models
(a) Single-site model
(b) Two-site model
10.2.3 Position of the plane of shear
10.3 Detection of Stern layer adsorption
10.3.1 Modificationsto the Stern layer model
10.4 The oxide-solution interface
10.5 Adsorption of multivalent ions
10.5.1 Adsorption of hydrolysable metal ions onto oxide surfaces
10.6 Surfactant adsorption
10.6.1 Ionic surfactant adsorption on a hydrophobic surface (Agl)
10.6.2 Ionic surfactant adsorption on hydrophilic (oxide) surfaces
10.6.3 Adsorption of ionic surfactants on other surfaces
10.6.4Adsorption of non-ionic surfactants
10.1 Introduction
We discussed the adsorption of uncharged molecules at the liquid-liquid interface
briefly in Section 2.4 and at the solid-liquid interface more extensively in Chapter 6. In
this chapter we will concentrate attention on the adsorption of charged species at
charged interfaces, in an attempt to reconcile the data obtained from equilibrium
(usually titration) studies with those obtained from electrokinetics.
The classical work in this field is on the mercury-solution interface, on which there
is much literature. Delahay (1965) (Chapter 5) lists (after Parsons 1958) the various
isotherms which have been used to describe the adsorption of a solute onto a charged
surface. At the mercurysolution interface the analysis can be taken quite a long way
and one can compare the various possible isotherms (Henry, Volmer, Langmuir,
Frumkin, Temkin, etc.) to determine which one best fits the data. From the
482
INTRODUCTION I483
corresponding equation of state for the ‘pressure’ of the adsorbed species (i.e. the
spreading pressure (Section 2.10.1)) one can then attempt to draw some inferences
about the nature of the interaction between the adsorbed molecules (both with one
another and with the surface).
Surveys of the isotherms commonly used in colloid and surface chemistry have been
given by, for example, Davies and Rideal 1963 (Chapter 4), Adamson 1969 (Chapter 8),
Vold and Vold 1983 (Chapter 4) and Dobias and Rybinski (1999). Some of these are
purely empirical relationshipsand some have a more or less solid theoreticalbasis. Here we
are concerned primarily with adsorption of ionic species onto surfaces which are usually
themselves charged by ionic adsorption or dissociation processes. Coulombic interactions
will always be important and are often difficult to separate from other ‘chemical’ bonding
interactions. The isotherm most commonly used is the Langmuir equation (6.4.6) and its
extension to charged interfaces, the Stern equation which was introduced in Section 7.4.3.
Adsorption from fairly dilute solutions onto the solid-liquid or liquid-liquid (emulsion)
interface is usually limited to, at most, a monolayer, and the Stern model can usually give a
good account of the behaviour. The Stern model is, however, most readily applied to a
metallic conducting interface where the surface charge can be regarded as smeared out
uniformly and the surface itself is at a constant potential, $0.
The classical silver iodide surface (Section 7.7) could be reasonably approximated by
a constant potential surface and one might expect the counterions to be restricted in
their approach to a very smooth AgI surface in somewhat the same way as is observed
at the mercury-solution interface. That surface is also hydrophobic like mercury. But
for most colloid surfaces the surface potential develops as a result of the presence of
individual ionic groups which can be quite well separated one from another.
Furthermore, the intrinsic surface roughness means that the locus of the centres of
counterions may be very close to that of the surface ionic charges. The first layers of
the interfacial region may then be likened to a very concentrated electrolyte solution
where ionic activities may be far removed from their bulk solution values and ion
pairing may be more the rule that the exception. The current models of the oxide-
solution interface attempt to take these effects into account by dealing specifically with
the chemical reaction equilibria which are involved. These are the bases of the site-
dissociation-site-binding models.
I!
+4
f 2
I I
9 10
PH
Fig. 10.1.I Sum of diffuse layer charges (oa)and Stern layer charge (q for i = Na+, Cl-) with the
measured surface charge (no) for a sample of titania. (From Foissy et al. 1982, with permission.)
T o test the model we require data on the surface charge (by titration, say) and the
components of charge in the diffuse layer and the Stern layer. The diffuse layer charge
can be determined using electrokinetic data, provided we can assume that 5 $d and
we will use that approach in the discussion below. That procedure was used by Foissey
et al. (1982) who determined not only the titration charge and the diffuse layer (5)
potential but also followed the adsorption of Na+ and C1- on a titanium dioxide
surface using radio-isotopes. They were able to build up a complete picture of the
double layer charge (Fig. 10.1.1) which showed that only about 10% was in the diffuse
layer. Na+ was adsorbed into the Stern layer at pH values above the p.z.c (about
pH 6.3) and C1- was adsorbed at lower pHs. The required adsorption potentials (eqn
7.4.14) were not very large (0 = 3.31 kT for Na+ and -2.85 kT for Cl-) and were
independent of pH (and hence of the state of charge of the surface). The Stern theory
was able to quantitatively account for the entire charge distribution with an outer layer
capacitance of 22 pF crnp2.The important point about this result is that there did not
appear to be any significant amount of ‘superequivalent’ adsorption (i.e. Stern layer
adsorption of an ion of the same sign as the surface charge). This behaviour is
consistent with the low adsorption potentials and contrary to that normally expected of
‘specifically adsorbed’ ions. It is referred to by Lyklema (1995 p. 3.64) as speczjic
adsorption of the second kind and provides a different, and much simpler, picture to that
suggested by the site-dissociation-site-binding models to be considered below (Section
10.2 and 10.4). It may well be much more common in colloidal systems than has
previously been recognized.
ADSORPTION OF POTENTIAL DETERMINING I O N S I485
The formulation in terms of the excess of neutral species is perhaps more appropriate
because the double layer region is neutral overall. In practice, however, we will
distinguish the actual potential determining ions, Ag+ and I- from their partners
because measurements are normally done in the presence of very small concentrations
of the p.d.i, with a swamping concentration of an (indifferent) electrolyte (KNO3 in
this case). The concentrations of the partner ions (K+ and NO, in eqn (10.2.1)) are
then negligible compared to their concentrations in the remaining indifferent
electrolyte.
In the simplest models of the silver iodideaqueous solution interface one would
assume that the excess of Ag+ or I- ions (depending on whether the surface was
positive or negative) would be uniformly distributed over the surface and would be
indistinguishable from the lattice ions. At the point of zero charge there would be large
but equal numbers of Ag+ and I- ions on the surface and as the surface potential was
built up by adding more of one or other species, the additional ions would remain a
negligible fraction of the total number of such ions on the surface (Exercise 7.7.1). In
such circumstances, the chemical activity of those surface ions will be independent of
the state of charge of the surface.+ It should be recalled that that was the crucial
assumption involved in establishing (Section 7.7.1) the Nernst equation (from eqn
(7.7.5)):
[7.7.5]
(or $0 (volt) % 0.0592 loglo([Ag+]/[Ag+],,) for the surface potential of silver iodide at
25 "C.
There is reason to believe that the charge is not confined to the surface but extends
some distance into the bulk solid, in the form of a diffuse defect structure (see
Sparnaay 1972). This diffuse structure in the solid has been invoked to account for
certain features of colloid behaviour (see, for example, Napper and Hunter 1972 and
Hunter 1981 pp. 54-5 for brief reviews of that work). The more recent consensus,
however, is that its effect is principally on the position of the point of zero charge,
which is usually taken to be an experimentally measured quantity (Levine et al. 1970).
We will, therefore, assume that the silver iodide surface can be treated as a non-
conducting hydrophobic plane with a potential given by eqn (7.7.5) and a uniform
distribution of charges confined to its surface.
For many colloidal systems, especially those for which H+ and OH- are the
potential determining ions there are very few charged groups present at the point of
zero charge. Most of the ionisable groups are in the form -A-H and it can no longer be
t If the lattice ions are considered to be able to establish an equilibrium with the
surface ions the point can be made much more strongly.
ADSORPTION OF POTENTIAL DETERMINING I O N S I487
assumed that the charges which bring about the surface potential enjoy a constant
chemical environment, independent of the state of charge of the surface. The analogue
of eqn (7.7.5):
11.0 = (2.303 kT/e)(pHpzc - PHI (10.2.3)
is then no longer valid and the more elaborate analysis of Section 7.9 and 7.10 becomes
necessary. (Semiconductor oxides such as RuOz seem to obey eqn (10.2.3) quite well
and Ti02 departs from it only to the extent of about lo%, but for SiOz it is a very poor
representation indeed.)
is not adequate to describe the behaviour of polymer latex systems. T o get an accurate
representation of the potentiometric (and conductimetric) titration curves for, say, a
carboxylate latex, it seems to be necessary to introduce the possibility of site-binding
i.e. the occurrence of reactions like:
-SO- + Na+ f t SO-Na’. (10.2.5)
Such a reaction makes possible the development of a significant surface charge without
pushing the electrostatic potential in the surface regions up to unrealistic values. The
counterions ‘screen’ the developing surface charge. The problem is that it now
becomes necessary to evaluate, in addition to K,, the equilibrium constant for eqn
(10.2.5):
{-SO -Na +}
KNa = (10.2.6)
{-SO -}{Na +}
Unfortunately, as it stands such a quantity will not be constant as the surface is charged
up, unless the quantities in braces are interpreted as electrochemical activities. The
problem is to find a sufficiently general procedure to describe the electrical and
chemical effects at the charged interface. The most common procedure has been to try
to separate out the chemical and electrical effects which contribute to this exchange
process by modelling it in terms of our current pictures of the structure of the
electrical double layer (Fig. 7.3.4). The ‘chemical’ aspect of the binding process is
subsumed into a so-called intrinsic surface binding constant aames and Parks 1982):
488 I 10: ADSORPTION AT CHARGED INTERFACES
. [-SO-Na+]
(10.2.7)
rG:= [-SO -]"a +Is
where the quantities in square brackets are now chemical activities and the subscript 's'
refers to the surface. In practice, these activities are usually replaced by concentrations,
which implies a simple random mixing model for the surface species [SOH] and
[SO-Na+]. For computational purposes it also turns out to be easier to replace eqn
(10.2.5) by an exchange reaction:
for which
(10.2.9)
The surface activities or concentrations of H+ and Na+ ions are estimated from the
Boltzmann equation:
where z is the valency of the ion and $ is the mean electrical potential which it
experiences in the surface. In the models discussed below we will find that $ is
sometimes taken as the same for all adsorbed ions, ($0 = $i = $d) and sometimes
equal to the electrokinetic potential, (5) and sometimes given different values for the
potential determining ions ($0) and the 'specifically adsorbed' ions (h).The most
elaborate (triple-layer) model, used by James et al. (1978) (Fig. 7.3.4) to describe the
titration data of Stone-Masui and Watillon (1975) on polystyrene uses the latter
procedure so that
(10.2.11)
where $ = e$/kT is the reduced (i.e. dimensionless) surface potential and b refers to
bulk concentration.
(b) Two-site model
In Sections 7.9 and 7.10 we also treated the two-site models, including the zwitterionic
and the amphoteric models.
In order to solve these models it is again necessary to have an alternative expression
for ao.If there is no inner (Stern) layer adsorption we have 00 = - a d and eqns (7.9.5)
and (7.9.6). T o close the equations we have from eqn (7.3.36) the relation a0 = Ki ($0
- $d) where K; is the (integral) capacitance of the inner (compact) part of the double
layer. The detailed procedure for fitting the electrokinetic data is given by Rendall and
Smith (1978) and outlined by Hunter (1981 pp. 2 7 6 8 ) . It need not concern us here.
Suffice it to say that the fit to 5 data for a nylon surface is very good (Fig. 10.2.1),and is
obtained with reasonable values of the integral capacitance, Ki (25 pF cmP2) and the
ADSORPTION OF POTENTIAL DETERMINING I O N S I489
- 6 o a
2 4 6 8 10
PH
Fig. 10.2.1 Comparison of experimental {-potential data with theoretical curves calculated from a
two site model for nylon at 25 "C.Electrolyte concentration:0 M; 0: M; A:5 x lop3
M; 0: lop2 M. Model parameters: Ns+ e = 0.62 pC cmp2; Ns- e = 1.42 pC atp2; i.e. p. = 5.4;
pKp = 5.5; KA = 8.6; K, = 13.2; Ki = 25 pF cm-'. (After Rendall and Smith 1978 with
permission.)
acid and base dissociation constants of the zwitterionic model. Unfortunately, it was
not tested directly with titration data to determine how well it fits the surface charge. It
is perhaps for this reason that it was not necessary for Rendall and Smith to invoke the
notion of site binding (but see Section 10.3 below). When an effort is made to reconcile
the titration and electrokinetic data on oxide surfaces with the site dissociation model,
it certainly seems to be necessary to invoke site binding (see Section 10.4.)
Helmholtz plane (Section 7.3) where the potential is $d then, for such low potentials
(compare with eqn (7.3.21)):
(10.2.13)
The quantity (d$o/dpAg), = N , is the Nernst factor (obtainable from eqn (10.2.3))
and is equal to 59.2 mV at 25 "C, whilst the quantity (d$d/d$o)K, in the absence of
specific (inner layer) adsorption, can be written in terms of the capacitances of the
inner and diffuse layers (Kiand c d respectively) (Exercise 10.2.1). Substituting in eqn
(10.2.13) then gives:
[ I:
S-'= N-' 1 + - exp(KA). (10.2.14)
Smith (1973) gives a plot of the experimental values of S-' against c d (= E K ) for AgI
in KN03(aq) at 25 "C. The plot is linear with the expected intercept at C d = 0 giving
N = 59.2 mV and from the slope, the value of K; is 30 pF cmP2. Since this latter
figure can be confirmed by direct titration measurements, the result implies that exp
( K A )is very close to unity. Smith concludes that A must be less than 0.1 nm in this
system. He has applied the same approach to a number of other systems (see for
example Smith 1976) with similar results. This conclusion is of very great
importance; if it is accepted, it makes the {-potential a much more valuable and
important quantity for testing double layer models, as we shall see in the next
section.
Exercises
10.2.1 Use the definitions of Ki and c d , for an electrical double layer with no inner
layer adsorption, from Chapter 7 and show that (d$d/d$o) = Ki/(Ki Cd). +
Show also that c d = EK at low potentials, where E is the permittivity.
00 = z; efi (10.3.1)
and a0 + a d =0 (10.3.3)
Now we define
b
pi pi - zi e$d (10.3.4)
nb is the bulk concentration of component i and p! is the standard state value of the
chemical potential.
Hall (1978) shows, by a general thermodynamic argument, that for all points with
the same ri, the value of pi - p! is constant. It follows then (Exercise 10.3.1) that at
constant 00:
zi e$rd = constant - 2.303 kT p X (10.3.6)
492 I 10: ADSORPTION AT CHARGED INTERFACES
where X is a potential determining ion and the constant is a function of 00. The
(integral) capacity of the compact layer is given by eqn (7.3.36) and for the case where
the surface potential is given by the Nernst equation (Exercise 10.3.1):
2.303 kT 2.303 kT
(10.3.7)
+d=[ zie zie PX
2.303 k T 00 kT
pX' ----ln(Y/Y')
Ki 2zie PX (10.3.8)
@d=[ zie
where Y = [( 1- a-)/a-][a+/( 1 - a+)]and X' and Y' are the values of X and Y at the
P.Z.C.
In the absence of inner layer adsorption, constant values of 00 correspond to
constant Od and, for systems which are expected to obey the Nernst equation (10.2.3),
eqn (10.3.7) can be tested by plotting values of $fd (= {) for a fixed a d against ( p x -
pX') = ApX. The resulting plot should be linear with the Nernst slope (59.2/2; mV)
and Hall and Rendall (1980) show that, for the limited amount of data available, that is
so for the calcite (CaC03) and Ca3(P04)2 surface. For the AgI surface it is true for low
to moderate electrolyte concentrations but some evidence emerges for inner layer
adsorption at higher electrolyte concentrations (> 0.1 M KN03).
For zwitterionic surfaces, this test is not so easy to apply since the function Y / Y'
must first be evaluated. That has been done for the nylon sol (Rendall and Smith
1978), and Hall and Rendall (1980) show that, using those values in eqn (10.3.8) gives a
plot of ( against pH which has the Nernst slope. The conclusion is that the nylon
system does not exhibit inner layer adsorption of the supporting electrolyte (NaCl).
The same conclusion is reached from the congruence test, which is an alternative
method of using eqns (10.3.7) or (10.3.8).
In the congruence test a plot is made of Od against the function
If eqns (10.3.7) or (10.3.8) are valid then all of the data at different electrolyte
concentrations should fall on a common curve and Hall and Rendall (1980) show that
this is so for the nylon sol and is also true for AgI provided the salt concentration is less
than 0.1 M as before. (Note that to apply this test to nylon it is not necessary to
evaluate the departure from Nernst behaviour.)
Using the Nernst equation, eqn (10.3.7) can be rearranged to read (Exercise 10.3.3):
-
-2 -1 l
t
I
1
ApAg+(zle<12.303 kT)
Fig. 10.3.1 Congruence test for electrokinetic data on silver iodide. Results from titration (open
symbols) and electrokinetics (filled symbols). The unbroken line corresponds to an integral
capacitance Ki of 26.4 pF cm-’. The broken line is for Ki = 18 pF cm-’. The linearity indicates
conformityto the Nernst equation (10.2.9). The ionic strength ranges from lo-’ ( 0 )to lop3M(0).
(Modified from Hall and Rendall 1980.)
Figure 10.3.1 shows a plot of od against ApAg%for the silver iodide surface, from which
the integral capacitance, Ki can be estimated. The full line, for positively charged surfaces
corresponds to Kj = 26.4 pF mp2. The broken line gives a rather better description of
the data for negative surfaces and corresponds to K ; = 18 pF cmp2. This was the value
used by Hunter (1981) to represent the {-potential data on silver iodide (Fig. 10.3.2)over
the range -100 mV < { < 100 mV. (The model used for that analysis corresponds exactly
with that given above.) The variation in capacitance of the inner layer for positive and
negative surfaces is a familiar feature of the mercurysolution interface (Fig. 7.4.5) and is
attributed to the closer approach distance of (unhydrated) anions compared to (hydrated)
cations. [Strictly one should not compare these integral capacitance values with the
differential capacitance but in regions where K; is constant they are identical (Exercise
7.4.5) and even when Ci is changing fairly quickly near the p.z.c (Fig. 10.3.3) they differ
by only a few pF cmP2 (Exercise 10.3.4)].
The data shown in Fig. 10.3.1 suggest that over the range 2 > (oo/pC cmp2) > -2,
different but constant values of Ki can be used on either side of the point of zero
charge. These values are compared in Fig. 10.3.3 with the estimates of the dzfferentiul
capacity of the inner layer, obtained from potentiometric titration data by Lyklema and
Overbeek (1961). It appears from this figure that a good description of the charge
behaviour at higher values of I a0 I could not be given with these same K; values.
Unfortunately there is no reliable electrokinetic data on this system for values of a0
outside the range indicated on Fig. 10.3.1 and 2.
494 I 10: ADSORPTION AT CHARGED INTERFACES
Fig. 10.3.2 {-potential of AgI in 0.001 M KNO3; KU = 1 and T = 293 K. 0 :calculated using the
Henry equation (8.2.25), from data of Osseo-hare et al. (1978). 0: obtained from the complete
numerical treatment of O'Brien and White (see Section 8.10). The full line is the theoretically predicted
value of $d, in the absence of specific adsorption, for & = 18 pF m-'. (From Hunter 1981.)
-KF
I l l 1 1 1 1 1
+2 0 -2 -4
uo(p c cm-')
Fig. 10.3.3 Comparison of inner layer differential capacitance, Ci, of AgI as determined by
titratable charge (Lyklema and Overbeek 1961) with estimates of Ki obtained (A) by Hall and Rendall
(1980) and (B) by Hunter (1981) using electrokinetic data. Values of Ci for the mercury-solution
interface are included for comparison.
DETECTION OF STERN LAYER ADSORPTION 1495
Fig. 10.3.4 The congruence test applied to the silica-solution interface. The catastrophic failure is
here attributed to counterion adsorption into the Stern layer. (From Hall and Rendall 1980.)
When the congruence test is applied to electrokinetic data on silica it fails in a very
obvious fashion (Fig. 10.3.4). No minor shift in the position of the plane of shear can
account for such behaviour. What Fig. 10.3.4 does not show, however, is the very large
increase which occurs in 00 as K increases (see Figs 10.1.1 and 10.4.1). T o reconcile the
resulting large 00 values with the quite modest measured values of { it is reasonable,
therefore, to assume that one (or perhaps both) ions of the electrolyte are, in this case,
adsorbed into the inner layer, even though the potentiometric titration curves exhibit a
common intersection point (c.i.p.). We would have to postulate that cations and anions
are specifically adsorbed in approximately equal amounts (or not at all) at the P.Z.C.in
order to understand why the c.i.p. occurs. Indeed it transpires that the parameters
introduced to describe the adsorption of these simple ions (Na+, K+, C1-, NO,) on
oxide surfaces do correspond to almost exactly equal adsorption of each ion type at the
P.Z.C.(Section 10.4).
3 4 5 6 7 8 9 10
Fig. 10.3.5 (a) The effect of introducing a charge free inner layer into the single site dissociation
model. (From Healy and White 1978, with permission.) Approach of a cation to the plane of the
surface charge need not be restricted to one atomic radius as in (b) but may occur as in (c). This
would correspond to very high values (-100-250 pF cm-’) of the inner layer capacitance Ki.
experimental data of Ottewill and Yates (1975) on a carboxylate latex show no such
reduction and can be best represented by a very high Ki value.
Considering eqn (7.3.36) in the form Ki = Ei/d and recognizing that the dielectric
permittivity Ei is, if anything, significantly less than the bulk value, it follows that the
thickness, d, of the compact layer must be very small (< 0.3 nm and probably
< 0.05 nm) in these systems. This means that the counter ions can approach very close
to the plane of the surface charge groups. They are evidently not subject to the same
restriction as applies on the mercury, nylon, and AgI surface. It should not be
DETECTION OF STERN LAYER ADSORPTION 1497
0
CH2-0-R’ + II
(CH,),N-CH
CHa-R”
II 0
II + I
R ” 4 4 H
CH2-0 -P --O -CH2
-CH2 4 H R -NH3
-NH3
I
0-
I R’-&CH,
Fig. 10.3.6 The common phospholipids. They are tri-esters of glycerol. The third acid residue is
an organic amine-phosphate. R ’ and R ” are long-chain fatty acid residues (CH2,+zCO- or
&Hz,CO-) with n = 13-20. The commonest acids are hexadecanoic (palmitic) (a = 15) and cis-9-
octadecenoic (oleic) (n = 17).
surprising that in these carboxylate latices the surface charge groups do not generate an
impenetrable plane with the counterions constrained to remain at least one ion radius
distant from the head-group plane (Fig. 10.3.5(b)). Rather one would expect that the
arrangement shown in Fig. 10.3.5(c) would be more energetically favourable. This
conclusion had already been reached by Davies and Rideal (1963) for spread
monolayers at the air-water interface. It appears to be a quite general phenomenon for
surfaces in which the charge is confined to particular chemical groups (although nylon
seems to be an exception). A model of this sort was used by Hunter (1966) to account
for the electrokinetic behaviour of surfactant stabilized emulsions. In that case $0 and 5
had been estimated independently (Haydon and Taylor 1960) and at ‘low’ charge
densities the two were identical, suggesting that the diffuse double layer could be
assumed to start from the plane of the headgroups. (‘Low’ in this context (-10 p C
cmP2) is actually higher than the maximum values attained by carboxylate latices.)
More recently, McLaughlin and his collaborators, in an interesting series of papers on
phospholipid vesicles (see for example Eisenberg et al. (1979)) have also used a model in
which the Poisson-Boltzmann equation is assumed to hold right up to the plane
containing the head group. In their systems the surface charge density on the vesicle can
be varied by varying the proportion of phosphatidyl choline (PC) to phosphatidyl serine
(PS), because PC is dipolar whilst PS is negatively charged (Fig. 10.3.6).
T o analyse their results, McLaughlin et al. introduce a modified form of Stern layer
which is essentially of zero thickness, by allowing counterions to adsorb onto the
headgroups in accordance with a Langmuir isotherm (compare with Sections 7.4.3,
10.1.1, and Exercise 10.3.5):
498 I 10: ADSORPTION AT CHARGED INTERFACES
On
(10.3.11)
where 00 is the maximum charge density (due to the total PS- concentration in the
surface), O, is the net charge after adsorption of Na+, and [Na+Is is the surface
concentration of the counterion. (ek
is the intrinsic association constant.) They then
assume that the plane of shear (Section 8.2.1) lies at a distance A ( m 0.2 nm) from the
charge groups and calculate the potential there ($) from (compare eqn (7.3.20):
The value of q0is calculated from the usual Gouy-Chapman expression (eqn (10.3.2))
assuming that the surface charge density which determines $0 is equal to a,.
+
The agreement between (at A = 0.2 nm) and the measured <-potential is excellent
(Fig. 10.3.7) and the model can be checked using two independent methods:
fluorescence probes (Section 9.5.2) adsorbed at phospholipid bilayers and electrical
conductance measurements. Both give values for $0 in close accord with those
estimated from the model.
<
The use of eqn (10.3.12) to estimate implicitly assumes the validity of the Poisson-
Boltzmann equation even in that first 0.2 nm between the head groups and the shear
plane. Although there is sufficient room for the necessary counterions to be
accommodated, it must be noted that the effective local average concentration in that
0.114
c
g -100
-120 0.015 -
- 140
0.0031
-160 -
/ 1 1 , 1 1 1 1 I I 1111,1I , I 1 1 ( 1 1 1 1 I
0.01 0.1 1 10
Surface concentration of PS (nrn-')
Fig. 10.3.7 {-potential of vesicles formed from phosphatidylcholine (PC) and phosphatidylserine
(PS) mixtures. The total negative charge, no, is determined by the PS content since PC is
zwitterionic. Some of the charge is balanced by Na+ ions in the surface layer and the zeta potential is
calculated some little distance from the head groups using the Poisson-Boltzmann equation with @o
estimated from the nett surface charge. (From Eisenberg et al. 1979.)
DETECTION OF STERN LAYER ADSORPTION 1499
layer can be very high -up to -9 M in the most unfavourable case (Exercise 10.3.6).
The simple Poisson-Boltzmann equation must be treated with considerable suspicion
at such high electrolyte concentrations.
The value of 4:; = 0.6 M-' required by Eisenberg et al. (1979)to fit their data
corresponds to a chemical free energy of adsorption of the sodium ion to the
headgroups of
(1 0.3.14)
leads immediately to the expression derived earlier for the Stern layer charge (eqn
(7.4.11)) but in this case it is actually the surface charge (Exercise 10.3.7):
where -S is a surface site and the 55.5 in the denominator converts from mole
fraction to molar concentration units. (Note that K A Nis~not an intrinsic dissociation
constant.) In this case the amount of ANS- bound to the surface of the liposome
(i.e. [-%ANSI can be estimated from the fluorescence intensity, I,so that (Exercise
10.3.8):
1
--
I - 4[-S -
1
ANS-] 1 (1 0.3.16)
For all of the salt systems studied (NaC1, KC1, CaC12, MgC12) the value of Q; was
-(16.3 f 0.2) kT o n phosphatidyl choline and -15 kT on phosphatidyl ethanolamine.
We may conclude then that in systems where the surface charge is produced by
groups which can be expected to be fairly mobile in the interface, the 'Stern Layer' is
500 I 10: ADSORPTION AT CHARGED INTERFACES
z -l
Fig. 10.3.8 Plot of I-' against the <-potential/concentration function suggested by eqn (10.3.16).
(From Kamo et al. 1978, with permission.)
very thin, the integral capacitance is very large and, to all intents and purposes, the
Poisson-Boltzmann equation appears to hold essentially right up to the plane of the
headgroups.
Exercises
10.3.1 Establish eqns (10.3.6) and (10.3.7).
10.3.2 Establish the modified form of eqn (10.3.7) assuming @O = 0 at the P.Z.C.
10.3.3 Establish eqn (10.3.10).
10.3.4 Fig. 10.3.3 shows a plot of the differential capacitance of the inner layer on AgI:
ci =
d(@O- @d)'
If it is represented by the empirical expression Ci = 3 1 5Do +
one can estimate the potential drop (@o- @d) when a0 = 1 pC cmP2 (by
integration). Use this estimate to calculate the integral capacity and compare it
with the average value of Ci over the range (31 < (Ci/pF cmP2) < 36).
10.3.5 For the reaction involved on the phospholipid vesicle surface:
to the head groups. (The ionic strength may be assumed to be equal to the Na+
molarity shown on the figure.)
[Note that the log scale in Fig 10.3.7 must be read as the reciprocal in nm2 per
molecule.]
10.3.7 For the adsorption of ANS- on liposomes, Kamo et al. (1978) consider the
equilibrium
where x is the mole fraction of ANS- in the bulk solution. Hence derive eqn
(10.3.15) assuming that AGS, = -kT In KANS= - (e@o - Oi).
10.3.8 Establish eqns (10.3.16) and (10.3.17).
. [-SO-][H+],
-SOH p - SO- + H+; KEt = (10.4.2)
[-SOHI
with the constants formulated as acid dissociation constants. Counterions are then
assumed to adsorb onto the resulting charged groups:
502 I 10: ADSORPTION AT CHARGED INTERFACES
The protons involved in these reactions are assumed to lie in the surface plane so that
[H+], is given by eqn (10.2.10) with $ = $0, whilst the counterions reside in the Stern
plane where the potential is $i.
The intrinsic constants can then be written (Exercise 10.4.1):
(10.4.7)
(10.4.8)
+
Ns = NA([SOW [SOH:] + [SO-] + [SOHlCl] + [SO-Na']), (10.4.9)
00 = eN~([soH,+]
- [SO-] + [SOH$Cl-] - [SO-Na+]), (10.4.10)
Combining these with the expressions for the integral capacities of the two parts of the
inner (compact) layer:
together with the relation (7.3.27) between $d and a d and (7.3.39) between the charge
densities provides the complete set of equations for the model. The six unknowns: $0,
THE OXIDE-SOLUTION INTERFACE I 503
y!ri, y!rd, 00, q,and Od can be solved for in terms of the independent variables, c and pH,
provided suitable values are chosen for the parameters:
* pNa,
t * P$, K F and K Z , N,, Kl and K2
with the proviso that Kal and Ka2 are linked to the P.Z.C. (Exercise 10.4.2):
(10.4.12)
There are several possible procedures for solving the model. The first step is the
estimation of the intrinsic complexing constants, *e,!,, which Davis et al. (1978) and
James and Parks (1982) estimate using a double extrapolation technique [extrapolating
to zero surface charge and zero electrolyte concentration]. Unfortunately, the
procedure is somewhat ambiguous in this case because of the possible co-existence of
sites of opposite sign at a given pH. Davis et al. assume that at any pH (not too close to
the P.z.c.) only one of the possible sites is present and the procedure is then essentially
the same as for a single site model. This assumption is, however, strictly correct only
when the acid dissociation constants of the surface sites are very different, that is for
rather large values of A p e ( > 4). The oxide which they treat in detail (TiO2) has the
smallest ApPft value (-2 according to Fig. 7.10.2). In fact, their procedure produces a
rather larger value of A p P t t (-4) which may well be an artefact. If one uses an
optimization procedure to evaluate the *Pt values without assuming that only one
sign of charged site is present at any pH, one obtains smaller values of Ap@ft for
TiO2, closer adherence to the Nernst equation, and larger numbers of bound
counterions (Koopal, personal communication).
The complete amphoteric site-dissociation-site-binding model has been solved
using a programme called MINEQL and applied to the oxide-solution interface (Davis
et al. 1978) and the polymer-latex-solution interface (James et al. 1978).The results of
a typical matching process for the Ti02 are shown in Fig. 10.4.1. Note that the
required values of p * p n tfor the cation (7.2) and the anion (4.2)correspond to values of
the chemical adsorption potential, f3ion, of (Exercise 10.4.3):
so that OK+ = -8.4 KT and @NO; = -7.5 kT. It is this relatively small difference in
adsorption potential between the cations and anions which leads to a common
intersection point. As Johnson (1984) has shown in his computer simulation studies,
quite large differences are required between 0, and 8- to produce a significant shift in
the P.Z.C.(These simple ions would, therefore, be regarded as specifically adsorbed at
the oxide-solution interface.) The amount of specific adsorption occurring near the
P.Z.C.is very large indeed. It is given by (Exercise 10.4.4):
lGl - c exp(-&/kT)
- (10.4.14)
N,e +
55.5 c exp(-&/kT)
504 I 10: ADSORPTION AT CHARGED INTERFACES
-20 c
Fig. 10.4.1 Surface charge density (Yates 1975) and <-potential (Wiese 1973) of a Ti02 dispersion
as a function of pH at various concentrationsof KNO3 and at 25 "C. Solid lines are calculated from
the site binding model with constants indicated on the figure. (From James and Parks 1982.)
so for c = 0.1 mol L-' and 13, M -8 kT,o*/eN, is about 0.84; i.e. 84% of all sites have
counterions of each sign associated with them!! At such levels this simple expression is
obviously inaccurate since it does not take account of the area occupied by both ions. A
more accurate (but still very approximate) estimate would include a factor of 2 in the
second term of the denominator (Exercise 10.4.4) and this would imply a figure of 46%
occupancy for each ion (which is at least possible). The point to be noted is simply that
on the Stern model of the double layer the amount of counterion adsorption is
extremely high. For c = 0.01 M, the fraction is 0.35 (0.26) and for c = 0.001 M it is
0.051 (0.049). The figures in brackets are obtained from the modified formula and
these lower values are likely to be much more reliable estimates.
It should be noted that these estimates of and the resulting I& I values are
much larger than those obtained by Barrow et al. (1980) using the model introduced by
Bowden et al. (1977). The discrepancy is, however, probably more apparent than real.
The Bowden model postulates that a neutral site should be defined by the group OH-
+
M-OHz so that only one charged site (either or -) can develop for every two surface
oxygens. Their N , value is, therefore, perhaps half of that used by James and Parks
(1982). Such a change would be expected (from eqn (10.4.14), to require a larger I f3+I
to achieve a given value of I o+ I . In the Bowden model, however, a separate estimate is
THE OXIDE-SOLUTION INTERFACE I 505
made of the maximum number, NT, of adsorption sites for each anion and cation
(based on the plateau of the adsorption when the electrostatic potential is favourable).
Some of the ‘chemical affinity’ is, therefore, subsumed into this revised estimate of NT
for each ion. (Recall Exercise 10.3.7 and note that the product N s K ~ occurs~ s in the
numerator.) Also, in the James and Parks model, ions such as Na+ and C1- can only
adsorb onto previously charged sites (- SO- and - SOH: respectively) which requires
a bigger I 8+ I to achieve the same level of saturation. This postulate can be regarded as
a crude means of separating out, to some extent, chemical effects from those related to
the state of charge of the surface; it should not, therefore, be lightly discarded.
The nature of the interaction between simple ions such as K+, Na+, C1-, and NO,
and the oxide surface is discussed in some detail by Sposito (19816). He regards them
as forming ‘outer sphere’ complexes with the appropriate surface site, rather than the
‘inner sphere’ complexes one would normally expect for a specifically adsorbed ion.
This means that the adsorbed ion retains its hydration sheath. In terms of the double
layer model developed on mercury, such an ion would not be regarded as specifically
(i.e. chemically) adsorbed, but on the oxide surface this is no longer the case. Even with
its hydration layer, the ion is involved in a ‘chemical complexation’ reaction, although
the large values of inner layer capacitance (Kl M 200 pF cmP2) suggest that these
complexed ions lie near to the plane of the surface charge.
The question whether ions form inner-sphere or outer-sphere complexes is not
simple. In their studies on the adsorption of multivalent ions onto phospholipid
vesicles, McLaughlin et al. (1981) used n.m.r. measurements to distinguish the two
possibilities for the Co2+ ion. They found that only about 10 per cent of the cobalt was
bound in inner complexes whilst the remainder was outer sphere. It should be noted,
however, that the distinction between the two is perhaps not so important on these
surfaces. As noted above (Section 10.3.1), the adsorbed counterions are able to sit in or
close to the plane of the head-groups whether or not they retain their hydration
sheaths. The difference between inner and outer sphere complexes should not,
therefore, be regarded as synonymous with adsorption in the inner and outer
Helmholtz plane (Fig. 7.3.4). The difference, presumably, lies principally in the
magnitude of the chemical adsorption potential (being more negative for inner sphere
complexes).
It should be made clear that the site-binding model estimates of adsorption of simple
ions like Na+ and C1- on oxide surfaces are demonstrably excessive. The radiotracer
studies of Foissy et al. (1982) discussed in Section 10.1.1 may not give exact values but
there can be no doubt that, if there were such large amounts of salt adsorption at the
P.z.c., those studies would have shown them. It seems much more likely then that the
salt adsorption is a fallacy and that the picture revealed by Fig. 10.1.1 is nearer the
truth. How then do we reconcile the results? The large values of 13i which appear in the
site-binding models occur because it is necessary to pull large amounts of counterion
charge into the inner layer to balance the high surface charge with a small <-potential.
Because the adsorption is confined to existing charged sites, it is necessary to postulate
a high 8 value (eqn 7.4.14) for the cation in order to get enough adsorption. This, in
turn, produces a large adsorption density of the cation at the p.z.c and some adsorption
of this same ion when the surface changes sign. In turn that requires more adsorption
of the anion to produce the same effect on the diffuse layer charge. Every increase in
the ‘specific’ adsorption potential of one ion produces a counterbalancing increase for
506 I 10: ADSORPTION AT CHARGED INTERFACES
Fig. 10.4.2 Plots of NS-' against C,j for: (A) AgI, (B) TiOz, and (C) SiOz using KNO3 as
indifferent electrolyte. Curve D shows the effect of putting A = 0.5 nm with the same initial slope as
(C). (From Smith 1976, with permission.)
the other in order to satisfy the requirement that there should be no diffuse charge at
the P.Z.C.These high values for both ions are thus artefacts of the fitting procedure.
The picture can be examined further using Smith's (1976) analysis of the
electrokinetic data on oxide surfaces. Using a somewhat more elaborate version of the
technique described in Section 10.2.3, Smith shows that, if inner layer adsorption is
occurring, then the expression for (d$ro/d$d)K becomes (Exercise 10.4.5):
(10.4.15)
It is, of course, no longer possible to identify (d$ro/dpX) in eqn (10.2.13) with the
Nernst slope, N, but instead one must use eqn (7.10.6).Smith then arrives at (compare
with 10.2.14):
where K, = 2N,e240/kT.
Plots of NS-' against c d (= E K ) are shown in Fig. 10.4.2 for AgI, TiOz, and SiOz.
For the AgI system, inner layer adsorption is absent near the P.Z.C.(a;= 0) and $0 is
large so that eqn (10.2.14) is recovered and the system shows the Nernst slope. For
Ti02 and SiOz the slope is different but the fact that the plots are still linear and
pass through NS-' = 1 suggests that A is small (< 0.5 nm). Smith analyses the case
KZ >> K; (corresponding to a shear plane close to the IHP but a significant
separation between this and the plane of the head groups). This leads to:
THE OXIDE-SOLUTION INTERFACE I 507
(10.4.17)
The linearity and the intercept at NS-' = 1 then would require that (da;/d$d) be
small compared with c d for d l c d (which seems difficult to understand when c d
approaches zero).
These results can be more readily reconciled with the ideas developed earlier in this
section if we assume that the capacitance of the inner part of the compact layer is high,
as is normally required in the models (Kl M 200 pF cmP2).In that case K;/K2 % 1 and
eqn (10.4.17) is replaced by:
NS- = 1
1 + (ki-+- k> ki:)cd exp @A). (10.4.18)
If in the neighbourhood of the P.Z.C. we set 0; equal to its maximum value (a;% ao)
(Fig. 10.1.1) and use the experimental values of dao/d$d % dao/d( we find that the
term K;'(dai/d$d) is very small compared with unity for both Ti02 and Si02
(Exercise 10.4.6). The linear relationship, the intercept, and the altered slope are then
all easily understandable. Note that the high slope for Si02 can be accounted for in
terms of a decrease in the apparent integral capacitance of the inner layer, KA:
This corresponds to a small value for K,, and hence a small q50 (from the definition of
K, in eqn (10.4.16)) and a large ApK!tt value as noted earlier (Fig. 7.10.2).
There are not many systems for which both titration (DO)and (-potential data are
available outside the Ti02 system shown in Fig. 10.4.1. T h e charge data is well
described in that figure but the agreement with the (-potential is rather poor. The
impression is often given (see for example, Morel et al. 1981) that any one of the
usual complexation models can give a satisfactory description of titration data, with
a suitable choice of parameters. That is certainly not the case, however, if one
demands a simultaneous description of 00 and (. Unfortunately, that more stringent
test has rarely been applied. Smit et al. (1978) have, for example, made some direct
radiotracer estimations of the amount of sodium adsorbed into the compact layer of
(non-porous) vitreous silica and, like Foissey et al. (1982) have been able to account
satisfactorily for both the electrokinetic (with ( = $d) and titration data on that
system. The new electro kinetic methods of measurement, involving high frequency
conductance and mobility, are also providing some fresh evidence on this question
as we noted in section 8.11.
I
Exercises
10.4.1 Verify eqns (10.4.6), (10.4.7), and (10.4.8).
10.4.2 Establish eqn (10.4.12).
Next Page
(Hint: assume UK = N A e [SO-K+] and @i M @o).What are the QN, and f3cl
values corresponding to the Barrow et al. (1980) estimates of Kit = 0.72 L
mol-' and K't:
= 0.1 1 L mol-' on goethite? Note that Davis et al. used pK'At =
4.9; pK'G' = 10.7 and p*K'g = 6.6 to describe their data on goethite. To what
value of 8- does this correspond?
10.4.4 Verify eqn (10.4.14). The more general expression for a;,taking some account
of surface coverage by both ions is:
with a similar expression for a-, where g* = exp (-A G!*,da,/kT) = exp[-
(x*e@i +8*)/kT] and x is the mole fraction of the ion in solution. Show that
this implies the correction factor 2 (at the P.z.c.) referred to in Section 10.4
under eqn (10.4.14).
10.4.5 Use the definitions of K1 and Kz:
with the charge neutrality condition, to show that for the double layer model
shown in Fig. 7.3.4
+
if c d = dad/d@d and KF1 = KT1 KF'. (Assume that K2 and Ki are
independent of @d when I 5 I is small.)
10.4.6 Use the data given in Fig. 10.4.1 to show that
for the Ti02 system. Take N, = 5 x lo'* m-', and get &J from Section 7.10
and the ApK value. [(dao/d<)s+o is even lower for the SiOz system.]
The Theory of Van Der Waals Forces
11.1 Introduction
11.1.1 Interactions between molecules
11.2 London theory
11.3 Pairwise summation of forces (Hamaker theory)
11.3.1 Interaction of a molecule with a macrobody
11.3.2 Interaction of two macrobodies
11.3.3 Effect of the suspension medium
11.4 Retardation effects in Hamaker theory
11.5 The Deryaguin approximation
11.6 Modern dispersion force theory
11.6.1 Interaction between two flat semi-infinite bodies across a vacuum
11.6.2 ~ ( wrevisited
)
11.6.3 The dispersion relation method
11.6.4 Modern theory for planar half-spaces
11.7 Numerical computation of interaction energy
11.7.1 Construction of .$<I
11.7.2 Representation of &
. ) for metals
11.7.3 Numerical evaluation of El32 (L)
11.8 Influence of electrolyte concentration
11.9 Theoretical estimation of surface properties
11.9.1 Surface and interfacial tension and energy
11.9.2 Contact angle of liquids on low-energy solids
11.IIntroduction
A treatment of the theoretical calculation of the van der Waals or Hamaker function
(A) will necessarily take us into mathematical areas that many will find unfamiliar. We
will make every effort to explain the analysis as clearly as possible but will concentrate
on those aspects that are most germane to the practical problem of calculating the value
of A for a particular material. It must be admitted that not every colloid scientist will
choose to undertake such a calculation but we will show how, at least for simple
533
534 I 1 1 : THE THEORY OF V A N DER WAALS FORCES
Many of the exercises in this chapter are designed to show the reader how much of
the mathematical manipulation should be accessible with only the tools of elementary
calculus and a little algebra. It is not the intention to expose the assumptions of the
macroscopic theory to rigorous analysis but rather to indicate the nature of the
experimental data that is currently used to evaluate the interaction energy. Hopefully,
a wider knowledge of the limitations of that data will lead to the collection of much
more useful information on the important colloid chemical materials.
--
The range of values of the Hamaker function (at short distances where retardation is
unimportant) for substances immersed in water can be summarized as follows:
These figures reflect the differences in the polarizabilities of the various materials, on
which A largely depends. For approximate calculations, the above information may be
sufficiently accurate. For more exacting analyses, the more detailed considerations of
this chapter must be invoked.
(p+$)(V-nb)=nRT (1 1.1.1)
in which the constant b accounted for the finite volume occupied by the molecules of the
gas and the constant a was directly related to the strength of the intermolecular attractive
forces. The success of the van der Waals equation in summarizing the properties of
gaseous systems and their phase behaviour stimulated theoretical attempts to discover
the origin of the forces responsible. It is interesting to note that, even in 1873 with only a
vague notion of the origins of intermolecular forces, van der Waals was already
separating the short-range repulsive forces that give rise to the excluded volume term b
from the long range attractive forces that account for the constant a.
It has been pointed out (Sparnaay 1983) that as long ago as 1686, Isaac Newton in his
Principia discussed the attraction between two bodies separated by a distance, R, in
terms of a force, proportional to R-", where n > 4.Though he could not discuss the
possible origins of such a force he could show that n must exceed four or the
interaction between a small particle and a large plate would become infinitely large.
In the early years of the present century, several workers (van der Waals 1909;
Reinganum 1912; Thomsonl914;Keesom 1921) sought an explanation of the 'van der
Waals' forces by postulating the existence of a permanent electric dipole in each atom
or molecule of the gas. By suitably averaging the permanent dipoledipole interaction
energy over all orientations of the dipoles at a fixed separation distance R, an attractive
potential energy proportional to 1/R6 was obtained.
536 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
This simple theoretical explanation suffered from the unfortunate fact that the
fundamental postulate could be tested experimentally and the necessary permanent
dipole moment was found to be absent in most of the simplest molecules. Debye (1920)
proposed a less easily tested model in which atoms and molecules were assumed to
possess a permanent quadrupole moment. An attractive interaction energy (al/R9)
was obtained by orientation averaging the permanent quadrupole-induced dipole
interaction energy.
It was, however, only with the advent of quantum mechanics that a satisfactory
explanation of the origin of van der Waals forces was forthcoming (London 1930).
a0 = e2 / 8 x ~ o h ~ (1 1.2.1)
where e is the proton charge, €0 is the permittivity of free space, h is Planck’s constant,
and u is a characteristic frequency (in Hz) associated with the electron’s motion around
the nucleus ( u = 3.3 x l O I 5 s-’ for the Bohr hydrogen atom). (Note that the value of a0
given by eqn (1 1.2.1) corresponds to the maximum value in the electron density
distribution, 111.0 I 2, in the electronic ground state of hydrogen, as calculated by
quantum mechanics.) The energy -hu is the energy of the electron in the ground state
(relative to the separated particles) and so is equal to the ionization potential of the H
atom.
Although the H atom has no permanent dipole moment it can be regarded as having
an instantaneous dipole moment, p, or order p l m aoe. The field of this instantaneous
dipole, at a distance R from the atom will be of the order:
(11.2.2)
If a neutral atom is nearby it will therefore be polarized by this field and acquire an
induced dipole moment of strength p2:
p2 = aE m aaoe/(4neoR3) (1 1.2.3)
where a is the atomic polarizability of the second atom. This measures the ease with
which the electron distribution can be displaced (Section 3.2) and is proportional to the
LONDON THEORY 1537
volume of the atom (a M 4n€oa;). The potential energy of interaction between the
dipoles $1 and p2 is then, using eqn (11.2.3):
Knt(R) = -$lp2/4n€oR3 M - ~ a ; e ~ / ( 4 n € o ) ~=
R ~-2a2hv/[(4nc,)2R6]. (1 1.2.4)
is the frequency (rad s-l) of electromagnetic radiation which would cause the
transition from the ground state to the excited state llr, in the isolated molecule. The
corresponding oscillator strength for this transition is given by:
(11.2.7)
where Ip, I is the magnitude of the transition dipole, h = h/2n and me is the electron
mass. fOm measures the probability of the transition occurring and is, therefore, the
measure of the intensity of the absorption band.
The similarities between eqns (11.2.5) and (11.2.4) are obvious. Apart from the
1/R6 dependence we see that the magnitude of the interaction energy is determined by
the ease with which the electrons in the two atoms are able to undergo displacements,
since this is, in effect, what the terms in the summation calculate. The connection
between Knt and the absorption spectrum of the material is already apparent and we
will return to that point again.
The oscillator strengths, fOm obey a useful sum rule (Dalgarno and Lynn 1957)
derived from quantum mechanics:
Elirn
m
=N (11.2.8)
where N is the number of electrons in the molecule. The London theory of the van der
Waals force does not require the existence of permanent molecular multipole moments
and demonstrates the universality of this long range attractive force.
If it can be assumed that only one frequency dominates in the interaction it is
possible to write the constant CAB in eqn (11.2.5) in the form:
(11.2.9)
538 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
( 1 1.2.10)
(1 1.2.11)
With these definitions the van der Waals interaction energy (eqn (1 1.2.5)) becomes:
3-
V,,(R) = - - h*(Ro/R)6 ( 1 1.2.12)
4
- - -
For R 0.4 nm at T 300 K, we see that Knt 2 kT,which emphasizes the fact that
van der Waals forces are thermodynamically important forces under normal
experimental condition. It should be remembered, however, that other long-range
interactions between molecules can be important, especially if the molecules have a
permanent dipole moment. The importance of the London interaction for colloid
science lies in the fact that, when interactions between large collections of molecules
are treated, it is the London dispersion force that dominates over the dipole-dipole
(Keesom) and dipole-induced dipole (Debye) forces (if the molecules are
uncharged). This is because the summation assumes that the interactions are
pairwise additive and this is a rather better assumption for the London forces than
for the others (Overbeek 1952, p. 265). The modern Lifshitz theory avoids this
problem and calculates the change in the (Helmholtz) free energy due to all of the
interactions at the same time.
PAIRWISE SUMMATION OF FORCES (HAMAKER THEORY) 1539
r
Exercise
11.2.1. Establish eqn (11.2.4). [Note that the polarizability is sometimes defined in
terms of the equationp = a@ so that it has dimensions of volume. (Atkins
1978, p. 751; 1982, p. 772).]
11.2.2 Establish eqn (11.2.12) and note the similarity to eqn (11.2.4). Only the
constant is missing from the simple formula.
The interaction energy of the N-body system, in the pairwise summation method, is
taken to be
(1 1.3.2)
where ?it (R)is the interaction energy of molecules i andj separated by distance Rq in
the absence of any other molecules.
For van der Waals forces derived from second-order quantum perturbation theory,
eqn (11.3.2) is only an approximation. The internal states of molecules i and j will be
modified by the presence of all the other molecules of the system and they do not,
therefore, interact with each other in the way that they would if the other molecules
were not present. Clearly the pairwise summation method will be least in error when
the molecules are far from one another, so that the individual pair interactions are
relatively unaffected by the presence of other molecules.
(a)
Fig. 1 1.3.1 (a) The geometry of a type 1 molecule interacting with a body comprised of type 2
molecules at density p2. (b) Coordinate system for a Hamaker summation in planar half-space. z is
the radius of a ring of the material in the solid.
interaction energy of molecule 1 with the molecules of the volume element in the
pairwise approximation is:
(1 1.3.3)
The total van der Waals interaction energy of the molecule 1 with the body 2 is
obtained by integrating eqn (11.3.3) over the volume of the body, viz.
For the case of usual interest (a molecule at a distance D from a planar half-space)+ we
may write, using Pythagoras' theorem (see Fig. 11.3.1(b) and Exercise 11.3.1):
0 0 0 0
Note how the summation process leads to a much longer range interaction (DP3)
compared with that for moleculemolecule interaction (rP6).
VA = p1
s
Body 1
A12
Body 1 Body 2
dV' (11.3.7)
In the limit as the bodies are separated by a distance large compared to the largest
dimension of each body, the distance Ir - r' I can be replaced by R, the distance
between the centres of mass of the two bodies, and eqn (11.3.7) reduces to
(11.3.9)
A12
vA(L) = -~ (11.3.1 0)
67c
Surface of body 1 0
with the aid of eqn (11.3.5). In this case the volume element has been chosen as dxdS
where x is a coordinate measuring normal distance into body 1from its surface and dS1
is an element in that surface. Performing the x integration and evaluating the
interaction per unit area of the surface of body 1 we have
Note that the range of the interaction energy is even longer in this case, compared with
that in Section 11.3.1.
542 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
The Hamaker summation for more complicated geometries is, as one might expect,
a mathematically involved operation. One other geometry leads to a relatively simple
formula, viz. the spheresphere interaction.
When body 1 is a sphere of radius a1 and body 2 is a sphere of radius a2, with
centre-centre distance R, eqn (1 1.3.7) reduces to (Hamaker 1937):
(1 1.3.12)
(1 1.3.13)
VA(H) = --[-+
A12 1 hl +2hlln(hl( 1+h2 )}] (11.3.14)
12h1 1 + h 2 1 + h i +h3 1 + hl + h3
In the limit as H << 5,we have, from eqn (11.3.14) (Exercise 11.3.3):
VA(H) = -&[I
12H +t(l -
The case of a sphere of radius a at distance H from a planar half-space is obtained from
eqn (11.3.15) by setting a1 = a and taking the limit as a2 + 00. We obtain (Exercise
11.3.3):
(1 1.3.16)
Several other geometries yield analytic expressions for VA. Among these are
multilayers, and parallel and crossed cylinders. (See Mahanty and Ninham 1976,
pp. 13-21; also Fig. 11.3.2.)
Two spheres
al,a2>>H
Parallel cylinders
VA K
-
-length
H5
length)
0.c)
--- Crossed
cylinders
VA
V A =- (al a2) A/6H (per unit = -At12 nL2
area)
Fig. 11.3.2 Distance dependence of the non-retarded potential energy between pairs of solid
surfaces of various geometries. The retarded interaction has a distance dependence 1/H times that
and the proportionality constant then changes. In some cases the integration process yields quite
messy results and these have not been included explicitly. Mahanty and Ninham (1976) remark that
numerical integration over the particle volumes is usually easier and more satisfactory than seeking
analytical solutions for all but the simplest geometries.
but by a fluid medium whose molecules interact with the other molecules of the
system. The presence of these extra interactions can drastically diminish the value of
VA,and even reverse its sign in some circumstances.
T o see why, consider bodies 1 and 2 immersed in suspension medium 3. If they are
brought from infinite separation to some distance D apart, the free energy change (the
interaction energy) is not as large as in the case where medium 3 is a vacuum. When
isolated, body 1 is interacting with its environment - a universe of medium 3. When
544 I 1 1 : THE THEORY OF V A N DER WAALS FORCES
Fig. 11.3.3 Thermodynamic path for calculating the interaction energy V132(D)of bodies 1 and 2 in
a fluid medium 3.
brought close to body 2 it is interacting with a very similar environment, the only
difference being that a number of molecules of medium 3 have been replaced by the
molecules of body 2. The change of energy is less than that of the vacuum situation
where the environment of body 1 originally contains no molecules and is then modified
by the addition of the molecules of body 2.
T o calculate the van der Waals interaction energy in the presence of a suspension
medium, we consider the following thermodynamic path (see Fig. 11.3.3). Consider an
initial state in which bodies 1 and 2 are infinitely separated in medium 3. We may
regard the molecules of 3 in the volume that will finally be occupied by body 1, to be a
‘body’ 3 as shown. Let us remove body 1 and body 3 from the medium to vacuum. The
change in free energy in this first step is just
where D is the distance from ‘body 3’ to body 2 and Fi is the interaction energy of the
isolated body i with a universe of medium 3. The energy change in removing body 3 is
+
not - F3 but -[F3 -V33 ( 0 ) V32 (D] since its environment is not all medium 3. The
energy V33 (D)- V32 (D)represents the change in the interaction energy of body 3 with
its environment, when the molecules of 3 that would have occupied the position of
body 2 are replaced by body 2. Note in this argument we are using the notation that
V&(0)represents the (vacuum) interaction energy of a body the size and shape of
PAIRWISE SUMMATION OF FORCES (HAMAKER THEORY) 1545
body 1 comprised of k-type molecules with a body the size and shape of body 2
comprised of j-type molecules at separation distance D.
The second step is to return body 3 to the medium into the hole occupied originally
by body 1 and to return body 1 to the hole originally occupied by body 3. The energy
change in this second process is
As before, the energy Vlz - V13 represents the change in the interaction energy of body
1 with its environment when the molecules of 3 that would have occupied the position
of body 2 are replaced by body 2. The interaction energy Vl32 (D)of bodies 1 and 2 at
separation D in bathing medium 3 is given by
By the nature of the pairwise summation method we may write (see eqn (11.3.7))
where V(D)is a positive function only of the geometry of the system and independent
of the composition of the bodies 1 and 2 and Ak, is the vacuum Hamaker constant given
by eqn (11.3.8). Thus eqn (11.3.19) becomes:
The presence of a bathing medium does not change the distance dependence of the van
der Waals force but alters its magnitude through A132. With an obvious notation, we
may write Alv2 = A12 when bodies 1 and 2 are separated by a vacuum and from eqn
(11.3.22), it is apparent that
provided A33 < A13 + A32 which is usually so. As Fig. 11.3.4 clearly shows, this
reduction in Hamaker constant in a bathing medium can be one of even two orders of
magnitude. Obviously for All = A33, the function A131 is reduced to zero.
Several other generalizations can be made. In particular we note that Ajkj > 0 so that
like particles always attract one another in any medium. That is not necessarily the case
for unlike particles since A132 can be negative.
A reasonable approximation for the interaction function between different materials
is:
(1 1.3.23)
546 I I I : T H E THEORY OF VAN D E R WAALS FORCES
h
0
N
0
r
2s.
Fig. 11.3.4 The effect of an intervening medium (3) on the interaction between two semi-infinite
bodies. The value O f A33 is taken as 3.70 x lopzoJ corresponding to liquid water. (See Exercise
11.3.4.)
and this can be used to express eqn (11.3.22) in the more convenient form (Exercise
11.3.8):
(11.3.24)
It should be noted, though, that the usefulness of such approximations for accurate
research purposes is fast disappearing with the ready availability of high speed
computers and the rise of modern dispersion force theory.
Exercises
11.3.1 Establish eqn (11.3.5)
11.3.2 Establish eqn (11.3.14) from (11.3.12). Also show that for two equal spheres of
radius a:
where s = R/a.
11.3.3 Establish eqn (11.3.15) and (11.3.16).
11.3.4 Verify Fig. 11.3.4 using A33 = 3.70 x J for water.
RETARDATION EFFECTS I N HAMAKER THEORY 1547
characteristic wavelength ho for the internal molecular motion, the finite propagation
time (retardation) must be allowed for and the interaction energy will be smaller than
that predicted by the electrostatic theory.
By modifying Vn, to account correctly for the finite propagation time of the
electromagnetic interaction between the molecular charge distributions, Casimir and
Polder (1948) calculated the correct behaviour of Vint(R)for all R. In particular for
R << ho the RP6 form (eqn (11.2.5)) is valid and for R >> ha:
(1 1.4.2)
The major consequence of the retardation effect is to limit the range of macroscopic
(i.e. ‘long-range’) van der Waals forces.
Overbeek (1952, p. 266) represents the Casimir and Polder correction function in
the following way (compare eqn (1 1.2.12)):
Correction factor
Fig. 1 1.4.1 Retardationcorrection factor to the London attraction energy between two flat plates of
infinite thickness at a separation L. Note that althoughfb) = 0.59 for p = 3, the total attraction
energy is down by a factor of about four at that point. (From Overbeek 1952, p. 269, with
permission.)
THE DERYAGUIN APPROXIMATION 1549
Exercise
1 1.4.1 Plot the correction factorfb) (eqn (11.4.3))as a function ofp for 0 < p < 10.
Note the difference between this and Fp of Fig. 11.4.1. 1
11.5 The Deryaguin approximation
When dealing with interactions between macrobodies, it is often the case that the range of
the interaction is such that the two bodies do not interact significantly until the distance of
closest approach is small compared to the radii of curvature of the bodies. Under these
circumstances, a very useful approximate expression for the interaction energy of the
bodies can be derived from the corresponding interaction energy per unit area of plane
parallel half-spaces. The approximation is due, originally, to Deryaguin (1934).
A suitable coordinate system is shown in Fig. 11.5.1. The bodies are assumed
smooth and quadratically curved in the neighbourhood of 0 and 0’. Consider an
element of area d S at the point P(x, y, z ) on the surface of body 1. Provided the radii of
curvature of bodies 1 and 2 are large compared to the separation distance H, then the
area element can be approximately regarded as a surface element of a plane half-space
parallel to and separated a distance D (see Fig. 11.5.1) from another planar half-space.
If E(D)is the energy per unit area of half-space 1 interacting with the half-space 2,
then, in this approximation,
d V = E(D)dSl (1 1.5.1)
550 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
Fig. 1 1.5.1 Geometry of two interacting bodies illustrating appropriate principal coordinate axes
for each body. The insert shows the orientation of the coordinate axes at 0 and 0’, looking along the
line of centres between the two bodies.
is the interaction energy. The distance D increases as the area element is taken further
from the origin 0 and the associated energy d V decreases appropriately.
In the Deryaguin approximation, the total interaction energy of bodies 1 and 2 is
obtained by integrating d V over the surface of body 1, viz.
V(H) = /
Surface of body 1
dSIE(D). (1 1.5.2)
The energy E(D) must decay sufficiently rapidly with distance D so that contributions
to V(H) are insignificant from area elements very far from the centre 0 where curvature
effects are large and eqn (11.5.1) is a poor approximation. Deryaguin’s original analysis
was applied to two spheres of large radius of curvature, a, and in that case:
00
(1 1.5.4)
H
White (1983) has shown that the general equation for V(H)where the approaching
surfaces may have arbitrary orientation and curvature (provided the latter is large) is:
THE DERYAGUIN APPROXIMATION I551
M
(1 1.5.5)
where h.lh.2=
(d,-+rd,)(d2-+- d;) +sin 4
(d, d&
--- ---
a) ; ; (11*5*6)
The R s and R's are the principal radii of curvature of the two surfaces respectively.
The angle (see Fig. 11.5.1) is the angle between the principal axes on the two surfaces.
Obviously, 4 is immaterial when either of the bodies is a sphere and for two equal
spheres h.1h.2 = 4/a2, which reduces eqn (11.5.5) to (11.5.4).
From the definition of the force F(H) exerted by one body on another we have
that
(1 1.5.7)
Equation (1 1.5.5) is a valid approximation for any type of interaction energy and need
not be restricted to the van der Waals interaction of this chapter. T o apply eqn (1 1.5.5)
with confidence to any particular interaction it is sufficient to require that:
where LOis the length scale on which the interaction decays to zero and Ro is the
smallest radius of curvature of the system.
One geometry of particular experimental importance is that of crossed cylinders of
radii a1 and 4.In this case R1 = a,, R2 = 00, R', = a2, Ri = 00 and 4 = n/2. For this
geometry
ala2= i/ala2 (1 1.5.9)
V ( H )= 2x&a2) 1
H
00
dL E(L). (1 1.5.10)
11.5.1. Use eqn (11S . 5 ) to show that the van der Waals interaction energy of two large
spheres of radii a1 and a2 in the limit H << ai is given by
Note that this corresponds to taking only the first term in the expression
( 11.3.15).
The fundamental idea that lies behind the method is that the interaction we are
trying to calculated is propagated as an electromagnetic wave from one body to another
over distances that are large compared to atomic dimensions. Furthermore, the
frequencies with which the electrons in these bodies are able to move most readily (wg
< lo1*s-l) correspond to wavelengths which are also large compared with atomic
dimensions. It should be possible, therefore, to analyse the propagation of these waves
by appeal to the classical equations of wave motion (i.e. Maxwell’s equations). The
medium could be treated as a continuum and its bulk properties introduced through
the permittivity (or dielectric response), ~ ( w )introduced
, in Section 3.2. The method
may break down at very close separations, where the graininess of the matter becomes
evident, but that turns out to be of little importance.
Our analysis of the London dispersion interaction (Sections 11.1-1 1.3) makes clear
the importance of the fluctuating dipoles in two interacting atoms. The macroscopic
approach draws heavily on the idea of correlated fluctuations, that is, the idea that the
charge movements that occur on one atom are influenced by the movements occurring
on neighbouring atoms and even those that are further away. The macroscopic body is
considered to be made up of many local oscillating dipoles that are continuously
radiating energy (as any vibrating dipole must do according to classical electromagnetic
(e.m.) theory). These dipoles are also continuously absorbing energy from the e.m.
field generated by all of their neighbours. Obviously, they are best able to absorb
energy at frequencies corresponding to one of their natural resonances (recall Exercises
3.2.5-7) and these are also the frequencies at which they are most easily polarized and
those at which they radiate energy. Two macroscopic bodies thus ‘see’ or ‘feel’ one
another across an intervening gap as a result of the electromagnetic waves which
emanate from one to the other.
Most of the energy of these correlated fluctuations remains inside the body and is
part of its cohesive energy, i.e. what holds it together. The part that is of most concern
to us is carried by the electromagnetic waves that escape from the surface of the two
neighbouring bodies. This is the energy contained in the surface vibrational modes, and
our problem is to determine how that electromagnetic field is propagated across the
gap between the two bodies. We will show how that is done for the very simplest
possible situation: two semi-infinite flat plates (i.e. two half-spaces) separated by a
vacuum. That calculation will introduce most of the important concepts, which can
then be generalized to more important situations.
It should be obvious that, in considering how one material responds to the
electromagnetic field generated by a neighbouring material, we will draw heavily on the
dielectric response function (Section 3.2). That quantity contains all the important
information about how the substance responds to an alternating electric field. It should
also be clear that such information is implicitly contained in the electromagnetic
absorption spectrum of the material because that measures exactly the same thing:
what frequencies of e.m. energy can be taken up by the material and to what extent?
We will find that the function C ( W ) can, in principle, be constructed from a
knowledge of the absorption spectrum, but that it is a difficult task requiring an
extraordinary level of information over the whole frequency range (from microwave to
far ultraviolet). Fortunately, the solution of the van der Waals problem requires much
less information -merely the values of the function C ( W ) for purely imaginay values of
the frequency (i.e. ~(zt)where t is a real number). At first sight this may seem a rather
554 I 1 1 : THE THEORY OF V A N DER WAALS FORCES
(1 1.6.1)
so that e(i() may be regarded as the measure of the response of the system to an
exponentially decaying (rather than an oscillating) field. 1/t is then the time-constant.
We assume that this is the only contribution to the energy change. Calculating the
internal energy change corresponds to London’s and Hamaker’s calculation. Equation
Fig. 1 1.6.1 Two semi-infinite materials separated by a vacuum. The approximate shape of the
lowest permitted frequency wave function is shown. (Note the change in slope at the interface due to
the change in permittivity.)
MODERN DISPERSION FORCE THEORY 1555
(1 1.6.2) is true only at the absolute zero of temperature. The temperature must be
introduced by calculating the free energy change A F rather than AU and we will
introduce that correction later.
As the gap width diminishes, there is a decrease in the number and frequency of the
modes which can satisfy Maxwell’s equations in the gap and hence the (zero point)
internal energy of the system decreases. It is shown in Appendix A1 that the
frequencies which have the necessary characteristics are defined by what is called a
dispersion relation? which takes the form:
+
where A,, = (€1 - E O ) / ( E I €0) and R is the magnitude of the wave vector (eqn (3.3.1)).
The values of w(= wj) which satisfy eqn (11.6.3) are the frequencies which can be
propagated across the gap.
For very large L values, the exponential term may be neglected and wj (00) must
then satisfy:
(1 1.6.5)
Thus the absorption mode gives rise to a permitted frequency that is just a little higher
than the absorption frequency. Furthermore, if we add more terms in the series to
represent more absorption frequencies in eqn (11.6.5), each one will give rise to an
additional wj value like that in eqn (1 1.6.6),since each of the terms is important only in
the neighbourhood of the absorption frequency. In this way we could construct all the
wj s required to calculate
well separated.
$xli wj(00), in eqn (1 1.6.2), provided the absorptions were
At finite distances the solution would be a little more difficult but two things can still
be said: to each absorption frequency there still corresponds an wj that will satisfy eqn
(1 1.6.3) but its numerical value is lower than wj(00) and this will ensure that VA is
negative. Also, it is not too difficult to see that for finite L, eqn (11.6.3) has solutions
only when €1 (w) is negative. From eqn (1 1.6.5) we see that this occurs at values of w
just above a resonance frequency (Fig. 11.6.2).
+The word ‘dispersion’ refers to the process whereby light of the different
wavelengths is dispersed on going through a triangular prism. Equation (1 1.6.3) gives
the condition for propagation of a wave without hindrance (as distinct from
absorption).
556 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
/
/
Fig. 11.6.2 The dielectric response function E ( W ) for a single undamped oscillator of resonance
frequency W O . We will discuss the function +$) later, but note how simple it is in form compared
with the behaviour of E' in the neighbourhood of an absorption.
Fig. 11.6.3 (a)A schematic plot of ~ ” ( was) a function of frequency. (b) Schematic plots of s ’ ( w )
(dotted curve) and €(it)(smooth curve) as functions of frequency. Note the coincidence of the two
functions in the ‘flat’ regions between absorption frequencies.
€:(Lo) = 1 +- (11.6.7)
0
Eqn (11.6.7) means that if E”(w) (i.e. the absorption spectrum) is known for all
frequencies then ~ ’ ( w can,
) in principle, be calculated from eqn (1 1.6.7) (see
Fig. 11.6.3 (b)).
+This integral diverges near x = w so that region has to be deleted from the
integration process.
558 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
We have already assumed, in writing expressions like eqn (3.2.39), that the function
can take complex values of its argument but the results of Exercise 3.2.8 show that this
+
is merely a mathematical manipulation of the function ~ ( w=) c' Ed' of a real variable.
For later use we must now formally extend the definition of c:
m(t)exp(iwt)d t [3.2.25]
(1 1.6.8)
by eliminating the unknown m(t ). This equation connects the function €(it)to the
imaginary part of the real (i.e. experimentally measurable) dielectric response.
Note the close similarity of eqn (11.6.7) defining E' (w) and eqn (11.6.8) defining
E(E6) as functionalst of E;(W). We will make use of this similarity to construct €(it)from
experimental data (see Section 11.7). For the present, it is sufficient to note that €(it)
defined by eqn (1 1.6.8) is an even function of t. The Kramers-Kronig relation has
served to extend the definition of e(i6) onto the negative 6 axis.
+A functional is a mathematical device that converts a function (in this case ~ ~ " ( x ) )
into a single number. In this case a new number is generated for each new value of 6
and those are the (real) values of the function e(i6).
$A brief outline of the formalism of vector calculus with definitions of the operators
V. and V2 is given in Appendix A3.
MODERN DISPERSION FORCE THEORY 1559
v .(P(4H) =0 (1 1.6.10)
+ E0
V2E w 2 p ( w ) ~ ( w )= (1 1.6.1 1)
c = (Eopo)-i (1 1.6.12)
a result we will use later.
Thus modern dispersion force theory starts with the concept of a randomly
fluctuating electromagnetic field pervading the material system. This field is driven by
the motions of the individual molecular charge distributions, propagates with
frequencies characteristic of these motions and is constrained to obey the macroscopic
Maxwell’s equations given above.
In the presence of dielectric boundaries (interfaces), where ~ ( win) particular suffers
a sudden change in value over a length scale (the width of the interface) small
compared to the wavelength of the e.m. radiation, the usual e.m. boundary conditions
must be satisfied (compare Section 11.6.1). That is, we impose continuity of (a) the
normal components of the dielectric displacement D(= EE)and of p H , and (b) the
parallel components of E and H across the interface. These boundary conditions serve
as constraints on the allowed frequencies of propagation. (A communications engineer
would regard the system as a dielectric waveguide and, using conventional waveguide
theory, would subsume the constraints imposed by the dielectric boundaries of the
system into a single dispersion relation.) That is, for a system of given geometry and
dielectric properties we can find a function D(w) such that the allowed frequencies of
e.m. propagation in the system are given by
D(w) = 0 (1 1.6.13)
just as we did for the simple problem in Appendix A l . In that case (Exercise A1.2) the
magnetic field vector was ignored and only the continuity of potential and dielectric
displacement were required. Clearly the roots of eqn (11.6.13), wj (j= 1, 2,. . .), are
functions of the geometry and dielectric properties of the system.
T o make the connection with quantum mechanics, the fluctuating e.m. field is
regarded as a collection of photons. If o,is an allowed frequency (D(wj)= 0) then the
560 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
+
energy associated with this mode is simply (n 1/2)h o j , where n is the number of
photons in the mode. As photons are emitted and absorbed by the molecules of the
system, the number of photons in any given mode fluctuates along with the energy
contained in that mode. The important quantity for present purposes is the Helmholtz
free energy of the mode, 4 - the energy available to do work on the system’s
environment. From statistical thermodynamics (Landau and Lifshitz 1969) we have
(11.6.14)
where Zj is the partition function for the modej (Exercise 11.6.3) and the summation is
over all of the permitted quantum states of the modej. (We use k g here for the
Boltzmann constant to avoid any confusion with the magnitude of the wave vector.)
The total free energy of the e.m. field is obtained by summing the free energies for
each mode
F = kg T cjln[2sinh @Wj/2kg T ) ] . (11.6.16)
However, to evaluate the wj values for eqn (1 1.6.16) explicitly is not feasible in general.
It is possible to circumvent this difficulty by using the same mathematical device that
was alluded to earlier (see Ninham et al. 1970). The free energy of the e.m. field can be
written as:
(11.6.17)
The summation is from n = 0 to 00 and the prime on the summation sign indicates that
the n = 0 term must be divided by two. Equation (11.6.17) is derived from (11.6.16)
using Cauchy’s theorem (Appendix A2) and replaces the difficulty of explicitly
obtaining the zeros of D(w) with the difficulty of evaluating the function ID(;() at an
infinite set of discrete values (Langbein 1974).
If the system comprises two dielectric bodies immersed in a third dielectric fluid
phase, the dispersion relation for the system will change as the separation of the bodies
is altered. The allowed modes of the system will change correspondingly, with a
consequent change in the e.m. free energy of the system. The van der Waals
interaction energy V(D)of the two bodies is just the work done to bring them from
infinite separation to the separation distance D. Assuming that the work done
manifests itself only in the change in e.m. free energy during the process, we can write:
where :
MODERN DISPERSION FORCE THEORY I561
00
(1 1.6.22)
All E and p functions are evaluated at o = iCn. Recall that for all except the
ferromagnetic materials we may make the replacement pj = PO with negligible error
and then DM = 1.
Let us examine the distance dependence of E132(L). Changing to variable x (where
+
x2 = 4k2L2 x;), eqn (1 1.6.20) becomes:
xln{DEDM}dx (1 1.6.25)
The function [&PI; yppearing in eqn (11.6.24) for xo is of order l / c (see equation
11.6.12).Thus L[EP]? is a measure of the time of propagation of an e.m. field over the
distance L. When this time is small compared to the period of the radiation ((J'
562 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
there will be negligible phase lag. In this limit (often called the c+ 00 limit) both L and
xo tend to zero and x1 = x2 = x. The non-retarded interaction energy (c = 00) is
(1 1.6.26)
(The transformation from the integral to the summation in eqn (1 1.6.27) is considered
in Exercise 11.6.4. The ~ivalues are again evaluated at zt,. In practice one usually
needs only a few terms in the series for s to obtain a very accurate assessment of the
integral.)
Thus the non-retarded dispersion energy has exactly the same form as obtained by
the Hamaker summation (eqn (1 1.3.11)). Lifshitz theory in the non-retarded limit still
produces a Hamaker constant, but one that has a more complicated density
dependence (via the €(it)) than that obtained in the simpler Hamaker theory. Note
also that, in modern theory, the bathing medium is automatically included.
Lifshitz theory and Hamaker theory agree to leading order in the density and will
give comparable results when the €(it)values for the material of the system are close to
unity. It is precisely in this dilute regime that pairwise summation would be expected to
be a valid approximation. Indeed, the importance of many-body effects can be gauged
directly by noting the significant deviations of c(ic) from €0 that occur in liquid and
solid dielectrics. In many systems of colloidal interest, Lifshitz theory and Hamaker
theory can predict Hamaker constants differing by an order of magnitude or more. It is
this quantitative discrepancy that has reduced Hamaker theory to a minor role in
modern colloid science.
We note that the n = 0 term in the expression (11.6.25) for El32 is always non-
retarded, since xo is zero for n = 0 regardless of the value of c. It follows that even for
finite c we can separate out an unretarded term like eqn (11.6.26) where the van der
Waals constant is the zero frequency contribution:
Various experiments have been performed to test the validity of modern dispersion
force theory (Sabisky and Anderson 1974; Israelachvili and Tabor 1972) with
considerable success. On the theoretical side, other geometries, e.g. spheres or
cylinders, have been examined (see Mahanty and Ninham 1976 for details). Generally
speaking, the explicit expressions so obtained for V*(H)are cumbersome and often
difficult to compute. The planar half-space problem dealt with here is, however, a
NUM E RICAL COMPUTATION OF INTERACTION E N E R G Y I 563
relatively straightforward calculation that all modern colloid scientists should be able to
appreciate (Section 11.7.2). Useful approximate results in other geometries at small
separation distances can be obtained by applying the Deryaguin approximation (eqn
11.5.3) to the planar result. A quasi-empirical procedure that yields a strikingly
accurate approximation to the interaction energy of two bodies, VA(H), over the entire
separation distance range, is the replacement of the Hamaker constant in the Hamaker
summation expression (eqn (1 1.3.21)) by the quantity (12nH2 E132(H)) calculated
from eqn ( 1 1.6.20) (Pailthorpe and Russel 1982).
An excellent compilation of Hamaker constants for a variety of materials and
combinations, especially those of interest in the application of atomic force
microscopy, has been given by Bergstrom (1997).
Exercises.
11.6.1 Verify that eqn (11.6.6) is a solution of eqn (11.6.4) if ~ ( wis) given by eqn
(1 1.6.5).
1 1.6.2 Consider the dielectric response function:
+
and convert it into E,(o) = E:(w) i~r(o)+ for real o.(E’ and E” must both be
real functions.)
11.6.3 Establish eqn (11.6.15) from eqn (11.6.14).
00
1 1.6.4 Use the series expansion for ln(1 - p) = c ( - p ’ / s ) for 0 < p < 1
1
to show that
7
0
xln[l - aexp(-x)]dx = - c00
s= 1
as/s3
Furthermore, we note from eqn(ll.6.8) that, since e:(w) is everywhere positive, e(i<)is
a monotonic decreasing function of .$.Thus e(i<) decreases steadily from the static
<
) = 0 to €0 at $ = 00 (see Figs 11.6.2 and 11.6.3). It is this
dielectric constant ~ ( 0 at
well-behaved nature of €(it) that enables its construction from the minimum of
experimental information as we shall show below.
It should be pointed out that the behaviour of e(z<)in the ultraviolet frequency range
(< > 10l6 rad s-') is of paramount importance in the calculation 0fA132 or E132(L). The
reason for this is that the frequencies tnoccur at equally spaced intervals of 2 n k T/h ~
-
(-3x 1014 rad s-' at T - 300 K). In the microwave (6 10" rad s-l) and infra-red (6 -
1014 rad s-l) regions, there are very few sampling points. For example, there are
approximately 30 terms in the frequency sum (eqn (1 1.6.20)) in the region < 10l6 rad<
s-l compared with 300 terms in the region 10l6 <<< 10'' rad s-l, where e(i.$)is still
reasonably large (Fig. 11.6.3). This picture is somewhat modified in cases when the
intervening medium 3 has dielectric properties and relaxation frequencies comparable
to one or both of the half spaces 1 and 2. Then the ultraviolet terms in the frequency
sum do not contribute as much to the sum because either or both of A13 and A23 are
small. In such cases, the ultraviolet representation is not quite as critical as it would be
if medium 3 were a vacuum. In all cases, however, the ultraviolet region makes a major
contribution to the total frequency sum.
The relative unimportance of the infra-red contribution to €(it)allows one to ignore,
in all but an average sort of way, the complicated fine structure of relaxations in this
region. Moreover, the UV absorption spectra of materials are usually simple if
somewhat broad. If these UV and IR spectra are known, then one can easily extract
from the data the relaxation frequencies Wk corresponding to the important absorption
peaks in the spectra. The function <(w) can then be represented in a simple way by a
discrete set of peaked functions of various heights (and widths) centred on each of the
frequencies
N
(11.7.2)
where Fk(w - wk) is a function peaked about wk that tends to zero for I w - wk I >> 0.
The area under each of these peaks is a measure of the strength of the absorption. We
NUM E RICAL COMPUTATION OF INTERACTION E N E R G Y I 565
would need to specify the functional form of Fk(W - W k ) in order to use the
fundamental equation (11.63) to construct ~(26).Since the function e(z6) is a simple
function, it is possible to ignore the detailed shape of the peak and to replace the
function Fk by the zero width, infinite height delta function without introducing too
much error. Thus, we make the approximation, that
/
00
Substituting eqn (1 1.7.5) into the fundamental relation (11.63) and making use of the
delta function property:
7
<wL!
6(W - W k ) G ( W ) d W = G(Wk) (11.7.6)
(11.7.7)
k=l
We note here the close relation between E ( z 6 ) and E:(w)as evidenced by eqns (11.6.7
and 11.6.8) connecting these functions to E:(w).If the representation (11.7.5) for E:(w)
is used in eqn (11.6.7) we obtain (Exercise (11.7.1):
(11.7.9)
566 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
These two functions E(i$) and ~ ' ( ware) plotted schematically in Fig 11.6.3(b). We see
that when the relaxation frequencies are widely spaced, both ~ ' ( w and ) ~(ie)are
characterized by large regions where those functions are essentially constant and
coincident. This would occur in cases where there was only one significant UV
relaxation and where IR relaxations were either unimportant (and so could all be
lumped together as a single relaxation at some average frequency containing all the
oscillator strength) or dominated by a single IR relaxation. Between the significant
relaxations E;(O) is essentially zero. There E:(w)is simply the square of the refractive
index n(w) (eqn (3.2.36)), which is usually tabulated in data handbooks together with
the dielectric constant.
If we denote by €,(before w ) and €,(after w ) the values of the relative dielectric
response (n2)in the flat regions before and after the relaxation w, then obviously:
E, (after WN) = r(oo)/€o= 1 ( 11.7.10)
(since OIN is defined as the largest relaxation frequency) and
If we evaluate eqn (11.7.9) in the flat region before the last frequency W N but after the
frequency W N - ~ ,we obtain,
>>
( w / w ~ ) ~1 (k = 1,2,3, ....., N - 1.) and (w/wN)~ << 1.
Therefore
CN = E , (before LON) - E, (after W N ) (1 1.7.13)
Proceeding in this manner it is a simple matter to show (for widely spaced relaxation
frequencies) that
Ck E, (before wk) - E , (after wk) (1 1.7.15)
where nl is the refractive index in the visible region. An inspection of the IR spectrum
suffices to estimate the relative strengths of the different bands among which this is to
be split. (See Hough and White 1980 for details.)
Unfortunately, experimental data in the literature for the UV region are more
scanty. Where evidence exists as to a complicated UV absorption spectrum the
technique of using more than one frequency with appropriate sharing of the available
oscillator strength (a2 - l), as outlined above, is clearly applicable in the absence of
direct knowledge of E ; ( W ) .
At this juncture it is worth making a few remarks about the microwave contribution
to €(it).In some polar substances there is a significant permanent dipole relaxation at
microwave frequencies (-10'' rad s-'). For example, in water the dielectric response
drops from about 80 at zero frequency to about 4 at IR frequencies rad s-').
This relaxation has sometimes been included in the ~ ( i $construction
) (eqn (1 1.7.7)) by
a Debye term (Ninham and Parsegian 1970) (compare eqn (3.2.38)):
Such a term is numerically dangerous, since it becomes the dominant term in E,(i() in
the far UV, where it has no right to exist. The microwave relaxation would typically
occur at 5x10'' rad s-' and would certainly be complete by 5 ~ 1 0 rad ' ~ s-'. As the
function E ( i ( ) is sampled first at $ = 0 rad s-' and next at $ = 3x 1014 rad s-', we need
only a construction that gives ~(20)and ~ ( i correctly (See Fig. 11.7.1). That ~ ( i $ )
is not equal to E ( i is immaterial provided that these values are correct. One would
-
need to be interested in a dipolar substance at T 2 K before the behaviour in the
microwave region was even faintly important. (Only then would the n = 1 sample point
fall in the microwave region.) The reader is therefore advised to omit such a term from
the E ( Z ( ) representation.
The value of the UV relaxation frequency is a critical parameter in the calculation of
A132 and E132(L). In the absence of UV spectral data, a frequency w = 11/h
corresponding to the first ionization potential 11 of the material is often taken. This can
be a serious error and should be avoided. In the absence of knowledge of <(w) over the
frequency regime w > 1015 rad s-l, it is still possible to obtain a reasonably accurate
construction of €(it)in the UV region provided the absorption is simple; that is, only
one relaxation is important.
568 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
c(rad sC1)
Fig. 1 1.7.1 A schematic representation of the function ~(26)for a substance with a single relaxation
in the microwave region (q,,ia0), a single IR absorption peak and a broad absorption in the UV. Note
the positions of the sampling points t,, for T = 300 K.
For ( in the visible region (provided ( >> w ~we) can write (from eqn (11.7.7)):
( 1 1.7.18)
Similarly, for ~ ' ( win) the visible region we can write, from eqn(11.7.9):
( 1 1.7.19)
Therefore a plot of [n2(w)- 11 against [n2(w)- 1]w2should yield a straight line of slope
(1 /-)2and intercept CUV.Experimentally it is relatively straightforward to measure
the refractive index n as a function of wavelength h (= 2nc/w) in the visible region and
such information is tabulated in the literature for most common substances. The plot
indicated in eqn (1 1.7.20)(called a Cauchy plot) is shown for some common materials
in Fig. 11.7.2).Table 11.1 also lists the wv, CUV,data so obtained along with IR and
zero frequency data. Table 1 1.1 enables an accurate construction of the e(i() function
in most cases, though for water the more elaborate procedure of Gingell and Parsegian
(1972)is preferable because of the significant IR contributions.
In the literature, use has sometimes been made of an ultraviolet interpolation to
connect the near UV representation of € ( i t ) (eqn 11.7.7) to the far UV (plasma)
representation. For a full discussion of this procedure see Hough and White (1980).
NUMERICAL COMPUTATION OF INTERACTION ENERGY 1 569
0 100
d(nZ-l) x - 2 0
Fig. 11.7.2 Cauchy plots (from Hough and White 1980),with permission. (a) Sapphire; (b) Calcite
(0-ordinary ray; e- extraordinary ray); (c) Crystalline quartz; (d) Fused quartz and fused silica;
(e) Calcium fluoride; (0 Water.
We advise strongly against the use of such procedures, which are unnecessary and
often misleading.
where pel is the number density of conduction electrons and me is the electron mass.
Typical wp values are rad s-l, so that €(it)for metals is extremely large until the
ultraviolet region is reached. Consequently, interaction energies are much larger than
for dielectric systems. The screening effect of an intervening dielectric is not as
important for the interaction of metal half-spaces as for dielectric half-spaces.
570 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
Table 11.1
Dielectric data f o r c(i6) construction f o r a variety of common substances.
Alkanes
n=5 1.844 1.819 0.025 5.540 0.819 1.877
6 1.890 1.864 0.026 5.540 0.864 1.873
7 1.899 0.025t 5.540 0.898 1.870
8 1.948 1.925 0.023 5.540 0.925 1.863
9 1.972 1.947 0.025 5.540 0.947 1.864
10 1.991 1.965 0.026 5.540 0.965 1.873
11 2.005 1.979 0.026 5.540 0.979 1.853
12 2.014 1.991 0.023 5.540 0.991 1.877
13 2.002 0.025t 5.540 1.002 1.852
14 2.011 0.025t 5.540 1.011 1.846
15 2.019 0.025t 5.540 1.019 1.845
16 2.026 0.025t 5.540 1.026 1.848
Water 80.10 1.755 Gingell and Parsegian (1972) 0.755 1.899
Xtalline quartz
ord. ray 4.27 2.350 1.92 2.093 1.350 2.040
Xord. ray 4.34 2.377 1.96 2.093 1.377 2.024
average 4.29 2.359 1.93 2.093 1.359 2.032
Fused quartz 3.80 2.098 1.70 1.880 1.098 2.024
Fused silica 3.81 2.098 1.71 1.880 1.098 2.033
Calcite
ord. ray 8.0 2.683 5.3 1.683 1.660
Xord. ray 8.5 2.182 6.3 1.182 2.134
average 8.2 2.516 5.7 2.691 1.516 1.897
caF2 7.36 2.036 5.32 0.6279 1.036 2.368
Sapphire 11.6 3.071 8.5 1.880 2.071 2.017
PMMA 3.4 2.189 1.2 5.540 1.189 1.915
PVC 3.2 2.333 0.9 5.540 1.333 1.815
PS 2.6 2.447 0.2 5.540 1.424 1.432
Poly-(isoprene) 2.41 2.255 0.16 5.540 1.255 1.565
PTFE 2.10 1.846 0.25 2.270 0.846 1.793
PMMA: Poly (methyl methacrylate); PVC Poly(viny1 chloride); PS: Poly(styrene); PTFE: Poly
(tetrafluorethylene).
+Assumedvalue.
I N F L U E N C EOF ELECTROLYTE CONCENTRATION I 571
11.7.3 Numerical evaluation of E132(L)
0.023 0.991
+
El/EO =1
+ [
5.54 tnx 1014I 2 + l + [ 1.877 x 10l6 l2
where t,, = [4n2 k~T/h]n= 2.4513 x lOI4 n rad/s.
Step 1. Set up the first column with n = 0 to 2000. Set up column 2 with values of
= 2.4513 n.
+ +
Step 2. Set up column 3 with E I / E O = 1 0.023/[1+(~012/5.54)~] 0.991/[1+(~012/
187.7)2].
Step 3. Set up column 4 as f = [(co13-l)/(c013+1)]~.
+ +
Step 4. Set up column 5 as f f 2/8+ f 3/27 f 4/64 (more than enough terms).
Step 5. Add all these terms in column 5, except the first one, then add it in after
multiplying by 0.5. This is the final frequency sum. T o get A121 you now only have to
multiply by 1.5 kBT and you should get 5.036 x J. Table 11.2 gives
5.04 x lop2' J.
(Exercises
11.7.1 Establish eqns (11.7.7), (11.7.8), and (11.7.9).
II
Table 11.2
Hamaker constants (x l@20)/Jforsome common materials.
Material (M) M I air IM M I water I M M I water I air M I air I water water I M I air
Alkanes
n =5 3.75 0.336 0.153 3.63 0.108
6 4.07 0.360 -0.368 x lo-’ 3.78 0.285
7 4.32 0.386 -0.118 3.89 0.423
8 4.50 0.410 -0.200 3.97 0.527
9 4.66 0.435 -0.275 4.05 0.624
10 4.82 0.462 -0.344 4.11 0.719
11 4.88 0.471 -0.368 4.14 0.751
12 5.04 0.502 -0.436 4.20 0.848
13 5.05 0.504 -0.442 4.21 0.855
14 5.10 0.514 -0.464 4.23 0.886
15 5.16 0.526 -0.490 4.25 0.923
16 5.23 0.540 -0.518 4.28 0.964
Fused quartz 6.50 0.833 -1.01 4.81
Xstal. quartz 8.83 1.70 -1.83 5.59
Water 3.70 0 0 3.70
Fused silica 6.55 0.849 -1.03 4.83
Calcite 10.1 2.23 -2.26 6.00
Calcium fluoride 7.20 1.04 -1.23 5.06
Sapphire 15.6 5.32 -3.78 7.40
PMMA 7.11 1.05 -1.25 5.03
PVC 7.78 1.30 -1.50 5.25
PS 6.58 0.950 -1.06 4.81
Poly (isoprene) 5.99 0.743 -0.836 4.59
PTFE 3.80 0.333 0.128 3.67
Mica (brown) - 1.98 - -
this question cannot be answered completely we can make a few observations; a more
complete description of the situation was given by Mahanty and Ninham (1976)
(Chapter 7) who made it clear that there are profound problems involved in the general
treatment of two charged interfaces separated by an electrolyte solution.
I N F L U E N C EOF ELECTROLYTE CONCENTRATION I 573
=-
4
7
0
dx x In (1 - 813823 exp(-s))
(1 1.8.1)
cj(0)x - ~(0)s
where s2 = x2 + ~ ( K L )and
~ 6. -
j3 - €j(o)X + €3(O)S
Once ~ K becomes
L greater than unity, the n = 0 term will be screened and will
approach zero. Thus the presence of the electrolyte ensures that the n = 0 term is no
longer the dominant term in the estimation of E132(L) for large L. Comparison of eqn
(11.8.1) with the first term of eqn (11.6.27) shows that the effect is to make the
argument of the logarithm function approach unity more quickly so that Ao is reduced
in magnitude. An effect of order 50% is quite possible.
The more realistic problem of two approaching charged surfaces has been solved
only under conditions that Mahanty and Ninham (1976) regard as too restrictive to
furnish unequivocal results. The theoretical point at issue is the question of whether it
is permissible to treat the attraction and repulsion potential energies as separable. The
success of the DLVO theory (which assumes such a separation is possible), and the
lack of any viable alternative, gives us no choice. It is not the first time (nor will it be
the last) where the limited insights of an approximate theory are used until a more
rigorous development becomes available.
TMahanty and Ninham point out that this statement may not hold for the proton in
water.
574 I 1 1 : THE THEORY OF V A N DER WAALS FORCES
(1 1.9.1)
Fowkes used the rather crude Hamaker theory to estimate surface free energies but
with the analysis given above we should be able to improve on that. In order to
calculate y& from the dispersion equations we have only to imagine the reverse
process to that shown in Fig. 2.10.l(a). That is, we bring together two semi-infinite
blocks of hydrocarbon (of unit area) from infinity until they are (almost) touching
(separation L,) and calculate the interaction energy per unit area, E ~ where
H H stands
for hydrocarbon and V for vacuum (or vapour), from eqn (11.6.26). Then
(11.9.2)
Unfortunately, to get a finite value out we must postulate a minimum (non-zero) value
for L, corresponding to the ‘distance’ between the bodies when they are in contact. We
noted in Section (11.6.4) that the divergent behaviour as L approaches zero is a
consequence of the failure of the heuristic method of van Kampen to take account of
the ‘graininess’ of matter and that it is absent from the exact Lifshitz theory.
Intuitively, we expect L, in the more simplified theory to be of the order of a molecular
radius. The values of A- calculated from spectral data for various hydrocarbons are
given in Table 11.2 (in the column headed M I air I M) and using those values in eqn
(1 1.9.2) with the experimental ~ / H Cvalues we find the values of L, listed in Table 11.3.
They are close to being constant and are certainly of the correct order of magnitude.
The slight decrease with increasing chain length reflects the increase in density of
hydrocarbons with molar mass.
Hough and White (1980) show that the critical length, L,, calculated in this way is
inversely proportional to the square root of the liquid density. This relationship can be
exploited to obtain the values calculated in the last column of the table with only one
adjustable parameter, the value of L, for some reference hydrocarbon, calculated from
the measured surface tension.
In Table 11.3 we have used the value for decane and the calculated values are then
given by:
for A in joules and p in kg m-3. Hough and White (1980) apply a less objectionable
procedure for avoiding the L + 0 problem but it is a good deal more difficult to apply
than eqn (1 1.9.3).
In order to estimate the dispersion contribution to y at other interfaces, Fowkes
argues as follows. When two immiscible liquids (say mercury and a hydrocarbon) are in
contact, the dominant interaction between them is the dispersion (van der Waals)
interaction and in the case of a hydrocarbon-hydrocarbon interface it may be regarded
as the only important interaction. The surface tension of the pure hydrocarbon is
reduced by the interaction that the hydrocarbon molecules have with the adjacent
mercury atoms. Likewise, the surface tension of the mercury is reduced from its value
in air ( y ~ to) some lower value due to interactions between the mercury atoms and the
hydrocarbon molecules. Fowkes estimates the reduction in each case from the
geometric mean of the dispersion components (compare eqn (1 1.3.23)):
0 d
For the hydrocarbon ~ H V M= yHv - (yHv y&)i
(11.9.4)
0 d
and for mercury = yM - (yHvy&)i.
~MVH
0 0
m c = YM + YHV d
- 2(yHVy&)'. (1 1.9.5)
Since y& = y h this equation contains only one unknown y$ and by comparing the
interfacial tensions of a series of hydrocarbons against mercury it is possible to estimate
the dispersion component of the surface tension of mercury. Fowkes finds a value of
y$ = 200 f 7 mJ mP2. The remainder of y~ (284 mJ mP2) would be attributed to
metal bonding in this case.
576 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
The same argument can be applied to the water-hydrocarbon surface and a value for
y i z 0 of 21.8 f 0.7 J m-’ is obtained (Exercise 11.9.2). The remarkable thing about
these results is that if we now apply them to the mercury-water interface, assuming
again that the only important interaction is the dispersion force, we find:
0 d d f
YMW = yM + YW0 - 2(YMYW)
= (484 + 72.8 - 2(200 x 21.8)i = 424.8 mJ m-’
which is very close to the experimental value of 426-427 mJ m-’. There is, therefore,
surprisingly little contribution from the permanent dipole moment of the water
molecule.
This empirical finding suggests that reliable estimates of interfacial tension can be
obtained directly from the dispersion energy, even for water films where hydrogen
bonding might be expected to play a role. Unfortunately, a more detailed
examination (Hough and White 1980) shows that one can be very easily led into
false conclusions regarding the behaviour of water films if only the dispersion
contribution is examined.
Normally the presence of the n, term makes theoretical estimation of the contact angle
from eqn (1 1.9.6) quite impossible. In the case of very-low-energy-solids like
polyethylene, however, it is found that ne = 0. That is to say, the vapour of high
surface energy liquids will not adsorb onto a low-energy solid since this will not lower
the surface energy. Neither will the vapour of a hydrocarbon liquid adsorb to any
appreciable extent. Low-energy solids are substances for which AG,d, is small for any
adsorbate. They are usually substances with little or no dipole moment and very low
polarizability so that they can interact with adsorbates only by van der Waals forces,
and then only slightly. The classic example is the non-stick frying pan coated with
poly-( tetrafluorethylene).
If we again assume that the interaction between the liquid and the low-energy solid
is predominantly due to dispersion forces, we can then substitute for y s ~from the
analogue of eqn ( 1 1.9.5) and obtain:
(11.9.7)
Fowkes (1965) shows that a plot of the cosine of the measured contact angle against
( y ; ) f / y for
~ a number of liquids on low-energy surfaces (like polyethylene and paraffin
wax): Fig. 11.9.1 does obey eqn (11.9.7), from the slope of which an estimate of the
dispersion contribution to the surface energy of the solid (y,”) can be obtained. Note
that again water appears to behave like other liquids in this plot.
THEORETICAL ESTIMATION OF SURFACE PROPERTIES I 577
YP
40 30 20 10
+1
20
30
40
50
60
70
80 0
d
m 2
$ 0
u 90 2
5
100 g
u
110
120
130
140
150
-1 180
0.1 0.2 0.3
(YLd)flYL
Fig. 11.9.1 Contact angle of a number of liquids on low energy surfaces. V: polyethylene;
0: paraffin wax; 0: c36H74; . fluorodecanoic
: acid monolayer on platinum. All contact angles
below the arrow are with water. (Copyright American Chemical Society. Reproduced with
permission from Fowkes 1965.)
-
The value of y,d for the four solids in Fig. 11.9.1 should be calculable from the van
der Waals constants using (compare eqn (1 1.9.2)):
(11.9.8)
50
-
rn
-
v
G)
M
*
:
cd
0
a4
0
5 10 15
No. of carbon atoms
Fig. 11.9.2 A comparison of the theoretical (a) and experimental (b) curve for contact angles of
alkanes on PTFE. Wetting would correspond to 6' = 0 and is expected to occur for n 5 5. (From
Hough and White (1980) with permission.)
where EsvH(O)is the van der Waals energy of interaction per unit area of the solid with
the liquid hydrocarbon across a gap of essentially zero width. This term is assumed to
represent the work of adhesion between the solid and the liquid. Again, we have to
make a cut-off approximation and set
(1 1.9.10)
where, for simplicity, we assume that the critical separation L, is the same as for the
liquid (eqn (1 1.9.2)).
Using eqns (11.9.2), (11.9.6), and (11.9.10) we have:
Hough and White (1980) recalculated cos 8 using their more accurate assessments of
A ~ V and
H A ~ forHa number of liquid hydrocarbons on PTFE and obtained the
results shown in Fig. 11.9.2.
A cautionary note is in order at this point. The use of L, as an adjustable parameter
is not always successful. Ninham (1980) points out that it fails for liquid argon. The
fact that it works reasonably well for liquid hydrocarbons must therefore be to some
extent fortuitous. Using the estimates of van der Waals constants given by Bergstrom
(1997) should allow techniques like those of Fowkes, even with their limitations, to
provide a more systematic approach to the problem of wetting and contact angle.
Progress involves the combined efforts of experimentalists and theoreticians but the
general procedures are now clear. There remain enough observed anomalies to keep
this area an active and fruitful one and to ensure the value of further research on both
experimental and theoretical fronts.
REFERENCES 1579
Exercises
1 1.9.1 Use the data in Table 11.2 to estimate theoretical values of cos 6 for
hydrocarbons with 6, 8, 10, 12, 14, and 16 carbons and compare your results
with those shown in Fig. 11.9.2.
1 1.9.2 Given the following data for the interfacial tension ( n z ) between water and
various hydrocarbons, estimate y:.
n= 6 7 8 10 14
References
Atkins, P.W. (1978). Physical chemistry (1st edn) (6th edn 1998). Oxford University
Press, Oxford.
Bergstrom, L. (1997). Adv. Colloid Interface Sci. 70, 12549.
Bhattacharjee, S. and Elimelech, M. (1997).J. Colloid Interface Scz. 193,273-85.
Bradley, R.S. (1932). Phil. Mag. 13,853.
Carrier, G.F., Crook, M., and Pearson, C.E. (1966). Functions of a complex
variable. McGraw-Hill, New York.
Casimir, H.B.G. and Polder, D. (1948). Phys. Rev. 73, 360.
Chan, D.C. and Richmond, P. (1977). Proc. Royal. Soc. Lond. A353, 163.
Christenson, H.K. (1983). Ph.D. thesis, Australian National University.
Clayfield, E.J., Lumb, E.C, and Mackey, P.H. (1971).J Colloid Interface Sci. 37,
382.
Dalgarno, A. and Lynn, N. (1957). Proc. Roy. Soc.Lond. A70,802.
Debye, P. (1920). Phys. 2. 21, 178.
Deryaguin,B.V. (1934). Kolloid-2. 69, 155.
Dzyaloshinski, I.E.. Lifshitz, E.M., and Pitaevski, L.P. (1961). AdnPhys. 10,
165.
Fowkes, F.M. (1965). Attractive forces at interfaces. In Chemistry and physics of
interfaces (ed. D.E. Gushee), pp. 1-12. American Chemical Society, Washington.
Gingell, D. and Pargsegian, V.A. (1972).J Theor. Biol. 36,41.
Girifalco, L.A. and Good, R.J. (1957). J Phys. Chem. 61, 904.
Gregory, J. (1969). Adv. Colloid Interface Sci., 2, 3 9 H 1 7 .
Gregory, J. (1981). 3’.Colloid Interface Sci. 83, 138.
Hamaker, H.C. (1937). Physics, 4, 1058.
Hough, D.B. and White, L.R. (1980). Adv. Colloid Interface Sci. 14, 3-41.
Hunter, R.J. (1963). Austral. 3’.Chem. 16, 774.
Hunter, R.J. (1975). Electrochemical aspects of colloid chemistry. In Modern aspects
of electrochemistry (ed. B.E. Conway and J.O’M Bockris) Vol. 11, Chapter 2,
pp. 33-84. Plenum Press, New York.
Israelachvili, J.N. (1973~).Q Rev. Biophys. 6, 341.
Israelachvili, J.N. (19736).J. Chem. Soc. Faraday Trans. 2 69, 1729
Israelachvili, J.N. (1974). Contemp. Phys. 15, 159.
580 I I I : T H E THEORY OF VAN D E R WAALS F O R C E S
Israelachvili, J.N. and Tabor, D. (1972). Proc. Roy. Soc. Lond. A331, 19.
Jackson, J.D. (1975). Classical electrodynamics (2nd edn). Wiley, New York.
Kallmann, H. and Willstatter, M. (1932). Naturwissenschaften 20, 952.
Keesom, W.H. (1921). Phys. 2. 22, 129 and 643.
Landau, L.D. and Lifshitz, E.M. (1960). Electrodynamics of continuous media.
Pergamon, London.
Landau, L.D. and Lifshitz, E.M. (1969). Statistical physics. Pergamon, Oxford.
Langbein, D. (1974). Theory of van der Waals attraction. Tracts in modern Physics.
Springer, Berlin.
Lifshitz, E.M. (1956). Sov. Phys.JETP., 2, 73.
London, F.(1930). 2. Phys. 63,245.
Mahanty, J. and Ninham, B.W. (1976). Dispersion forces. Academic Press,
London.
McClellan, A.L. (1963). Tables of experimental dipole moments. Freeman,
San Francisco.
Ninham, B.W. (1980). J. Phys. Chem. 84, 1423.
Ninham, B.W. and Parsegian, V.A. (1970). Biophys. J. 10,646.
Ninham, B.W., Parsegian, V.A., and Weiss, G. (1970).J. Statist. Phys. 2, 323.
Overbeek, J. Th. G. (1952). In Colloid Science (ed. H.R. Kruyt) Vol.1, p. 265.
Elsevier, Amsterdam.
Pailthorpe, B.A. and Russel, W. (1982). J. Colloid Interface Sci. 89, 563-6.
Parsegian, V.A. (1975). Long range van der Waals forces. In Physical chemistry:
enriching topicsfrom colloid and surface science (ed. H. van Olphen and K.J. Mysels)
Chapter 4. IUPAC Commission 16. Theorex, La Jolla, California.
Pitzer, K.S. (1959). Adv. Chem. Phys. 2, 59.
Reinganum, M. (1912). Ann. Phys. 38,649.
Richmond, P. (1975). In Colloid science (ed. D.H. Everett) Vol. 2, Chapter 4.
Chemical Society, London.
Sabisky E.S. and Anderson, C.H. (1974). Low temperature physics - LT13
Proceedings of the 13th International Conference on Low Temperature Physics
(1972), Vol. 1, pp. 206-10.
Sparnaay, M.J. (1983). J. Colloid Interface Sci., 91, 307.
Thomson. J.J. (1914). Phil. Mag. 27, 757.
van der Waals, J.H. (1873). Ph.D. Thesis, University of Leiden.
van der Waals, J.D. Jnr. (1909). Amsterdam Acad. Proc. pp. 132, 315.
van Kampen N.G., Nijboer, B.R.A., and Schram, K. (1968). Phys. Lett. 26A,
307.
Vincent, B. (1973). J. Colloid Interface Sci. 42, 270.
Visser, J. (1972). Adv. Colloid Interface Sci. 3, 33143.
White, L.R. (1983). J. Colloid Interface Sci. 95, 286-8.
Wilson, J.N. (1965). J. chem. Phys. 43,2564.
Double Layer Interaction and Particle
Coagulation
12.1 Surface conditions during interaction
12.2 Free energy of formation of a double layer
12.3 Overlap of two flat double layers
12.3.1 Approximate relations for overlap o f flat double layers
12.4 Interaction between dissimilar flat plates
12.4.1 Approximation formulae for the interaction between dissimilar
plates
12.5 Interaction between two spherical particles
12.5.1 For large values of Ka
12.5.2 For small values of Ka
12.6 Total potential energy of interaction
12.6.1 The Schultz-Hardy rule
12.7 Experimental studies of the equilibrium interaction between diffuse
double layers
12.7.1 Adsorbed liquid films
12.7.2 Soap films
12.7.3 Swelling of clays
12.7.4 Direct force measurements between mica surfaces
12.8 Kinetics of coagulation
12.8.1 Rate o f rapid coagulation
12.8.2 Aggregate formation
12.8.3 Coagulation under shear
12.8.4 Rate o f slow coagulation
12.8.5 The fractal character of aggregates
12.9 Effect of polymers on colloid stability
12.9.1 Steric stabilization
12.9.2 Polymer bridging
12.9.3 Charge effects
12.9.4 Depletion interactions
581
582 I 1 2 : D O U B L E L A Y E R I N T E R A C T I O NAND PARTICLE COAGULATION
In this chapter we consider the various forces that come into play when two colloidal
particles or two double-layer systems approach one another. The behaviour is
determined to some extent by how rapidly the approach occurs, since the double layers
may take a significant time to adjust to the new situation. Equilibrium behaviour may
then be somewhat different from transient behaviour.
The repulsive interaction that occurs between double layers of like sign can be
analysed by two alternative procedures by considering:
The case of large flat plates can be treated fairly rigorously but some useful
approximation formulae are also derived for the case of low potentials or for small
degrees of double layer overlap. We then look at situations where the potentials on the
two approaching surfaces are different. This is relevant to such processes as flotation,
mixed colloidal suspensions, and the drainage of wetting films on solids. The theory for
spherical particles is more complicated algebraically and will not be discussed in great
detail.
The experimental evidence in support of the DLVO Theory (Section 1.6) comes
from a variety of sources but the most definitive results have been obtained since about
1976, using atomically smooth sheets of mica, interacting in an aqueous electrolyte
solution. That work clearly establishes the general validity of the theory presented
below, at the same time suggesting that when particles approach one another very
closely there are quite dramatic new effects that come into play.
The chapter concludes with a brief description of the kinetics of coagulation, the
fractal dimensions of colloidal aggregates, and the effect of polymers on colloid
stability.
y/,c .........\..*
I"
Fig. 12.1.1 The overlap of two diffuse double layers. The potential distribution in the
neighbourhood of a single double layer is shown The full line is the anticipated potential
distributionfor the pair of particles. llr,is the potential at the minimum (which in this case is also in
the median plane). The high potentials between the plates (llr,> 0) lead to high counterion
concentrationsand this in turn leads to a large osmotic pressure, tending to push the particles apart.
the approach. On the other hand, if the particle charge is caused by built-in crystal
defects, as in some clay minerals, it might be more sensible to assume that the
surface charge is constant during approach. In the case of oxide surfaces, the
interaction may itself influence the degree of dissociation of surface groups
(Section 7.10) so that neither $0 nor 00 is constant. T h e condition known as charge
regulation (Ninham and Parsegian 1971) may then be more appropriate. There are
other possibilities if specific adsorption in a Stern layer is involved. In general,
however, these three possibilities can adequately cover most experimental
situations.
The constant potential case was extensively studied by the early workers in the
field and is the basis of the DLVO theory of colloid stability (Section 1.6) described
in detail in the monograph by Verwey and Overbeek (1948). T h e other possibilities
are best understood as fairly straightforward generalizations of the constant
potential case.
We will examine first the case where the approaching surfaces are identical and
subsequently deal with the case of unlike systems (heterocoagulation). The
repulsion can be calculated either from the osmotic pressure, as noted above, or
from the increase in Helmholtz free energy, AF, that occurs as two double layers
overlap. T he osmotic pressure method cannot be applied to the approach of
spherical particles because another factor (the Maxwell stress, Section 12.3) enters
in that case. T he free energy method is rather more general in its application. We
will begin, therefore, with an examination of the free energy of a single (i.e.
isolated) double layer.
584 I 1 2 : D O U B L E L A Y E R I N T E R A C T I O NAND PARTICLE COAGULATION
(This inherently assumes that A F is independent of the state of charge of the surface.
It is strictly true only at low charge density and corresponds to the condition
introduced in Section 7.7.1 to derive the Nernst equation. Chan and Mitchell (1983)
show how to remove this restriction to generate a more general expression for A F but
we will not need that in the present treatment.)
During the charging process, as the double layer builds up, the electrical work done
can be calculated by imagining the transfer of infinitesimal amounts of charge d o (per
unit area) through a potential vo
where Yo increases from zero to its final value $0
when the last piece of charge is installed. The electrical work done is then:
AFelec= 1
0
$;do. (1 2.2.2)
(1 2.2.3)
F R E E ENERGY OF FORMATION OF A DOUBLEL A Y E R I585
(Note that the chemical term is always larger in magnitude than the electrical term so
that AF is negative since 00 and $0 always have the same sign.)
Equation (12.2.3) may also be derived directly from the Lippmann equation
(7.2.14), expressed in the form (ay/avo), = do. Then:
/
11.0
AF=
AF = yo
y o-
- yy==
/ d/ yd =
y - j=
o h-d0I &ohdI&.
.
0
The free energy of a single (isolated) diffuse double layer is readily calculated from eqn
(12.2.3) because 00 can be expressed as a function of $0 using the appropriate form of
eqn (7.3.27). Then (Exercise 12.2.2):
(12.2.4)
DO = -E(d$/dX),,o. (12.2.5)
As the particles approach, the surface potential can only remain constant if the
potential determining ions are gradually driven off the surface until the double layer
finally disappears on contact. Since the double layer was initially formed
spontaneously, this forcible discharge of the plates results in an increase in free
energy. The repulsive potential energy VR,due to approach of the two plates is thus
given by:
where AF(o0) is the free energy (per unit area) associated with the isolated double
layer and AF(D) is the corresponding free energy when the plates are at a distance D
apart.
The corresponding expression for the repulsive potential energy, evaluated from
the osmotic pressure, 3, between the plates is:
D
VR= - S p d D (12.2.7)
586 I 1 2 : D O U B L E L A Y E R I N T E R A C T I O NAND PARTICLE COAGULATION
Fig. 12.2.1 As two plates approach at constant potential the slope d$/dx at x = 0 decreases in
absolute value, indicating a progressive discharge of the surface potential-determiningions.
where ji is the dtflerence in the osmotic pressure midway between the plates and in the
bulk solution.
Both the free energy and the osmotic pressure method can be used to calculate VR
for flat plates and each has advantages in different situations. Given VR,the total
potential energy of interaction VT can then be obtained by adding VA(Chapter 11) and
VT can then be used to describe the stability behaviour (Section 1.6).
Exercises
12.2.1 Show that the free energy change involved in arranging the diffuse counter
charge is zero if the ions obey the Boltzmann equation.
12.2.2 Establish eqn (12.2.4) using the appropriate form of eqn (7.3.27).
12.2.3 Show that at low potentials, where the Debye-Huckel approximation holds, the
free energy of a flat double layer is A F = - ~ c K & and of a sphere is -2ma
(1 + .a,&.
0 xe$ d2$
2n xesinh- - C- = 0. (12.3.1)
KT dx2
O V E R L A P OF TWO FLAT DOUBLE L A Y E R S I587
This equation can be integrated (for a fixed value of D) as before to obtain (compare
eqn 7.3.18):
=p , (a constant). (12.3.2)
The first term on the left is related to the osmotic pressure between the plates. That
pressure varies from point to point because of the variation in $ (and, hence, of the
local ion concentration). It can be most readily evaluated at the midplane between the
particles, because at that plane (d$/dx) = 0 and so the osmotic pressure is equal to 9,:
p , = 2nOkTcosh
kT
The higher osmotic pressure at other points between the plates is partly offset by the
other term on the left of eqn (12.3.2) (= $ IE I ’), which is called the Maxwell stress.
The actual force per unit area exerted on the plates is given by the difference in
osmotic pressure between the solution in the midplane between the plates and that
outside (in the reservoir of electrolyte):
10 f 268.3 228.2 192.6 160.0 127.1 75.4 44.1 25.4 14.1 7.36 3.42 1.26 0.26 0.06 0.015 0.0023
KD 0.0000 0.00868 0.01672 0.0268 0.0408 0.0874 0.1626 0.286 0.488 0.824 1.380 2.296 3.924 5.442 6.880 8.732
9 f 161.5 135.2 115.2 95.6 76.3 44.3 25.4 14.1 7.36 3.42 1.26 0.26 0.06 0.015 0.0023
KD 0.0000 0.0146 0.0276 0.0442 0.0674 0.1442 0.268 0.472 0.806 1.358 2.278 3.906 5.424 6.862 8.714
8 f 96.52 80.56 68.56 56.60 44. 8 25.4 14.1 7.36 3.42 1.26 0.26 0.06 0.015 0.0023
KD 0.0000 0.0242 0.0454 0.0728 0.111 0.238 0.442 0.776 1.330 2.248 3.878 5.396 6.834 8.686
7 f 57.13 47.46 40.18 32.89 25.8 14.17 7.36 3.42 1.26 0.26 0.06 0.015 0.0023
KD 0.0000 0.0398 0.075 0.120 0.183 0.392 0.728 1.282 2.202 3.830 5.348 6.786 8.636
6 f 33.27 27.47 23.04 18.66 14.38 7.39 3.42 1.26 0.26 0.06 0.015 0.0023
KD 0.0000 0.0654 0.1236 0.198 0.3018 0.646 1.202 2.122 3.752 5.270 6.708 8.560
5 f 18.83 15.32 12.69 10.07 7.52 3.43 1.26 0.26 0.06 0.015 0.0023
KD 0.0000 0.1082 0.2036 0.3264 0.4976 1.066 1.990 3.622 5.140 6.580 8.430
4 f 10.13 8.07 6.51 4.97 3.50 1.26 0.26 0.06 0.015 0.0023
KD 0.0000 0.1782 0.336 0.5384 0.821 1.768 3.404 4.924 6.362 8.214
3 f 4.962 3.793 2.913 2.061 1.291 0.26 0.06 0.015 0.0023
KD 0.0000 0.2942 0.5548 0.891 1.362 3.036 4.560 5.996 7.848
2 f 1.993 1.413 0.966 0.584 0.265 0.06 0.015 0.0023
KD 0.0000 0.487 0.9286 1.502 2.356 3.916 5.360 7.216
following method was described by Chan et al. (1980). Writing eqn (7.3.15) in terms of
the dimensionless variables y = x e @ / k T and X = KX we have (Exercise 12.3.2):
d2y/dX2 = sinhy. (12.3.5)
where the coordinate system is placed with the origin for X in the midplane. The
function sgn(y) = y / Iy I (= f l ) gives the algebraic sign to be attached to dy/dX so
that there is a minimum in I 11.1 at the midplane.
Now note that (Exercise 12.3.3):
(12.3.8)
This equation can be used to calculate the relation between X and ym provided the
a
value ofQon the surface, is known. The variable Qis defined by eqn (12.3.6) and if
the surface potential, 11.0, is known then
Qs = [ ~ ( c o s -
~~coshy,)]f
o (12.3.9)
where yo = xe@o/kT. Given any arbitrary ym(with I ymI C I yo I ) one can integrate
(numerically) eqn (12.3.8) from the midplane where X = 0 and Q = 0 to obtain
corresponding values of Qand X. When Qreaches the value a,
the corresponding X
value corresponds to half the separation between the plates (i.e. X (Q= Qs)= K D / ~ )
for the given initial value of y,. By choosing different ym values from near zero
(corresponding to large separations) up to near yo (small separations) one can obtain y,
as a function of separation D for the given surface condition (Qs). Since the force per
unit area between the plates is then given by eqn (12.3.4), the potential energy of
repulsion can be calculated by numerical integration using eqn (12.2.7).
If the interaction occurs under conditions of constant surface charge, a0 then Qs is
given by:
(from eqn (12.2.5)). Again, the integration of eqn (12.3.8) is pursued until the value of
Qs is reached and the corresponding value of X is recorded for subsequent evaluation
of V, from eqn (12.2.7).
590 I 1 2 : D O U B L E L A Y E R I N T E R A C T I O NAND PARTICLE COAGULATION
The same procedure can be used for the condition of charge regulation. In that case
the surface groups undergo some form of dissociation and the degree of dissociation is
influenced by the interaction. Using the single-site model of Section 7.9.1 it is
necessary to ensure that the surface condition satisfies eqn (7.9.4). For a positive
surface that releases a negative ion (A-) into solution the corresponding relation is:
Hence, when two such identical surfaces interact this condition must be fulfilled at all
separations. In order to apply the numerical algorithm described in the previous
section the finishing condition Qmust always comply with eqn (12.3.11). Calculation
of the repulsive interaction with this surface condition gives the regulating case -
where both surface potential and charge will vary with separation in order to satisfy the
surface equilibria. The precise regulation of the interacting surface will depend on the
concentration of adsorbing ions, the dissociation constant, and the site density. An
example is shown in Fig. 12.3.1. It is found that the regulation case must always lie
between the limits of constant charge and potential (Chan and Mitchell 1983). For
some systems, the difference in interaction energy between the range of surface
conditions is slight. However, for spherical colloids near the point of coagulation the
effect of regulation can be of the ultmost importance. Under these conditions it is
necessary to obtain values for the surface dissociation parameters, e.g. site density and
dissociation or binding constants.
1.0 -
0.8 -
VR
mJ m-’
Fig. 12.3.1 Comparison between the constant surface potential (lowest curve), constant surface
charge (uppermost curve) and charge regulation (CR) calculation of VR.The calculation is for a
surface of ApK = 6 (see Section 7.10) at pH 7 and M in 1:l electrolyte. (From Healy et al. 1980,
with permission.)
O V E R L A P OF TWO FLAT DOUBLE L A Y E R S I591
This same numerical procedure can also be used for calculating the interaction
between two dissimilar plates. Two separate integrations are done up to a 1 and a 2
and the resulting values of D1 and D2 are added together to obtain the total separation
between the plates.
(12.3.12)
$, = 2$(x = D/2).
Under these conditions, eqn (7.3.22)is a very good approximation for $ and so:
$m = [8kT/xe]Zexp(-~D/2) (12.3.13)
/ /
D D
V$ = - jdD = - 64nokTZ2exp(-KD)dD
00 00 (12.3.14)
-
64n0k TZ2
exp( -KD).
K
This is one of the most widely used approximate expressions for VR.It is also valid for
conditions of constant charge, since little discharge occurs if the degree of double layer
overlap is small.
If the potential in the region between the plates is small everywhere (i.e. I yo I < 1) so
that the DebyeHuckel equation (7.3.11) holds, then it can be shown (Exercise 12.3.5)
that the midplane potential is given by:
Then using the appropriate form for the osmotic pressure when I $,,, I is small (eqn
(12.3.12))it can be shown that (Exercise 12.3.6):
K +
yo 1 exp(-KD) *
(12.3.16)
592 I 1 2 : D O U B L E L A Y E R I N T E R A C T I O NAND PARTICLE COAGULATION
Comparing this with eqn (12.3.14)it would seem that a very good approximation, valid
for all potentials provided the interaction is not too strong, would be (Verwey and
Overbeek 1948, p. 97) (Exercise 12.3.7):
32n'kT
VZ = y Z2[1 - tanh(~D/2)]. (12.3.17)
Figure 12.3.2 shows a comparison between the 'exact' value of VR calculated from
Table 12.1 (or by numerical integration) with that obtained using eqn (12.3.14). The
log-linear relation between VRand D is evident over a considerable range of KDvalues,
though the approximate equation tends to overestimate the value of VR.
The corresponding expression for VR when the plates interact under conditions of
constant charge Vg rather than constant potential is given by Parsegian and Gingell(l972):
vg = sinh KD
(12.3.18)
Note that under low potential conditions where 00 = EK$O and small degrees of double
layer overlap (exp(-KD) << 1) this expression reduces to:
which is the same as the result obtained under constant (low) potential conditions
(Exercise 12.3.9).This equality of V i and Vg for low potentials holds only at large D.
At small D values the repulsion energy is different in the two cases (though the force is
the same).
t2o
llRV'O
5
0.5
0.2
0.1
0.05
I I I
0 1 2 3 4 5 6
KD-
Fig. 12.3.2 Repulsive potential energy on a logarithmic scale against distance separating the plates.
Full curves:Table 12.1. Dotted curves;eqn (12.3.14).The energy scale is in mJ mp2for K = lo9 mpl
and in pJ m-' for K = lo6 m-'. (After Venvey and Overbeek 1948, p. 85, with permission.)
OVERLAP OF TWO FLAT DOUBLE LAYERS 1593
A more elaborate expression, valid for higher surface potentials, has been derived by
Gregory (1973a). He calculated the volume density of charge between the plates by
adding together the effects of (i) the compression process limiting the available volume
and (ii) the presence of the other plate with its counterions. The potential in the
midplane can then be written:
where again, yo = ze+o/kT and $0 is the potential when the plates are infinitely far
apart. The repulsion potential energy under constant charge conditions is then:
(12.3.2 1)
+
where B = [l y; cosech2 (~D/2)]f.This gives values in close agreement with the
exact calculation.
Exercises
12.3.1 Calculate the potential energy of repulsion V, as a function of separation
(0 5 D 5 20 nm) for two flat plates of surface potential 51.4 mV in a 2:2
electrolyte at a concentration of lop4 M in water at 298 K.
12.3.2 Establish eqn (12.3.5) and integrate it to obtain eqn (12.3.6) using the
boundary condition dy/dX = 0 at X = 0 and y = ym.
12.3.3 Establish eqns (12.3.7) and (12.3.8)
12.3.4 Expand the cosh term in eqn (12.3.4) as a power series to establish eqn
(12.3.12). Hence establish eqn (12.3.14).
12.3.5 Integrate the Debye-Huckel equation d2@/dx2= K ~ @with the appropriate
boundary conditions to obtain d@/dx = K(@ - em)lf2. Use the substitution
@ = &,cosh w to perform the second integration and hence show that @O =
@m cosh K D / ~ .
12.3.6 Use eqn (12.3.12) along with the result ofExercise (12.3.5) and eqn (12.2.7) to
establish eqn (12.3.16).
12.3.7 Calculate the value of VRfrom eqn (12.3.17) for the same conditions as in
Exercise 12.3.1 and plot the results on the same graph.
12.3.8 Show that for small @O (in the DebyeHuckel approximation) the potential
profile between two flat plates may be written (for large D):
coSh[~(D/2- x)]
where x = 0 on one plate.
@(') = @O COS~(KD/~)
12.3.9 Show that if the Debye-Huckel approximation holds throughout the region
between two approaching flat plates then VR= 2r~&exp(-~D)if the surface
potential is constant during the interaction.
594 I 12: DOUBLE LAYER INTERACTION A N D PARTICLE COAGULATION
12.3.10 Use eqns (8.7.3) and (8.7.4) to derive eqn (12.3.2) for the osmotic pressure
between two identical charged flat plates in the form: p-p, = (~/2)(d@/dx)~
where p , is the pressure in the midplane and p is the osmotic pressure
elsewhere.
I\ I . /-
Fig. 12.4.1 (a) The potential profile between two dissimilar plates A and B of surface potential @01
and $002. Note that the minimum no longer occurs in the midplane. (b) Forceedistance relation for
the interaction between two plates of unequal (but constant) potential; both potentials have the same
sign.
INTERACTION BETWEEN DISSIMILAR FLAT PLATES I 595
It is also possible to calculate the 'exact' value of VR from the information in
Table 12.1 by using the method of isodynamic curves as suggested by Deryaguin
(1954). The method is described in some detail by Hunter (1975). It depends on the
fact that the quantity
is constant everywhere between the plates (eqn (12.3.2)). Here p = 2nokT (cosh
ze$/kT - 1) a n d j is the value o f p for $ = $m (as defined by eqn (12.3.4)).
The potential profile between the plates A and B is simply a part of the profile
between two plates of equalpotential, if the second plate is placed at the appropriate
position (C in Fig. 12.4.la). A plate B of surface potential $02 placed as shown in
Fig. 12.4.la is repelled by the same force as would be exerted on a plate, C, of
potential $01 placed further away, at a distance D from the first plate. The total
potential energy of repulsion between the plates A and B is then given by:
and the quantities on the right depend only on interactions between identical plates.
If the second plate were placed at the position E in Fig. 12.4.l(a) it would still
experience the same force if its potential were $02 but this would correspond to a plate
with a charge of opposite sign to the first plate (A). [Note that (d$/dx) at the plate E
has the opposite slope to that at plate B.] Thus it is possible for plates of opposite sign
of charge to repel one another, if they are interacting under constant potential
conditions.
The force between the two plates A and B at separation h is determined by
adjusting the position of the plate C (called the control plate) until the potential
profile between A and B passes through $02 at a distance h from plate A. Then,
from eqn (12.4.2):
The force between the two plates, for any particular values of $01 and $02 has the
same magnitude at two different values of h (corresponding to position B and E in
Fig. 12.4.1(a)). The plot of force as a function of h is shown schematically in
Fig. 12.4.l(b). Notice that it rises to a maximum which must correspond to the
situation when plate 2 is a distance D/2 from the first plate and $02 = lcr, (where lcr, is
the potential midway between two plates of potential $01 and separation 0). Note also
that when plate B is in that position (Fig. 12.4.l(a)) the function d$/dx = 0 at the
surface of B and so B must be uncharged (from eqn (7.3.26)). The maximum pressure
is then given by eqn (12.3.4).
This at first sight rather surprising result (that the force is a maximum when the
plate B is uncharged) can be readily understood when one realizes that constant
potential plates, as they move through the position of the maximum, change their
surface charge from positive to negative values (assuming $01 and $02 are both positive
to begin with). Indeed, if the charge on the second plate is made sufficiently negative
then the repulsion force diminishes to zero. This corresponds to a situation where the
596 I 1 2 : D O U B L E L A Y E R I N T E R A C T I O NAND PARTICLE COAGULATION
potential on the second plate at distance ho is exactly equal to the potential at a distance
ho from an isolated plate of potential $01 (see Exercise 12.4.5):
The potential profile between the plates is then identical to that of an isolated plate and
there is no force (Fig. 12.4.1(b)). The charge on the second plate compensates for all
the charge which would have been associated with plate 1 in the region ho < x < 00.
Finally, it should be noted that if the second plate is moved even closer to the first
plate, then its potential can only be kept constant by piling even more negative charge
onto it. (Note the higher slope of (d$/dx) for x < ho.) This gives rise to an attractive
force between the plates even though they have the same sign of potential. The
converse behaviour is not possible -if both surfaces have the same sign of charge then
the interaction is always repulsive.
The above analysis assumes that the surface potential is constant on each plate. For a
discussion of the constant charge behaviour see Jones and Levine (1969). The charge
regulation behaviour is treated in Chan et al. (1976)
A F = -$000/2. (12.4.5)
Hogg et al. (1966) used this procedure with two dissimilar double layers of (constant)
potential to obtain:
where the surface charges 01 and a 2 are, of course, functions of the separation distance
D.Then if we assume that the DebyeHuckel equation (7.3.11) applies:
(12.4.10)
which describes the potential distribution as a function of distance x from the $01
potential surface. Using eqn (12.4.10) we can now obtain by differentiation the
potential gradient at each surface (Exercise 12.4.2):
(12.4.11)
and 9
1 = -K($OI cosech KD- $02 coth KD).
dx x=D
(12.4.14)
Thus we obtain an explicit expression for the potential energy of repulsion between
two planar surfaces of constant unequal (low) potential. When $01 = $02 and the
interaction is weak, this equation reduces to the relation derived in Exercise 12.3.9.
(But see Exercise 12.4.6.)
Equation (12.4.15) allows computation of the effect of unequal potentials on the
interaction, which corresponds to the case of heterocoagulation. Under conditions of
constant dissimilar potential we noted in Section 12.4 above that it is possible to obtain
attractive double-layer forces below some particular separation. The separation (in
terms of K - l ) at which the interaction energy VRbecomes zero can easily be obtained
for particular values of $01 and $02 using eqn (12.4.15) (Exercise 12.4.4). For example,
for $01 = 25 mV and $02 = 50 mV, VR = 0 at 0.2 K - ’ . (The separation at which an
energy maximum occurs can similarly be found from eqn (12.4.15).)
The separation at which the force between the plates changes from being positive
(repulsive) to negative (attractive) is a rather more significant quantity and that is examined
in Exercise 12.4.5. As noted above (Section 12.4) there is no force between the plates when
the potential of the second plate satisfies $01 = $02 exp(-KD) so that the profile between
the plates is the same as for an isolated double layer at low potential. For an extension of
this treatment to higher potentials, see Ohshima et al. (1982, 1983).
In summary, if the charge on the plates is fixed during an interaction then the
interaction free energy is always positive (repulsive) if the signs are like and negative
(attractive) if they are unlike. If the potential is constant during an interaction and is
different on each plate then an attraction can occur at small separations, even when the
potentials have the same sign.
Exercises
12.4.1 If the approaching plates are of opposite sign of potential the control plate must
have potential -@o and the potential in the midplane is then zero. Discuss the
resulting forces in terms of the Maxwell stress in the region between the plates
(refer to Hunter 1975).
12.4.2 Verify that eqn (12.4.8) is a general solution for eqn (12.4.7). (Compare the
result of Exercise 12.3.5.) Derive eqns (12.4.9-1 1).
12.4.3 Establish eqns (12.4.13-15).
12.4.4 Verify that @ = 0 for @01 = 25 mV and @02 = 50 mV for KD 0.2.
12.4.5 The force between two flat plates is given by F = -d&/dD. Use eqn (12.4.15)
to show that:
+ G2)
F = - ~ E K ~ [ ( G ~ cosech KD- 2@01@02
coth KD]/ sinh KD
and hence prove that F = 0 when @02 = @01 exp ( ~ K D(i.e. ) plate 1 cannot ‘see’
plate 2 if the potential there is the same as it would be for an isolated plate).
12.4.6 Show that for any value of KD, eqn (12.4.15) reduces to
(12.5.1)
Introducing eqn (12.3.16) as an approximation for @ (valid for low potentials) then
(Exercise 12.5.1):
00
The Deryaguin method may be used provided a is reasonably large (say K a > 10) and
the results of ‘exact’ and approximate calculations of VRare shown in Fig. 12.5.l(a).
When the double layer around the particles is very extensive (Ka C 5) the Deryaguin
procedure begins to break down and an alternative approach is necessary. Verwey and
Overbeek (1948) developed an approximate procedure for low surface potentials but
the resulting expressions are not simple. They did, however, show that an approximate
value for VRcan be calculated from:
600 I 1 2 : D O U B L E L A Y E R I N T E R A C T I O NAND PARTICLE COAGULATION
Z2VR
(pJ $l)
0.3
0.2
L \\\ //LKa=2.0
0.1
0.0
0 1 2
KH
Fig. 12.5.1 (a) The repulsive potential energy, VRbetween two large spherical particles when the
exact expression for VR(flat plates) is used in eqn (12.5.1). Broken lines refer to the approximate
expression (eqn (12.5.4)). (Adapted from Venvey and Overbeek 1948, p. 141, with permission.) (b)
The potential energy of repulsion for approach at constant surface potential when a is small.
(Adapted from Overbeek 1952. p. 260, with permission.)
TOTAL POTENTIAL ENERGY OF INTERACTION I601
if an error of up to about 40 per cent can be tolerated. The results of the more complete
calculation for V i are shown in Fig. 12.5.l(b).
This expression is valid only for low surface potentials but a number of other
approximate expressions have been developed by Honig and Mu1 (1971) for small and
large separations, and moderate potentials or charges, on the basis of both the constant
potential and constant charge assumption. An alternate procedure is the surface
element integration method of Bhattacharjee and Elimelech (1997) referred to briefly
in Section 11.5. That can give reasonable estimates of interaction energies for small
values of KU with no restrictions on double layer potential.
[ Exercise I
I 12.5.1. Establish eqns (12.5.2) and (12.5.3).Also establish the identity of the alternate
expressions in eqn (12.5.4). I
12.6 Total potential energy of interaction
We have already discussed, in a qualitative way, the total potential energy of interaction
between surfaces in Fig. (1.6.2). The total potential energy of interaction between a
pair of approaching particles is:
where VAis obtained using the appropriate expression for the non-retarded van der
Waals interaction (Chapter 11) between two infinite flat plates (compare eqns (1 1.3.11
and 11.6.26)):
The repulsive term due to the double layer interaction between identical surfaces, VR,
is given, approximately, for the case of weak interaction (KD> l), by eqn (12.3.14).
The combination of these two functions is the essence of the DLVO theory. (The
particular (approximate) expressions used here give rise to the curves shown in
Fig. 1.6.2. It should be noted that the van der Waals attraction always dominates at
both large and small separations; in the former case however, it may be too weak to be
of significance. At small separations VR must approach a finite magnitude, whereas
I VA I increases very markedly and hence is expected to pull the surfaces into a deep
attractive well, called the primay minimum.This well is not infinitely deep, as expected
from the equation for VA,because of a very steep, short-range repulsion between the
atoms on each surface (see Fig. 12.6.1).The secondary minimum that occurs at larger
distances (- 7 ~ is ~responsible
~ ) for a number of important effects in colloidal
suspensions and these will be alluded to later.
Experimental investigation of the coagulation properties of a wide range of colloidal
solutions suggest that not all systems can be explained using the DLVO theory. Many
602 I 1 2 : D O U B L E L A Y E R I N T E R A C T I O NAND PARTICLE COAGULATION
Fig. 12.6.1 Total potential energy of interaction VT,where Vs is the potential energy of repulsion
due to the solvent layers. Vs is assumed to be negligible until D < -10 nm.
experiments now indicate that there is an extra, so-called structural term that must be
included in eqn (12.6.1). This term arises because of the influence of a surface on
adjacent solvent layers. Depending on the type of surface this can give rise to either
repulsive or attractive forces. We may therefore generally define the total interaction
as:
the VAand VRterms for small values of D. The discrepancy, for 0 < (D/nm) < 5 is
then attributed to Vs. In aqueous media, positive values of Vs indicate the presence of
'hydration forces', whereas negative, attractive values are believed to be due to the
'hydrophobic interaction'. Both types of interaction apparently decay exponentially
with decay lengths typically of the order of one nanometre (Fig. 12.7.10).
(12.6.4)
(12.6.5)
where NAis the Avogadro number and E, is the relative dielectric permittivity. The
quantities on the right are in SI units. (For c.g.s. units the constant is 1.07 x lo5 and
4n€o = 1 statfarad cm-'.) At 25 "C in water (taking E, = 80) if the potential is high
(2 = 1):
where A is in joules.
The observed variation of C.C.C.with valency does depend approximately on the
inverse sixth power of z. The data for As2S3 in Table 1.2 for example, give the ratio
(for z = 1, 2, and 3):
potential surfaces. If the more realistic assumption of low potentials is used, for
example, the result obtained is (Exercise 12.6.2):
where now we have a less dramatic dependence on the valency and a strong dependence on
the surface potential. The experimentaldata could easily be explained by this result if +o a
l/x. That is, if the surface potential was reduced by, for example, adsorption of oppositely
charged ions, where the adsorption energy increases with valency of the ion. At present
there is some interest in applying ion adsorption models to quantitative coagulation studies
in a similar manner to the model used in the study of surface regulation (Section 12.3).
This raises the question of which potential is to be used in the equations for the
repulsive potential energy. Since the interaction occurs between the diffuse layers it
would seem to be more reasonable to assess the magnitude of VRfrom the value of the
diffuse layer potential, +d (Section 7.3). That potential can be estimated from the
electrokinetic (or zeta-) potential ({), which was discussed in Chapter 8. A large body
of experimental evidence shows that when rapid coagulation occurs in a colloidal sol,
the {-potential is commonly around 25-50 mV, which is indeed too small to assume
Z M 1 if we define Z in terms of +d (eqn (12.3.13)). Such a result is, however, quite
consistent with eqn (12.6.7) above, since { would measure the potential, +d, after any
counterion adsorption in the Stern plane (Section 7.4)
Exercises
12.6.1 The condition for the potential energy barrier to just disappear is that
( VA+VR)= 0 and d( VA+VR)/dD = 0 simultaneously. Show that this occurs
when KD= 2. Hence establish the relations (12.6.4) and (12.6.5).
12.6.2 Use the approximate expression (12.5.5) together with an appropriate attractive
energy for spheres of radius a (Chapter 11) to establish the relation (12.6.7) for
the C.C.C.under conditions of low surface potential.
12.6.3 Calculate the repulsive potential energy (under constant potential conditions)
between two spherical particles of radius 0.5 pm, of surface potential 35 mV,
when the electrolyte concentration is (a) lop4 M NaCl and (b) lo-' M NaC1.
(Use H values from 0 to 20 nm.)
12.6.4 Calculate the attraction potential energy between the particles in Exercise
12.6.3 (assuming that A = 5 x lop2' J) as a function of Hfor 0 < H < 20 nm.
Combine this with the curves for V, found in Exercise 12.6.3 to produce
curves for V, and comment on the result. (Ignore Vs.)
systems that are sufficiently well-defined and controllable to provide quantitative tests
of the interaction force or energy as a function of distance between the approaching
surfaces. In this section we will examine some of the experiments conducted on double
layers that are approaching sufficiently slowly to enable equilibrium to be maintained.
We will also restrict attention here for the most part to the interaction of macroscopic
surfaces. Studies of the equilibrium interaction between microscopic (colloidal)
particles can be conducted by a variety of methods [osmotic pressure measurement
(Barclay and Ottewilll970; Barclay et al. 1972), light scattering and neutron scattering
(Ottewill 1982), centrifugation (El-Aaser and Robertson 1971, 1973)] and general
agreement is obtained between the experimental results and the DLVO theory. A
proper discussion of those experiments requires, however, a knowledge of the average
particleparticle distances in concentrated suspensions. The treatment uses the notion
of distribution functions which are discussed in Chapter 13 and their measurement
using scattering theory (Chapter 14). The aim in such experiments is to compare the
theoretical shape of the energy barrier with the experimental observations.
One can also study the kinetics of the coagulation process, which are particularly
sensitive to the height of the barrier, but little influenced by other features of its shape.
That material is discussed briefly in Section 12.8. In Chapter 15 the DLVO theory is
applied to the interpretation of rheology (flow) experiments on coagulating colloidal
suspensions. Kinetics experiments are not always conducted under equilibrium double
layer conditions. Whether double layer equilibrium can be assumed during the
interaction depends on the velocity of approach (e.g. the shear rate in a rheology
experiment) and the rate of adsorption/desorption of the potential determining ions.
The general validity of the DLVO theory was established in the period from 1940 to
1980 by the study of (a) adsorbed liquid films, (b) soap films, (c) swelling of clay
minerals, and (d) interaction between immersed solid bodies. Clay minerals are, of
course, microscopic particles but the distance between them can be estimated from
macroscopic observations (assuming that they are aligned parallel to one another) and
confirmed by low angle X-ray diffraction without recourse to radial distribution
functions.
Fig. 12.7.1 Study of thin liquid film by the captive bubble method.
Soap solution
I
Fig. 12.7.2 (a) Soap film stabilized by cationic surfactant. (b) Soap film produced by drawing a wire
frame out of a solution.
evidence for the validity of the double layer model. Scheludko, Lyklema, and
Mysels continued this work to investigate more comprehensively the effect of salt
concentration on the equilibrium film thickness. That work is reviewed by
Lyklema (1967). Th e agreement between calculated and experimental film
thickness (Fig. 12.7.3) is very reasonable.
Donners et al. (1977) have used light scattering from soap films in order to
investigate surface forces. By measuring the power spectrum of light scattered
from ripples or fluctuations at the soap-film interface, information can be obtained
both from single interfaces (i.e. surface tension and viscosity data) and from
interacting charged soap layers. Their results indicate that for CTAB (cationic
cetyl trimethylammonium bromide) films the great majority of the head group
charge (98 per cent) is balanced by counterions in a compact layer, and the
appropriate potential determining the magnitude of VR is the diffuse layer
potential, +,J.
Next Page
Fig. 12.7.3 Calculated and experimental thickness of soap films as a function of ionic strength
(OAmerican Chemical Society). (From Lyklema and Mysels 1965, with permission.)
12.7.3Swelling of clays
The structure of clay minerals was described in Section 1.4.5 where we noted their
significance as model systems for the study of double-layer interaction. The actual
swelling behaviour of clay is, of course, very important in agriculture, and in civil
engineering (dam, road, and building construction) and in the making of ceramics and
other clay products.
In montmorillonite and vermiculite the alumino-silicate sheets are separated by
water layers whose thickness varies with the concentration and type of electrolyte. In
o Increasing 1
0 Decreasing J pressure
h
I 0.03 M LiCl
z.
E
z. lo.':
lo.'
v
tiL
Y - O \ -
DLVO theory
5ti
P
0
4n loJ:
loJ;
4
1
4 6 8 10 12
Plate separation (nm)
Fig. 12.7.4 Swelling of lithium vermiculite under pressure in 0.03 M LiCl. The theoretical line is
calculated from the total crystal charge, which is almost certainly much higher than the diffuse layer
charge in this system. (From Norrish and Rausell-Colom 1963, with permission.)
Introduction to StatisticaI Mec ha nics
of Fluids
13.1 Introduction
13.2 Molecular interactions
13.3 The structure of liquids
13.4 The potential of mean force
13.5 Time-dependent correlation functions
13.6 Applications of the pair distribution function
13.7 Measurement of correlation functions
13.7.1 Static structure factor
13.7.2 Dynamic structure factor
13.8 Calculation of distribution functions
13.1 Introduction
Many of the properties of colloidal systems can be quite accurately described by
essentially continuum theories. In such theories, the properties of the solvent or
dispersion medium are conveniently characterized by some bulk macroscopic
parameters such as the dielectric permittivity or shear viscosity coefficient. Where
the molecular nature of the solvent has to be included, it is introduced through the
use of ‘cut off parameters or similar conceptual devices (like the quantity L, in
Table 11.4). While it may be aesthetically and philosophically more satisfying to
describe a physical system in terms of the molecular properties of the constituent
molecules at the outset, such an approach will be impossible to implement in
practice and the details may even obscure the more interesting features of the
system. Nevertheless it is clear that experimental techniques have become
sufficiently refined directly to detect, for instance, molecular granularity in surface
force measurements (Fig. 12.7.8) and the theoretical framework required to
understand such results already exists in the field of liquid state physics. Also in
studies of concentrated dispersions (e.g. polymer latices, silica particles, micelles,
and microemulsions) by conventional and dynamic light scattering as well as
638
MOLECULAR I N T E R A C T I O N S I639
t Note that in many cases this ‘excess’ quantity can be negative so that the relevant
property of the real gas is smaller than that of the ideal gas.
640 I 1 3 : INTRODUCTION T O STATISTICAL MECHANICS OF FLUIDS
total potential energy of the system can be simply written as a sum of interactions
between every pair of molecules in the system, that is
Fig. 13.2.1 The interaction energy between two monoatomic molecules as a function of the
separation between the molecules.
THE STRUCTURE OF L I Q U I D S I641
The two parameters 6 and E provide a measure of the size and strength of the
interaction between the molecules (see Exercise 13.2.2). [Note that E in this chapter is
an energy and not the permittivity. The permittivity will always be written with a
subscript.]
The pair potential illustrated in Fig. 13.2.1 shares certain features with the DLVO
potential between two spherical colloidal particles (see Chapter 12) where the attraction
is again due to van der Waals interactions. However, the effective size of the colloidal
particles is determined by the distance at which the repulsive interaction exceeds a few
LT. For colloidal systems at low salt concentrations this effective size can be much larger
than the actual physical size of the particles (for instance when K a << 1). The presence
of the primary minimum (Fig. 12.6.1) can be ignored provided the height of the
primary maximum is much greater than KTbecause in that case particleparticle contact
becomes highly improbable.
The interaction between two polyatomic molecules will depend on the separation as
well as the orientation and conformation of each molecule. Apart from the obvious
steric effects, interactions involving dipoledipole, dipole-induced dipole, etc., are
responsible for the orientational dependence in the pair potential (Hirchfelder et al.
1954). As we shall see, the effective size or excluded volume between molecules plays
an important role in determining the structure of liquids as well as concentrated
colloidal dispersions.
Exercises
13.2.1 The de Broglie wavelength for translational motion of a molecule was defined as
h = I z / ( r n I ~ 7 ) ~What
/ ~ . is the corresponding translational velocity of the
molecules? What is the kinetic energy (in terms of K T ) ?
13.2.2 Verify that the depth of the energy minimum in Fig. 13.2.1 is E where E is
defined in eqn (13.2.2). Where is the minimum, relative to S ? Plot the function
u(r) for 0 5 r 5 1 nm using 6 = 0.3 nm, E = 2 KT.
where the factor N arises because we can select from any one of N molecules. The
above expression is valid provided we are dealing with a uniform system. For non-
uniform systems such as a liquid near the interfacial region, the average number of
molecules at position r is given by p(')(r) d V where d V is the differential volume
element located at r. The quantity p(')(r) is called the one-particle number density. In a
crystalline solid p(')(r) is specified by the lattice structure but in a bulk liquid p(')(r) =
p = N / V (a constant).
T o characterize the structure in a bulk liquid, one would have to speak of joint
probabilities of finding two or more molecules at different points in the liquid. In the
absence of intermolecular forces, each molecule moves independently of the remaining
molecules in the system. The joint probability of finding one particular molecule in the
volume element AV1 located at 11 and another particular molecule in the volume
element A V2 at 1-2 is just (A Vl/ V)(A V2/ V). Thus the number of molecules in these
volume elements is N(N-1) x ( A Vl/ V)(A V2/ V). The factor N(N- 1) follows from
the fact that there are N choices for the first molecule and (N-1) choices for the
second. In the thermodynamic limit (N + 00, V + 00 but (N/ V ) = p, a constant)
this number becomes p2 A V1 A V2.
When molecular interactions are taken into consideration, the presence of a
molecule at one point will influence the probability of finding another molecule at a
nearby location. We quantify this effect as follows:
joint probability of observing molecules in dV1 at rl and in dV2 at r 2
= P2g(rl r2)d Vl d Vz
7 (1 3.3.2)
The function g(r1, r2) is the pair correlation function or pair distribution function (this is a
standard notation). As the name suggests, it accounts for the influence of
intermolecular forces on the likelihood of simultaneously observing two molecules at
positions rl and 1-2.From eqn (13.3.3) we can see from the discussion in the previous
paragraph that ( p dV1) is the probability of observing a molecule in the volume
element dV1 located at rl. Therefore (pg(r1, r2) dVz) is the conditional probability of
observing a molecule in the volume element dVz located at rz, given that a molecule is
already in dV1 at 11. In a bulk liquid the pair distribution function can only be a
function of the distance between the points rl and r2, hence
Therefore pg(r) may be regarded as the local density of molecules given a molecule is
located at the origin of the axis system, or in other words 4n?pg(r) dr is the number of
+
molecules located within a spherical shell of inner and outer radii Y and r dr centred
THE STRUCTURE OF L I Q U I D S I643
about a given molecule. As a consequence g(r) is also known as the radial distribution
function.
Since pg(r) is the local number density, it must approach the bulk value, p, as r
becomes large since the effects of the molecule at the origin must become negligible far
from this molecule. Thus we must have the result
g(r) + 0 as r +0 (1 3.3.6)
which simply says that excluded volume effects prevent two molecules from being at
the same location.
From the very general physical constraints on the behaviour of g(r) given in eqns
(13.3.5) and (13.3.6) we can make a qualitative argument about what g(r) should look
like for a liquid. We know that the number density of molecules of a material in the
liquid state is high, similar to that in the solid state, since the change in density upon
-
melting is 10 - 20 per cent. Therefore the average distance between molecules is close
to the molecular size. (We take 6 to be a rough measure of molecular size.) Due to the
strong repulsion that exists between molecules when their separation is less than 6, the
probability of finding a pair of molecules separated by much less than 6, (i.e. the value
ofg(r)) must be very close to zero. But in order to maintain the high molecular number
density of a liquid this geometric constraint implies that there must be a ‘shell’ of
molecules at a distance slightly greater than 6 from any given molecule. The existence
of such a shell prevents there being any other molecules with centres between one and
two diameters from any given molecule.
By a similar argument, this shell ofJiYSt nearest neighbour molecules also forces the
existence of a shell of second nearest neighbour molecules. However, as one goes
further from the central molecule, correlations between the centres of other molecules
and the central molecule diminish rapidly. At positions corresponding to the location
of the shells of neighbouring molecules, the local density is higher than the average bulk
density; hence at these points g(r) > 1. Between these shells, the local density is lower
than the bulk density and hence g(r) C 1. The magnitudes of these local density
fluctuations decrease with increasing distance from the central molecule i.e. as r
increases; g(r) must therefore be an oscillatory function of r where the amplitudes of
the oscillations decrease with increasing r and g(r) + 1 as Y + 00.
It is important to remember that in the above discussion on the form of g(r), we are
dealing with the structure of the liquid at a given instant in time. The ‘shells’ of
molecules about a given central molecule are not meant to imply the existence of
permanent structural entities. Indeed molecules belonging to different ‘shells’ are in
constant dynamical exchange. In a monatomic liquid the lifetime over which any given
pair of molecules can be regarded as nearest neighbours is of the order of a few
picoseconds. (See Section 13.5.)
Some examples of g(r) are given in Figs 13.3.1 and 13.3.2. The result for argon
(Fig. 13.3.1) is typical of that for simple monatomic liquids. For a liquid of polyatomic
644 I 1 3 : INTRODUCTION TO STATISTICAL M E C H A N I C S OF FLUIDS
molecules, the pair correlation function will depend on the separation as well as the
orientation of each of the molecules. In the case of water (Fig. 13.3.2(a)) this
orientation dependence is equivalent to saying that there are three different atomic
correlation functions between two molecules: the oxygen - oxygen, goo; the oxygen -
hydrogen, and the hydrogen - hydrogen, gHH. However, g o o is very nearly equal
to the pair correlation function between the centres-of-mass of two water molecules.
Note that there are no density correlations around a given water molecule beyond
about 0.8 nm. The pair correlation for a hard sphere (Fig. 13.3.2(b)) fluid serves to
illustrate the observation that the characteristic form of g(r) is due to the repulsion
between the molecules. The sharp discontinuity of g(r) at Y = 6 is a consequence of the
abrupt nature of the hard sphere interaction, namely
u(r) = 0 for r > 6 and u(r) = 00 for r < 6 (13.3.7)
From a knowledge of g(r) one may try to infer the coordination number, N, (the
number of nearest neighbours) of a given molecule. However, there are a number of
ways of defining N, in terms ofg(r), and these estimates can differ by up to 30 per cent
(Pings 1968). For instance, in liquids of noble gases below the critical temperature, N,
varies from about 3.7 f0.5 at the critical point to about 10 f2 at the triple point. The
value of N , varies with the density but is fairly insensitive to temperature variations.
For liquid water, the coordination number is about four and is almost independent of
temperature in the range 0-200 "C (Narton and Levy 1972).
Another familiar example of pair correlation functions is that between ions in a bulk
electrolyte. Even if the solvent is regarded as a continuum, we must have at least a two
component system, namely anions and cations. The ionic pair correlation function
between species of type i a n d j is written as gc(r). In the linear Debye-Huckel theory g i
(r) is approximated by (for r >6, the ionic diameter) (Exercise 13.3.2):
gq = exp[-zje&(r)/kT] M 1 - [z+$i(r)/kT]
= 1 - z.z.
L ( )exp[-K(r - a)]
e2 (13.3.8)
+
~
For r < 6 we have gc(r) = 0. Notice that in this example the pair correlation function
does not have the characteristic oscillatory form of that in a dense liquid. The reason is
Fig. 13.3.1 The pair correlation function for liquid argon at 143 Kand 0.91 g crnp3,which is about
half way between the critical and the triple point near the vapour pressure curve (Pings 1968).
THE STRUCTURE OF L I Q U I D S I645
that the electrolyte, which may be viewed as an ionic liquid, is at a very low density and
interacts via long range Coulombic forces.
Although we have only considered pair correlation functions so far, it is
straightforward to generalize to the concept of triplet, four-body, etc., correlation
functions. For instance, the triplet correlation function g(3)(rl, r 2 , q ) is the joint
probability of observing three molecules at positions rl, r2, and r3. In the next section,
we shall have occasion to use this quantity.
2-
h
v
L.
$ 1
0 Y...............J
2 3
r/6
Fig. 13.3.2 (a) The oxygen-oxygen correlation function for water at 20" C. (After Narten and Levy
1971.) (b) The pair correlation function for a hard sphere fluid at a volume fraction of 45%.
646 I 1 3 : INTRODUCTION T O STATISTICAL MECHANICS OF FLUIDS
r
Exercises
13.3.1 Departures from ideal behaviour in gases can be represented by a virial
expansion, as for example:
where B, Care the second and third virial coefficients, which are functions only
of the temperature. Show that if the van der Waals equation (see Exercise
(12.9.1) is written as a virial expansion then B (b - a/RT) i.e. the second
virial coefficient incorporates both the attractive and repulsive components of
the departure from ideal behaviour.
13.3.2 (a) Use eqn (7.11.7) for y9i to establish eqn (13.3.8).
(b) The number density of ions of typej at r, given there is an ion of type i at the
origin is nj (r) = nj g i (r)where nj is the average number density of ions of typej
in the system. The excess density of ions of typej is then Anj (r) = nj ( r )- nj =
nj [gii ( r ) - 11. Therefore, the number of excessj ions in a spherical shell of
radius r and thickness dr centred on ion i is:
Anj(r)41tr2dr = F(r)dr.
where W(r)is the reversible work needed to move two molecules through a liquid from
some initial large separation (say infinity) to a separation r. In a dilute gas this is just the
work done against the interaction potential between two molecules since the effect of a
third molecule on this process is negligible at low densities.
THE POTENTIAL OF M E A N FORCE I647
s
v'=v
where the integral is to be taken along any path between some point far away (r' = 00)
and the point r' = r. Since the process is reversible (i.e. the moving process is not
dissipative) the integral is independent of the actual course of the path taken.
An equivalent statement of eqn (13.4.3)is that the x-component of the average force
between two particles is
with similar expressions for they- and z-components. Thus W(r) has the role of a
potential energy from which the components of the average force may be obtained by
taking the appropriate spatial derivative. W(r)is therefore known as the potential of
averageforce or potential of mean force. It should be evident from the above discussion
that W(r) is in fact the change in thefree energy in bringing two molecules from infinity
to a separation r. For example, the total interaction free energy between two colloidal
particles (van der Waals and electrical double layer interaction) is a potential of mean
force.
In a dense liquid, the potential of mean force W(r) depends on the temperature and the
density since it is associated with the average force between particles, averaged over all
other particles in the system. To see how this comes about in more detail, consider the
x-component of the average force on a molecule at rl due to the presence of another
molecule at r2. From eqn (13.4.4) this force is
There are, of course, similar equations for the y- and z-components of the average
force. The first term on the right hand side of eqn (13.4.5) is the 'direct' force on the
molecule due to the pair interaction with the molecule at r2. The second term accounts
for the effects of all other molecules in the system. This term says the following: let the
conditional probability of finding a molecule at r3, given two molecules are already at rl
and r2 be P(r3 I 11, r2). Under the same conditions, the number of molecules in a
volume element d r 3 , located at r3, is therefore pP(r3 lrl,r2)dr3 and each of these
648 I 1 3 : INTRODUCTION T O STATISTICAL MECHANICS OF FLUIDS
molecules exerts a force on the molecule at rl, the x-component of which is (-du( 1 rl -
13 I)/dxl). The total effect of all other molecules in the system can be obtained by
integrating (or summing) over all volume elements. The conditional probability, P(r3 1
rl,r2) is related to the triplet and pair distribution function by:
g(3)(rl,r2, r3) = r2)P(r3Irl, r2) (1 3.4.6)
which states that d3)(rl,r 2 , r3), the joint probability of observing molecules at rl, 1-2,
and r3 is equal to the probability of finding molecules at rl and rz (namely g(r1, r2))
multiplied by the conditional probability P(r3 1 rl, r ~of) observing a molecule at r3
given that molecules already exist at rl and rz. By combining eqns (13.4.2), (13.4.5),
and (13.4.6) we find (Exercise 13.4.1):
where Y = Irl - r 2 I. This exact result was first obtained by Born, Green, and Yvon
(BGY) and is known as a member of the BGY hierarchy equation, which provides a
relation between the pair and triplet correlation functions. The equation may be solved
(though not easily) for the pair distribution function g(r) provided we have some
additional information about the triplet distribution function g(3)(rl,r - 2 , ~ ) .
Unfortunately little is known about g(3)(rl,rz, r3) except that, like g(r), it can also be
written as
r ~ r, 3 ) i k ~ ]
g(3)(rl,r2, r3) = e x p [ - ~ ( ~ ) ( r2, (1 3.4.8)
with m3)(r1, 1-2, 13) being the triplet potential of mean force for assembling three
molecules at rl, 1 2 , and 1 3 from infinity. One possibility, known as the superposition
approximation, is to approximate M3)by a sum of potentials of mean force between
each pair of molecules in the triplet, that is,
Note that this remains an approximation even for systems in which the molecules
interact solely via two-body potentials. Using eqns (13.4.2) and (13.4.8), eqn (13.4.9) is
equivalent to the assumption of independent probabilities
Fig. 13.4.1 The pair potential, the potential of mean force, and the average force for a hard sphere
fluid at a volume fraction of 45%.
on all sides, and consequently the average force between the two molecules is zero.
Consider now the configuration in which the two molecular centres are 1 to 1; diameters
apart. In this situation there is insufficient room for a third molecule to fit in between the
two molecules (see Fig. 13.4.2). As a result collisions with all other molecules in the
system which are nearly in line with the intermolecular axis will have a net effect of
pushing the two molecules together. In other words, the molecules experiencean effective
attractive force (< 0) between them.
When the two molecular centres are 1; to 2 diameters apart, collisions with other
neighbouring molecules will now tend to force the two molecules apart in order to fit in
a third molecule. As a consequence the two molecules experience an effective repulsive
650 I 1 3 : INTRODUCTION T O STATISTICAL MECHANICS OF FLUIDS
Fig. 13.4.2 A schematic illustration of the distribution of hard spheres in the neighbourhood of two
spheres held at a fixed separation.
force (> 0) between them. Thus even when there is no direct interaction between two hard
sphere molecules the presence of other molecules can generate an eflective attractive and
repulsive interaction between the two molecules. While the above discussion only exploits
the excluded volume and associated geometric constraints, the presence of attractive
interactions in real liquids will only modify the details of the effective interaction. The
general structure of a liquid is determined almost entirely by the repulsive (excluded
volume) forces in the system. [Note the similarity between this description and that of
the depletion effects of non-adsorbing polymers (Section 12.9.4).]
A striking illustration of the form of the potential of mean force across liquids may
be found in direct measurements of the force between crossed mica cylinders across
organic liquids (Horn and Israelachvili 1981). The apparatus used in these experiments
is that shown in Fig. 6.2.1. In the example given in Fig. 13.4.3 the liquid is octamethyl
cyclo-tetrasiloxane, whose molecule is nearly spherical in shape with a mean diameter
of 1 nm. Direct measurement of the force between the cylinders as a function of
separation can in this case reveal the graininess of the liquid because the molecules are
so very large. Recall that it has more recently proved possible to detect the graininess of
water molecules in this apparatus (Fig. 12.7.10) (Israelachvili and Pashley 1983).
A more familiar example of a potential of mean force may be found in the theory of
the electrical double layer at a charged surface. The density of ions of species i near a
flat surface may be written as
Pi(x) = pi exp[-W,(x)lkTI. (13.4.11)
(In the notation of Section 7.3 pj(x) = ni and pi = np.)
In the Gouy-Chapman theory the potential of mean force of ionic species i, Wi(x)
is
approximated by the charge on the ion (xje) multiplied by the mean electrostatic
potential +(x) (Section 7.3), i.e.
W;:(x)M zje$(x). (13.4.12)
The more elaborate treatments referred to in Section 7.6 attempt to improve on eqn
(13.4.12) by introducing corrections for the various effects: the finite ion size, the work
done in moving water molecules aside to make way for the ion, the effect of the ion on
its neighbours and on the structure of the surrounding (dipolar) liquid, the effect of
the local field on the volume occupied by the water molecules (electrostriction), etc.
THE POTENTIAL OF M E A N FORCE I651
T
I
I
I
I
I
I
I
I
I
I
I
I
I
T
I
I
I
I
I
I
I
I
I
I
I
I
I
I I I I I I I , ,
2 3 4 5 6 7 8 9
Distance (nm)
Fig. 13.4.3 Measurement of force as a function of separation between two mica plates in
octamethyl-cyclotetrasiloxane showing the solvent structure. (After Horn and Israelachvili 1981 with
permission.)
Exercises
13.4.1 Establish eqn (13.4.7).
652 I 1 3 : INTRODUCTION T O STATISTICAL MECHANICS OF FLUIDS
The self part, G&, t)is obviously related to processes such as self diffusion (Section 4.9.2)
while the distinct part, Gd(r,t ) Will give us a measure of the life time of liquid structures.
In fact we must havetfor an isotropic liquid:
In Figs 13.5.1 and 13.5.2 we show some examples of the distinct part, Gd(r, t ) of the
time-dependent pair correlation function. These results are obtained by computer
simulation of a system of particles or molecules that interact via some model potential
(see Section 13.8). The parameters are chosen to mimic the properties of argon
(Rahman 1964) and a suspension of polystyrene latices (Gaylor et al. 1980, 1981). The
rate at which the structure in Gd(Y,t ) decays in time is indicative of the lifetime of the
local environment around a given molecule and this changes as a result of the thermal
motion of the molecules. With the passage of time or with increasing separation, we
have the expected limit Gd(Y, t ) + p as t + 00 or Y + 00 which implies that all
Some authors have defined G(r,t) with the bulk density normalized out so that
Gd(r,t=O)=g(r).
TIME-DEPENDENT C O R R E L A T I O NF U N C T ION S I653
Fig. 13.5.1 The distinct time dependent correlation function G&, t ) for a Lennard-Jones liquid
( c / k = 20 K 6 = 0.34 nm) which closely models liquid argon at 94.4 K and 1.374 g cmp3. (c is the
well depth shown in Fig. 13.2.1.) (After Rahman 1964.)
2-
h
r=
v
$1-
'Q
I I
0 500 1000
Fig. 13.5.2 The distinct time dependent correlation function, G&, t) for a colloidal dispersion of
sphericalparticles (radius 23 nm, surface potential 150 mV, M 1:l electrolyte,volume fraction 4.4 x
lop4)interacting in accordance with the linear DebyeHuckel formula (section 12.5) for electrical double
layer interaction (Gaylor et al. 1980). Full line t = 0; . . . t = 5 x 10p4s;- - - - t = 2 x 10p3s.
654 I 1 3 : INTRODUCTION T O STATISTICAL MECHANICS OF FLUIDS
correlations eventually fade away and there is equal probability of finding a second
molecule anywhere. [Note the very large difference in the time scales for these two
systems.]
(13.6.1)
The first term is the kinetic energy due to translational motion of the molecules. The
notation in eqn (13.6.1) means that the two summations run over all values of i = 1,
2,. . ., N a n d j = 1,2,. . ., N , but all terms with i = j are omitted (since a molecule does
not interact with itself). The factor 1/2 is to correct for double counting in the two
summations. T o obtain the average internal energy ( U ) however, we need to average
over all possible configurations of the molecules. We can accomplish this as follows:
choose a molecule and set up a coordinate system at the centre of this molecule. The
number of molecules in a spherical shell of radius Y and thickness dr around this
molecule at the origin is 4x2pg(r) dr. The total interaction energy between the central
molecule and all other molecules may be obtained by multiplying the number in the
spherical shell by u(r) and integrating over I from r = 0 to r = 00. The total interaction
energy between all molecules is therefore
00
3
( U ) = - NkT
2
+ $Np (1 3.6.2)
0
The factor N arises because there are N choices for the molecule at the origin and the
factor 1/2 corrects for the double counting, as each molecule has been counted once as
the central molecule and again as one of the molecules in the shell.
One can also obtain a similar expression for the pressure by differentiating the free
energy function (p = -(W/~V)T)but we will not go into the details here. The result
(McQuarrie 1977, p. 262) is an expression in terms of g(r):
2x
p = pkT - -pz
3
7
0
M
d4r)
y 3 - g ( ~ )d ~ .
dr
(1 3.6.3)
APPLICATIONS OF THE PAIR DISTRIBUTION FUNCTION I655
Equations (13.6.2) and (13.6.3) thus provide possible links between statistical mechanics
and thermodynamics. For example, ifg(r) and hence C U> are known functions of p and
T, all other thermodynamic quantities can be calculated in the usual way.
Since the pair correlation function, g(r) is related to the local density, it is possible to
show from the theory of linear response (see Section 3.2.3) that g(r) also characterizes
the change in the local density as a consequence of an external perturbation (Hansen
and McDonald 1976). This is explained as follows. Consider a bulk liquid which has
been subjected to an external field of the form u(ri).That is, the external field acts
on each molecule of the liquid and gives it a potential energy, u, which depends only on
the position, r;, of that molecule. The potential u(r)may be the gravitational field, or
the external radiation field in a scattering experiment (light, X-rays, or neutrons) or it
may even be that due to the effect of a colloidal particle on the neighbouring dispersion
medium, or on the ions in that medium.The change in local density A&) is then given
by (Exercise 13.6.1):
where we have used the standard notation h(r) = g(r) - 1. This result is strictly valid
for weak external fields (u/KT C 1) since terms of order [ u / k g 2 and higher powers
have been omitted from the general expression for A&).
The notion ofg(r) as a response function which characterizeschanges in local density as
a result of an external perturbation can be extended by introducing the idea of the
susceptibility, 2 of a system. [A familiar example of a susceptibilityis the conductivity of a
circuit element, K(w),which relates the response [namely, the current I(@), through the
circuit] to the perturbation [in this case the applied voltage, V(w)]:
K ( 4 = I(@)/ V(w) (= 1/Z(4)
where Z is the impedance, introduced in Exercise 3.2.5. [In a biological situation one
would speak of the susceptibility of an organism to a stimulus so producing a response.
In linear theory the response is proportional to the stimulus and the proportionality
constant is the susceptibility.]
Ideally, we would like to know the value of the ratio g(r)/u(r) but the form of eqn
(13.6.4), in which both g and u appear in an integral, poses a problem. Fortunately, the
difficulty can be circumvented by introducing a mathematical device called a Fourier
transform which is described briefly in Appendix A4. The Fourier transform of eqn
(13.6.4) is given by (see Exercise 13.6.2):
(1 3.6.5)
- kT [ + 2[
r
1
00
00 1
r sin(Qr)h(r)dr] .
(1 3.6.6)
The function inside the square brackets is of fundamental importance since it can be
shown that it is directly related to the isothermal compressibility of the fluid:
(1 3.6.8)
[In view of the probabilistic interpretation of g(r) one might think that the integral
4n ?g(r) dr should be equal to N-1 rather than be related to the compressibility as
indicated in eqn (13.6.9).This result for g(r) is valid befire the thermodynamic limit ( V
+ 00, N + 00 with N / V = p) is taken. However, to relate the integral in eqn (13.6.7)
to a thermodynamic quantity, one must first obtain the function h(r) in the
thermodynamic limit before performing the integral. This somewhat subtle point is
discussed in detail in Hill 1956.1
I I
Exercises.
13.6.1 Equation (13.6.4) has a rigorous derivation based on statistical mechanics.
Here we consider a simplified heuristic derivation. The local density at r in an
external field u(r)can be written as
P ( 4 = P e v - W(r)/KT)
MEASUREMENT OF CORRELATION FUNCTIONS I 657
where W(r)is the free energy [relative to a point where u(r) = 01 of putting a
molecule at r. The quantity W(r)has two contributions: (a) the direct
interaction between the molecule and the external field, i.e. u(r) and (b) the
effect which the presence of a molecule at r will have on the local density of
molecules around r. The resulting excess of molecules (which may be positive
or negative) due to (b) also interacts with the external field and contributes to
W(r).For a weak external field this second contribution has the form
where [pg(r - r‘) - p] dr‘ is the excess number of molecules at the volume
element dr‘ located at r‘ given that there is a molecule at r.
Start with the above argument and derive eqn (13.6.4) assuming that u/kT << 1.
13.6.2 Obtain (13.6.5) using the three dimensional convolution theorem (Appendix
-
A4). (The symbol above a function denotes its Fourier transform.)
13.6.3 Derive eqn (13.6.6).
Hints : /f(r)dr = 7 7[
0
d@
0 0
sin 8d8 g f ( r ) - dr
1
exp(-kr) = exp(-zkr cos 8) and sin x = -[exp(zx) - exp(-zx)].
2i
from Fig. 3.3.2 with kj M ks.The angle 8 through which the incident beam has been
deflected is known as the scattering angle.
At an observation point, r far from all scattering centres (a condition well satisfied in
an experimental setup), the scattered field from the ith particle can be shown to be:
(13.7.2)
where A is the scattering amplitude and c j j the phase. (Compare this with equation
(5.7.15) where the light scattered by a single colloidal particle is calculated.) The total
scattering field from all particles, assumed identical and hence having the same
scattering amplitude, is
N A N
(1 3.7.3)
i= 1
MEASUREMENT OF CORRELATION FUNCTIONS I 659
Since the observation point is far from the scattering volume, we can, to an excellent
approximation, replace II - ri I by I I I = r. T o obtain the final expression we have used
the expression for the phase difference to measure all phases relative to particle 1. The
scattered intensity is (Exercise 13.7.1)
(1 3.7.4)
I J
In the above discussion, the positions of the particles in the system are assumed to be
fixed in a given configuration. This is quite reasonable as the time taken for radiation to
pass through the entire scattering volume is much smaller than the time scale of
molecular motion (Exercise 13.72). However, experimental measurement of the
intensity is really averaged over many particle configurations. In other words, we
require the average (I,)which is given by:
r')] +p s I'
dRh(R) exp(iQ.R)
(1 3.7.5)
Since the quantities to be averaged involve only pairs of particles, the pair correlation
can be used to effect this average (compare Section 13.6). The first term on the right
hand side corresponds to scattering from N independent particles. The second term
corresponds to scattering from the surface of the scattering volume. For a macroscopic
sample the magnitude of this term is negligible except in the forward direction (Q = 0)
which is difficult to study with light waves and X-rays (though it is studied by low
angle laser light and small angle neutron scattering (LALLS and SANS)). This term
will, therefore, be omitted from the remainder of this chapter. The third term accounts
for interference effects due to correlations in the positions of the scattering molecules,
distance R apart.
So far, we have treated the particles as point scatterers. The general result for the
average scattered intensity from particles of finite size should read
in which f(0) depends on the polarisation of the incident and scattered radiation and
although it may depend on 0 it will be unity for the usual laser light system (see
Sections 3.3.1 and 3.3.2). Theform factor P(QJ is a property of the shape and size of the
660 I 1 3 : INTRODUCTION T O STATISTICAL MECHANICS OF FLUIDS
scattering particle and was introduced in eqn (3.3.14). The (static) structure factor S(Q)
is [compare with eqn (13.6.6)]:
M
On the other hand, in the limit of large Q, it can be shown from eqn (13.7.7) that S(Q)
+ 1 as Q+ 00 because h(r) + 0 (i.e. g(r) + 1) as r+ 00. Physically we can see how
the large Qlimit comes about. From eqns (13.6.6) and (13.7.7) we see that S(Q) is
proportional to the susceptibility X ( Q ) which characterizes the density response of the
system to an external perturbation of spatial frequency Q o r wavelength h = 2n/Q. As
Q+ 00 (A + 0), the wavelength will become much smaller than the interparticle
spacing and the external field would have undergone many oscillations between every
pair of particles. Consequently there can be no correlations in the local density in
response to such a perturbation, as each particle is in effect subject to uncorrelated
perturbations.
At intermediate values of Q, the structure factor S(Q) is an oscillatory function. The
maxima of S(Q)are analogues of Bragg peaks in the crystal diffraction pattern and the
locations of these maxima are roughly at values of Qfo r which the product Qd is an
integral multiple of 2n, where d is the mean spacing between particles (Brown et al.
1976). Since S(Q) is proportional to i ( Q ) , the susceptibility of the local density to an
external perturbation of spatial frequency Q (wavelength h = 2n/Q),we therefore
expect that external fields of wavelength d will generate large density fluctuations as the
positions of the particles in the system are ‘in phase’ with the applied field.
In the dilute limit ( p + 0) we have S(QJ = 1 [see eqn (13.7.7)] and this limit is used
to determine the prefactors of S(Q) in eqn (13.7.6) which are independent of number
density.
As discussed in Section 13.3, the form of the pair distribution g(r) of molecules in a
liquid or that of particles in a stable colloidal dispersion is controlled mainly by the
excluded volume effect due to the repulsive interaction between the particles or
molecules. The form of the structure factor S(Q) is therefore also determined by
similar factors. Indeed experimentally determined S(Q) for a variety of systems can be
MEASUREMENT OF CORRELATION FUNCTIONS I 661
Fig. 13.7.1 A comparison of the structure factor of a Lennard-Jones fluid at ~6~ = 0.844, kT/e =
0.72, which is near the triple point (full line) with the structure factor of a hard sphere fluid (points).
(After Verlet 1968.)
fitted to the structure factor of a hard sphere fluid, provided a suitable hard sphere
diameter is used as illustrated in Fig. 13.7.1. In reality this only poses a new question as
to how to determine a priori the appropriate hard sphere size. A corollary of this
observation is that attempts to fit experimental structure factors do not provide very
stringent or sensitive methods of determining the interparticle interaction.
An analytic expression for S(@ for a hard sphere fluid has been obtained (Baxter
1968) in the Percus-Yevick approximation (see Section 13.8) which for most practical
purposes is sufficiently accurate:
1
~
S(Q)
=1 + 244
~
x3
{a(sinx - xcosx) + b[(2/x2 - 1)xcosx + 2 sinx - 2/x]}
(1 3.7.9)
+ 12cp2a
~
x3
{ 24/x3 + 4( 1 - 6 / x 2 )sin x - (1 - 12/x2 + 24/x4)x cos x]}
where 6 is the hard sphere diameter, p the number density, x = QS; 4 = npS3/6 and
(13.7.10)
the intermediate scattering function, F(Q,t) which is related to the time dependent (van
Hove) pair correlationfunction, G(r,t) by
and F(.Q t= 0) = S(QJ (Exercise 13.7.3). However the experimental results are
presented in a normalized form in terms of the field autocorrelationfunction g(')(Q, t )
(compare eqn (5.7.15):
is the diffusion coefficient of the free particle of radius a in a medium of shear viscosity
r]. In general when Ink(') (a t)]is plotted as a function of (@t) the rate of decay is
slower than that of DO due to interparticle interaction (Fig. 13.7.2) and a central
problem in studying the dynamics of colloidal systems is the interpretation of the form
of g(')(Q, t ) (Pusey and Tough 1982).
OC
Fig. 13.7.2 The field autocorrelation function g(') (a t ) at Q = &, the first maximum of the
structure factor, for a dispersion of deionized polystyrene latex spheres (radius 25 nm, volume
fraction -lop3.) (Redrawn from Pusey and Tough 1982.)
C A L C U L A T I O N O F DISTRIBUTION FUNCTIONS I663
I
Exercises
13.7.1 Justify the final form for I, in eqn (13.7.4) from the previous equation.
13.7.2 Compare the time scale of molecular motion with the time taken for light to
pass through a scattering cell.
13.7.3 Show that F ( Q t = 0) = S(QJ using eqn (13.7.11).
In order to solve this single equation, which contains two unknown functions, we
need some extra information. It turns out that from a detailed analysis of the
statistical mechanics of liquids, it is possible to give a formal prescription for
the direct correlation function c(r) in terms of the intermolecular potential, the
temperature, density, and the total correlation function h(r). This prescription
follows from a formal density expansion of the potential of mean force in powers of
the density (Hansen and MacDonald 1976). Unfortunately, like the virial equation
of state, the expression for c(r) cannot be written down in closed form. However, by
omitting unmanageable terms, one can obtain an approximate expression for c(r).
This procedure generates a class of approximation schemes with names such as the
Hypernetted Chain (HNC), the Percus-Yevick (PY), the Mean Spherical
Approximation (MSA), etc. One simple rigorous result about the function c(r) is
664 I 1 3 : INTRODUCTION T O STATISTICAL MECHANICS OF FLUIDS
that for systems that are characterized by a pair potential u(r), it has the limiting
form
Indeed the familiar deb ye-Huckel theory of electrolytes is equivalent to assuming that
eqn (13.8.2)is valid for all r and not just in the limit r + 00.
T o see the origin of the various approximations mentioned above, we begin with the
formally exact result (compare eqn (13.3.8)):
The function B(r) represents contributions from so called Bridge functions due to
their appearance in a graphical representation. They are the terms that are difficult to
handle. In the hypernetted chain approximation one assumes B(r) = 0 and obtains
c ( r ) = -u(r)/kT. (1 3.8.6)
It is easy to see that for hard sphere interaction (c.f. eqn (13.3.7))the Percus-Yevick
equation implies that c(r) = 0 for r > 6 while the hard sphere exclusion imposes the
exact condition
For hard sphere fluids, the PY approximation turns out to be quite accurate in
addition to the fact that it has an analytic solution. [See eqns (13.7.9 and lo)].
Unfortunately, this appears to be a coincidence due to a cancellation of errors. In
general, it is not possible to obtain a good a priori estimate of the errors in the
approximations described above. However, experience has shown that for systems
involving Coulombic interactions, the hypernetted chain approximation (which is in
fact a slightly more sophisticated version of the Poisson-Boltzmann theory) seems
quite accurate.
A feature of these approximate integral equation methods that should be
remembered is that because the pair distribution functions are not given exactly, the
different possible methods of obtaining the thermodynamics (namely via the energy
equation (eqn (13.6.2))or the pressure or virial equation (13.6.3)or the compressibility
equations (13.6.8)and (13.7.8)will in general not yield identical results. For instance if
C A L C U L A T I O N O F DISTRIBUTION FUNCTIONS I665
p" - +
p k T - (1 -4)3
" (compressibility equation) (13.8.8)
9" - 1 + 2 4 + 3 4 2
(virial equation) (13.8.9)
PkT- (1 -4)2
where 4 = nps3/6 is the hard sphere volume fraction. It turns out that for the hard
sphere fluid the compressibilityresult is an overestimate of the pressure while the virial
result is an underestimate. An heuristic equation constructed by Carnahan and
Starling (see Hansen and McDonald 1976)
(13.8.10)
gives excellent agreement with the exact results for hard spheres. For a general liquid,
experience has shown that the energy tends to give the most accurate results. For a
hard sphere system, the energy equation cannot be used (Exercise 13.8.2).
A more systematic way of determining the structure and thermodynamic properties
of a simple liquid is by perturbation methods. These methods are based on the
observation that at liquid densities the repulsive interaction between the molecules
plays a dominant role in determining the basic structure of a liquid. The idea then is to
separate the intermolecular potential, U ( Y ) into a sum of attractive, UA(Y) and repulsive,
U R ( Y ) parts
(13.8.12)
0
There are many other theories that exploit essentially the same ideas though differing
in the choice of the decomposition of the intermolecular potential (see eqn (13.8.11)) as
well as in the detailed treatment of the attractive and repulsive part of the pair
potential.
A direct method of determining liquid structure is by computer simulations. Given
the form of the intermolecular potential, one carries out a computer 'experiment' to see
how such a 'model' liquid would behave. There are two methods of simulation. In the
666 I 1 3 : INTRODUCTION TO STATISTICAL M E C H A N I C S OF FLUIDS
Monte Carlo method, the idea is to generate a large number (-lo6) of possible
configurations of the system (positional, and where necessary orientational,
coordinates of the molecules) and the configurations are weighted according to the
Boltzmann factor, (exp(-U/k7)) of the whole system. (In practice, this weighting is
done by a clever method called significant sampling, the details of which need not
concern us here.) These configurations are then used to obtain averages of equilibrium
properties. In implementing this scheme, one chooses a cubical cell containing 102-103
particles. T o simulate an infinite system, this cell is replicated as neighbouring cells of a
central cell. The number of particles in the cell is limited by computer memory, time,
and cost. However, for short-ranged potentials (e.g. the Lennard-Jones potential) and
for state points that are not too close to the critical point where one might expect long
range fluctuations, the above constraints are not serious. Indeed, Monte Carlo studies
have been carried out to simulate a dispersion of colloidal particles interacting via
DLVO potentials (Snook and van Megan 1976). However, for the case of an electrolyte
solution with ions and solvent molecules, computer simulations still pose an enormous
problem since to simulate a 0.1 M electrolyte with 200 ions, one would need several
thousand solvent molecules.
In order to obtain time dependent properties, a different simulation method
known as molecular dynamics has to be used. In this case the idea is to treat the
collection of particles in a cell as a many-body problem in classical mechanics and to
solve Newton’s laws of motion to determine the position and velocity of each particle
as a function of time. The effective size of the cell is increased enormously by the use
of ‘periodic boundary conditions’. In effect, this means that when the trajectory of a
particle takes it towards a cell boundary it is allowed to leave the system and a new
particle is introduced on the opposite side (Fig. 13.8.1) with the appropriate velocity.
This method is a more efficient simulation technique and equilibrium properties can
also be extracted by using a time average. Examples of results of molecular dynamics
simulations have been given in Section 13.5. In this regard we note that in simulating
the dynamical properties of a colloidal system it is obviously not feasible to study the
time evolution of the colloidal particles as well as the solvent molecules. Due to the
very large difference between the mass of a colloidal particle and a solvent molecule
the characteristic time scale associated with the evolution of each species is very
different. Indeed one is only interested in the solvent in that it provides the ‘driving
force’ for the random Brownian motion of the colloidal particle. Consequently one
regards a colloidal dispersion simply as a collection of colloidal particles in which the
motion of each particle will have a random component of the appropriate statistical
property to simulate the effects of thermal agitations due to the solvent.
Hydrodynamic interactions due to the motion of the colloidal particles are treated by
classical continuum hydrodynamics (compare Section 4.9). This simulation method is
known as Brownian dynamics and its potentialities have yet to be fully explored. For a
resumi of advances in the study of concentrated dispersions see the 1983 Faraday
Discussion of the Royal Society of Chemistry (No. 76).
Exercises
13.8.1 Verify eqn (13.8.6).
13.8.2 Why is it not possible to use the energy equation to obtain the thermodynamics
of a hard sphere fluid?
13.8.3 Plot the functionspC/pkTandpU/pkTfor0 5 4 5 0.5 and compare them with
the Carnahan and Starling eqn (13.8.10) which represents the hard sphere fluid
very well.
References
Atkins, P.W. (1978). Physical Chemistry. Oxford University Press.
Baxter, R.J. (1968). Aust. J Phys. 21, 563.
Brown, J.C., Goodwin, J.W., Ottewill, R.H., and Pusey, P.N. (1976). In Colloid
and Interface Science (ed. M. Kerker) Vol. 4, p. 59. Academic Press, New York.
Egglestaff, P.A. (1967). A n introduction to the liquid state. Academic Press,
New York.
Gaylor, K., Snook, I., van Megan, W., and Watts, R.O. (1980).3. Chem. Soc.
Faraday 2, 76, 1067.
Gaylor, K., Snook, I., and van Megan, W. (1981).J. Chem. Phys. 75, 1682.
Hansen, J.-P. and McDonald, I.R. (1976). Theory of simple liquids. Academic
Press, New York.
Hill, T.L. (1956). Statistical mechanics. McGraw-Hill, New York.
Hirchfelder, J.O., Curtis, C.F., and Bird, R.B. (1954). Molecular theory of gases
and liquids. John Wiley, New York.
Horn, R.G. and Israelachvili, J.N. (1981).3’.Chem. Phys. 75, 1400.
Israelachvili, J.N. and Pashley, R.M. (1983). Nature 306,249,
McQuarrie, D.A. (1977). Statistical mechanics. Harper and Row, New York.
Narten, A.H. and Levy, H.A. (1971).3. Chem. Phys. 55,2263.
Narten, A.H. and Levy, H.A. (1972). In Water: a comprehensive treatise
(ed. F. Franks). Plenum, New York.
Ottewill, R.H. (1982). Concentrated systems. In Colloidal dispersions
(ed. J.W. Goodwin). Royal Society of Chemistry, London.
668 I 1 3 : INTRODUCTION TO STATISTICAL M E C H A N I C S OF FLUIDS
14.1 Introduction
We begin this chapter with a discussion of the structure of concentrated dispersions
and how that can be elucidated using the technique of neutron scattering. We define a
colloidal dispersion as concentrated if its properties are influenced by interactions
between the constituent particles. In the case of a sterically stabilized suspension, for
example, the primary interaction may be due to excluded volume, which merely
reflects the fact that particles cannot pass through one another. This is the so-called
‘hard-sphere’ interaction of Chapter 13.
We can obtain a feeling for the magnitude of the effect of this interaction by
computing the osmotic compressibility, K T , of a hard sphere suspension, which from
eqns (13.7.8), (13.7.9), and (13.7.10) is given in the Percus-Yevick approximation by+
PkTKT = [(I @z/(l 24912
- + (14.1.1)
where p is the number density of particles of diameter 6 in the suspension, and
4 = xpS3/6 (14.1.2)
is the fraction of the total volume occupied by particles (i.e. the volume fraction of the
suspension). The right-hand side of eqn (14.1.1) is unity at infinite dilution, and has
only fallen to half of its infinite dilution value at a volume fraction of 8.8%, which is a
669
670 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
fairly high solids content in suspension. In the case of hard spheres, therefore, our
concept of ‘concentrated’ parallels the intuitive idea of a high concentration of
dispersed material. For many of the most interesting colloidal systems, however, this is
not the case.
Dispersions of sulphonated polystyrene latex particles, which we shall consider in
some detail later, provide an important class of examples. These particles interact
through a screened Coulomb potential whose range is the Debye length, h.d = K1,of
the suspension (Section 7.3).
As a first crude estimate of the effect of adding such a potential to the excluded volume
interaction already present, we may treat the particles as if they had an effective diameter
of (6+2hd). The effective volume fraction of such a suspension is, from eqn (14.1.2),
(14.1.3)
so that if h.d is only two particle diameters, the behaviour of the 8.8% hard-sphere
suspension considered earlier will occur in this suspension at 4 = 0.07%. In this
example, the suspension shows concentrated behaviour, even though it contains very
little material.
We shall see in Section 14.2 that the effective hard-sphere approximation used
above is in most cases not even qualitatively correct as a means of representing
dispersions with real finite-ranged potentials. The physics displayed in the above
example is quite general, however. This may be seen by the following thought
experiment, which also allows us to quantify the notion of ‘concentrated’. By
definition, the average volume per particle is l/p, where p is the number density of
particles in the suspension. Let this volume be represented by a cube of side d, so that d
would be the average separation between particles if they were equally spaced. Then
since p = l/d3, we may write eqn (13.6.8) as (Exercise 14.1.1):
The osmotic compressibility is thus a measure of the resistance of the dispersion to any
attempt to squeeze its particles closer together, and anything which increases that
resistance will lower the compressibility. Clearly, adding a repulsive potential has this
effect, and the longer the range of the potential, the lower the density at which the
compressibility will begin to diminish. Imagine first starting with a hard-sphere
system, in which the range of the interaction is the diameter, 6, of the spheres. From
the above discussion, we expect this system to exhibit features associated with
concentrated behaviour once d becomes comparable with the range of the potential, or
when 6/d M 1. Now let the spheres develop a charge, for example by ionization, with
the resulting total ionic strength of the suspension corresponding to a Debye screening
length Ad. In this case, the onset of concentrated behaviour will occur when
(14.1.5)
Equation (14.1.5) shows that in a charged colloidal dispersion there may, in fact, be
three regimes of concentration, in each of which we may expect different behaviour: Ad
>>6, Ad 6; and Ad <<a.
INTRODUCTION I671
Fig. 14.1.1 Stereo pair showing a colloidal crystal structure with face centred cubic symmetry.
(Stereo viewers are usually available in the crystallography department.)
672 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
Fig. 14.1.2 Stereo pair showing a snapshot of a colloidal dispersion of polystyrene latex particles at
13% volume fraction in M salt; the surface potential is 50 mV. This structure is almost at the
point of crystallizing. A small increase in surface potential would convert it to the structure in
Fig. 14.1.1.
INTRODUCTION I673
How, then, can we know that Fig. 14.1.2 represents a structure with nearly as high a
degree of organization as Fig. 14.1.1?The answer is to use some method which tells us
directly about correlations in the system. One method is to construct g(r), by
measuring all possible distances r between pairs of particles in the box and plotting the
number of times a given distance occurs as a function of that distance, normalized to
4nP. (To obtain reasonable accuracy, we should either have a very large number of
particles in the box, or we should repeat the procedure for a large number of boxes
observed at different times.) The g(r) so obtained is shown in Fig. 14.1.3, which should
be compared with g(r) = 1 for a random structure. A more efficient method, applicable
to real samples, is to measure directly the amplitude density of particle-particle
separations, by performing a scattering experiment to measure the Fourier transform
(see Appendix A4) of g(r), namely S(Q) (see eqn (13.7.7)). Anticipating the results of
the next section, this quantity modulates the intensity of radiation scattered by the
dispersion, and we expect to see maximum scattered intensity at Q m 2n/ro, where YO is
the most probable pair separation between particles.
Now there is no preferred direction in our isotropic dispersion, so S(QJ will depend
only on the scattering angle, 8,and not on the scattering direction (recall that Q= (4n/h)
sin Q / 2 at wavelength A). We thus expect to see a DebyeScherrer cone of maximum
intensity, and this will indeed be observed in a normal scattering experiment. In contrast,
the crystal structure of Fig. 14.1.1 will scatter sharply defined Bragg spots in directions
determined by the orientation and symmetry of the crystal. On some distance scale,
however, there must be strong local symmetry even in the isotropic sample. Consider a
pair of particles held fixed and imagine bringing a third particle into their proximity; the
preferred position will be on the perpendicular bisector of the line joining the first two,
Fig. 14.1.3 The pair correlation function, g(r), as a function of r (in units of diameter, 6 ) calculated
from many snapshots of the dispersion shown in Fig. 14.1.2.
674 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
since the interactions are screened Coulomb (Exercise 14.1.2). It is easy to see that the
approach trajectory of a new particle entering the space near any accidental grouping of a
few particles will be strongly influenced by the instantaneous orientation of the group.
Furthermore, this may be true for many groups at any given time, but the orientation of
one group relative to another will be random (which is why the dispersion is isotropic as a
whole).
The lifetime of these locally symmetric groups is governed by the time it takes a
particle to move a distance of order the size of the cluster. In a simple (monatomic)
liquid, this is only a few picoseconds, but in a colloidal dispersion it may be many
LASER
Fig. 14.1.4 (a) The experimental apparatus of Clark et al. (1983). (b) Scattered intensity pattern
(averaged over 30 ms) observed using a large scattering volume. (c)-(e) The patterns observed at
three different times when only -25 particles were illuminated.
INTRODUCTION I675
milliseconds (Exercise 14.1.4) or even seconds. Clark et al. (1983) recognized that these
times were sufficiently long to perform a laser-light scattering measurement, and
undertook a series of experiments which demonstrate graphically the above concepts.
The experimental arrangement (Fig. 14.1.4(a)) was extremely simple: a laser was shone
on a colloidal latex dispersion which was almost at the point of crystallization, and the
scattered light was detected by a T V camera. The image was recorded on videotape,
which could be played back frame-by-frame, each frame corresponding to an average
over 30 milliseconds of observation. As expected from the previous discussion, a cone
of scattering was observed in each frame (Fig. 14.1.4(b)).
Next, the laser was focused tightly, so that only a small volume of sample, containing
about 25 latex particles, was illuminated. While the real-time image on the T V screen
was unchanged (still corresponding to the same cone of scattering), the single frame
images were dramatically different (Figs 14.1.4~-e), showing well-defined diffraction
spots being scattered from the liquid. These spots have symmetries and orientations
which vary with time, and correspond to the observation of just one fluctuating cluster
in the dispersion; they give a net cone of scattering when averaged over time (by
viewing the tape at normal speed) or over a large number of clusters (using an
unfocused laser beam). This experiment shows how a liquid structure can exhibit
different types of behaviour when viewed on different space-time scales, and
emphasizes the need to decide which features of the structure are appropriate to
observe when designing experiments to measure the structure of concentrated
systems. We shall return to this in Section 14.3, after discussing the theory necessary
for an understanding of scattering experiments.
Exercises
14.1.1 Establish eqn (14.1.4).
14.1.2 Consider three latex particles interacting through a screened Coulomb pair
potential U = [exp(-rij/hd)]/rc, where rc is the distance between particles i and j.
Fix two of the particles, and let the third move on a locus at any fixed distance
measured from the midpoint of the first two. Show that the sum of the pair
potentials is a minimum when the centres of the three particles form an isosceles
triangle, and is a maximum when the particles are colinear.
14.1.3 Use the equipartition theorem to estimate the average thermal velocity, u (m/s), of
particles moving in a given direction in a dispersion in terms of their mass, m (kg)
and the temperature, T (K).
14.1.4 Use the previous result to find the time taken by a polystyrene sphere of
diameter 234 nm to travel 1 diameter in a suspension at T = 298 K, assuming it
moves in a straight line. (Take the density of polystyrene to be 950 kg/m3.) Now
calculate the time the particle takes to travel 5.3 pm, assuming it takes a random
walk with step-length equal to: (a) the diameter; and (b) 1/4 diameter. Discuss
your answers with reference to the time-scales which occur in simple liquids,
and compare them with those observed in the experiments of Clark et al. (1983),
to which these numerical values correspond.
676 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
Since the order in which pairs are taken is irrelevant (;jis equivalent toji), there will be
s(s+1)/2 simultaneous equations to be solved for the s(s+l) correlation functions hi(+
cij(r). This is clearly not possible (since there are fewer equations than unknowns), and
+
eqns (14.2.1)must be supplemented with another set of s(s 1)/2 equations, known as
the closure conditions, before the system can be solved. [Note: the symbol d3t indicates
a vector integration in three dimensions. Exercise 13.6.3 indicates how this is done.]
The OZ equations essentially define the cc(r) in terms of the hg(r), and do not in
themselves contain any specific information about the potentials between particles in
the fluid. The need for closure relations provides the opportunity to add this
information about any particular system. Note, however, that the OZ equations do
require the number densities to be specified, and hence require as input the average
interparticle distances. Solving for h(r) will then tell us how the particle positions
fluctuate about this mean value (compare with Section 13.3).
If we could calculate, say, the cq(r) by some independent method, the hi(r) would
follow directly from eqn (14.2.1). Unfortunately, however, there is no simple physical
interpretation of the direct correlation function, cq(r), for values of r inside the particle.
Statistical mechanics at present yields an a priori estimate of this function only outside
the particle. A common feature among the various types of commonly used closure
schemes which were introduced in Section 13.8 is therefore to partition space into two
<
regions, r > and r < <,where $ is an appropriate length, and to assume some
knowledge of cij(r) in the first region and of h&) in the second. For the latter, the most
678 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
convenient assumption is to specify a distance of closest approach, say &, for any given
pair of particles; by definition we then have
For the remaining spatial region, from the choices presented in Section 13.8, we shall
restrict our discussion to the generalization of eqn (13.8.6), the mean spherical
approximation (MSA):
cq(r) = -Uq(r)/kT, forr > <q. (14.2.4)
From a purely theoretical viewpoint, there are better approximations available than the
MSA, and the choice of the latter is mainly pragmatic. The MSA (and certain other
approximations based on the MSA) may be solved analytically for several realistic
model potentials, and these analytic results have provided the first formal framework
within which to analyse scattering experiments on colloids. (The term ‘analytic’
solution is often used by liquid-state theorists to mean the reduction of the OZ integral
equation to a set of simultaneous, non-linear algebraic equations which must still be
solved numerically. We shall call the latter ‘semi-analytic’ solutions, and reserve the
term analytic for solutions specified in closed form in terms of standard functions.)
T he Percus-Yevick hard-sphere system is a special case of the MSA, in which
Uq = 0 unless the particles are in contact. Th e solution to the OZ equation for this
case has been given for monodisperse spheres in Section 13.7. An analytic solution
for a mixture of two sizes of hard spheres has been given by Lebowitz (1964) and
Ashcroft and Langreth (1967). Th e general polydisperse hard-sphere case has been
considered numerically by van Beurten and Vrij (1981). A remarkable feature of the
binary mixture case is the large effect which the presence of a small concentration
of a second component can have on the structure of a hard-sphere fluid. This is a
purely geometric effect, related to the fact that in a neutral system, a hole formed in
a cluster of large spheres may be filled by a smaller sphere.
In a dispersion of charged spheres, the smaller sphere is unlikely to approach a
cluster of similarly charged spheres in the first place, and quite different packing
results. For this and other reasons (some of which we have already considered), hard-
sphere fluids are somewhat unique, and we shall need to employ a rather more realistic
potential to handle the general case of dispersions of charged particles (or, more
succinctly, charged dispersions).
For a dispersion of uniformly charged spheres, Newton’s theorem lets us write the
electrostatic pair potentials exactly as the Coulomb interactions between equivalent
point charges, so that the MSA becomes
R EL A T IN G POTENTIALTO STRUCTURE I679
together with eqns (14.2.1-14.2.4). Here, xie is the charge on species i, and EOE, is the
dielectric permittivity of the dispersing medium. In the simplest case of (partially)
ionized monodisperse colloidal particles (z = 1) and counterions (z = 2) without added
salt, there will thus be a system of three equations to solve. This is the so-called
primitive model (PM), which may be solved semi-analytically (Blum and Heye 1977).
It shows a very interesting limiting feature (Medina-Noyola and McQuarrie 1980):
when the counterions are taken to have negligible size, the potential of mean force
(compare with Section 13.4) between two isolated colloidal particles may be explicitly
calculated as
(14.2.6)
which is just the pair potential derived initially by Verwey and Overbeek (1948) on the
basis of the Poisson-Boltzmann eqn (7.3.11) (see Exercise 14.2.3).
The above result suggests a further simplification of the system of OZ equations for
monodisperse charged dispersions. Although there are several lengths in the problem
(ad and the &), an obvious feature of the colloidal state is that the colloidal particles are
usually very much larger than any other ion in solution, so that, among the particle
dimensions, 81 is by far the most significant (the colloidal particles in a monodisperse
system will always be labelled i = 1, with z = 2 labelling the counterions and i 2 3
referring to any added salt). Whenever this is the case, we may concentrate our
attention on the particles alone, and treat them as if they were the only macroscopic
component present (Hayter and Penfold 1981a,b; Hayter 1983). This is the ‘one-
component macrofluid’ approach (OCM), in which all correlations are neglected
except those between the macroions, which are taken as interacting through the
potential of eqn (14.2.6). The counterions thus provide a neutralizing background
whose quantitative effect on the structure now appears indirectly, through the
screening length (Ad) in the potential. The main features of the OCM approach are that
it preserves those length scales which dominate the physics, and the correlations
calculated are those of most interest, namely those between the colloidal particles
themselves.
The OCM is thus equivalent to the primitive model (PM) in the limit 82 + 0, and
since it is numerically no more difficult to solve the P M for finite counterions than in
this limit (see, for example, Naegele et al. 1985), we must look at the underlying mean
spherical approximation (MSA) to see why the OCM is of interest.
Consider the apparently straightforward problem of using the MSA to determine
correlations in a dilute system. In the limit Pk + 0, eqn (14.2.1) reduces to
(14.2.7)
From eqns (14.2.4) and (14.2.5), the MSA in this case thus amounts to
so that whenever Uq(r) > KT, g&) < 0, which is not permitted for a probability
function. In the low-density limit, therefore, the MSA yields unphysical results, and it
680 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
is clear that this becomes worse as the potential becomes more strongly repulsive.
Numerical calculations show that this breakdown of the MSA may occur at volume
fractions as high as 20% in many representative charged dispersions, so that the P M is
not generally useful. We shall now see, however, that the OCM may be reformulated in
a way that avoids this problem without adding any further mathematical complications.
T o this end, we return to our earlier discussion of the scaling of physical lengths in the
system.
In the case of hard spheres, we have seen that apparently different suspensions have
equivalent structures at equal volume fractions, as long as the sphere diameter is used as
the unit of length when describing the structure. This means that, if we know g(r) for a
fluid of hard spheres of diameter 60, we may trivially calculate the structure of a fluid of
hard spheres of diameter 61 at the same volume fraction by simply rescaling g(r) to g(rSo/
61). In the case of charged hard spheres, a second length, h d , enters the physics, and as
we saw in Section 14.1 this introduces concentration regimes in which the behaviour
may cross over from being dominated by excluded volume to being dominated by
electrostatics. The former must clearly be the case at high density, simply because the
particles are crowded together and their size plays an important role.
At the other extreme of low density but long-ranged interaction, however, the size
of the particles must be irrelevant (except in so far as it influences the net charge), since
at long range the interactions will appear to be between point charges. Such a system
will have the characteristics of a one-component plasma (see, for example, Baus and
Hansen 1980), in which the density (i.e. particle concentration) rather than the particle
diameter dominates the structure, so that the appropriate length-scale for correlations
is the ion-sphere radius, a, defined by
a = 6/(2&). (14.2.10)
Consider next how the potential scales with particle size. Rewriting eqn (14.2.6) in
terms of x = r/6 and dividing by k T (omitting subscripts, which will not be needed for
the OCM), we obtain the general dimensionless form for the potential
The potential (in k T units) when two particles are in contact (x = 1) is given by
y exp (-K’). For x < 1, we may complete the potential by writing
When this form of the pair potential is plotted for several volume fractions (Fig. 14.2.1), a
remarkable feature emerges: apart from the position of the hard core, eqns (14.2.12)
describe a universal potential for an infinite family of colloidal systems having common
values of y’ and K” at their particular volume fraction.
We now ask, under what conditions will members of this family have a universal
structure (when measured with a ruler of length a (eqn (14.2.10))? This question is
equivalent to asking when the hard core can influence the structure, since all other
aspects of the potential are common to all members of the family. Now, by
‘structure’ we mean an average representation of the configurations which the
particles in the dispersion can adopt (compare with Section 13.1). Since
interparticle distances y for which the Boltzmann factor exp[-V(y)/kT] is small
are improbable, the presence of a hard core will have negligible influence on the
structure provided YO, = $‘/3) >> 1, since particles will rarely climb high enough
up the potential curve to sample it.
Fig. 14.2.1 Plot of the potential function V(y)/kTagainst y for three dispersions (volume fraction
41 > 4 2 > 43) with common values of y’ and K ” .
682 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
which simply states that counterion positions are random with respect to either
colloidal particle or other counterion positions. The main correlations of this type
present in the real system are those required to satisfy the Poisson-Boltzmann
equation, and are already included in the Debye length in the potential.
These additional closure conditions are defined throughout space, and so two of the
three OZ equations are solved (by definition). The OCM thus reduces to the MSA
defined by eqns (14.2.14) for the sole case i = j = 1. This may be solved in closed
analytic form, but still suffers from the defects of the MSA discussed earlier. Now,
however, we have a model system to which our scaling argument applies, since all
interactions are repulsive. Remembering that the MSA gives good results at high
density, we thus solve the low density case simply by constructing an equivalent high
density system, using eqn (14.2.12) to specify the potential in the new system. The
analytic solution to the MSA gives the structure accurately at the new density, and we
simply rescale the length-scale of this structure to obtain the desired result. This is the
rescaled MSA (RMSA) proposed by Hansen and Hayter (1982), and it provides a
powerful means of computing correlations in charged colloidal systems of any particle
concentration.
One final refinement is needed to relate it to real systems, however. The closure
(14.2.14) is not quite the same as taking the PM with 82 + 0, because it allows the
counterions to appear anywhere in space, including inside the colloidal particles. This
R EL A T IN G POTENTIALTO STRUCTURE I683
2.0
1.0
h
k
v
a0
-1.0
0 10 20 30 40 3
r/6
Fig. 14.2.2 Comparison of g(r) against r/6 calculated by two different methods for a latex
dispersion at volume fraction qj = 1.2 x lop4 in lop6 M salt. The latex diameter is 6 = 90 nm and
surface potential is 230 mV. Data computed by (m) Brownian dynamics; ( ) RMSA. The
simulation took about 6 hours on a desk top computer (IBM-XT), compared with 2 seconds for the
RMSA.
to allow for the charge compensation by the neutralizing background. T o keep the
structures in the effective and real systems equivalent, we must still work at fixed
Coulomb coupling (eqn (14.2.13)), which yields the effective screening corresponding
to eqn (14.2.15) (Exercise 14.2.6):
Equations (14.2.15) and (14.2.16) complete the specification of the use of the
analytic RMSA calculation to compute correlation functions in the OCM
approximation. The validity of this approximation has been thoroughly tested by
comparison with computer simulation results, which are essentially exact, and it
proves to be extremely accurate; the agreement shown in Fig. 14.2.2 is typical
throughout the entire range of particle concentrations and potentials found in
charged colloidal systems.
684 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
Exercises
14.2.1Why must van der Waals attraction always dominate screened Coulomb
repulsion at sufficiently large distances? Is this still the case for a bare
(unscreened) Coulomb potential? At large enough distances, both potentials
may be very small compared with kT,in which case the question is of little or no
significance; can you give an example of a system where it matters in practice?
14.2.2F. Zernike was awarded the Nobel Prize in Physics in 1953 for the invention of
phase-contrast microscopy, but is probably better known for his contribution to
liquid state physics. The Ornstein-Zernike eqn was published in 1914, but not
solved for any liquid until 1962. The original paper, which is reprinted in
Frisch and Lebowitz (1964), was published, inter aha, to correct some
misconceptions which Einstein had about correlations in the liquid state! Look
it up, and compare the original derivation with that outlined in Section 13.8.
14.2.3Show that eqn (14.2.6) is identical to the repulsion energy calculated by Verwey
and Overbeek (1948) for approaching spheres in the low potential
approximation.
14.2.4Derive eqn (14.2.10) from eqns (14.1.2) and (14.2.9).
14.2.5Find expressions for y (equivalent to eqn 14.2.11b) when the inter-particle
potential is expressed in terms of (a) the surface potential or (b) the surface
charge density, rather than the total charge.
14.2.6Use eqns (14.2.10), (14.2.11(a)), and (14.2.13) to show that eqn (14.2.16) is
implied by eqn (14.2.15) when the coupling, r, is held fixed.
major problem in dense samples, unless they are so thin that they may exhibit artefacts
induced by interactions with the walls of the container, and hence no longer represent
the bulk dispersion. Since the mid-l960s, therefore, concentrated colloidal dispersions
have been increasingly studied by a form of radiation which appears to interact only
very weakly with matter, namely neutrons.
Neutron scattering has proved to be such an important tool in materials science in
general that most neutron scattering centres in the world now make their facilities
available to scientists from other laboratories on a routine basis, and neutrons are more
accessible to the colloid chemist than, for example, synchrotron X-rays. The technique
is, however, much less readily available than a light scattering or X-ray camera in one’s
home laboratory, and its popularity is due to several unique scientific features which
we shall explore in the remainder of this chapter.
Neutrons are produced by well-known processes in reactors (fission) or
accelerators (spallation), and the interested reader is referred to any nuclear
physics text for details. As produced, they are too energetic for use in scattering
experiments, and must be slowed down in a moderator until their kinetic energy
corresponds to about room temperature (thermal neutrons) or less (cold neutrons),
at which point their de Broglie wavelengths h = h/mnv are comparable with typical
interatomic distances (Exercise 14.3.1); here, h is the Planck constant, m, is the
neutron mass, and v the neutron velocity. In the study of colloids, we shall be
interested in correlations over many atomic distances, or conversely in Fourier
components of the structure having small Qvalues, so we shall mostly be interested
in cold neutron scattering. Th e neutron has a magnetic moment, but no measurable
charge or electric dipole moment, so it has negligible interaction with electrons
unless they are unpaired. Th e only scattering we will consider will be due to
interactions with the nuclei of the atoms in the dispersion. These interactions are in
fact very strong, but of very much shorter range m) than interatomic
distances.
Now, it should already be clear from the previous discussion (Sections 14.2 and 14.3
and Chapter 13) that scattering is concerned with phase shifts between scattered
waves, and the way in which waves of various phases arrive at different atoms in the
sample. This means that we do not need to know the amplitude and phase of any wave
right at the point of scattering, as long as we can calculate these quantities when the
wave arrives at another atom. In the case of neutrons, this is at a large distance from the
point of scattering, compared with the range of the scattering interaction. We may thus
use a device which is common in scattering theory. We replace the real interaction
potential which scatters the neutron by any fictitious potential which gives the correct
answer for scattered amplitude and phase at any distance which will appear in the
scattering calculation. Such potentials are known as pseudopotentials, and the
appropriate pseudopotential for neutron scattering is
(14.3.1)
where fi = h/2n, Rj is the position of thejth nucleus, and the amplitude bj has the
dimensions of length. (Note that 6 is here the Dirac delta function as defined in
Chapter 11 (eqn (11.7.3)).)
686 I 14:SCATTERING STUDIES OF C O L L O I D STRUCTURE
Table 14.1
Neutron scattering amplitudes, b, for selected nuclei. (See Lovesey 1984 for a more extensive listing.)
Element ‘H *D C N 0 Na A1 Si C1 Ti
b(lOPl5m) -3.741 6.674 6.648 9.36 5.805 3.63 3.449 4.149 9.579 -3.30
Table 14.1 summarizes some neutron scattering amplitudes for nuclei typically
found in colloidal dispersions. Two features are noteworthy. First, there is no
systematic variation with atomic number (unlike the case for X-rays), and light atoms
may be as easily studied as heavy atoms. Second, certain values are negative, so that
cancellation of amplitudes may occur. (Recalling that exp(in) = -1, a negative value
merely indicates a phase shift of n on scattering.)
The refractive index, n, of any material for a particular radiation is defined as the
ratio of the momentum of the radiation in vacuo, po, to its momentum, p , in the bulk
material. Since po = mnvO and the corresponding kinetic energy is m,vi/2, we may
write the momentum in vucuo in terms of the kinetic energy: po = d 2 m n E o ) .At any
+
point, r, in the material, the energy will be EO U(r),where U(r)is given by eqn
(14.3.1). We may calculate the mean value, U, of this potential by averaging over the
volume, V, of the sample:
U = (1/V)
s U(r)d3r = -x-
2d2
mn
bj
where the integration follows immediately from the properties of the &function. The
(14.3.2)
quantity Cbj/V is the scattering amplitude density, B, in the material (Exercise 14.3.3)
and corresponds to electron density in light or X-ray scattering. Clearly, the sum need
only be evaluated over a representative formula unit, using the corresponding volume,
V = 1/ N , where N is the number density of such units in the material, so we may write
B = N C b, provided it is understood that the sum is over atoms in the formula unit
whose density is N .
Since the scattering amplitude of vacuum is null, we thus have
Note that an- 1 is very small [since U << Eo, and may have either sign (compare with
Exercise 14.3.3)].
In terms of the de Broglie wavelength, A, an equivalent form for the refractive index
is :
n, = 1 - BA2/n. (14.3.4)
Two materials with scattering amplitude densities B1 and B2 will have the same
refractive index if B1 = B2. The difference B1 - B2 is called the contrast between the
two materials, and refractive index matching thus corresponds to null contrast. This is
USE O F SCATTERING TO MEASURE STRUCTURE I687
Values for water, DzO, and a number of polymers are given in Table 14.2.
Thermal neutrons, by definition, have energies comparable with those of dynamic
thermal excitations in the sample, and a major use of neutron scattering is to measure
such excitations. An introduction to the measurement of time-dependent phenomena
with neutrons is given by Lovesey (1977). In this chapter, we shall only be concerned
Fig. 14.3.1 The rotating helical monochromator. Only neutrons with a particular velocity can
traverse the slot and they form the incident beam.
+ Often called the scattering length density. We consider only the coherent scattering.
An incoherent scattering component is also present but it is isotropic and can be
subtracted from the scattered signal (eqn 14.3.10). It is cause by the presence of
different nuclear isotopes of the component atoms which have different nuclear cross-
section.
K Lower monochromator
housing
Upper monochromator
housing
Graphite
monochromatingcrystals
Cold beryllium filter
J
Collimator and neutron
beam guide
Sample chamber
B 58 cm gate valve
20 m by 152 cm dim
vacuum flight path
28 cm wood shielding
Detector carriage
112 cm diam two-
dimentional position
sensitive detector (64 by
64 cm active area)
Data acquisition system
Fig. 14.3.2 T h e 30 m SANS camera layout at the High Flux Isotope Reactor (HFIR) at Oak Ridge National Laboratory. Neutrons from the reactor (off the
figure at top left) are reflected by two sets of graphite crystal monochromators A and B into the neutron guide/collimator system, C and then onto the sample at
F. Scattered neutrons are detected by the 64 x 64 cm2 two-dimensional position-sensitive detector K, which is mounted on a chariot, J. The chariot moves on
rails in the evacuated tube H allowing the detector to be positioned anywhere from 1.5 m to 20 m from the sample.
USE O F SCATTERING TO MEASURE STRUCTURE I689
with time-averaged structures, and hence with static experiments, in which no energy
analysis is performed.
Neutrons of a particular wavelength [or velocity] are selected from a beam, either by
Bragg reflection from a suitable crystal (such as graphite) or by passage through a
rotating helix whose speed and pitch are adjusted to allow transmission only of
neutrons of the desired velocity (Fig. 14.3.1). The monochromatic beam is then
collimated by appropriate slits or diaphragms before impinging on the sample, and the
intensity scattered in a given direction is measured by a detector of known efficiency.
In the case of scattering from colloids, we are interested in scattering at small Q
(corresponding to large distances), which means small angles and long wavelengths.
Typically, 0.4 CA/(nm) C 1.2 and 0.05 C B/(deg) C 5.0, corresponding to 0.01 C Q/
(nm-’) C 3. This type of experiment is known as small angle neutron scattering or
SANS. SANS cameras of all types share several common features, exemplified in the
practical instrument shown in Fig. 14.3.2.
The basic optical geometry is that of a pinhole camera, in which the source (usually a
circular hole at the entrance to the collimating system) is imaged on a two-dimensional
position-sensitive detector through the ‘pinhole’ at the sample position. The latter is
usually at least 1cm in diameter, to allow studies of bulk samples, and the overall size of
the instrument is sufficiently large (several tens of metres long) to maintain apparent
pinhole focusing. The beam is transported in vacuo where possible, to minimize air
scattering, but windows may be used as dictated by different sample containment
requirements, since many common materials (e.g. quartz or aluminium) are effectively
totally transparent to neutrons of long wavelength.
The sample-to-detector distance may be varied by moving the detector on rails
inside the evacuated tube after the sample. Optimal focussing requires a fixed ratio
between the source-to-sample and sample-to-detector distances. This is maintained by
transporting neutrons to the source via a variable-length neutron guide, which
operates in a manner exactly analogous to a light-pipe. Reference to eqn (14.3.4) shows
that the neutron refractive index is less than unity for any material with a net positive
mean scattering amplitude density, so that vacuum is optically dense relative to such a
material. In this case, total external reflection is possible at the interface and neutrons
are transported inside the evacuated guide tube by successive reflections.
One of the most important features of SANS instruments is that they are situated
at large central facilities, and are used on a shared basis by many groups, most of
whom are not specialists in scattering. This has led to the development of highly
automated equipment, with standard procedures for correcting and normalizing the
data, so that the outcome of a SANS measurement is the absolutely scaled intensity,
I(!.?),scattered coherently by correlations in the sample being studied. We may thus,
without loss, bypass technical details of the corrections (in particular, the removal of
solvent scattering and the corrections for single-atom, or incoherent, effects), and
proceed directly to a consideration of the information obtainable from the corrected
coherent intensity, I(!.?),which corresponds to the scattering from the colloidal
particles alone.
We have seen that the refractive index, which corresponds to scattering at Q = 0,
a
depends on the quantity N C b. At finite we must allow for the phase at each atom
when summing the scattered amplitudes. The square of the resultant total amplitude is
the scattered intensity (compare with Section 13.7) for any given configuration of the
690 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
Table 14.2
Coherent neutron scattering amplitude (length) densitiesfor various substances.
Molecule Water D20 h-PS d-PS h-PMMA d-PMMA h-PAN Cl2H26 C12D26
[B/lO"]mP2 -0.56 6.40 1.42 6.47 1.07 7.03 2.27 -0.46 6.71
atoms. Averaging over all possible atomic configurations then yields the measured
intensity (Exercise 14.3.7):
(14.3.5)
(14.3.6)
Here, Rj is the position vector of thejth atom in the sample, and ( ) represents the
configurational(ensemble) average. Equation (14.3.6) shows that whenever atomsj and
k are well-separated, the phase factor will oscillate through many cycles and will
average to zero over small variations in R, unless I Q I is small; conversely, the phase
factor will remain essentially unity (the refractive limit) at small Q unless large
correlation distances are involved. SANS thus has the desirable feature of being
sensitive to the colloidal structure while not inundating us with detail on the scale of
atomic distances (which we usually know from separate, high Q diffraction
experiments).
We are considering scattering by the colloidal particles alone, the solvent scattering
having been measured separately and subtracted from the data. (We shall return to this
point shortly.) The double sum in eqn (14.3.6) is thus over all possible pairs of atoms in
the dispersed particles. Any given pair of atoms either lies in the same colloidal
particle, say particle n, or there is one atom from the pair in each of two colloidal
particles, say particles n and m. The summation can be broken up into a sum term
involving atoms of n and another term involving the ensemble average of scattering
from pairs of particles with different centre-to-centre distances. The resulting
scattering thus depends intimately on the probability of a given interparticle distance,
and hence on the pair distribution function, g(R,) (Sections 13.8 and 14.2).
Before further discussion of this point, however, it is worth examining that part of
the scattering which does not depend on interparticle correlations. T o this end, let us
consider the dilute case, when particles n and m are uncorrelated, so that only the
scattering from a single particle need be considered. Then [compare with eqn (14.3.6)]
USE O F SCATTERING TO MEASURE STRUCTURE I691
The form factor gives the amplitude and phase of a wave scattered by a given particle.
The scattered intensity is thus the square of the modulus of the form factor, and the
total scattered intensity in this case is just the sum of the intensities scattered by the
individual particles. Form factors are readily calculable for various common
geometrical shapes (Exercise 14.3.8).
Since SANS is insensitive to detail on the atomic scale, individual atomic scattering
lengths may be replaced by local averages over, for example, functional chemical
groups of several atoms.
This lets us work in terms of the local scattering amplitude density B(r)at position r
in the sample, and has the considerable advantage that we need only know the local
chemical composition, without knowing the detailed local atomic structure. In this
case, it is convenient to replace the sum over atoms in a particle by an integration of the
scattering amplitude density over the volume of the particle, so that we shall generally
write
s
F(Q) = [B(r)- Bm]exp(zQ.r)d3r (14.3.9)
where we have now explicitly allowed for the fact that scattering from the same volume
of the suspension medium (of scattering amplitude density B,) has been subtracted.
The form factor will in general depend on the relative orientation of the particle and
the momentum transfer, Q. When B(r) depends only on Y = Irl we can use the
technique outlined in Exercise (13.6.3) to transform eqn (14.3.9) into the Debye result,
familiar in the theory of crystallography (see Atkins 1982 p. 755) (Exercise 14.3.8):
sin(@)
[B(r) - Bm]r2- dr. (14.3.10)
Qr
In the particular case where the particle is a sphere of uniform scattering amplitude
density, Bo, the integral can be evaluated explicitly to give (Exercise 14.3.8(b)):
where Y is the volume of the sphere. In the case of scattering from an isotropic
dispersion, the intensity must also be averaged over all possible particle orientations:
I(Q)= 1 4n 1
2n n
sI(Q)sinB dQd$
(14.3.12)
0 0
where (B,$) are the polar angles defining the particle orientation relative to Q. (This is
irrelevant for spheres, of course, since they have no orientation.)
692 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
The form factor used in eqn (14.3.11) is related to the earlier form factor introduced
in connection with light scattering (Section 3.3) simply by P = F2 and so we see the
relation between this and the earlier eqn (3.3.14). Some information about the shape of
anisotropic particles is lost in the angular averaging, but that can be avoided. The
analogue of the Rayleigh-Gans-Debye equation for neutron scattering is then:
where the B factor takes the place of the refractive index [eqn (14.3.4)] and Npis the
number of particles of volume V,.
If a model for the particle shape is available, scattering measurements on a dilute
dispersion will provide detailed quantitative information on the geometry. For
cylinders of length 21 aligned parallel to the direction of Q for example the form factor
is
where 4 is the particle volume fraction. The value of B, can be varied over a wide
range using mixtures of D20 and H20 and using eqn (14.3.15) in the form [compare
Cebula et al. (1978)l:
we see that a plot of J I ( 0 )against B, (Fig. 14.3.3) will pass through zero at the point
where B, = Bp allowing us to check experimentally the estimates in Table 14.2. The
slope and intercept of these plots is given by -[$A V,]Jand Bp [4AVp]4 respectively, so
the ratio:
which gives a further check on Bp.Furthermore, if the instrument constant A, and the
volume fraction are known then obviously one can evaluate V, and, hence, the radius
of the particles. An alternative formulation allows one to calculate the molecular weight
(via the known chemical composition and scattering amplitude values). For this reason,
the ability to measure absolute intensities easily is an important feature of SANS.
A second general result for dilute systems is Guinier’s law which we met in
Section 3.3 in connection with light scattering. In the limit of small Q (see
Exercises 3.3.4 and 3.3.5):
USE O F SCATTERING TO MEASURE STRUCTURE I693
\Hydrogenated PS
I I
0 1 2 3 4 5 6 7
B, W 2 ) xi014
Fig. 14.3.3 The function d I ( 0 )plotted against scattering length density of the dispersion medium,
B, for polystyrene and deuterated polystyrene in H20/D20 mixtures. (After Ottewill.)
Thus a characteristic dimension of the particle may be obtained from the initial slope of
a plot of ln[l(QJ] vs @. (This is called a Guinier plot as we noted in Section 3.3.)
(4 (b)
Fig. 14.3.4 Illustrating the technique of contrast matching. (a) B, = Bp (b) B, = B, (c) B, # B,
# B,. Subscript is medium, ,is particle, and is shell.
where Vp is the volume of the core particle, Vis the total volume of the core plus shell,
Bp, B,, and B, are respectively the scattering length densities of the core particle,
shell, and solvent medium, and (from eqn.14.3.11):
which reduces to eqn (14.3.15) for [f(Q=O)] = 1. The same result is obtained ifwe set
B, = B,. Thus if the shell and the solvent are matched we see only the central core
particle.
If the core is contrast matched by setting Bp = B, we obtain:
(14.3.22)
This formulation is particularly useful for studying the adsorption of surfactants onto
the surface of polymer latex particles which we will now examine.
mixed systems, since as Q+ 0 the functionf(x) + 1. Combining this result with that
in equation (14.3.23) gives:
V, = (b - a)S, (14.3.25)
where b is the (core + shell) radius and S, is the surface area of the particle. We can also
set B, = b,/ V, where b, is the scattering length density of the adsorbed molecule and
V, is its molar volume. The number of adsorbed molecules per particle, n is V,/ V,
and if r is the number adsorbed per unit area then:
1 2 3
~2 I 10-4 A-2
Fig. 14.3.5 Guinier plots of (a) bare polystyrene latex; (b) latex at 4 x 10p3M total d23-
dodecanoate concentration; (c) latex at 1.2 x 10p2M total dz3-dodecanoate concentration. (After
Harris et al. 1983.)
696 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
whilst the corresponding scattering function for the uncoated particle is, from eqn
(14.3.24):
'
'-
T - (%)I=AN, BpVp. (14.3.30)
(14.3.31)
This relation has been used by Harris et al. (1983) to measure the adsorption of d23-
dodecanoate ions onto a polystyrene particle at pH 8.1. A 0.2% dispersion of the latex
Fig. 14.3.6 Adsorption isotherm of d23-dodecanoic acid on polystyrenelatex at pH 8.1 and 22" C at
a total electrolyte concentrationof 2.2 x lO-*M. A: Calculated by fitting a spherical cell model, with
error bars. 0 :Data obtained by Guinier analysis (Fig. 14.3.5). (After Harris et al. 1983.)
USE O F SCATTERING TO MEASURE STRUCTURE I697
was made up in 8% D20/92’/0 H20 (for which B, = 0). The scattering behaviour at
low angle was as shown in Fig. 14.3.5 and by using the Guinier Law it was possible to
extract I(0)values as a function of added surfactant. T h e resulting isotherm is shown
in Fig. 14.3.6.
Exercises.
14.3.1Given the neutron rest mass is m, = 1.67494 x lopz7kg, calculate the velocity
of a neutron whose thermal energy corresponds to a temperature (a) T = 293 K,
(b) T = 20 K. Are relativistic effects important? Calculate the de Broglie
wavelengths corresponding to these velocities, and compare the neutron
energies with those of X-ray photons at the same wavelengths.
14.3.2Calculate the electron densities (e-/m3) in (a) CsH170H, (b) CsH18, (c)
CzHsOH, (d) A1203,(e) SiO2, (f) H20, and (g) D20. Take the respective
densities as 825, 704, 790, 3970, 2655, 1000, and 1100 kg/m3.
14.3.3Use Table 14.1 to calculate the neutron scattering amplitude densities (m/m3)
corresponding to the electron densities found in the previous exercise.
Compare the relative light scattering contrasts (scattering density differences)
between the various materials with those for neutron scattering. How does the
mean potential (14.3.2) compare with kT for each material at room
temperature?
14.3.4Use the results of Exercises 14.3.3 to evaluate eqn (14.3.9) at Q= 0 for a
uniform silica particle in H20. Show how a measurement of this intensity can
be used to find the mass of the particle. Does the result depend on particle
shape? [Note that this procedure can be used to evaluate the molar mass of any
particle, e.g. a protein in dilute solution.]
14.3.5What mixture of H20 and D20 would make the A1203 invisible to neutrons in a
mixed A1203 dispersion, so that SiOz-Si02 correlations could be studied
directly? (Assume ideal mixing and ignore hydration effects.) Is there an
equivalent solvent mixture for light scattering?
14.3.6The refractive index, normally thought of as a ‘wave’ property, was derived in
eqn (14.3.3) using a purely particle kinetic energy. See, for example, Bacon
(1975) for a derivation based entirely on wave mechanics. (This is a rather
striking example of wave -particle duality.) Many examples of ‘neutron optical’
experiments, such as interferometry, are given by Klein and Werner (1983).
14.3.7Satisfy yourself that the eqns. (14.3.5) and (14.3.6) are equivalent.
14.3.8(a) Show that, when B(r) depends only on Y = Irl ,eqn (14.3.8) reduces to the
Debye result
s
F ( P ) = 4n [B(r)- Bm]3sin (Qr)/(Qr) dr.
[Hint: Choose spherical polar coordinates centred on the particle and evaluate
the anpdar integrals using the hints suggested in Exercise (13.6.3).]
698 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
(b) Use part (a) to derive the form factor for a sphere of radius a and uniform
scattering amplitude density Bo:
F(Q) = 3 V(BO - B,)[sin(Qu) - ~u cos(~a)]/(~a)~
where the form factors now only depend on the magnitude of Q since the particles are
isotropic. If all the particles are not the same size, there will be an obvious correlation
between the distance of closest approach of any pair of particles and their form factors,
which depend on the particle sizes. Calculation of the ensemble average will require
knowledge of all of the partial pair distribution functions g,, (compare with eqns
14.2.2-3)). This calculation is possible numerically for hard sphere potentials (see van
Beurten and Vrij 1981), but becomes intractable for more general potentials. If the
dispersion is monodisperse (or may be approximated as monodisperse), however,
Fn(QJ no longer depends on n, so that eqn (14.4.1) reduces to
where F(QJ is the form factor of any sphere. The expression in [ ] has already been
discussed at length in Section 13.7; it is the static structure factor, S ( g , for the
monodisperse, isotropic dispersion. Thus the scattered intensity in this very particular
case reduces to
In the case of a non-interacting (dilute) dispersion, S(QJ = 1, and this result reduces to
the monodisperse form of eqn (14.3.7), as expected. In the general case, the Q + 0
limit is of special interest. Recalling the discussion at the start of this chapter, and the
relation (13.7.8) between osmotic compressibility and S(O), the sign of the mean
STRUCTURE OF CONCENTRATED ISOTROPICDISPERSIONSOF SPHERICAL PARTICLES I 699
interparticle potential may be obtained by comparing the measured scattered intensity
from a concentrated suspension with that calculated from a measurement on a dilute
dispersion, scaled by the concentration ratio. The scattering from the concentrated
system will be lower than the scaled value if the potential of mean force is repulsive,
and higher if it is attractive.
S(Dfor other values of Q in a monodisperse system can readily be evaluated in a
similar way. Since S(@ + 1 in dilute systems, we can use eqn (14.3.13) to write for
concentrated and dilute dispersions:
(14.4.6)
Figure 14.4.l(a), which shows S(@ for a number of charged latex dispersions,
illustrates these features. The data were obtained from SANS intensities via eqn
(14.4.3). In Fig. 14.4.l(a), the Debye length and particle size are held constant, while
the dispersed volume fraction is increased from 1% to 13%. As anticipated in the
presence of a repulsive interaction, the compressibility [reflected in S(O)]diminishes
with increasing concentration, while the degree of structural order [as evidenced by
the height of the peak in S(@] increases. Similar effects are observed when the range
of the repulsion is increased (by lowering the ionic strength) at constant
concentration (Fig. 14.4.1(b)).
Fig. 14.4.1 (a) Plot of S(QJ against Qfor a polystyrene latex in lop4 mol dmp3 NaCl solution, at
volume fractions of: ( 0 )1%; (0) 4%; (A)8%; and (0) 13% (Goodwin et al. 1985).
700 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
Fig. 14.4.1 (b) Plot of S(QJ against Q for a polystyrene latex of 4% volume fraction, at ionic
strengths corresponding to: ( 0 )5 mM NaCI; (A) 1 mM NaU, (0) deionized suspension (Goodwin
et al. 1985).
The opposite effect may be seen in Fig. 14.4.2, which plots the observed SANS
intensity from a non-ionic micellar dispersion, in which the short-ranged interaction
between particles becomes increasingly attractive as the temperature is raised.
Reference to eqn (14.4.3) shows that the scattering at high Q will become the same as
I - - - L
Fig. 14.4.2 Plot of scattered intensity against Q for a dispersion of non-ionic micelles at the
indicated temperatures. The solid lines are fits to eqn (14.4.3) using an attractive short-ranged
potential to generate S(QJ(Hayter and Zulauf 1982).
STRUCTURE OF CONCENTRATED ISOTROPICDISPERSIONSOF SPHERICAL PARTICLES I 701
that from a non-interacting system, since S(QJ then tends to unity; this is evident in the
fact that the scattering becomes independent of the temperature as Qincreases. At low
on the other hand, increasing the attraction by raising the temperature leads to
much higher forward scattering, since the compressibility increases. (The uppermost
curve is very close to the lower consolute temperature, at which the phase transition
will be accompanied by infinite compressibility).
It should be emphasized that eqn (14.4.3), which is exact, is valid only under the
restrictive conditions posed, namely monodisperse, spherical particles, and is usually
not even a rough approximation otherwise. The reason is that we have used the
condition of spherical symmetry in its derivation, and ‘slightly broken’ symmetry is
usually meaningless.
There is one case, however, in which an approximation with some physical validity
is possible, and that is the case of a dilute dispersion of charged particles at low ionic
strength. The average interparticle distance in such a dispersion is large, relative to the
particle diameter, because of the long range of the potential under conditions of low
screening (see Section 14.1). Since the interaction between charged spheres at a
distance is identical to the interaction between equivalent point charges, and the
spheres in this case will effectively never come into contact, this type of suspension is
in the plasma-like limit (discussed in Section 14.2). If the charge on each sphere were
independent of diameter, the diameter would not be a relevant physical length in the
problem. In this limit, we may use the so-called decoupling approximation.
The basis of the decoupling approximation is to assume that the correlations in eqn
(14.4.1) do not depend on the particular pair of particles involved (i.e. that gmn(r)does
not depend on rn, n). In turn, this means that the form factors are uncorrelated, and so
they may be averaged over sizes, independently of the centre-to-centre correlations
involving R,,. Denoting a size average by a horizontal bar, eqn (14.4.1) becomes in this
approximation
(14.4.7)
which again reduces to the size-averaged dilute result when S(QJ = 1. Analytic
expressions are available (Hayter 1985) for the size-averaged form factors for sphere/
shell particles in the case of a Schulz size distribution (eqn (5.4.10)). In physical terms,
the decoupling approximation is equivalent to constructing snapshots of the dispersion
by the following two-step process: (i) place points in space such that their pair
correlation function is the same as would be calculated for a monodisperse suspension
of spheres of the average size; (ii) randomly place the actual polydisperse spheres on
these points. It is easy to see that the procedure is reasonable only if there is effectively
no chance of placing two spheres such that they overlap, so that it can only be
attempted under conditions of low particle concentration and long ranged repulsive
interaction. Even when these conditions are satisfied, it must still be remembered that
the arguments leading to it are based on the assumption that the diameter plays no role.
How good an assumption this is depends on how much the interparticle potential
varies with the particle diameters.
702 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
so that .Ye,(&)) may be obtained by measuring the intensity scattered per particle by
concentrated and dilute dispersions, respectively, and taking the ratio [compare eqn
(14.4.6)].
Consider the data of Fig. 14.4.3,which shows Se,(QJ measured at two volume
fractions of a dilute latex dispersion at low ionic strength. (To emphasize our earlier
comments about the universality of the general scattering formalism these are in fact
light scattering data, due to Brown et al. 1975.) The ionic strength and volume
fractions are known experimental parameters, and the surface potential in this system
was evaluated by extensive computer simulation modelling of the data (van Megan and
Snook 1977). Further, the particle size distribution was obtained from electron
microscopy, so that all parameters of the RMSA are known. The solid lines are direct
RMSA calculations of S(QJ,assuming a monodisperse system of the mean particle size
and surface potential. The data at low Qare not well predicted. These dispersions are
precisely those where the decoupling approximation may be expected to work, but
before applying it we must decide how to choose the mean interaction potential which
will be used to calculate S(QJ for the equivalent monodisperse system in eqn (14.4.8).
From Section 14.2we expect the polydisperse and equivalent monodisperse structures
to correspond when the Coulomb coupling, r, is the same in the two systems. Under
conditions of small screening (K M 0), eqns (14.2.11)and (14.2.13) reduce to three choices
for the size- averaged coupling, depending on whether we average at constant charge
density, 00, constant surface potential, $0, or constant total charge, Z:
-
r c( Z2 [T c( 22/s] Constant total charge (14.4.94
-
r o< $is2 [T c( +is] Constant surface potential (14.4.9b)
-
r IX [T c( 00231 Constant chargedensity (14.4.96)
s
where is the mean particle diameter. The potential used to calculate S(QJ in eqn
(14.4.8)thus depends on different moments of the size distribution, depending on
which aspect of the potential is held constant during the interaction. The best
match to the data of Fig. 14.4.3 is obtained using eqn (14.4.96)in the decoupling
approximation (dashed curve). Since there are no free parameters, this is an
experimental indication that these latex systems are characterized by constant
surface charge density.
There are very few polydisperse systems for which the decoupling approximation is
a reasonable approximation, and it should only be applied with extreme care, keeping
STRUCTURE OF CONCENTRATED ISOTROPICDISPERSIONSOF SPHERICAL PARTICLES I 703
Fig. 14.4.3 S,c (@ measured by light scattering for deionized latex dispersionsat volume fractions:
(a) 1.66 x lop4; (b) 4.85 x lop4 (Brown et al. 1975). The mean particle diameter is 90 nm, with a
standard deviation of 19%. ( ~ )Calculated S(@ assuming a monodisperse suspension.
(- - --) calculated in the decoupling approximation at constant surface charge density.
in mind the conditions of its derivation. For systems which are (at least nearly)
monodisperse, however, eqn (14.4.3) gives an excellent description of the scattering
from a wide variety of concentrated colloidal dispersions, and allows direct observation
of the dispersion structure factors (eqn 14.4.6).
Mixed monodisperse colloids of different sizes provide a particular limiting case of
polydispersity. In the special case of a binary mixture of colloidal particles, the
reduction of eqn (14.4.1) to an analytic form analogous to eqn (14.4.3) has been given
by Ashcroft and Langreth (1967). Designating the smaller particles as type 1 (diameter
Sl), and the larger as type 2 (diameter 62), the result is
(14.4.10)
where x is the relative number concentration of particles of type 2, F,(QJ is the form
factor for a particle of typej (j= 1,2), and the Sg (QJ are the partial structure factors
for correlations between particles of type i and typej, corresponding to the partial pair
correlation functions hy(r) of eqn (14.2.1). The Sg (a
may be evaluated from eqns
(14.2.1-4) in the case of hard sphere potentials, using a solution due to Lebowitz
704 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
(1964) (see also Baxter 1970), so that eqn (14.4.10) is fully analytic in the case of neutral
colloids.
The most remarkable feature of these mixed systems is the large effect which a small
amount of a second component may have on the scattering from the first, especially at
low Q(Fig. 14.4.4(a)). This is particularly pronounced in the case of neutral colloids,
where very strong cross-correlations may develop between the sizes and positions of
the particles. This is evident in Fig. 14.4.4(b), which shows that the addition of the
1.2
0.8
h
0)
v
0.4
0
L /
/-
-0.41-< I I I I I I I I
0 1 2 3 4 (x 10-2)
Q
Fig. 14.4.4 (a) Plot of I(QJ against Q for a 30% volume fraction dispersion of 25 nm particles
( ~ ), compared with that from the same dispersion containing an added 2% volume fraction of
50 nm particles (- - --). (b) Plot of the particle structure factors corresponding to the mixture
shown in (a) the 25 nm particles are labelled type 1 and the 50 nm, type 2. Sll(QJ is shown both in the
presence (A) and in the absence (B) of the second component.
NEUTRON REFLECTIVITY I705
second component of Fig. 14.4.4(a) has very little effect on the correlations (Sll)which
were present between particles in the original (unperturbed) structure, and that the
added particles remain almost uncorrelated with one another (SZZ), but that strong
correlations (S12) develop between the two different types of particle. One would
expect the small particles to tend to fit into the spaces between the larger ones.
We have discussed scattering techniques which are either time independent, or
which observe structural features which vary on macroscopic time scales, for
example under the influence of viscous shear. The extension of SANS to time-
dependent measurements on the microscopic (t > 1 ns) timescale has become
possible in recent years using a technique known as neutron spin-echo (NSE),
which is the neutron analogue of photon correlation spectroscopy. T h e interested
reader is referred to Hayter (1985) for an introduction to the application of NSE to
colloidal systems.
DzO. Reflection is therefore only possible at glancing incidence and much of the beam
still goes into the substrate from which some neutrons will be scattered back into the
detector. The background scattering is, fortunately, almost independent of Q and can
be readily corrected for. Figure 14.5.1 shows the calculated reflectivity of 0.4 nm
wavelength neutrons from a 2 nm (full line) and a 5 nm surfactant layer in a system
where the contrast has been adjusted so that only the surfactant is visible to the
0 2 4 6
Distance normal to surfacehm
Fig. 14.5.1 (a) Calculated neutron reflectivity (continuous line) for a typical 2 nm thick surfactant
layer at the air-water interface (h. = 0.4 nm). The horizontal dashed line is the background scattered
from the bulk solution. The dashed reflectivity profile is for a 5 nm layer of constant scattering length
density (Fig. 14.5.1 (b)). The underlying liquid has been adjusted to zero reflectivity [the NRW (null
reflecting water) condition]; this makes the air-water interface invisible but some scattering still
occurs from the bulk water. (From Thomas 1999 with permission.)
NEUTRON REFLECTIVITY I707
neutrons. Note the constancy of the background scattering and the interference effects
for the thicker layer.
T he advantage of observing the interference fringes directly is that the thickness
of the layer can be determined directly from eqn (14.5.1) independently of any
information about the composition of the layer. Even for the thinner (-2 nm) films
the shape of the reflectivity profile can be used to estimate the ‘thickness’ of the
adsorbed film but the method is unable to distinguish between different atom
density profiles (say, the constant density profile of Fig. 14.5.1 and a Gaussian) until
the thickness exceeds about 3 nm.
The striking ability of the method to distinguish different aspects of the adsorbed
layer is shown in Fig. 14.5.2. The reflectivity profile from a surface layer of fully
deuterated CTAB is there compared with profiles from CTAB in which deuterium
replaces hydrogen in the head group, and in the chain. In this way one can determine
the thickness of the chain, the head group and the overall molecule. The values
obtained (1.65, 1.4, and 1.9 nm, respectively) are far from simply additive, suggesting
that the structure differs from the simplistic picture usually displayed. Thomas shows
how Lu e t al. (1995) used progressive substitution of the chain with deuterium to
provide a more plausible description involving the notion of ‘roughness’ of the
adsorbed film, with the chains tilted from the vertical (but see Section 14.5.2). This
sort of technique can obviously also be applied to study the interaction between a
surfactant and its co- surfactant or between two different surfactants using differential
deuterium labelling.
Fig. 14.5.2 Neutron reflectivity profiles of three labelled cetyl trimethylammonium bromides:(0)
[C16D33N(CD3)31+Brp; (+) [C16D33N(CH3)31+Brp; and (X) [C16H33N(CD3)31+ Brp. All
measurements done in NRW with the surfactants at their c.m.c. (Data from Lu et al. 1994).
708 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
(14.5.2)
where, in this case, the factor, F, is the Fourier transform of the mean scattering length
density distribution perpendicular to the interface (compare eqn 14.3.7):
F(Q) = 1
00
-ca
B(z) exp(-zQz)dz. (14.5.3)
For a uniform monolayer of surfactant (s) between media 1 and 2 this becomes:
R = [16d/Q4]{(& - +
B I ) ~ (B2 - B,)2
+2(B, - Bl)(B2 - B,) cos Qd} (14.5.4)
where d is the layer thickness. If the two media are matched by deuterium substitution
to the same refractive index (B1 = B2 = B) this reduces to (Exercise 14.5.1):
with b the scattering amplitude and p the number density of atoms; r is, as before, the
surface coverage. This result makes it possible to derive the surface coverage
independently of the thickness, by extrapolating the reflectivity function @R to Q= 0.
The theory behind the analysis of the deuterated CTAB example shown in
Fig. 14.5.2 follows along similar lines. Consider a molecule which is made up of two
parts, A and B which can be separately labelled. The molecule is assumed to be
adsorbed at the water (W) surface. The scattering length density profile along the
normal to the surface can be written:
NEUTRON REFLECTIVITY I709
and substituting this into eqn (14.5.2) gives, for the reflectivity (Thomas 1999):
where the h factors are the self and partial structure factors, defined by:
and, again, the F factors are the one-dimensional Fourier transforms of the number
densities along the normal to the surface. Then the reflectivities of each of the labelled
species: dA-OB, OA-dB and dA-dB at the NRW surface will be:
respectively. Measurement of these three reflectivities can therefore give all the necessary
information to recover the Fourier transform of the scattering length density for each
sample. One can then deduce the actual density profile with a minimum of uncertainty.
Thomas (1999) discusses a number of aspects of surfactant systems, including
some rather surprising effects of certain large counterions (tetramethyl ammonium
and p-toluene sulphonate), at the water-air interface and the reader is referred to that
work for details.
Fig. 14.5.3 Reflectivities of C12E6 on the bare quartz surface. The compositions are: (0)
hC12hE6
in D20; (O)dCl~hE6in water, and (+)hC,zhE6 in water. The continuous lines are calculated for a
bilayer 4.3 nm thick. The points at a reflectivity of lo-’ correspond to negative values after
background subtraction and should be taken to indicate that the reflectivity is lower than about
2 x lo-’. (From Thomas 1999, with permission.)
Exercise
14.5.1 Establish eqn (14.5.5).
References
Ashcroft, N.W. and Langreth, D.C. (1967). Phys. Rev. 156,685.
Atkins, P.W. (1982). Physical chemistry 2nd (student) edn,. p. 755.
Oxford University Press, Oxford.
Bacon, G.E. (1975). Neutron dzffraction (3rd edn). Clarendon Press, Oxford.
Baus, M. and Hansen, J.-P. (1980). Phys. Rep. 59, 1.
Baxter, R.J. (1970). J. Chem. Phys. 52, 4559.
Blum, L. and Torruella, A.J. (1972).j? Chem. Phys. 56,303 (and references therein).
Blum, L. and W y e , J.S. (1977).J. Phys. Chem. 81, 1311.
Brown, J.C., Pusey, P.N., Goodwin, J.W., and Ottewill, R.H. (1975).J Phys. A,
8, 664.
Cebula D.J., Thomas, R.K Harris, N.M, Tabony, J., and White J.W. (1978).
Faraday Soc. Disc. 65,7691
Clark, N.A., Ackerson, B.J., and Hurd, A.J. (1983). Phys. Rev. Lett. 50, 1459.
Crowley,T.L. (1993). Physica A, 195, 354.
Ducker, W. and Grant, L.M. (1996). J. Phys. Chem. 100, 3207.
Fragneto, G., Thomas, R.K., Rennie, A.R., and Penfold, J. (1996). Langmuir 12,
6036.
Frisch, H.L. and Lebowitz, J.L. (1964). The equilibrium theory of classicalfluids.
Benjamin, New York.
Goodwin, J.W., Ottewill, R.H., Owens, S.M., Richardson, R.A., and Hayter, J.
B. (1985). Makromol. chem. Suppl. l O / l l , 499.
Hansen, J.-P. and Hayter, J.B. (1982). Mol. Phys. 46,651.
Harris, N.M., Ottewill, R.H. and White, J.W. (1983). In Adsorptionfrom Solution
(R.H. Ottewill, C.H. Rochester, and A.L. Smith (eds) 1982 Symposium Proc.,
pp.139-154. Academic Press, London.
Hayter, J.B. (1983). Faraday Discuss. Chem. Soc. 76, 7. This volume is devoted to
concentrated colloidal dispersions and contains many papers of interest.
Hayter, J.B. (1985). In Physics of amphiphiles: micelles, vesicles and microemulsions
(ed. V. Degiorgio and M. Corti), p. 59. North Holland, Amsterdam.
Hayter, J.B. and Penfold, J. (1981~).Mol. Phys. 42, 109.
Hayter, J.B. and Penfold, J. (1981b).J. Chem. SOL.Faraday Trans I , 77, 1851.
Hayter, J.B. and Zulauf, M. (1982). Colloid Polymer Sci. 260, 1023.
Klein, A.G. and Werner, S.A. (1983). Rept. Prog. Phys. 46, 259.
Lebowitz, J.L. (1964). Phys. Rev. A, 133, 895.
Lebowitz, J.L. and Lieb, E.H. (1969). Phys. Rev. Lett. 22, 631.
Lee, E.M., Thomas, R.K., C u m i n s , P.G., Staples, E.J., Penfold, J. and
Rennie, A.R. (1989). Chem.Phys. Letters 162, 196.
Livsey, I. and Ottewill, R.H. (1991) Adv. Colloid Interface Sci. 36, 173-184.
Lovesey, S.W. (1977). In Dynamics of solids and liquids by neutron scattering
(ed. S. W. Lovesey and T. Springer), Chapter 1. Springer, New York.
Lovesey, S.W. (1984). Theory of neutron scattering from condensed matter. Clarendon
Press, Oxford.
712 I 14: SCATTERING STUDIES OF COLLOID STRUCTURE
Lu, J.R., Hromadova, M., Simister, E.A., Thomas, R.K., and Penfold, J.
(1994).J. Phys. Chem. 98, 11519.
Lu, J.R., Li, Z.X., Smallwood,J., Thomas, R.K. and Penfold, J. (1995)J. phys.
Chem. 99, 8233.
Markovi;, I. and Ottewill, R.H. (1986). Colloid and Polymer Sci. 264, 65-76.
Medina-Noyola, M. and McQuarrie, D.A. (1980). J. Chem. Phys. 73, 6279.
McDermott, D.C., Lu, J.R., Lee, E.M., Thomas, R.K., and Rennie, A.R.
(1992). Langmuir 8, 1204.
Naegele, G., Klein, R, and Medina-Noyola, M. (1985).J. Chem. Phys. 83,2560.
Ornstein, O.S. and Zernike, F. (1914). Proc. K. Ned. Acad. Wet. 17,793.
(Reprinted in Frisch and Lebowitz, 1964.)
Schulz, J.C., Warr, G.G., Butler, P.D., and Hami1ton.W.A. (1999). Adsorbed
layer structure of cationic surfactants on quartz. Submitted to Phys. Rev. Let.
Thomas, R.K. (1999). Neutron reflectivity at interfaces” In Modern characterization
methods of surfactant systems (ed. B.P. Binks). Marcel Dekker, New York.
van Beurten, P. and Vrij, A. (1981).J. Chem. Phys. 74,2744.
van Megan, W. and Snook, I.K. (1977). 3. Chem. Phys. 66, 813.
Verwey, E.J.W. and Overbeek, J. Th. G. (1948). Theory of stability of lyophobic
colloids, pp. 205. Elsevier, Amsterdam.
Rheology of Colloidal Dispersions
15.1 Introduction
15.2 Behaviour of time-independent inelastic fluids
15.2.1 Bingham plastic behaviour
15.2.2 Pseudoplastic behaviour
15.2.3 Dilatant behaviour
15.3 Behaviour of time-dependent inelastic fluids
15.4 Visco-elastic fluids
15.4.1 Macroscopic flow behaviour of elastic fluids
15.4.2 Describing the dynamic (oscillatory) behaviour of visco-elastic fluids
15.4.3 Steady shear flow of visco-elastic fluids
15.4.4 Comparison between dynamic and steady shear flow properties of
visco-elastic fluids
15.5 Measurement of rheological properties of inelastic fluids in Couette flow
15.5.1 The stress-strain rate relationship
15.5.2 Sources of error in the Couette viscometer
(a) End effects
(b) Wall slip effects
(c) Temperature control
(d) Taylor vortex development
15.6 Capillary viscometer
15.6.1 Flow rate versus pressure drop
15.7 Cone and plate or cone and cone viscometer
15.8 Time-dependent inelastic behaviour
15.9 Microrheology
15.10 Microscopic basis of rheological models
15.10.1 Flow behaviour of a dispersion of hard spheres
15.10.2 Flow of systems with anisometric particles
15.10.3 Kinetic interpretation of non-Newtonian flow
15.10.4 Flow of coagulated colloidal sols
15.10.5 Time-dependent systems: kinetic interpretation of thixotropy
15.10.6 Elastic behaviour of concentrated sols
713
714 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
15.1 Introduction
Rheology is the study of the deformation and flow of materials under the influence of
an applied stress. Some of the basic elements of this study were introduced in Section
3.4. Colloidal materials, especially gels and pastes, commonly exhibit both solid-like
(elastic) and liquid-like (viscous) behaviour and a complete treatment ought to take
both such aspects into account simultaneously. Unfortunately, the theoretical analysis
of such materials requires a rather extensive array of mathematical tools and is, even so,
restricted in most cases to the limit of small strains and strain rates (the region of linear
visco-elastic theory).
In order to make any real progress towards the solution of practical problems it will,
therefore, be necessary to adopt a very pragmatic approach. In some cases the
theoretical constructs will guide us in the choice of suitable measurement procedures
whilst in other cases the phenomenology may be so complex that the best one can hope
for is to be able to provide some understanding of the macroscopic behaviour in terms
of what can be assumed to be occurring at the microscopic (particleparticle
interaction) level.
We have already seen (Section 4.10.4) that a considerable departure from
Newtonian liquid behaviour is expected to occur when anisometric particles are
immersed in a Newtonian liquid. Such behaviour is predicted (Fig. 4.10.1) even when
the particles are in such dilute suspension that their mutual interactions may be
ignored. How much more complex, then, is the range of anticipated behaviour when
the particles can interact with one another. For modest concentrations of spherical (or
near spherical) particles whose long range van der Waals attraction has been
suppressed, the effect is principally on the viscous behaviour (Fig. 3.4.9) and such
systems show little elasticity.
As the number of particle contacts increases, however, so that contact is essentially
continuous, the colloidal sol begins to take on the properties of a gel; its elasticity then
increases markedly and it begins to behave as a (possibly rather easily distorted) solid.
For extremely anisometric particles (e.g. the long, thin rods of the clay mineral
attapulgite, vanadium pentoxide, or laponite (synthetic aluminium silicate)) such gel
formation can occur at very low particle concentrations (< lob). If van der Waals
attraction is allowed to dominate over repulsive forces, the systems can show gel-like
and elastic properties even if the primary particles are spherical; long chains of spheres
are probably involved in the gel structure in that case.
Linear visco-elastic theory, appropriate to the treatment of small strains, has been
widely applied to the behaviour of polymer solutions and has yielded some useful
insights. In the treatment given here we will make some reference to that area but will
concentrate most attention on the behaviour of suspensions undergoing continuous
strain (i.e. flow) where elastic effects, though important, play a secondary role in many
respects. Even with this limitation, the difficulties of providing a fundamental
theoretical base remain formidable and we shall often have to settle for a semi-
empirical approach.
The first problem which must be addressed is the question of how to collect
unambiguous rheological information. Once one introduces the notion of a variable
viscosity, the discussion of viscometer devices in Section 4.7 is no longer adequate. For
time-independent systems in which the viscosity, q, depends only on the shear rate,
BEHAVIOUR OF TIME-INDEPENDENT IN EL A ST IC F L U I D S I715
the Couette viscometer can give reliable data if the gap width is small (recall Exercise
4.7.4) but for the Ostwald (capillary) viscometer, in which the shear rate necessarily
varies from zero in the centre to a maximum at the wall, an alternative approach is
necessary. We will discuss the usual approach to such problems in Section 15.6.
where D: is the deviatoric stress tensor, r] is the viscosity, and eq is the rate of strain
tensor. For the simple flow regimes with which we will be primarily concerned here
(Fig. 15.2.1) the shearing stress is applied in only one direction and the observed flow
occurs in the same direction. In such cases a set of coordinates can be chosen so that the
velocity field takes the simple form (Boger et al. 1980):
u1 =f(coordinate 2 only); uz = u3 = 0 (15.2.1)
where subscript 1 indicates the direction of flow and 2 is at right angles to that direction
and also to the boundary surface. The more common flow regimes of this sort are
illustrated in Fig. 15.2.2, and they are all exploited in different forms of viscometer;
they are, therefore, referred to as vzscometrzcflows. The general relation between the
deviatoric stress tensor, u:, and the rate of strain tensor, eij, can then be greatly
simplified. For an inelastic fluid, in these simple flow conditions, the normal stresses
( ~ i i ) are
, all equal to the local hydrodynamic pressure, so that
011 = 0 2 2 = 0 3 3 = -p (15.2.2)
716 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
and the only non-zero components of the shearing stress are q 2 and a21 (which are also
equal) so that
(1 5.2.3)
where S2l is the externally applied stress. This was the situation treated in Section 4.7.
Similarly, the only non-zero component of the rate of strain tensor is el2 where:
avl avz
2e12 =-+-. (1 5.2.4)
ax2 axl
In these viscometric flows, it is conventional to replace 2e12 by the rate of strain, j
which is also a function of only one coordinate (though it does not appear so when
expressed in terms of Cartesian coordinates as in eqn (15.2.4)). We can, therefore,
write:
Fig. 15.2.2 (a) Poiseuille flow in a tube. (b) Couette flow in the annulus between two cylinders. (c)
The cone and plate configuration. Note that Om, is normally only a few degrees (or less) so that the
system is much more like two parallel plates. (From Boger et al. (1980) with permission.)
S =S B + VPL j. [3.4.23]
Concentrated slurries of coal dispersed in water exhibit this type of behaviour (Rowel1
1983 personal communication) essentially exactly, as do some food products (Halmos
and Tiu 1981). It is of more general significance because many engineering designs are
based on this expression since the departure of the system from this ‘ideal’ behaviour is
often sufficiently small to be neglected.
It is, however, more usual to observe a non-linear relation between shear stress and
shear rate for shear stresses above the yield value. The simplest model used to fit such
systems is the Herschel-Bulkley model (Exercise 15.2.1):
S=So+Kj” (1 5.2.6)
718 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
where So is referred to as the primary yield value. This behaviour would be detected
by a linear plot of log (S- So) against log exhibiting a slope, n, different from
unity. This is, of course, purely an empirical relation which has no underlying
theoretical base. It is, therefore, unwise to use it outside the region over which it has
been established experimentally. Th e physical behaviour of these fluids can be
understood in terms of a three dimensional structure which has sufficient strength
to prevent flow if the applied stress is less than So. For S > So the structure may
collapse suddenly to produce flow units which are not subsequently affected by the
shear regime. This would give rise to ideal Bingham plastic flow and then So = SB.
More usually the breakdown is progressive and the flow units become smaller, and/
or more compact, and/or more closely aligned with the stream lines as the shear rate
increases: the result in any case is a decrease in differential viscosity (dS/dy) with
increasing shear rate.
An intermediate behaviour pattern has been observed in the flow of coagulated
dispersions of small particle size (< 1 pm) (Hunter 1982). Such systems show a
non-linear S versus p relation at low shear rates (1-500 s-l) which becomes strictly
linear at higher shear rates (500 - 3000 s-'); (see Section 15.10.4). This is, however,
a relatively restricted range of p and the behaviour outside that range has not been
well characterized. At very low shear rates (< 1 s-l) the shear stress is very low but
it may not be zero when = 0 (curve 5 of Fig. 3.4.9). T h e intercept on the stress
axis from the linear part of the curve at high shear rate (SBin Fig. 3.4.9) is again
called the Bingham yield value since eqn (3.4.23) above is obeyed for high values of
-
p (> 500 s-').
Yield strength is a very important characteristic of such widely different materials as
oil-well drilling muds, paint, toothpaste, paper pulp, crude oil and many food
products, pharmaceutical preparations, and cosmetics.
S=Kj." (15.2.7)
known as the Ostwald-de Waele model (Skelland 1967). Obviously the Herschel-
Bulkley eqn (15.2.6) is a simple extension of this expression. The main advantage of
eqn (15.2.7) lies in its simple two-parameter form and the ease with which it can be
differentiated and integrated. Even if it applies only over a limited range of 9, it can be
used to extract meaningful rheological data from viscometric flow situations provided
that it holds for all shear rates being experienced by the material. Once the S vs. p data
have been acquired they can then be fitted with a more elaborate expression. There are
a number of two, three, and even four parameter expressions which have been
developed on the basis of more or less elaborate microscopic models. Some of them are
listed in Table 15.1, which is modified from the compilation provided by Skelland
(1967).
BEHAVIOUR OF TIME-INDEPENDENT IN EL A ST IC F L U I D S I719
Table 15.I
Models relating shear stress (S)to shear rate y, forJuids without a yield stress.
Adapted from Skelland (1967). "Ostwald (1926), Reiner (1949); bReiner (1960); 'Eyring (1936),
Prandtl(l928); dChristiansenet al. (1955); eBoger et al. (1980), Meter (1964).
T he more elaborate Ellis and Meter models recognize the fact that the shear
stress-shear rate curve is often linear at very low and again at very high shear rates
so that it is possible to define two limiting viscosities:
S
lim = qo = zero shear viscosity
Y+O y
(15.2.8)
and
. s= roo= infinite shear viscosity
lim
Y+m y
and the apparent viscosity falls smoothly from ~0 to roo as p increases. Since the
dimensionless parameter a in these models is often close to 2, the parameter S; can
be roughly identified as the shear stress at which ~0 has fallen half way to its final
value. Figure 15.2.3 shows how the introduction of these extra parameters improves
the fit for an extensive range of viscosity data compiled by Boger (1977). Both of
these models have a form similar to that used by Krieger (1972), to which we alluded
in Section 4.10.8 (Exercise 15.2.2):
~
r-roo
ro- roo
= [ 1+- 71-l
(15.2.9)
where r] is the apparent viscosity. Note that this corresponds to a = 2 in the Ellis and
Meter models. Boger found a = 2.43 to give the best fit to his data but no doubt a
reasonable fit could be obtained using a = 2, which reduces the Meter model to only
three parameters. An alternative representation can be given in terms of the shear rate
(Carreau 1968):
(1 5.2.10)
720 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
Meter model
\
\ Ellis
‘,model
1u-- 1U” 10- IU’
Shear rate (s-I)
Fig. 15.2.3 Viscosity as a function of shear rate for a poly(acry1ic acid) solution in water (from
Boger 1977, with permission.) Note that only the Meter model fits the data over the entire range of
shear rates.
where h is an empirical (time) parameter. This has some advantages since, in the
normal Couette viscometer in which these systems are studied, it is the shear rate
rather than the shear stress which is the independent variable.
The de Haven model is obviously identical to the Ellis model with appropriate
relations between the parameters (Exercise 15.2.2). The Eyring models are based on
the concept of activated transport and are more likely to be relevant to the viscous
behaviour of molecular fluids, including (possibly) dissolved polymers, but they have
been used in the description of the visco-elastic behaviour of latex suspensions at very
low electrolyte concentrations (Goodwin et al. 1982). We will return to discuss the
microscopic origins of these sorts of expressions in Section 15.10.
r
Exercises
15.2.1 The following shear stress-shear rate data were obtained on a meat extract at
77 "C (Boger et al. 1980). (The first element of each number pair is shear rate
(s-l) and the second is shear stress (N m-2).):
0.111, 18.04; 0.140, 18.32; 0.176, 18.65; 0.222, 19.09; 0.279, 19.62;
0.352, 20.31; 0.443, 21.16; 0.557, 22.24; 0.702, 23.60; 0.883, 25.30;
1.112, 27.45; 1.400, 30.16; 1.762, 33.56; 2.218, 37.85; 2.793; 43.25;
3.516, 50.05; 4.426, 58.60; 5.576, 69.41; 7.015, 82.94; 8.831, 100.01;
11.117, 121.50; 14.00, 148.60.
Plot S versus 1; on linear paper and on log-log paper and hence determine the
characteristics of this fluid in terms of eqns (15.2.6) and (15.2.7).
15.2.2 Show that the Meter model ofTable 15.1 corresponds to eqn (15.2.9) with a! = 2.
What are the relations between the de Haven and Ellis parameters?Why does the
Ellis model break down at very high shear rates?
15.2.3 The following shear stress-shear rate data were obtained for an aqueous
solution of methyl cellulose at 18 "C (Boger et al. 1980). (The first figure is
shear rate (s-') and the second is shear stress (N mP2).)
0.1400, 0.117; 0.1762, 0.141; 0.2218, 0.169; 0.2793, 0.211; 0.3516, 0.281;
0.4426, 0.352; 0.5572, 0.446; 0.7015, 0.563; 0.8831, 0.687; 1.1117, 0.847;
1.400, 1.076; 1.762, 1.305; 2.218, 1.625; 2.793, 2.010; 3.516, 2.53;
4.426,3.08; 5.572,3.79; 7.015,4.68; 8.831,5.41; 11.117,6.53; 14.000,8.11;
17.620, 9.46; 22.18, 11.49; 27.93, 13.52; 35.16, 16.22; 44.26, 18.92;
55.72, 22.10; 70.15, 26.13; 88.31, 30.00; 111.17, 34.80; 140.00, 40.0;
176.2, 45.7; 221.8, 52.5; 279.3, 57.6.
See how well they fit a power law (log-log) plot over limited ranges of shear
rate. State the shear ranges over which a reasonable straight line fit is obtained
and estimate the parameters K and n applicable to those regions.
+
15.2.4 Recast the Ellis model in the form 1; = a(S bSm)and show that it gives a good
fit to the data of Exercise 15.2.3 over the whole range of 1; using the parameters
~0 = 0.794 N mP2 s, S i = 21.554 N mP2 and a! = 2.027.
Compare this with the fit using eqn (15.2.9) with a suitable choice of roo.(That
is the Meter model with a! = 2.)
the time over which the system is being studied. In other words, the ‘structure’ in
the system, which determines the variable shear viscosity, is being altered at a rate
which is observable over the time period of the measurement. As we noted earlier,
it is always possible to encounter this sort of phenomenon if one chooses the right
time scale but it is sometimes very difficult to avoid. If the system has a short
relaxation time (say <lop3s) it will be possible to construct a curve like Fig. 15.2.3
with little or no evidence of time dependence (except perhaps at the very highest
shear rate). But when the relaxation time is of order 10-103 s or even higher (as it
can well be with suspensions of highly anisometric particles) it becomes much more
difficult to acquire meaningful data.
T he most common phenomenon of this type is called thixotropy where the
apparent or differential viscosity at a particular shear rate decreases with time. Less
frequently, the opposite behaviour occurs and that is called negative thixotropy. In
thixotropic materials, shearing at a given shear rate produces a gradual breakdown
in the ‘structure’ with time. If it is allowed to proceed for long enough one can
obtain an ‘equilibrium’ or steady-state curve which is usually pseudoplastic in
nature (Fig. 15.3.1). More commonly, the time constraints make it impossible (or
inconvenient) to wait for the steady-state behaviour to be established. One often
finds that a system exhibits an hysteresis loop (Fig. 15.3.2). T h e area inside the loop
depends on the past shear and thermal history of the sample and on how rapidly the
shear rate is increased and decreased. Th e more time that is spent at each data
point, the closer the approach to the steady-state and the smaller the area of the
loop.
A related behaviour occurs in some systems at low shear rates where the rate at
which the structure is rebuilt can be increased by gently shearing the system. This is
referred to as rheopexy. It can be recognized by the fact that the system ‘recovers’ some
of its pre-sheared viscosity at a somewhat faster rate when it is sheared very slowly
compared to the rate of recovery when it is left to stand. Note that this is not the same
as negative thixotropy though the two can be confused if only a limited number of
measurements is made at one (low) shear rate.
Fig. 15.3.1 (a) Relation between shear stress and shear rate for a thixotropicfluid. t is the time taken
between successive changes in the shear rate. The material is assumed to have a relaxation time of
about 10-20 seconds and it is assumed that after each successive change of shear rate, the shear stress
is measured after an interval o f t seconds and the resulting points are joined by a smooth curve.
BEHAVIOUR OF T I M E - D E P E N D E N T I N E L A S T I CF L U I D S I723
0
Time
Fig. 15.3.1 (b) The dependence of shear stress on time at a given shear rate for a thixotropic
material.
Fig. 15.3.2 Hysteresis loops for thixotropic fluids (both positive and negative).
724 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
is, however, important in the starting up phase of a pipeline flow. (See Govier and Aziz
1972 for further discussion of this point.)
Obviously, the collection of meaningful rheological parameters in thixotropic
materials is a difficult task. At the very least it is important to ensure as far as
possible that these parameters are measured on a sample which has been subjected
to a shear and thermal history which is appropriate to the problem being studied. If,
for example, one is concerned with the slow flow of a suspension from a storage tank
into a pipe one would examine the suspension after storage under zero shear
conditions, making sure that the sample was not sheared during transfer to the
measuring apparatus. If, on the other hand, the material was to be pumped from the
storage tank, the infinite shear or steady state (long shear time) data would be more
appropriate.
Newtonian Viscoelastic
u Newtonian Viscoelastic
Fig. 15.4.1 Flow behaviour of visco-elastic fluids. (After Bird and Curtiss 1984.) (a) The
Weissenbergeffect; (b) Couette flow; (c) Die swell; (d) and (e) influence of a rotating disc on the flow
pattern; and (f) the open channel syphon.
q* = q’ + iq“ [3.4.16]
where q’ is called the dynamic viscosity and is a function of frequency, w , in the same
way as the steady shear viscosity, q, is a function of shear rate. Recall that from eqn
(3.4.17)for a linear visco-elastic material, q’ = G”/w where G” is called the loss modulus
of the material.
T he imaginary part of q*, (q”) measures the elastic response of the material and
we will see that it is related to the normal stress differences which give rise to the
Weissenberg effects and the other peculiar manifestations of visco-elastic flow. It is
also related to the storage (or dynamic rigidity) modulus, G’ by eqn (3.4.17):~”
G‘/w.
!
I
= s12 P22 (1 5.4.2)
0 P33
PI1 + +
P22 p33 = 0. (15.4.3)
Thus, only two of the normal stresses are independent. The flow of an incompressible
visco-elastic fluid can therefore be characterized by a measurement of the shear stress
and two of the three deviatoric normal stresses. In practice, one measures the shear
stress, S l 2 and the two normal stress differences: ( p l l - p 2 2 ) and ( p 2 2 - p 3 3 ) of which the
first is the more important.
VISCO-ELASTICF L U I D S I727
T o understand how these normal stresses cause the Weissenberg (or climbing rod)
effect note that with reference to Fig. 15.2.1, the axis of rotation of the rod is
perpendicular to the 12 plane. If the component of the normal stress, 011, in the
direction of flow is greater in magnitude than the components perpendicular to it (022,
033) there will result a tension in the direction of flow which increases towards the axis
of rotation (as the shear rate increases). This tension is equivalent to a pressure acting
uniformly on each concentric fluid layer, increasing towards the axis of rotation and
forcing the liquid to climb the rod.
G”
~0 = lim - = lim v’ (15.4.4)
-0 w w-0
and =
P l l -P22 . 2G‘
= limp=
. 24‘
lim-. (15.4.5)
YL wow2 w o w
Figure 15.4.2 shows some comparative data for a sample of polystyrene in which eqn
(15.4.4) appears to be satisfied. Equation (15.4.5) is presumably only satisfied for
frequencies and shear rates below the experimentally accessible range.
The first normal stress difference is relatively easy to measure at low shear rates and
is always positive and, typically, for polymer solutions and melts it is greater than the
shear stress at the same shear rate. For most colloidal suspensions, W1 is much smaller
than the shear stress; only in gels and pastes does one expect the development of high
degrees of elasticity.
Even for polymeric liquids it seems that the second normal stress difference (pzz -
p 3 3 ) is small, and usually negative, compared to (pll - pzz). It is, therefore, usually
possible to assume it to be zero (‘the Weissenberg hypothesis’) and this will certainly be
true for colloidal suspensions.
Although ( p l l - p 2 2 ) and S12 are relatively easy to measure, results for polymer
solutions have been limited to the region below Y = - 10 s-l in conventional
viscometers, presumably because of the development of flow instabilities in the
viscometer (Boger et al. 1980). This, of course, is much lower than the shear rates
commonly used in polymer technology, so some estimate is needed of these parameters
at higher shear rates. Two important approximate procedures used to obtain such
estimates are: (1) the Cox-Mertz rule; and (2) the determination of the first normal
stress difference from simple shear data.
The Cox-Mertz rule states that the curve of versus 9 should be superposable on
the curve of ( q ’2 + q ’’2)1/2 against w. (That is r] (Y) and qrn (w) are essentially
identical.) Bird and Curtiss (1984) have used a very general statistical mechanical
728 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
lo5- - 106
-
-
k
104- -lo5
E
‘
g
is
v 2
c M
lo3- -lo4
<
lo2 I I I I I lo3
Fig. 15.4.2 Comparison of dynamic (q’, G’) data with steady shear data [q, (all - p 2 2 ) ] for a sample
of polystyrene (Styron 686 from Dow Chemical Co.) Mass average and number average molar
masses were 2.89 x lo5 and 1.02 x lo5 daltons respectively (Section 5.3.6). (From Han et al. 1975,
with permission.)
model to derive a constitutive eguationt for a polymer solution or melt and established
that the Cox-Mertz rule should be true at least for small shear rates (hj.or hw < 50
where h is a characteristic relaxation time of the material). It is, however, used over a
much wider range of j . (and w ) than can be strictly justified theoretically.
The estimation of the first normal stress difference from simple shear measurements
also depends on the postulation of a particular sort of constitutive equation. One
popular procedure is that described by Abdel-Khalik et al. (1974) based on the
Goddard-Miller (1966) constitutive equation. There is, however, little point in
pursuing these developments until we have constitutive equations which are more
suited to the description of colloidal dispersions.
comparisons must be made at the same place. For the coaxial cylinder (Couette)
viscometer (Section 4.7) this is normally the wall of the inner cylinder.
Before examining the situations in detail we recall that the shear rate in a viscometric
flow is not necessarily the same as the velocity gradient. In the Couette viscometer, for
example, we can see from eqn (4.7.1) that for a Newtonian fluid the shear rate is given
by
(15.5.1)
and not simply by duQ/dr. (Here w(r)is the angular velocity = v ~ / r .The
) second term
on the left (= w ) accounts for that part of the motion which is simply a rigid body
rotation and does not contribute to the shearing process.
T = 2Rr2Ls. [4.7.2]
This torque is constant across the gap and equal to the external torque, TO,on the
stationary cylinder (Section 4.7). The determination of the shear rate is, however, a
little more difficult since the velocity distribution across the gap depends on the
properties of the fluid. Fortunately, one can determine those properties from the
experimental data using a procedure developed by Krieger and Maron (1952, 1954)
based on earlier work by Mooney (1931) and Saal and Koens (1933) (also see Reiner
1960).
Since eqn (15.2.4) still holds for the rate of strain of an inelastic non-Newtonian
fluid, we can still use eqn ( 1 5 5 . 1 ) for the shear rate. We assume that this shear rate is
some arbitrary function of the stress:
and obtain a general relation between the rate of rotation and the measured torque:
Assuming that the outer cylinder is stationary, and integrating from the bob, where
S = S b (= S;) to the cup where S = S, (=So),and from w = Q to 0, we obtain
(Exercise 15.5.1):
(1 5.5.6)
This equation can be handled in a number of ways to yield solutions for coaxial
cylinder viscometers and these are discussed by van Wazer et al. (1963 pp. 57-61). The
simplest solution applies for the case where the outer cylinder is of infinite radius (i.e.
the fluid is driven by the motion of a bob immersed in a large volume of liquid, as in the
Brookfield viscometer). In that case S, is zero, and differentiation of eqn (15.5.6) with
respect to S b gives
f(&) = 2(dQ/d In &). (1 5.5.7)
Thus f (&) can be obtained directly from a plot of Q versus In S b or Q versus In TO.
This solution is valid for any substance which does not have a yield value. For
substances with a yield value there will be no motion of the fluid at distances r > ro
where the stress has fallen below the yield value. Since this then becomes the effective
outer cylinder wall, eqn (15.5.7) breaks down.
When the fluid has a yield value (Fig. 3.4.9 curve (4) or (5)) the simplest
procedure is to use a viscometer with a narrow gap width (Rb/R, 1) so that the
shear rate remains essentially constant across the gap. There are, however,
situations in which that is not possible and one must then resort to one of two
possible procedures: (1) assume a particular (empirical) relationship for the fluid
and analyse the flow regime in the gap to arrive at the S(p) relation from the
measured To(S2)curve; or (2) use a general asymptotic estimate of the shear rate at
the surface of the bob, valid for any fluid. The best known example of the first
procedure is the Reiner-Riwlin (1927) equation based on the assumption that the
fluid is an ideal Bingham plastic (Table 15.1). They derive (Exercise 15.5.2):
(1 5.5.8)
provided that all of the material in the annulus is undergoing shear. If the gap width is
fairly large, this will not be the case at lower shear rates. There will develop a region
near one of the walls in which the material moves as a rigid body with the same velocity
as the wall so that the effective gap width is reduced. The effect of this is to reduce the
slope of the Q versus To curve which may be erroneously interpreted as due to an
increase in the differential viscosity at low shear rate. van Wazer et al. (1963 Fig. 2.5)
give a specific example of how an ideal Bingham plastic material can appear to be non-
ideal (Fig. 3.4.9 curve (5)) or even pseudoplastic (curve (2)), if examined over a narrow
MEASUREMENT OF RHE O LO G ICAL P R O P E R T I E S O F INELASTIC F L U I D S I N COUETTE F L O W I731
range of shear rates in a wide gap viscometer. They also show how the more elaborate
models of Table 15.1 can be used to derive reliable rheological data on non-Newtonian
fluids.
The second procedure is more general and depends upon the use of more or less
exact estimates of the shear rate at the stationary cylinder. Various procedures have
been developed by Krieger and Maron (1952, 1954), Krieger and Elrod (1953), and
Calderbank and Moo-Young (1959) with different levels of approximation using
either: (1) a multiple bob method; (2) a two-bob approximation of (1); and (3) a single
bob method which depends upon eqn (15.5.6). These methods are discussed in some
detail by van Wazer et al. (1963 pp. 5 7 4 1 ) and we will not examine them further.
Suffice it to note that they make it possible to obtain reliable values of S versus 9
-
(within 1-1S0/o) even with viscometers of large gap width, although the errors do
increase with increase in the ratio R,/Rb(= R,/Ri). (See Fig. 15.5.1.)
It should be pointed out that there is a limit to the extent to which the gap can be
reduced. T o treat the fluid between the cylinders as a homogeneous material it is
necessary to have a gap which is much larger (-1OOx) than the size of the largest
particles or aggregates in the suspension. T o satisfy this requirement, most
commercial viscometers tend to have rather large gaps so that the ratio R,/RI,
departs significantly from unity. Rayner et al. (1979) solved the problem of
measuring the flow behaviour of coarse coal slurries (particle size -50 pm) by
simply increasing the size of both Rb and R, so that the rheological problem was
replaced by an engineering one.
4 8 12 16 20 24 28 !
Angular velocity, 8 (s-l)
Fig. 15.5.1 The shear stress-shear rate curve in a wide gap Couette viscometer for an ideal
Bingham plastic in the case where some of the material in the gap is not undergoing shear. The stress
required to establish flow across the entire gap is shown on the shear stress axis and in the inset
figure. (After van Wazer et al. 1963, p. 64.)
732 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
noted in Section 4.7 this is no longer true if the rotational velocity becomes too
high. Just as the flow in a pipe becomes turbulent if the Reynolds number (Re)
(Section 4.8.1) exceeds about 2000, so too does it become turbulent in a Couette
viscometer when Re exceeds a certain figure (Exercise 15.5.6). If the outer cylinder
is rotating and the inner one is stationary, this instability does not develop until Re
exceeds about 50 000 so it is seldom of any consequence. For many commercial
instruments, however, the opposite configuration is used and instability can then
occur at quite low speeds (as little as a few hundred reciprocal seconds). This
instability is caused not by turbulence but by the onset of a radial flow pattern
(generated by centrifugal effects) and was studied by Taylor (1923). T h e
phenomenon is ascribed to the occurrence of ‘Taylor vortices’ which are circular
flow patterns occurring in certain parts of the annulus. T h e critical Reynolds
number at which this is expected to happen is given approximately by (van Wazer
1963 p. 85):
Considering the low shear rates at which this behaviour can occur (see Exercise 15.5.6)
it is surprising that so many commercial viscometers use this configuration.
Exercises
15.5.1 Establish eqns (15.5.4) and (15.5.6).
15.5.2 Assuming that for an ideal Bingham fluid one can write shear rate = - Y d d d r
= (S- SB)/~PL derive the Reiner-Riwlin equation (15.5.8) assuming that the
fluid is moving with angular velocity Q at the inner cylinder (Y = Rb) and zero
only at the outer cylinder (Y = Rc).
15.5.3 Show that if the Bingham yield value, SB,is such that only some of the fluid is
moving (S, < SB< S b (where and b refer to outer (cup) and inner (bob)
cylinders) then the critical radius beyond which the fluid is stationary is Rcrit=
Rb(Sb/&$. Hence show that the critical stress &it (measured at the bob)
which must be exceeded so that flow occurs throughout the annulus is given by
Scrit = SB(Rc/Rb)2.
15.5.4 Show that the Reiner-Riwlin equation (15.5.8) can be written in the form
where k1 and k2 are apparatus constants. (The data in the linear regime (where
TOcx Q is treated by plotting this function against S b (= To/2nR;L) and
determining the slope (which is q p ~ ) . )
15.5.5 For the power law fluid S = Kl(-rdw/dr)” in a coaxial cylinder viscometer we
have
To = S.2ny2L = K1(-rdo/dr)”.2nr2L.
734 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
Integrate this expression between the limits w = 0 and Q and Y = R, and Rb (i.e.
stationary outer cylinder) to obtain:
ay + Y
-0
(1 5.6.2)
CAPILLARY VISCOMETER 1735
For the inelastic fluids with which we are chiefly concerned, the pressure is a function of z
only and p, = pee = p, = 0 so that these equations reduce to (Boger et al. 1980):
YdP -
or, = -- - YAP. (15.6.3)
~
where a is the capillary radius and L is its length. Hence, from eqn (15.6.2a):
ar, = S,r/a (15.6.5)
which shows that the shearing stress increases linearly from its value of zero at the tube
axis to a maximum value, S, at the wall. This relationship is valid regardless of the flow
regime (laminar or turbulent) and is equally valid for inelastic and viscoelastic fluids
and for fluids having a yield stress. The relation between wall shear stress and pressure
drop (eqn 15.6.4) permits the development of a general relation between volume flow
rate and the pressure over the capillary length (Rabinovitch 1929).
15.6.1 Flow rate versus pressure drop
The following analysis follows that of Boger et al. (1980).
The volume flow rate, is related to the fluid velocity, v,, by (compare eqn 4.7.13):
d Q = v,2nrdr (15.6.6)
which can be integrated by parts to give:
Assuming no slip at the tube wall (see Section 15.5.2b) so that v, = 0 for Y = a:
a
Q = -n/r2dv,. (15.6.8)
0
The shear rate for fully developed laminar flow in a tube is in this case (- dv,/dr)
(compare eqn (15.5.1)) and this will be a function of the shear stress:
(15.6.10)
736 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
Just as in the case of the Couette viscometer (Section 15.5) this expression can be used
in two different ways: (1) it can be integrated for a particular fluid model (Table 15.1)
to obtain the volumetric flow rate/pressure drop relationship for flow through a
circular pipe; or (2) it can be differentiated to obtain an expression for shear rate at the
wall, independent of the fluid model.
If, for example, we choose a power law model for the fluid:
~ Q - (15.6.12)
where dp/dz = A p / L is the pressure gradient along the length of the pipe.
Corresponding expressions for other fluid models are given in Table 15.2. Notice that
the quantity 8V/D (= 4Q/nu3) is equal to the shear rate at the wall (S,/q) for a
Newtonian fluid and that it appears as a natural parameter for the other flow models. It
is common practice to plot this function (or its logarithm) against the pressure drop in
order to represent the flow behaviour (see eqn (15.6.15) below).
It turns out (Exercise 15.6.3) that, for a power law fluid (especially if n is small), the
pressure drop is much less sensitive to flow rate than it is for a Newtonian fluid. For
this reason, viscometers which rely on the measurement of differential pressure drop
across capillaries are not very satisfactory for examining the behaviour of pseudoplastic
fluids. The flow profile for such fluids is very different from that shown by a
Newtonian fluid (Fig. 15.6.1(a)). As n decreases below unity the velocity becomes
almost constant across a large part of the tube. For dilatant systems (n > 1) the profile is
Table 15.2
Relation between Qand S, or Ap for laminarjow of various time-independent inelasticjuids in tubes.
Fluid Relation
For a Bingham plastic, the shear stress near the axis of the tube ( r << a) does not exceed the yield
stress (SB)so that all of that material moves as a solid (that is, as an unsheared plug) with a resulting
velocity and shear rate profile as indicated in Fig. 15.6.l(b). Wazer et al. (1963 pp. 194-7).
CAPILLARY VISCOMETER 1737
8
9
Y
.-
n
Velocity ( V )
Fig. 15.6.1 Velocity profiles and approximate shear rate distribution for power law fluids. (a)
Velocity profiles corresponding to the same average velocity. (b) Reduced shear rate as a function of
reduced radius. (After van Wazer et al. 1963, p.198.)
sharper than the parabolic result obtained for a Newtonian fluid. The corresponding
distribution of shear rate is also shown in Fig. 15.6.l(b).
The expressions in Table 15.2 need to be manipulated a little to obtain values of, for
example, apparent viscosity as a function of shear rate. For the power-law model that
problem is examined in Exercise 15.6.5.
Returning to the second procedure for dealing with eqn (15.6.10), we can multiply
both sides by S: and differentiate with respect to S, to obtain (Boger et al. 1980):
wheref(S,) = (-duz/d&a is the shear rate at the wall (from eqn (15.6.9)).
738 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
(15.6.14)
The value of b is first evaluated from the slope of the log-log plot (eqn (15.6.15)). It
may be necessary to determine several different b values for different ranges of shear
rate. One can then evaluate YW and S, and hence the apparent viscosity from the ratio
of the two. Note that this procedure requires no assumptions about the viscous
behaviour of the fluid. For a power law fluid (eqn (15.2.7)) it is apparent from eqn
(15.6.15) that b = l/n (Exercise 15.6.4).
Capillary viscometers may be constructed of glass with the flow occurring solely
as a consequence of a difference in hydrostatic head between inlet and outlet
(Section 4.7.2). For concentrated colloidal suspensions or polymer solutions,
however, it is usually necessary to employ gas pressure or a moving ram to drive the
fluid through. There are various problems associated with this procedure and a
number of significant sources of error which are discussed in the standard texts (see,for
example, van Wazer et al. (1963 pp. 199-215) and Sherman (1970 pp. 3849)). Some of
these errors can be estimated and corrected for directly whilst some can be minimized
by proper design. End effects can, for example, be corrected by making measurements
in capillaries of different length and using an extrapolation procedure. The kinetic
energy effect, at least for near-Newtonian liquids, can be minimized by the use of
capillaries with tapered ends (Caw and Wylie 1961). We will not examine these
questions in any further detail. The problems caused by particle migration away from
the tube walls are, however, briefly considered in Section 15.8.
15.6.1 Derive eqn (15.6.2(a))assuming that dp/dx is constant for fully developed flow.
Rewrite eqn (15.6.la) in terms of the deviatoric stresses p , and $00 and hence
show that (Boger et al. 1980):
i
P, - 8 + d o , 4 - (Pll -poo)dlnr
0
= 0.
This equation (due to Philippoff 1942) can be used to correct data (qa) based on
uncorrected shear rates to those ( q ) based on true values (van Wazer et al.
(1963) p. 193).
which the upper plate is attached to a torsion bar and the lower cone is rotated. The
cone is able to move downward in response to the normal stress effect and the
instrument measures the force, F, necessary to return the cone to its initial position
relative to the upper plate. This reaction force can be directly related to the first normal
stress difference (Lodge 1964):
(15.7.1)
where R is the plate radius. If two flat plates are used, instead of a cone and plate, the
force required to keep the gap space constant is given by:
(15.7.2)
For parallel plate geometry, the shear stress and shear rate are given by:
(15.7.3)
where A is the gap space between the plates (assuming it is small) (compare with eqn
(4.7.4) and Exercise 4.7.4).
Thus a comparison of the cone and plate and the parallel plate behaviour enable a
complete characterization of the fluid, assuming that it obeys the Weissenberg
assumption ($22 % $33).
*
i2
c
rn
1
8
c
m
Shear rate (9 )
Fig. 15.8.1 (a) Possible hysteresis behaviour for the case where (i) yield stress is fixed and (ii) yield
stress is reduced by shearing. (b). Shear stress-time behaviour for a material subjected to increasing
strain rates at points 1,2,3, etc. The stress required for that strain rate rises sharply at first and then
falls exponentially to a steady state value as the ‘structure’ of the fluid breaks down to some extent.
This method is one of a number introduced by Oersterle.
15.9 Microrheology
T o understand the macroscopic rheological behaviour of a colloidal dispersion it would
be helpful to know what happens at the microscopic level when a colloidal dispersion
742 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
undergoes a shearing process. We discussed this question briefly in Section 4.10, for
the case of spherical and spheroidal particles at low concentrations and for spheres at
higher concentrations. There is a very large amount of literature on this subject, most
of it couched in the language of tensor calculus so we will be able to discuss only a small
part of it here. A much more detailed discussion is provided by van de Ven (1989) and
by Russel et al. (1989).
The shear behaviour of spheroidal particles at low concentration has been
investigated both theoretically and experimentally in considerable detail, especially for
larger particles which are microscopically visible and, hence, essentially unaffected by
Brownian motion. The theoretical equations of Jeffery (1922) which describe the
detailed trajectories of the particles and the motion of their axes of symmetry have been
amply confirmed by the elegant experimental investigations carried out by Mason and
his collaborators. The early work is described in an extensive review by Goldsmith and
Mason (1967) from which a few illustrative examples will be described below. We will
then describe a few of the many later investigations in this area by the same group.
When a dilute suspension of rods or discs is subjected to a Couette type flow, the
individual particles describe a rather complicated pattern of rotations induced by the
shear field. The motion of the axis of revolution of the spheroid appears very irregular but
it can be analysed exactly using Jeffery’s theoretical equations. Using the coordinate
system shown in Fig. 15.9.1, the rate of change of the angle $1 with time is given by:
Here p is the shear rate and q is the (equivalent) particle axis ratio. Integration of this
expression with respect to time (assuming p is constant) gives (Exercise 15.9.1):
Thus when viewed along the XI direction, the axis of revolution of the particle rotates
with a period of rotation given by:
From eqn (15.9.1) it can be shown (Exercise 15.9.2) that d$l/dt is a maximum for ($1
= 0, n)or for ($1 = n/2,3n/2) depending on whether the spheroid is prolate (q< 1) or
oblate ( q > 1). That is, of course, what one would expect. The rod ( q < 1) will be
twisting more quickly when it is oriented at right angles to the streamlines and least
quickly when it is parallel to the streamlines. The same is true for the disc, but for the
disc the broad face is oriented across the stream lines when the axis of revolution is
oriented along the streamlines ($ = n/2, 3n/2). The excellent agreement between
theory and experiment for this simple situation is shown in Fig. 15.9.2.
The end of the rod when viewed in the direction of X2 appears to oscillate backwards
and forwards along an arc of an ellipse which may be more or less eccentric. Again
Jeffery’s equations are able to quantitatively describe the behaviour but a curious feature
arises. If Brownian motion and particleparticle interactions are unimportant, and the
MICRORHEOLOGY 1743
Fig. 15.9.1 Orientation of a spheroidal particle in a shear field. AB is the major semi-axis of the
spheroid and the shear field is represented by a linear increase in the velocity in the X3 direction as
X2 increases (i.e. simple Couette flow).
suspending liquid is Newtonian, the actual characteristics of this ellipse are quite
arbitrary; different particles will have ellipses of different eccentricity, depending on their
initial orientation, but will retain them so long as the flow is slow and steady. Interactions,
or Brownian motion or non-Newtonian characteristicsof the suspending fluid, all tend to
push the particle into orbits which exhibit the minimum rate of energy dissipation.
This is but one example of a large number of similar studies undertaken by Mason
and his colleagues to relate the theoretical and experimental behaviour of rods, discs,
and spheres in shear flow and sometimes in the presence of other (e.g. electric) fields.
tlT
Fig. 15.9.2 Measured 41 as a function of time for a rod (q = 0.122) and a disc ( q = 4). The points
are experimental, and lines are calculated from eqns (15.9.3) and (15.9.4). (From Goldsmith and
Mason 1967 with permission.)
744 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
(1 5.9.5)
where h is the ratio of the viscosities of suspended phase to continuous medium and 4e
is the volume fraction of spheres. Thus q varies from the Einstein value for rigid
+
spheres (h + 00) to qo(1 @J as h approaches zero. This result is valid only if the
interface is clean. In the presence of surfactants (or surface active impurities) the
rigidity of the interface prevents transfer of momentum to the fluid inside the drops
and the system behaves as a suspension of rigid spheres (Nawab and Mason 1958).
The stress imposed by the shear field on a fluid drop can cause it to be deformed and
if the deformation is sufficiently pronounced the drop may split into two, usually with
formation of some very small satellite drops. The conditions under which such
‘bursting’ occurs are discussed by Goldsmith and Mason (1967) who also discuss the
way in which a uniform shear field can bend a flexible rod or a linear set of loosely
interconnected spheres.
When the shear field is non-uniform, as occurs in Poiseuille flow through a capillary,
the detailed rotational behaviour of the individual particles is similar to that observed
in Couette flow, if the flow is slow and the particles are small compared to the tube
radius. Isolated rigid spheres, for example, rotate with a constant velocity equal to half
the local shear rate just as they do in Couette flow (Exercise 15.9.3). Since the shear
rate varies linearly across the tube, from its value of zero on the axis, the rate of rotation
(my = y/2) likewise varies linearly with I as is shown in Fig. 15.9.3. Such spheres,
provided they are small enough, show no tendency to depart from the stream lines of
the fluid.
When the particle radius, R, becomes significant compared to its distance from the
wall ( R / ( u y )> 0.1 for spheres), the particle rotates more slowly near the wall (as also
MICRORHEOLOGY 1745
1.2
l.o[
rlR
Fig. 15.9.3 Measured angular velocity of neutrally buoyant small rigid spheres in a suspension
undergoing Poiseuille flow, as a function of the radial distance from the tube axis. The lines drawn
are calculated from the equation wo = 2Qr/na4 where the volume flow rate Q is doubled in going
from curve 3 to 2 and doubled again from curve 2 tol. (From Goldsmith and Mason 1967, with
permission.)
do rods and discs). Furthermore, for particle Reynolds numbers? (Section 4.8.1) even
as small as lop4,the interaction between the particle and the wall results in a migration
of the particle (Exercise 15.9.4) away from the wall, creating a particle free zone which
is of lower viscosity and which can lead to errors in the estimation of viscosity using the
capillary flow technique (Section 15.6). Even rigid spheres will migrate across the
stream lines to regions of lower shear rate, especially if the suspension concentration is
higher so that particle interactions are common: deformable particles or drops of any
size will migrate away from the wall, even at low concentrations, and will continue to
migrate, at gradually decreasing velocity, until they reach the axis. For more recent
developmets in this area see Davis (1993).
The next important step (Mason 1977) was the study of the trajectories of pairs of
particles in a shear field. This was discussed briefly in Section 4.10.5 where we introduced
the idea of open and closed orbits. According to the analysis of Batchelor and Green
(1972) two spheres approaching one another from infinity (in the absence of Brownian
motion, external forces or many-body interactions) are unable to penetrate into the region
of closed stream lines (Fig. 4.10.2 with Ri/a2 C 0). They are, therefore, unable to
approach closer than some minimum surface-to-surface distance, H, which depends upon
the relative size of the spheres. For spheres of equal size the minimum value of H/a
(where a is the particle radius) is 4.2 x so for the particles usually studied (a 3 pm),-
the minimum approach distance is well inside the double layer region. A very small
+ In this case the particle velocity is zero with respect to the surrounding fluid so the
Reynolds number of Section 4.8.1 does not apply. The internal (particle) Reynolds
number is now R2Ppo/qo where R is the radius.
746 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
4
Weak attraction
Fig. 15.9.4 Observed trajectories of equal sized latex spheres (in Poiseuille flow) in 50% aqueous
glycerol. (a) 4.0 p m spheres in 1 mM KCl. (b) 2.6 p m spheres in 10 mh4 KCl. (From van de Ven
1982 with permission.)
particle (<< 1 pm) cannot approach so close but even the smallest particle can move to
within a few hundred nanometres of this ‘large’ (3 pm) particle.
When double layer forces, Brownian motion, or third body interactions, are present
(van de Ven 1982) the trajectory is slightly modified and careful measurements (see
Takamura et al. 1981a,b) can distinguish the occurrence of double layer repulsion and
attraction (Fig. 15.9.4).Indeed, the measurements can be sufficiently precise to enable
an estimate to be made of the van der Waals constant and the effective surface potential
(Fig. 15.9.5). The former turns out to be concordant with theoretical estimates and the
latter agrees with the measured (-potential on the particles.
It is impossible to do justice here to the enormous body of work in this area, especially
that from Mason and his colleagues. The more recent material on interacting spheres is
reviewed by van de Ven (1982, 1989) who discusses in some detail the difference
between ‘primary’ and ‘secondary’ doublets. Spheres which are interacting at the
primary minimum can be distinguished from those interacting at the secondary potential
energy minimum (see Fig. 1.6.2) by examining the rate at which they rotate around each
other in the shear flow. It is even possible to distinguish between non-touching and
I I I 1
(4 v0=-45 mV
8- &=lo0 nm i o Z 1 J;
~(
I I I I
-5 0 5 10
I I I I
(b)
A
&=lo0 nm
*0(mV)
8 .- A=6x1dZ1J
4-
\tr '? .i:
\ I I 1 I I I 1
0 5 10
Reduced time
Fig. 15.9.5 Plots of trajectories of the spheres comprising a doublet showing the fit to theoretical curves as the various parameters are varied. Note that for
Doublet 8 (figs a, b, and c) a good fit is obtained with ho = 100 nm, @O = 4 5 mV, and the Hamaker constant, A = 6 x J. &(h) is the total interaction
here, calculated from DLVO theory, whilst ho is the retardation parameter of Fig. 11.4.1. (From Takamura et al. 1981a, with permission.)
748 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
d2
‘PI
-d2
d2
cp1
I I
-d 2
0 ,
-5 0
Reduced time
Fig. 15.9.6 Trajectories of colliding polystyrene latex spheres in 50% aqueous glycerol forming
permanent doublets. The upper curve is for a system with added cationic flocculating agent and the
asterisk marks the point at which it is suggested that the spheres are pulled into permanent contact
by a polymer bridge. The system shown in the lower figure has no added flocculant and the doublet
here is ascribed to van der Waals forces. (From Takamura et al. 1981b, with permission.)
Exercises
15.9.1 Integrate eqn (15.9.1) to establish eqn (15.9.2).
15.9.2 Establish that the maximum values of d&/dt occur at 41 = 0 and n if q < 1 (a
cigar shaped particle) and at n/2 and 3n/2 if q > 1 (a disc).
MICROSCOPIC BASIS OF RHEOLOGICAL MODELS 1749
15.9.3 Show that a rigid sphere in Couette flow rotates with a constant velocity equal
to half the local shear rate.
15.9.4 Estimate the particle Reynolds number for a system of neutrally buoyant
spheres of 100 nm radius suspended in water in a capillary viscometer of length
20 cm, for which the efflux time is 100 seconds. Problems arise in the
measurement of viscosity of such systems (due to creation of a particle free zone
near the wall) if this Reynolds number exceeds about lop5.At what particle size
does this occur?
t N ~IT~I~LZ~/KT (15.10.1)
and if this is long compared to l / p the system will show shear thinning behaviour
(Goodwin 1982). This effect has been demonstrated by Krieger who has written an
excellent review of the area (Krieger 1972) drawing attention to the utility of
describing the flow behaviour in terms of various dimensionless groups. In addition to
750 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
Fig. 15.10.1 Partitioning the shear stress between viscous flow and structural breakdown. On the
curvilinear part of the S-9 relation at, say, G, the stress S, required to support the viscous flow is
subtracted from the total stress to obtain the broken line OBC which therefore represents the stress,
ST involved in structural breakdown, as a function of 9.
for all values of i. and this is represented by the line OC. The total stress is then
(1 5.10.5)
where ST(00)is the value of STas + 00 and C is a measure of the curvature of the
S - function at low shear rates. For C = 0 the expression obviously simplifies to the
Bingham relation (eqn (3.4.23) and Table 15.1) with &(00) = Sp,whilst for ST(00) =
0 it represents simple Newtonian behaviour.
The next step was to find some rationale for the function C and this was provided by
Goodeve (1939) with his impulse theory which was modified by Gillespie (1960) to
incorporate the effect of Brownian motion. Essentially these developments sought to
describe the structural effects, ST( y), by examining the rate of 'link' formation and
breakdown during shear. A similar approach was adopted by Casson (1959), by Denny
and Brodkey (1962)' and later by Cross (1965). The arguments are treated at some
length by Sherman (1970) so we will describe only the Cross development as indicative
of the procedure.
Cross assumed that the suspension consisted of chains of particles with an average of
L links per chain. The links were formed by Brownian motion with a rate constant kz
and were ruptured by both Brownian motion (rate constant ko) and by shear (with rate
constant k l j " ) . The rate of change in the number of links is then given by:
where P is the number of particles. [Note that buildup of structure by the shearing
process (rheopexy) is denied in this approach.] In the steady state, dL/dt = 0 and so
+ v
L = k2P/(ko kly).If L = LOwhen = 0, then LO= k2P/ko and so:
( 15.10.7)
where K = kl/ko. In order to connect the number of links, L, to the viscosity, Cross
appealed to Bueche’s (1952) analysis of the viscosity of a polymer solution and wrote:
r]=r],+BL ( 15.10.8)
( 15.10.9)
Although couched in terms of the shear rate rather than the shear stress this is clearly
equivalent to eqn (4.10.15) or the Meter model of Table 15.1. The only drawbacks with
this approach are (i) there is insufficient theoretical basis for one to be able to estimate
what value the shear rate dependence parameter should have and (ii) there is no way in
which one can introduce the energy involved in breaking links. Nor is it possible to
incorporate any knowledge of the actual structure of the suspension into the analysis.
All suspensions are assumed to consist of chains of particles (rather like linear
polymers). It is clear from Fig. 15.2.3, however, that this type of equation can
represent the viscosity behaviour of some pseudoplastic systems over a very wide range
of shear rates; it has also been found to represent a very wide range of systems over a
rather more limited range of shear rates.
The relation between eqn (15.10.9) and the various models of Table 15.1, has been
explored by Oka (1971) and that work is described by Goodwin (1975). Starting from
the usual assumption that the number of bonds between particles decreases with
increasing shear rate or shear stress, Oka shows that
(15.10.10)
where k’, a,and B are material constants. Integration of this expression for n < 1 and
v
with = 0 when S = 0 gives the ‘generalized Casson equation’ (Exercise 15.10.2):
d = ko + kl+ (15.10.12)
i.
when a = 0, B = 0, and n = This equation has been used to describe the flow
behaviour of blood and other materials like printing ink.
MICROSCOPIC BASIS OF RHEOLOGICAL MODELS 1753
s= .[(+].Y+B
(15.10.13)
This will be indistinguishable from eqn (15.2.7) if a and /3 are not too large. For
suspensions with a yield value (S = SBwhen Y = 0, and n = l), integration of eqn
(15.10.10) produces the generalized Herschel and Bulkley equation (Exercise 15.10.3)
(15.10.14)
A rather more elaborate version of the kinetic equation of flow (15.10.9) has been
derived by Cooper et al. (1978). By analogy with the exponential effect of temperature
on rate constant, they propose that the rate constant for bond breakage is a similar
exponential function of the shear rate. This leads to a modified version of eqn (15.10.9).
rl-rlcc
VO - rlco
= [I + K F exp(-k/j)l-' (15.10.15)
which they found gives a much better description of their experimental data than the
Cross equation. It has, of course, an extra parameter and that will be expected to allow a
better fit.
There is, thus, no difficulty in providing a mathematical basis for the various
expressions in Table 15.1 but providing an underlying physico-chemical model which
relates the material parameters a, B, and n, to other known properties of the system is
not so simple.
Fig. 15.1 0 -dsic shear diagram for a coagulated PMMA latex sol in 31.2% glycerol water. yc
(=580 sP1) is the critical shear rate, above which the plastic (differential) viscosity, q p is
~
constant at 5.50 mPa s. The Bingham yield value, Se = 2.75 N m-2. (Particle radius = 125 nm,
ionic strength = 0.01 mol LP1,5 = -21 mV.)
with the plastic viscosity, VPL being determined by the volume fraction ofJilocs (&)
rather than that of the particles ($p). Thus, at low concentrations we might use the
Einstein equation (4.10.9) with 4 = q5f but at higher particle concentrations a more
appropriate expression would be (Krieger 1972):
VPL
[
= TO 1 -- (15.10.17)
where [q] = the intrinsic viscosity (defined in eqn (4.10.10) and equal to 2.5 for
spheres). The packing parameter p % 0.6-0.7 and is chosen to match the
asymptotically limiting v value when the particle concentration approaches close
pack. This expression is used to evaluate & and hence to obtain the floc volume ratio
C F (=~ 4f/&) which is a measure of the degree of openness of the flocs. The other
two parameters which characterize the high shear behaviour (Fig. 15.10.2) are the
critical shear rate, pc and the value of the shear stress (SB)obtained by extrapolating
the high shear behaviour down to p = 0. SBis called the Bingham yield value (even
though the system is not an ‘ideal’ Bingham material) and Firth and Hunter (1976)
have presented a model (the ‘elastic floc’ model) which makes it possible to describe
these three parameters in terms of the shear history (in particular the initial high shear
rate pmaxwhich determines the value of CFP)and the colloidal properties.
The important colloidal properties of the system are the volume fraction of particles
(4p),the particle radius ( r ) and the electrokinetic (or {-) potential of the particles.
According to the model, this latter parameter measures the magnitude of the maximum
attractive interaction force between two particles as a particleparticle ‘bond’ is
stretched by the shear field. The relation between 5 and the force is derived directly
MICROSCOPIC BASIS OF RHEOLOGICAL MODELS 1755
from the DLVO theory of colloid stability (Chapter 12). Since the system is
coagulated, there is no potential energy barrier separating the particles. All particle-
particle interactions are attractive and range from their maximum value (when { = 0)
to zero as I { I becomes sufficiently large to ensure dispersion (-45 mV for the poly-
(methylmethacrylate) latex particles used for most of the studies). The maximum
(attractive) force between two particles occurs when they are in their minimum
separation position (with surface separation dl) and is given by [eqns (11.3.15) and
(12.5.2)]:
aF c( y-0.4 (15.10.19)
and varies from -2 p m at ymaxto 4 p m at yC.Since the particles are about 0.1 p m in
radius, each floc contains many hundreds of particles which makes it possible to
describe the behaviour using rather crude averaging procedures. The model is not
expected to work for much larger particles (>1 pm) because the shear forces then
ensure that there are very few particles in each floc.
756 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
500
400
300
200
100
0
7 8 9 10 11 12
P"
1.2
1.0
0.8 a
0
0.6
0.4
B
0.2
0.0
7 8 9 10 11 12
PH
Fig. 15.10.3 The (a) measured and (b) normalized static shear yield stress (ty= So) of AKP-30
alumina suspensions as a function of both volume fraction and pH. 0 ,#J = 0.200; 0,#J = 0.225; A,
#J = 0.250; A, = 0.275; +,
#J = 0.300. (From Johnson et al. 1998, with permission.)
758 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
1.2
1.o
0.8
0.6
0.4
0.2
0.0
0 500 lo00 1500 2000 2500
c2b V 2 )
<*
Fig. 15.1 0.4 The normalized static shear yield stress (ty= So)as a function of for several A D -
30 alumina suspensions. The two central line plots correspond to d1 = 2.3 nm (- - --) and
dl = 2.4 nm ( ~ ) respectively. Line plots correspondingto dl = 1.4 and 3.4 nm are also given to
show the sensitivity of the gradient to the choice of interparticle separation. (4 values as for Fig.
15.10.3.) (From Johnson et al. 1998, with permission.)
As the pH is varied around the i.e.p. the data for a variety of volume fractions can be
placed on a master curve using a model which was developed by Kapur et al. (1997) (Fig.
15.10.3).The relative value of the yield stress, compared to the maximum value obtained
c2
at the i.e.p., can then be plotted against and the result is reasonably linear (Fig. 15.10.4).
For alumina the approach distance turns out to be about 2.4 nm compared with the value
of 0.6 nm found for PMMA, possibly an indication of the anisometric nature of the
particles. The same group has also shown that adsorption of bulky specifically adsorbed
molecules on the oxide surface not only shifts the i.e.p. but also pushes out the distance of
approach. The whole yield stress versus pH curve is therefore shifted to the new i.e.p. and
the maximum in the stress is reduced (Fig. 15.10.5.)
k1
unbroken bonds 2 broken bonds (15.10.20)
kz
MICROSCOPIC BASIS OF RHEOLOGICAL MODELS 1759
500 I I I I
v
400
300
cd
@
r"
200
100
0
2 4 6 8 10 12
PH
Fig. 15.10.5 The static yield stress (ty= SO) as a function of pH for a system with various additions
of phosphate. The phosphate is specificallyadsorbed to the surface and it moves the i.e.p and at the
same time increases the separation distance between the surfaces and so lowers the interparticle force
and energy. (From Leong et al. 1993, with permission.)
to represent the structural effects, they write for the rate of structural breakdown:
d(unbroken)
- = k',(unbroken)" - kz(broken)"
dt
and so
(15.10.2 1)
VO - Vco VO - Vco
wherep is a constant which reflects the 'shear sensitivity' of the material. Under steady
state conditions, dy/dt = 0 and, for n = m = 1, the Cross (1965) version of the Meter
model is generated (Exercise 15.10.4):
(15.10.23)
The novelty of the Denny and Brodkey approach lies in the solution for non-steady
state conditions. Choosing possible values for m and n (e.g. m = 2 and n = 1) they
obtain analytical solutions for the integrated form of eqn (15.10.21).Given the values
of the parameters m,n, y, qo, roo,and p derived from steady state measurements it is
then possible to evaluate k1 and k2 for a thixotropic material. For the heavy
thixotropic mineral oil used in their studies they found kl = 1.21 x s-' and k2
= 0.94 x lop3s-' so that the 'equilibrium constant' for eqn (15.10.20), K = k l / k 2 =
1.29 x lo-". The very low value of kl indicates why the material is thixotropic: the
rate of breakdown is so slow that it requires a considerable time to establish the steady
state structure at a given shear rate. The hopeful aspect of the exercise was the
agreement observed between the estimate of K from the time-dependent
(thixotropic) data and the value obtained in the steady-state (pseudoplastic) regime
(0.89 x lo-"). Unfortunately, the analysis is rather lengthy and requires a large
amount of good data to enable reasonable values of the parameters to be extracted.
The importance of the exercise lies in the demonstration of the link between
thixotropy and pseudoplasticity.
If the time for the establishment of the steady-state structure at a given shear rate is
long compared to the measuring time, then the system will exhibit thixotropy.
Pseudoplastic behaviour then appears as the limiting form of thixotropy when the time
between successive measurements is long compared to the relaxation time for the
structure in the system, which can then always exhibit its steady-state behaviour.
based on pair-wise additivity of the double layer repulsion forces. They show that G',
is related to the potential energy of interaction, VT by:
(1 5.10.24)
(15.10.26)
Values of the theoretically estimated G',/I+$ can be plotted against the measured G',
assuming a particular sort of packing (and, hence, a). The result is a good linear
correlation (Fig. 15.10.6) from the slope of which one can estimate I $d I e 50 mV.
This is a very reasonable value, in good agreement with (-potentials measured on
comparable sols. Some further experimental data is examined by Goodwin et al. (1982)
with similar very encouraging results.
~ o - ~ G(N: m-'
~ v2)
lvdz
Fig. 15.10.6 Plot of the experimental limiting shear modulus C6, against G&(theor)/& for
various volume fractions, 4, of a polymer latex in 5 x lop4 M NaCl solution. (&,(theor)/& values
calculated for Ka = 2.48, (11= 0.833, and R = 2a (0.7443. The slope correspondsto I @d I 50 mV.
(From Buscall et al. 19828, with permission, but corrected following recalculation by R.Williams
[1991 personal communication.].)
762 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
The experimental determination of Gb, in these cases was done with a pulse
shearometer (Rank Bros), the operation of which is described in an accompanying
paper (Buscall et al. 1982~).This instrument effectively measures the velocity, v , of a
shear wave (of frequency about 200 Hz) through the sample; the G’ value is then given
by G‘ = p v2 where p is the density. The same procedure was used by Hunter and
Everett (1988) in the experiments described at the end of Section 15.10.4.
The ordered latices used in the studies of Buscall et al. show quite pronounced
rigidity (G’ values of order 102-103 N m-’) and extremely high zero-shear viscosities
(- lo6 times higher than that of water). The rigidity can also be studied as a function of
compressive stress, as Barclay et al. (1972) and Buscall have done. Buscall et al. (19826)
show that the shear modulus and the bulk (compressibility) modulus, K , are closely
related to one another:
Exercises
15.10.1 Show that the time taken for a particle of radius a to undergo Brownian
diffusion through a distance equal to its radius is about 37cyoa3/kT (refer to
Chapter 1.)
15.10.2 Integrate eqn (15.10.10) with S = 0 when p = 0 to establish eqn (15.10.11)
and show how it reduces to eqn (15.2.7).Also show that for substances with a
yield value (S= SBwhen p = 0) eqn (15.10.10) with n = 0 integrates to give
the equation for Bingham flow.
15.10.3 Establish eqn (15.10.14) by the appropriate integration procedure and
determine the conditions under which it yields the equation for the Herschel-
Bulkley model.
15.10.4 Use eqns (15.10.21) and (15.10.22) for the Denny and Brodkey model to
establish eqn (15.10.23) when dy/dt = 0 and n = m = 1.
15.10.5 Establish eqn (15.10.26) for the shear modulus. (Note that +I$ is rendered as
@d in the original paper.)
References
Abdel-Khalik, S.I., Hassager, O., and Bird, R.B. (1974). Polym. Engr. and Sci.
14, 859.
Adams, M.J., Briscoe, B.J., and Kanjab, M. (1993).A h . Colloid Interface Sci. 44,
143-82.
REFERENCES I763
Arp, P.A. and Mason, S.G. (1977). 3.Colloid Interface Sci. 61, 21-43.
Barclay, L., Harrington, A., and Ottewill, R.H. (1972). Kolloid 2. 2. Polymere
250, 655.
Batchelor, G.K. and Green J.T. (1972).3 Fluid Mech. 56, 375,401.
Bird, R.B. and Curtiss, C.F. (1984). Physics Today 37, 3 W 3 .
Bird, R.B., Stewart, W.E., and Lightfoot, E.N. (1960). Transport phenomena.
John Wiley, New York.
Bird, R.B., Armstrong, R.C., and Hassager, 0. (1977). Dynamics of polymeric
liquids, Vol. I Fluid mechanics. John Wiley, New York.
Boger, D.V. (1977). Nature 265, 126.
Boger, D.V., Tiu, C., and Uhlherr, P.H.T. (1980). Introduction to theflow
properties of polymers. R.A.C.I. and Brit. SOC.Rheology, Australia.
Bueche, F. (1952).J. Chem. Phys. 20, 1959.
Buscall, R., Goodwin, J.W., Hawkins, M.W., and Ottewill, R.H. (1982a, b).
3. Chem. SOL.Faraday Trans. I 78,2873-88; 78,2889-99.
Calderbank, P.H. and Moo-Young, M.B. (1959). Trans. Inst. Ch. Eng. (London)
37, 26.
Carreau, P.J. (1968). Ph.D. Thesis, University of Wisconsin.
Casson, N. (1959). In Rheology of disperse systems (ed. C.C. Mills). Pergamon Press,
London.
Caw, W.A. and Wylie, R.G. (1961). Brit. 3. Appl. Phys. 12,94.
Christiansen, E.B., Ryan, N.W., and Stevens, W.E. (1955). A.I.Ch.E.3 1, 544.
Cooper, P.G., Rayner, J.G., and Nicol, S.K. (1978).3 Chem.SOL.Faraday Trans. I
74,785-94.
Cross, M.M. (1965). 3. Colloid Interface Sci. 20, 417.
Davis, R.H. (1993). Adv. Colloid Interface Sci. 43, 17-50.
de Haven, E.S. (1959). Ind. Eng. Chem. 51,63A46A; ibid. 813-16.
Denny, D.A. and Brodkey, R.S. (1962). 3. Appl. Phys. 33, 2269-74.
Dougherty, T.J. (1959). Ph.D. Thesis, Case Institute of Technology. (See Krieger
1972 and 1985).
Ekdawi, N. and Hunter, R.J. (1983). 3. Colloid Interface Sci. 94, 35541.
Eyring, H.J. (1936). 3 Chem. Phys. 4, 283.
Firth, B.A. and Hunter, R.J. (1976). J. Colloid Interface Sci. 57,266-75.
Fredrickson, A.G. (1964). Principles and applications of rheology. Prentice-Hall,
New Jersey.
Gillespie, T. (1960). 3 Colloid Sci. 15, 219.
Ginn, R.F. and Metzner, A.B. (1969). Trans. SOL.Rheol. 13, 48.
Goddard, D. and Miller, C. (1966). Rheol. Acta 5, 177.
Goldsmith, H.L. and Mason, S.G. (1967). The microrheology of dispersions. In
Rheology (ed. F.R. Eirich) Vol. 4, Chapter 2, pp. 85-250. Academic Press,
New York.
Goldsmith, H.L. and Mason, S.G. (1962). 3. Colloid Sci. 17,448.
Goodeve, C.F. (1939). Trans. Faraday SOL.35, 342.
Goodwin, J.W. (1975). In Colloid science Vol. 2, Chapter 7, pp. 246-93. Chemical
Society: London.
Goodwin, J.W. (1982). In Colloidal dispersions (ed. J.W. Goodwin) Chapter 8,
pp. 165-96. Royal Society of Chemistry, London.
Goodwin, J.W., Gregory, T., and Stile, J.A. (1982). Adv. Colloid Interface Sci. 17,
185-95.
Govier, G.W. and Aziz, K. (1972). Flow of complex mixtures inpipes. Van Nostrand-
Reinhold, New York.
764 I 1 5 : RHEOLOGY OF COLLOIDAL DISPERSIONS
Green, H. (1949). Industrial rheology and rheological structures, pp. 5242. John
Wiley, New York; Chapman and Hall, London.
Halmos, A.L. and Tiu, C. (1981). J. Texture Studies 12, 39.
Han, C.D. (1976). Rheology in polymer processing. Academic Press, New York.
Han, C.D., Kim, K.U., Siskovic, N., and Huang, C.R. (1975). Rheol. Acta 14,533.
Huilgol, R.R. (1975). Continuum mechanics of viscoelastic liquids. Halsted Press, John
Wiley and Sons, New York.
Hunter, R.J. (1982). Adv. Colloid Interface Sci. 17, 197-212.
Hunter, R.J. (1985). In Modern trends of colloid science in chemistry and biology (ed.
H.-F. Eicke) pp. 184-202. Association of Swiss Chemists, Birkhauser Verlag,
Basel.
Hunter, R.J. and Everett, D.W. (1988). Proc. Xth World Congress of Rheology.
Sydney.
Hunter, R.J. and Frayne, J. (1980).J. Colloid Interface Sci. 76, 107-15.
Jeffery, G.B. (1922). Proc. Roy. SOL.A102, 161.
Johnson, S.B., Russell, A S . , and Scales, P.J. (1998). Colloids and Surfaces Series A
PhysicoChem. and Eng. Aspects 141, 119-30.
Kapur, P.C., Scales, P.J., Boger, D.V., and Healy, T.W. (1997). AIChEJ. 43,
1171.
Krieger, I.M. (1972). Adv. Colloid Interface Sci. 3, 111-36.
Krieger, I.M. (1985). In Polymer colloids (ed. R. Buscall, T. Corner, and J.F.
Stageman). Chapter 6, pp. 21946. Elsevier Applied Science, London.
Krieger, I.M. and Dougherty, T.J. (1959). Trans. SOL.Rheol. 111, 137-52.
Krieger, I.M. and Elrod, H. (1953). J. Appl. Phys. 24, 134.
Krieger, I.M. and Maron, S.H. (1952).J. Appl. Phys. 23,147; (1954) 25,72; (1959)
30, 1705.
Leong, Y.K., Scales, P.J., Healy, T.W., Boger, D.V., and Buscall, R. (1993). J.
Chem. SOL.(Faraday Trans.) 89, 2473-8.
Lindsley, C.H. and Fischer, E.K. (1947). J. Appl. Phys. 18,988.
Lodge, A S . (1964). Elastic liquids. Academic Press, New York.
Mason, S.G. (1977). 3. Colloid Interface Sci. 58, 275-85.
Mason, S.G. and Bartok, W. (1959). In Rheology of disperse systems (ed. C.C. Mill)
Chapter 2. Macmillan (Pergamon), New York.
Meter, D.M. (1964). Thesis. University of Wisconsin, Madison, Wis. USA.
Mooney, M. (1931). J. Rheol. 2, 210.
Mooney, M. (1951). J. Colloid Sci. 6, 162.
Mooney, M. and Ewart, R.H. (1934). Physics 5, 350.
Nawab, M.A. and Mason, S.G. (1958). Trans. Faraday SOL.54, 1712.
Nguyen, QD. and Boger, D.V. (1983). Acta Rheologica 27, 321; 29(1985), 335.
Noll, W. (1958). Arch. Rat. Mech. Anal. 2, 197.
Oka, S. (1960). Rheology: theory and application (ed. F.R. Eirich) Vol. 3. Academic
Press, New York.
Oka, S. (1971).Jap. J Appl. Phys. 10,287.
Ostwald, W. (1926). Kolloid 2. 38, 261.
Philippoff, W. (1942). Viscositat der Kolloide. In Handbuch der Kolloidwissenschaft
9. Theodor Steinkopff, Dresden.
Prandtl, L. (1928). 2. Angew. Math. Mech. 8, 85.
Rabinowitch, B. (1929). Z. physik. Chem. (Leipzig) 145A, 1.
Rayner, J.G., Cooper, P.G., and Nicol, S.K. (1979). Rheol. Acta 18, 297-302.
Reiner, M. (1949). Deformation andjow. Interscience, New York.
Reiner, M. (1960). Deformation, strain, andjow. Interscience, New York.
REFERENCES I765
Refer to Figure 11.6.1 which represents the lowest energy vibrational mode which
can satisfy Maxwell’s equations in the vacuum between the two half spaces. The
electrical potential can be represented as a function of the form 4 (x,y, 2) $(t) = 4
(x,y, x)exp (-iot)and Maxwell’s equations in a charge-free region with c = 00
require that the spatial part of the wave function (4)must satisfy Laplace’s equation
(see Appendix A3):
024 = 0. (Al. 1)
A suitable solution of this equation is (Exercise Al.1):
(A1.3)
so that 4 will satisfy eqn (Al.1). The solution to eqn (A1.3) is:
f ( x ) = A exp(kx) for x < 0 and f ( x ) = Bexp(-kx) for x > L.
and, between the plates:
f ( x ) = Cexp(kx) + Dexp(-kx) for 0 < x < L.
(These functions are chosen to ensure thatf(x) is well behaved for large x.)
The boundary conditions at the respective surfaces require that both the potential,
4, and the dielectric displacement (Section 3.2)’ (= E(w)E = -ed$/dx), must be
continuous. Hence (Exercise Al.2):
A-C-D=O
exp(-kL)B - exp(+kL)C - exp(-kL)D = 0
(Al.4)
+
C I A- €oC COD= 0
€1 exp(-kL)B +
€0 exp(kL)C - €0 exp(kL)D = 0.
767
768 I APPENDICES
These equations are soluble only if the determinant of the coefficients is zero and this is
so only if (Exercise A1.2):
(A1.5)
I
Exercises
A1 . I
Al.2
Verify that eqn (A1.2) is a solution of eqn (Al.l) iff(x) is defined by eqn
(A1.3).
Establish the equations (A1.4) using the stated boundary conditions and hence
obtain the dispersion relation (Al.5) by evaluating the relevant determinant. I
APPENDIX A 2
Cauchy’s integral theorem states that the value of a functionf(w) of a complex variable
+
at a point (say a = q Zt) can be obtained from its values on the curve C using the
expression:
(A2.1)
where C is any closed curve surrounding the point a, and f (w) is analytic in the region
containing C.
The solutions, wj of the dispersion equation (1 1.6.3) must satisfy:
a(@)
= &(w - Wj) = 0 (A2.2)
adsorption 259
at gas-liquid interface 63
at gas-solid interface 277
at mercury-solution interface 309 482
at solid-liquid interface 287 356 482
694 709
chemical 336 499 503
521
competitive 287
discreteness of charge in 338
equilibrium 278 290 300
from dilute solution 288
inner (Stern) layer 490
mechanism 277
negative 349
of anions on mercury 336f
of complex adsorbates 298
of counterions, on oxides 503
of hydrolysed metal ions 513
of multivalent ions 509
of organics on mercury 341f
of p.d.i. 344 356 485
of polymers (neutral) 40 293
of surfactant 518
on porous solids 284
physical, energy of 277
positive, for area determination 348
potential, chemical 327 337
preferential, on crystals 203
specific, detection 329
bilayers 16f
(s.a. vesicle)
bile salts, aggregation of 447
binding constants see equilibrium
constants
binding of ions 504
to micelles 459t
Bingham flow 153 717 730
733e 751f
biotite 21
biphenyl, residence time in micelles 470
block copolymers in stabilization 294
blood and blood substitutes 5
Bohr theory of hydrogen 536
boiling chips 76
boiling point, influence of bubbl size 79t
Boltzmann constant, estimation of 31 120
Boltzmann equation 319
boundary condns. 767
in fluid mech. 179 398 403
666
Boyle point 635e
(s.a. theta-point)
Bragg diffraction 672
Bragg-Williams equation 297
Bredig arc 7
bridge functions 664
bridging by polymers in flocn 632
Brij 35 micelles
bromide ion (s.u. halides)
capacitance (Cont.)
of Agl interface 352 493
impedance of 132e
capacity of system to do work 780
capillary condensation 87
electrometer 313
flow in — 173 737
hydrodynamic fractionation, see CHDF
rise 73f 84
in powder 91
viscometer (s.u. Ostwald)
wetting of 84f
capillary pressure 84
effect on melting point 78
captive bubble method 606f
carbon tetrachloride, viscosity of 166t
carboxylates
c.m.c. of 439t
dimers of 447
Casson eqn 752
cation exchange capacity, see c.e.c.
Cauchy plot 569
Cauchy’s theorem 560 768
c.c.c., see critical coagulation concn
c.e.c. 24e
(s.a. ion exchange)
cell,electrochemical 316 347
models 423
centrifugal forces in Couette 172
in fractionation 249
coagulation 34 604
by electrolyte 36
by potential control 35
fractals produced in 625
in shear 619
kinetics of 616
coal slurry, flow of 717 731
coalescence of emulsion drops 18
coefficient of variation 220
cohesion between surfaces 92e 101
(s.a. work of—)
collision frequency 617
colloid vibration current (potential) 422
colloids 2
association 10
classification 2 5
examples 3t 6
monodisperse 8
preparation, chemical 6
significance of 4
size limits 2
stability 33
(see coagulation)
structure 669
(s.a. dispersions)
combinatorial entropy 297
common intersection point 349 510
compact double-layer 324
compensation plot for micellization 456f
complexes, formn. on oxides 487 505
(s.a. equilibrium constant)
compliance, creep 150
components influence on
degrees of freedom 70
composites 3
compressibility factor 656
computer modelling of surface
dissociation 503
concentration
effect in electrokinetics 423
on surfactant props. 14
gradient, detection of 230
variation across surfaces 59f 614
condensation, capillary 89f
from vapour, effect of curvature 74
in dispersion preparation 7
condenser, see capacitance
conductance, electrical 379 420
of plugs 427
cone and plate viscometer 175 739
congruence test 493f 495f
conservation
of ions, in electrokinetics 397
of matter in fluid mechanics 158
conservative field 776
constitutive equation 147 728
contact angle 111f
and wetting 100
determination of 112
on rough surfaces 108
theoretical estimates 576
time-dependence of 112
contact dissimilarity energy 297
contact value theorem 614
cylinder
double layer around 369
micellar 16f
cylinder-in-cylinder viscometer 170
cylindrical metal analyser 267f
diffusion (Cont.)
spheroids, rotational 185
translational 187 617
gradient in conc suspensions 181
diffusivity, see diffusion coefficient
dilatant flow 153 720
dimension, fractal 626
dimensionless variables 322f 407 412
680 750
dimer formation by surfactants 447
dipole-dipole interaction 535
dipole moment, induced 536
dipole potential, see chi-potential
dipole strength 189 193 421
disc centrifuge 229
discreteness of charge effect 338
discs (s.a. spheroids)
light scattering from 141f
rotational diffusion 185 743f
disperse phase 2
dispersing agent, see surfactant
dispersion force theory 533 552 767
(s.a. London, van der Waals)
dispersion medium 542
dispersion relation 555 558
roots of 559 767 768
dispersions 2
concentrated 181 193 254
423 698 760
mixed 704
model 8
preparation 6
dispersions (Cont.)
stabilization 33 582
structure of 669 760
transport properties of 188 749
(s.a. colloids)
displacement, mean square 31
displacement vector, see dielectric
dissimilar surfaces, interaction of 594
dissipation factor 237
( s.a. damping, viscosity, friction)
dissociation, degree of, see equilm. const.
dissymetry ratio 140 141f
distance of approach, s.u critical sepn
distribution function 216 663
adsorbed polymer 299
bimodal 224f
Gaudin-Schuman 224
ions near surfaces 350f
log-normal 217 223
Lorentz 241f
molecules at surface 614f
normal (Gaussian) 221
pair (s.u. correlation function)
particle orientation 185 742
quencher in micelles 472e
Schulz 225
size 213
disturbance velocity in sedimentation 179
div (divergence) 773
grad 324 774
of a tensor 777e
dividing surface between phases 59f
DLA 625
DLVO theory 39f 581 747f
761
experimental confirmation 604
in suspension flow 755
DME 310f
capacitance 332
Dodecyl compounds 521t 696f
in micelles 441f 459t
domains of close approach 630
Donnan e.m.f. 354
Doppler line broadening 241f 393
dot product of vectors 772
double bonds, effect on c.m.c. 438
double layer, see electrical––
doublets of spheres in flow 746
Dougherty-Krieger relation 425f
drag (viscous), force 28 121 168
drainage of wetting films 606
drop of liquid, flow in 744
free energy of formation of 95
motion in flow 744
on surface, shape see sessile & pendant
vapour pressure of 74
dropping mercury electrode, see DME
drop weight in surface tension measure 312
Dukhin number 407
dye, fluorescent, in micelles 470
dynamic light scattering 241 392
mobility 254
SIMS 268
electrolytes, indifferent 36
influence on attractive forces 571
simple, effect on coagulation 603
(s.a. ions)
electromagnetic theory 124 552
electrometer, Lippmann 313
electron beam 262
binding energy 267
electrochemical potential of 307
response of-to a.c.field 263f
electron loss spectroscopy 263f
electron microscope 207
electron probe microanalysis 209
electro-osmosis 375 390
in plugs 377
electrophoresis 380 390 412
electrostatic patch interaction 633
potential of a phase 305
ellipsoidal particles (s.u. spheroid)
elliptic integrals 587
Ellis model of shear 719t
e.m.f, Donnan 354
(s.a. potential, electrical)
emulsification, spontaneous 18
emulsions 16
sedimentation of 119
viscosity 744
end correction in viscometry 173
end-to-end length 295
energy dissipation see viscosity
interfacial, see surface energy
flux (Cont.)
in sedimentation 182
of vector field 773f
foam 3 65
fog, formation of 96
force between macrobodies 541
particles 38
spheres 646
two plates 39f
body, in sedimentation 179
diffusion 28
image, of ions 339
long- and short-range 160
of adhesion, measurement 99
on fluid element 160
van der Waals 39f
(s.a. dispersion force)
Viscous 28 117f 121
167
forces, additivity of 539 654
form factor 139 659 691
708
formamide, effect on aqueous micelles 442
micelles in 436
Fourier transform 655 660 673
777 708
fractal nature of aggregates 625
fractionation, see CHDF
free energy, definition 780
elastic, in polymer mixing 631
electrical 584
gradient (Cont.)
of chemical potential 28
of concentration 29
grad of a scalar field 770
of a vector 776
graft copolymers in stabilization 628
Grahame isotherm 337
Grahame model of double layer 326f
graphite 292
gravitational field 116
forces in soap films 606
in moving fluid 169
thermodynamic effects of 782
group, hydrophilic or hydrophobic 10
surface, dissociation of 356 501
growth of crystals 93
Guinierp lot 140 692
gyration, radius of 140 693
Ludox 10
s.a. silica
luminescence quenching 462
lyophilic solute 441
effect on surface tension 68f
lyotropic series 615
micellization (Cont.)
in non-aqueous media 457
phenomenology of 438
microemulsions 17f
microfluidity in micelles 470
micropores 284
micropotential for adsorbed ions 339
microrheology 741
microscope 204
atomic force, see AFM
electron 207
(s.a. SEM and STM)
field emission 264f
field ion 262
optical 205
scanning, see SFM & STM
total internal reflectance, see TIRM
ultra- 206f
microtubule 15
microviscosity 470
microwaves 567
Mie scattering 141
mill, colloid 7
mineral, clay —, (s.u. clay)
ores (s.u. flotation)
oxides (s.u. oxides)
MINEQL 503
mixtures, coagulation of 596
mobility, dynamic 254 422
electrophoretic 380
mode, vibrational, of e.m. radiation 560
potential energy
barrier to coagulation 621
of interaction 585 639f
approximation formulae 591
total 601
potential, gravito-chemical 119
potentiometer in electrochemistry 310
pottery, see ceramics
powders, determ. of contact angle of 91
power, dissipation 132e
law model of shear 720f 733e 736f
753
precision of distribution 27f
pressure average, in flow 165 189
difference across surface 52
drop in capillary 173 379 736
dynamic 51f
effect on micellization 440
effect on surface tension 49
excess in small solid particles 80
hydrostatic 173 313 607
inside bubble or drop 52
molecular basis of 654
static 51f
stat. mech. calcn. 665
primary minimum 602f 610
yield value 718 756
primitive model 679
principal radii of curvature 53
probability
for particle pairs (s.u. correlation)
paper 223
Stirling approximation 26
STM 268
Stokes equations 178
in electrokinetics 397
Stokes settling radius 116
strain 145f
complex 152
rate 151
see shear rate
streaming current 389 420f
potential 378 389
streamlines 117
orientation of particles along 190 741
stress 144
applied 146f 716
normal 168 715
oscillating 151
relaxation 148f 149f
tensile 86
stress in a moving fluid 145f 160
at equilibrium 162
stress tensor 161
macroscopic 166 188
in suspensions 188
structural effects in flow 751
forces in stability 602f
structure factor, static 657
makers & breakers 442
measurement of 669
suction pressure 86
sulphate ion as head group, see SDS
negative adsorption on Agl 351
surfactant (Cont.)
non-ionic, adsorption 526
scattering 700f
susceptibility 655
suspension effect 354
suspension medium, effect on van der
Waals force 542
suspensions: see colloids; dispersions
swelling of clays 608
symmetry, axial, of surfaces 104
of stress tensor 716
syneresis 741
UCFT 629t
ultracentrifuge 230f
ultrahigh vacuum 261f
ultramicroscope 206f
ultrasonic absorption 250
ultrasonication 7
ultraviolet interpolation 568
vacuum 261f
valency, counterion 37t 321
influence on coagulation 603
vanadium pentoxide 204
vanadyl ion, mobility on micelles 472
van der Waals equation 535
and Boyle temperature 628
van der Waals force 39f 533
calculation of 563 571 769
contribution to work of adhesion 574
effect of retardation 547
exptl study of 601 611f
relation to hydrophobic force 615f
van Hove correlation function 662
van Kampen method 552
vapour, definition 88
vapour pressure, effect of temperature 76
lowering, inside bubbles 75
of small particles 74
variance 221
vector calculus 770
vector field 770
vector, unit 160
XPS 266
X-ray adsorption in sedimentation 226
X-ray photoelectron spectroscopy 266
xylose, effect on micelles 442